Art and Abstract Objects

Art and Abstract Objects
This page intentionally left blank

Art and Abstract
Objects
EDITED BY
Christy Mag Uidhir
1
3
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
# the several contributors 2012
The moral rights of the authors have been asserted
First Edition published in 2012
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
British Library Cataloguing in Publication Data
Data available
Library of Congress Cataloging in Publication Data
Data available
ISBN 978–0–19–969149–4
Printed in Great Britain by
MPG Books Group, Bodmin and King’s Lynn
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Contents
List of Contributors vii

Acknowledgements viii
Introduction: Art, Metaphysics, and the Paradox of Standards 1

Christy Mag Uidhir
I. General Ontological Issues

1. Must Ontological Pragmatism be Self-Defeating? 29
Guy Rohrbaugh
2. Indication, Abstraction, and Individuation 49
Jerrold Levinson
3. Destroying Artworks 62
Marcus Rossberg
II. Informative Comparisons

4. Art, Open-Endedness, and Indefinite Extensibility 87
Roy T. Cook
5. Historical Individuals Like Anas platyrhynchos and ‘Classical Gas’ 108
P.D. Magnus
6. Repeatable Artwork Sentences and Generics 125
Shieva Kleinschmidt and Jacob Ross
III. Arguments Against and Alternatives To

7. Against Repeatable Artworks 161
Allan Hazlett
8. How to be a Nominalist and a Fictional Realist 179
Ross P. Cameron
vi CONTENTS
9. Platonism vs. Nominalism in Contemporary Musical Ontology 197

Andrew Kania
IV. Abstracta Across the Arts

10. Reflections on the Metaphysics of Sculpture 223
Hud Hudson
11. Installation Art and Performance: A Shared Ontology 242
Sherri Irvin
12. What Type of ‘Type’ is a Film? 263
David Davies
13. Musical Works: A Mash-Up 284
Joseph G. Moore
Index 307
List of Contributors
ROSS P. CAMERON, Leeds University

ROY T. COOK, University of Minnesota, Twin Cities
DAVID DAVIES, McGill University
ALLAN HAZLETT, University of Edinburgh
HUD HUDSON, Western Washington University
SHERRI IRVIN, University of Oklahoma
ANDREW KANIA, Trinity University
SHIEVA KLEINSCHMIDT, University of Southern California
JERROLD LEVINSON, University of Maryland, College Park
P.D. MAGNUS, University at Albany, State University of New York
CHRISTY MAG UIDHIR, University of Houston
JOSEPH G. MOORE, Amherst College
GUY ROHRBAUGH, Auburn University
JACOB ROSS, University of Southern California
MARCUS ROSSBERG, University of Connecticut, Storrs
Acknowledgements
The idea for this volume grew out of a panel on the titular subject matter
put together by myself, Roy T. Cook and Marcus Rossberg and presented
to the 2009 Annual Meeting of the American Society for Aesthetics held in
Denver, Colorado. I count myself stupendously lucky to have been able to
amass such a Murderers’ Row of frighteningly talented philosophers to
contribute to this volume; I am grateful to them all for their patience in
working with a first-time editor. I’d also like to extend special thanks to
Wylie Breckenridge for lending a critical eye, to L.A. Paul and Amie
Thomasson for their words of encouragement and support, to Donald
Baechler for so generously providing the fantastic cover image, and to
Peter Momtchiloff for his tireless editorial assistance.
Introduction: Art, Metaphysics,
and the Paradox of Standards
CHRISTY MAG UIDHIR*
I think it safe to say that philosophical aesthetics has had a less than stellar
record of its principal work being actively and substantively informed by
work in philosophical areas outside itself.1 Although I’m not entirely sure
what might fully explain the fact that aesthetics has for so long cultivated
a disturbingly insular character (or at least why it has achieved such a
reputation), I am quite certain that this protracted insularity has not only
effectively hobbled progress and productivity within philosophy of art
but also ostensibly poisoned any substantive and informative relationships
aesthetics might cultivate with outside areas (further fueling moves toward
insularity).
Of course, while I may share—or at least regard as neither hasty
nor terribly uncharitable—the view that aesthetics has to some extent
heretofore been a comparatively dim, unproductive, and deleteriously
insular area of philosophical enquiry, this should by no means suggest that
I also share the sadly not altogether uncommon outside sentiment that
* I would like to thank P.D. Magnus and L.A. Paul for helpful comments on earlier drafts.
1 After all, one needn’t look too hard to find a standard position within philosophical aesthetics that
runs directly counter to the relevant standard (if not received) positions in some other field (e.g. dominant
theories of art interpretation sharply diverging from the dominant theories of interpretation within
philosophy of language, well-established conceptual accounts of art and the aesthetic being predicated
upon highly controversial if not largely discredited theories of concepts within philosophy of mind,
object-kinds considered standard for art ontology being conspicuously absent from and utter alien-
looking within contemporary metaphysics). By contrast, consider the comparative ease in showing
the principal work within meta-ethics routinely and productively to engage with, and be informed by,
areas well outside itself (e.g. metaphysics, epistemology, philosophy of mind and language, philosophy
of science).
2 CHRISTY MAG UIDHIR
aesthetics is an at best second-rate philosophical field. In fact, I think such

attitudes more often than not reflect in their holders either a thinly
veiled contempt for the subject matter of philosophical aesthetics—and by
extension the field itself—or at least a casual indifference towards the results
of any philosophical enquiry conducted therein. I find the prevalence of
such attitudes incredibly frustrating not just because I self-identify as
a philosopher of art. To judge aesthetics but philosophical trifle requires
either an utterly brazen and reckless philosophical arrogance or an abyssal
(and a touch ironic) degree of philosophical ignorance—neither of which
any self-respecting philosopher ought to abide. Likewise, to grant aesthetics
philosophical legitimacy yet nevertheless remain indifferent to the results
of its enquiries is to be guilty of the very same sort of counterproductive
and philosophically irresponsible neglect of which aesthetics has been
traditionally accused.
To be sure, I do not expect every philosopher to find art and its relata as
philosophically captivating as I do; however, I do nevertheless expect them
to take philosophy of art seriously. The problem, of course, is that
the farther removed philosophers of art view their enterprise to be from
philosophy’s putative core, the fewer number of areas there are for those
operating within that core to seriously and productively engage with
philosophy of art. So, for the demand that philosophy of art be taken
seriously to be other than empty requires that philosophy of art, at least
in part, adopt a broadly philosophically informative approach to its central
issues (e.g. shaping and testing putative art-theoretic commitments against
the standard models employed within those relevant intersecting areas of
enquiry such as philosophy of mind and language, metaphysics, philosophy
of science, ethics, and so forth). The burden, however, ought not
exclusively rest with philosophical aesthetics.
Although this volume chiefly focuses on the relationship between
philosophy of art and contemporary metaphysics with respect to the overlap
issue of abstracta, my aim is that what follows should nevertheless provide,
at least in some small measure, a general methodological blueprint
from which both those from within aesthetics and those from without
can begin building responsible, and therefore mutually informative and
productive, relationships between their respective fields.
INTRODUCTION 3
I. Two Opposing Views

When philosophical aesthetics tends toward insularity, we shouldn’t be
surprised to find standard art-ontological categories incongruous with
those standardly employed in contemporary metaphysics. Of course,
when contemporary metaphysics tends to ignore aesthetic and art-theoretic
concerns, perhaps we likewise shouldn’t be surprised to find the climate of
contemporary metaphysics inhospitable for a theory of art.2 While this may
seem to suggest at least a prima facie tension between our basic art-theoretic
commitments considered from within philosophical aesthetics and
our standard ontological commitments considered from without, I think
any perceived tension or antagonism largely due to metaphysicians and
aestheticians at least implicitly assuming there to be but two available
methodological positions with respect to the relationship between meta-
physics and philosophical aesthetics in the relevant overlap areas. I call these
two opposing views Deference and Independence.
The Deference View: In all cases of relevant overlapping areas, aesthetics

ought to defer to contemporary metaphysics.
The Independence View: Art-ontological categories cannot be (or at
least we shouldn’t expect them to be) adequately carved out using only
the tools provided by contemporary metaphysics.
According to deference, the tools provided by contemporary meta-
physics are (at least prima facie) both necessary and sufficient for carving
out art-ontological categories, and as such, there ought be no sui generis art-
ontological kinds. So, for example, repeatable artworks can be plausibly
construed as abstract objects only if doing so entails no addition of a sui
generis abstract kind. According to independence, however, art onta need
not and do not neatly conform to the ontic kinds standardly in play for
contemporary metaphysics, such that, philosophers of art ought to expect
that adequately, let alone fully, capturing the operative constraints, interests,
2 For example, if art-theoretic considerations originally prompt metaphysical enquiry into the nature
of works of certain sort (e.g. music, film, literature), then the ontological conclusions drawn from such
enquiry ought to be consistent with those art-theoretic considerations. So, should these art-theoretic
considerations be promptly forgotten or ignored, we shouldn’t be at all surprised when the resultant
metaphysical enquiry answers with ontic models for works of that certain sort being at least indifferent to,
if not fundamentally at odds with, being an artwork.
practices unique to art-relevant domains requires carving out (perhaps

from whole cloth) befittingly unique (sui generis) ontological categories.
I suppose that one quite understandably might think preference for either
deference or independence tracks preference for either revisionary or
descriptivist art ontology respectively. For example, should we approach
the ontology of art thinking that contemporary art practices and conven-
tions are such that they cannot be adequately captured using only
the ontological tools provided by contemporary metaphysics, we might
understandably also think that the extent to which one endorses deference
is the extent to which one endorses a revisionary art ontology. So, perhaps
those already harboring strong descriptivist loyalties may understandably
thereby find themselves more inclined toward independence. After all, to
the extent philosophy of art seems to be relegated to the margins
of philosophy, philosophers of art needn’t be concerned with garnering
approval from the implicitly if not explicitly uninterested and distanced
core (no quantification without representation!). Therefore, philosophers
making art-ontological enquiries should take themselves primarily tasked
with engaging in conceptual analysis aimed at best capturing the art-relevant
practices and conventions and as such needn’t be concerned whether or
not the putatively available art-ontological terrain can be located on any
contemporary metaphysical map.
By contrast, should we approach the ontology of art with the default view
of the sufficiency of the metaphysician’s ontological toolbox and the con-
tents therein, we might understandably view endorsing independence as
simply trading one revision for another, protecting art ontology against
being revisionary with respect to art in a manner ultimately ensuring its
being revisionary with respect to ontology. That is, independence may well
be better suited to preserving our art practices and conventions (e.g. their
semantic or referential coherence) but only in virtue of entailing some
metaphysically revisionary program or other (e.g. positing otherwise
sui generis ontic models). If fitting artworks into the world in a manner
consistent with their surrounding practices and conventions requires
revising the world, at least in terms of the categories into which its inhabi-
tants may be placed, then we ought to expect any art-ontological account
to be to that extent revisionary, such that, insofar as one wants artworks
in the world, any art-ontologically required revision had better aim at
(and be wholly exhausted by) those art-relevant surrounding practices and
INTRODUCTION 5
conventions. Contemporary metaphysics being unable to provide ontic

models sufficient for descriptivist art ontology results not from some failure
within contemporary metaphysics but rather from a failure within the
philosophy of art: either of the descriptivist project itself or of art-realism
tout court. Simply put, regardless of the descriptivist’s concerns, merely
insofar as we have any non-passing sympathy for realism about artworks,
perhaps we ought to endorse deference.
Notice, however, at least in principle, that endorsing either deference
or independence itself says nothing about whether one prefers their
art-ontology revisionary or descriptivist. For example, one could coherently
both be a descriptivist and adopt deference; however, in so doing,
one presumably would find any required revisions to art practice and
convention to count against art-realism itself. Likewise, I suppose one
could coherently both be a revisionist and adopt independence; however,
in so doing, one presumably would regard any required metaphysical
revisions to be motivated by purely philosophical concerns (e.g. parsimony,
logical consistency, art-theoretic coherence) rather than the preservation
of the relevant surrounding practices and conventions. To be sure, the
principal conflict between deference and independence may be best
located at the level of the principal conflict between revisionary and
descriptivist art ontology. However, what matters for my purposes here
not only has little to do with the revisionary/descriptivist debate, but,
is ultimately philosophically and methodologically prior to such debates.
That is, I am principally concerned with the view we ought to take with
respect to the relationship between contemporary metaphysics and the
philosophy of art.
Given the above, I suppose that any perceived tension between
philosophical aesthetics and contemporary metaphysics looks quite simple
to explain: Most metaphysicians at least implicitly (or dispositionally) align
with deference while most aestheticians at least implicitly if not explicitly
align with independence. The problem with this dichotomy I take to be
the following. The degree to which philosophical aesthetics aligns with
independence is the degree to which philosophical aesthetics distances
itself from contemporary metaphysics and thereby the degree to which
philosophical aesthetics insulates itself from the core of philosophy
(i.e. neither properly informing nor being informed by that core). By
contrast, the degree to which philosophical aesthetics aligns with deference
is the degree to which philosophical aesthetics holds its development and

productivity hostage to a field in the main (if not wholly) unconcerned with
art-theoretic considerations and thereby the degree to which philosophical
aesthetics secures its status as a second-class philosophical field (i.e. one to
which the core presumably contributes but from which presumably such
contributions fail to be reciprocated). So, should metaphysicians largely
ignore philosophy of art (though perhaps recently less so at least for a few
peripherally related areas), philosophers of art may understandably see little
reason to take their metaphysical strictures seriously for art ontology.
Consequently, should the metaphysician take the occasional look at
the art-ontological goings on within philosophy of art, she’ll likely be
confronted by all manner of non-standard ontic models if not also perhaps
a few outright metaphysical monstrosities (e.g. causally efficacious abstracta,
created things that necessarily cannot be destroyed, haphazardly stitched
together object-kinds) and as a result see little reason to regard the philosophy
of art as anything more than an island of misfit ontologies—thus beginning
the cycle anew.
All things considered, I think philosophy of art does itself no favors either
from without by endorsing independence or from within by endorsing
deference. For interactions and exchanges between metaphysics and
philosophy of art to be more than sporadic and able to yield broadly
productive results demands a re-evaluation of the way in which we view
their relationship.
II. Art-Abstracta and the Paradox of Standards

In what follows, I target as an example the tension between philosophical
aesthetics and contemporary metaphysics with respect to the broad account
of abstract objects as standardly employed respectively therein. I show that
the issue of art-abstracta appears to give rise to what I refer to as a paradox
of standards: an art-ontological assumption; a metaphysical assumption;
and an art-theoretical assumption, that though each be standard, if not
foundational, within their respective domains of enquiry, when taken
together nevertheless form an inconsistent set.
INTRODUCTION 7
I. There are such things as art-abstracta.

II. Abstracta are non-spatiotemporal and causally inert.
III. An artwork must be created.
What I call the paradox of standards arises when from any two of the above
standard views we expect to arrive at the standard third view only to arrive
instead at its denial (either explicitly as such or implicitly via or some
inconsistent view non-standard within the same domain). Simply put, the
paradox of standards occurs when we expect standard and standard to begat
standard but instead find it to birth only non-standard.
Note, however, that my analysis of the constituent assumptions largely
concerns their status as standard rather than their individual truth or falsity.
That is, I take the philosophical point of the paradox of standards to be that
the issue of art-abstracta represents an informatively illustrative point about
the relationship between the philosophy of art and metaphysics, and not just
as a potential source of philosophical tension but also as a potential source
of productive philosophical exchange.
To that end, I provide a brief sketch of the three standard assumptions
relevant to the paradox of standards. I then discuss the means by which it can
be resolved (and the consequences thereof) for both those adopting defer-
ence and those adopting independence, showing the available methods of
resolving the paradox according to either view incapable of doing so in a
manner commensurate with a minimally responsible art-realism. From this,
I suggest an alternative account as how best to view the relationship between
philosophical aesthetics and contemporary metaphysics (what I call the reci-
procity view) and then discuss what consequences the relationship being so
viewed may have for the ontology of art. However, before I begin, perhaps
I should say a bit more about why I think it productive and illuminating to
view the issue of art-abstracta in terms of what I call the paradox of standards.
Art ontology looks to be the natural intersection between philosophy of art
and metaphysics. Any general art-theoretic enquiry can be broadly under-
stood as investigating the nature of art in terms of the conditions under which
something is and is not (or can and cannot be) an artwork. Likewise, any
general ontological enquiry can be broadly understood as investigating the
nature of the world in terms of the categorization of its onta. Presumably then,
any art-ontological enquiry can be broadly understood as investigating the
nature of the artworld in terms of the categorization of its onta: artworks (i.e.
things in the world satisfying (or having the capacity to satisfy) the conditions
for being art, whatever those may in fact be). From this, presumably one
would likewise expect the standard assumptions operative for any broad art-
ontological enquiry to be derived from, predicted by, or at least consistent
with the standard assumptions operative for any broad art-theoretic or general
ontological enquiry. After all, if the ontology of art lies at the intersection of
philosophy of art and metaphysics, then we should expect to locate the
standard views within the former where the standard views within the latter
two intersect—from the standard views within any of the two, we should be
able to arrive at some standard view within the third. The paradox of
standards (or at least the appearance of such) occurs when from the standard
views within any two domains we seem unable, upon pain of inconsis-
tency, to arrive at anything other than some decidedly non-standard view
within the third domain. Of course, the precise implications of the
paradox of standards (and the available resolutions thereof ) vary according
to the way in which one views the relationship between philosophical
aesthetics and contemporary metaphysics.
Consider again the following assumptions that within their respective
domains of enquiry appear to be not just standard but also at least prima facie
plausible (if not prima facie evident).3
I. There are such things as art-abstracta.
II. Abstracta are non-spatiotemporal and causally inert.
III. An artwork must be created.
Let’s begin by considering (I): The standard art-ontological assumption that
there are such things as art-abstracta.4
3 Predicated on these basic, standard assumptions are several further specified and standard positions,
out of which one can further construct several inconsistent sets. The main point here is that this threat of
inconsistency in standards can be avoided only by either outright denying one of the more basic, standard
assumptions at issue or implicitly doing so via adopting some non-standard position that itself cannot be
coherently predicated upon one of those more basic, standard assumptions.
4 The basic art-ontological argument for this I take to be pretty straightforward:
There are such things as artworks (Art-Realism).
Artworks are either repeatable or non-repeatable.
Repeatable artworks (or at least those of certain sorts) cannot be coherently or viably construed as
concrete things.
So, if there are such sorts of artworks, then those artworks must be abstract things.
There are such artworks.
So, there are such things as art-abstracta.
INTRODUCTION 9
The general impetus for and broad attraction to positing artworks of a

certain sort as abstracta ought to be relatively easy to grasp. Should one be
a realist about artworks, supposing putatively repeatable artworks to be in
some way or other abstract seems to preserve one’s art-realism in a manner
largely congruous with the principal metaphysical and semantic assumptions
as informed by the relevant history, practices, and conventions surrounding
such artworks. In fact, most positions within art ontology are at their
foundations predicated upon the same basic art-ontological assumption
and united under the same broad art-ontological consensus, specifically
that repeatability and non-repeatability more or less track the broad ontic
kinds abstracta and concreta respectively. As such, a basic art-ontological
assumption is that the principal ontological joint at which the artworld
must be carved lies squarely between traditionally repeatable (multiple-
instance) works of art, construed as abstracta, and traditionally non-repeatable
(single-instance) works of art, construed as concreta. Subsequently, for
the art-realist, commitment to art-abstracta looks to be not just a standard
position within art ontology but a basic working assumption of any prima
facie viable art-ontological enquiry.5
Of course, anyone holding prior commitments to some sweeping general
anti-realism (or at least harboring such anti-realist sympathies with respect
to art) will obviously have little trouble denying that there are such things as
art-abstracta. Likewise, those already in possession of general nominalist
commitments (or sharing such thoroughgoing sentiments) should find
themselves no more inclined to endorse abstracta for novels, poetry, or
concertos than for numbers, propositions, or concepts.6 However, what
5 Of course, within art ontology compete a wide variety of models of abstracta (e.g. unstructured
universals, indicated/initiated types, action-types, etc.). However, as to the general art-ontological
commitment to abstracta, any substantive debate has heretofore been largely and conspicuously absent.
6 Others may find art-abstracta inconsistent with the peculiarities of some pet theory of art with
which they have previously aligned (e.g. a particularly spartan aesthetic theory of art according to which
aesthetic properties are strictly perceptual, strictly supervene on the physical, and are strictly uninher-
itable from token to type). Others still may simply suspect such construal to have little to do with objects
as art. That is, while some members of the broad class of repeatable works appear patently to be artworks
(e.g. Moby Dick, Hamlet, The Magic Flute, Piano Concerto No. 9), the class of repeatable works
nevertheless remains a class for which being a member itself neither entails nor suggests being an artwork
(e.g. cookbooks and stereo-instruction manuals, Muzak and doorbell chimes, office memos and grocery
lists). So, the worry would then be what may be ontologically fertile for repeatable works simpliciter may
turn out to be decidedly toxic for those works as art or perhaps more simply that all of the philosophical
gravitas with which ontological debates about repeatable works of art (e.g. poems and symphonies) are
conducted would quickly turn into abject philosophical absurdity for any similarly conducted debate
matters for present purposes is that for those already located on the realist
side of the debate, commitment to art-realism standardly entails commitment
to art-abstracta.
Now consider (II): The standard metaphysical assumption that abstracta
are non-spatiotemporal and causally inert. To be sure, there are lively debates
within contemporary metaphysics about the precise nature of abstract
objects (including the method by which their natures ought to be charac-
terized). However, should any general characterization of abstracta have
a plausible claim to being standardly held, it clearly must be that abstracta
are non-causal (especially given the standard, broad characterization of
concreta as causally-efficacious material inhabitants of space-time).7
Lastly, consider (III): The standard art-theoretic assumption that an artwork
must be created. Presumably this can be best viewed as the standard expression
(or derivation) of one of the most basic and fundamentally intuitive necessary
conditions for something’s being art: An artwork must be the product of intentional
action.8 As such, any art-theoretic enquiry—either in terms of the supporting
claims therein or the resultant claims thereof—to be even prima facie viable as
such must be consistent with artworks being created things. Furthermore,
creation looks to require causation in the form of a causal chain from creator
to created—the standard sense of ‘create’ and its cognates is one indicating a
causal (if not causal-intentional) relation. As such, for any created thing, the
manner in which that thing comes into existence ought not be metaphysically
mysterious—in the case of artworks, they are the causal (causal-intentional)
products of some agential action.9
The source of the tension between the above three standard assumptions
should now be obvious if not also familiar.10 The standard art-ontological
about repeatable works of non-art (e.g. ringtones, wedding invitations, the Big Mac, the Corvette, and
so forth).
7 For instance, Gideon Rosen (2001) considers the view of abstracta as causally inert to be more or
less the standard view and so would presumably take his own contrarian account of abstracta (Burgess and
Rosen 1997) to be to that extent non-standard.
8 For a more detailed discussion of the nature of art’s intention-dependence, see (Mag Uidhir 2010).
9 For example, (Mid-Atlantic Ridge volcanoes were created by divergent tectonic plates) is stan-
dardly taken to entail Mid-Atlantic Ridge volcanoes being caused by divergent tectonic plates. Likewise,
(Sara created a doghouse) is standardly taken to entail Sara having successfully engaged in such-and-such
activities directed by so-and-so intentions, the causal product of which was a thing satisfying the
conditions for being a doghouse, whatever those may be.
10 A standard debate within the ontology of music centers around how to resolve the paradox of
creation: (i) musical works are abstracta; (ii) musical works are created; (iii) abstracta cannot be created.
INTRODUCTION 11
story tells us that given but the simplest investigation of and reflection
upon the nature of putatively repeatable artworks and the relevant sur-
rounding practices and conventions (both linguistic and otherwise), any
minimally defensible and coherent art-realism must construe repeatable art
as abstract such that any minimally adequate art-realism must entail that
there are such things as art-abstracta. However, from the standard meta-
physical story about abstracta it follows not only that such things must be
causally inert but also that if such things exist, then there can be no time at
which those things do not exist—if they exist, they exist eternally. Fur-
thermore, from the standard art-theoretic story about artwork creation, it
follows that if something is a created thing, then there must be a time at
which that thing did not exist—a created thing cannot exist prior to its
creation and so cannot exist eternally. Further follows yet another standard
assumption about creation, namely that creation requires causation—
created things must stand in a causal (if not causal-intentional) relation
to their creators.11 However, if abstracta must be causally inert, then as
such, abstracta must be things incapable of standing in any causal relation
whatsoever, let alone one in the robust causal-intentional sense of creation
standardly assumed to underwrite the art-theoretic commitment to art-
works being the products of intentional actions.12 Accordingly, there can
be no such thing that is a created abstracta (an exclusionary fact which the
standard characterizations of abstracta and creation clearly overdetermine).
So, if there are such things as artworks and artworks must be created
things, then artworks cannot be abstracta.
Clearly an inconsistency is afoot. More importantly, however, the
rather surprising, if not also disturbing, philosophical consequence of
this inconsistency is that what actually lies at the intersection between
our standard art-theoretic commitments and our standard metaphysical
commitments is not some standard art-ontological commitment but
instead the explicit negation of such.13 Again, the most interesting feature
11 It would then also follow that when considered relationally, creating any one kind of artwork
(e.g. those of the standardly repeatable variety such as novels, symphonies, operas) ought to be
metaphysically indistinct from creating an artwork of any other kind (e.g. those of the standardly non-
repeatable variety such as painting, sculpture, drawing).
12 This would also be the view of creation standard within contemporary metaphysics (French and
Vickers 2011).
13 Consider the following crude but useful analogy. Suppose we have what appears to be a perfectly
standard and straightforward recipe for gazpacho soup. However, when we follow this recipe
of the paradox of standards lies not in the nature of the inconsistency itself
but rather in what that inconsistency—both itself as well as the available
means by which it can be resolved—reveals about the current state of art
ontology.
Ultimately, the available means by which one may resolve the para-
dox of standards (and the implications thereof) depends entirely upon
how one chooses to view the relationship between philosophy of art and
contemporary metaphysics. In what follows, I discuss the resolutions to
the paradox of standards consistent with preserving the standard art-
ontological commitment to art-abstracta presumably available to those
adopting deference and to those choosing independence. I then show
that neither view can offer any adequate resolution to the paradox
consistent with commitment to art-abstracta at least proportional to
what is philosophically at stake (e.g. resolving the paradox by either
bankrupting art-realism or by making all artworld onta hopelessly sui
generis). As such, I claim we ought to adopt a third view regarding the
relationship between philosophy of art and contemporary metaphysics—
what I call the reciprocity view. This view allows us to dissolve
rather than resolve the paradox of standards as from reciprocity it
looks as if putatively standard art-ontological enquiry concerns neither
art nor ontology and therefore to that extent is an ‘ontology of art’ in
name only.
III. Resolutions from Deference and Independence

How might those inclined toward deference resolve the paradox of
standards so as to preserve standard art-ontological commitments? More
importantly, given the nature of deference, what might we reasonably
expect to follow art-theoretically given commitment both to there being
perfectly step-by-step the result is not as expected—a chilled vegetarian soup—but instead something
surprisingly far from it—a piping-hot beef stew. Furthermore, despite the fact that each step in our recipe
for gazpacho soup at no point seems anything other than perfectly standard and straightforward,
we nevertheless find that actually making gazpacho soup requires us at some point to deviate from
that recipe, specifically by replacing one of its seemingly standard and straightforward steps with a
decidedly non-standard and counter-intuitive alternative (or simply by skipping some such standard step
altogether).
INTRODUCTION 13
art-abstracta and to abstracta being causally-inert non-spatiotemporal

things?
For example, Julian Dodd’s sonicist view14 of musical works can
be characterized as a deference resolution to the paradox of standards that
preserves standard art-ontological commitment to art-abstracta. According
to Dodd, a musical work just is an unstructured universal, and such univer-
sals are nothing more than the standard sorts of abstracta—non-spatiotem-
poral and causally-inert eternal existants—with which we all ought to be
already familiar. As such, Dodd concedes that it follows from this that
musical works cannot be created things. Clearly what principally motivates
Dodd’s sonicism is not just commitment to the standard art-ontological
assumption that musical works must be abstracta but also a strict alignment
with deference. That is, should preserving our standard art-theoretic
commitments (i.e. artworks must be created things) conflict with the
preservation of our standard metaphysical commitments (i.e. abstracta
must be non-spatiotemporal causally-inert things), then insofar as we take
Dodd to adopt deference, the standard metaphysical commitments must
win out, regardless of how large the art-theoretical bullet we may be
required to bite.
The deference resolution to this paradox of standards is to reject
our standard art-theoretic commitment to art’s intention-dependence,
heading-off any threat of inconsistency or incoherence by rejecting
a standard, basic art-theoretic assumption presumably in favor of some
non-standard alternative according to which artworks needn’t be inten-
tion-dependent whatsoever, let alone in a manner consistent with their
being robustly created things. Of course, despite our frequent willingness to
attempt as much, not all bullets can be bitten, especially in cases where the
bullet in question is sufficiently non-standard so as to be Howitzer-sized.15
As such, standard art-ontological enquiry guided by deference looks
to be metaphysically responsible at the cost of being art-theorectically
irresponsible. Consequently, for those wishing to preserve their standard
art-ontological commitments, the deference resolution to the paradox of
14 Most notably defended in Dodd (2007).

15 The assumption that art needn’t be intention-dependent is art-theoretically non-standard in pre-
cisely the same way the assumption that murder is wholesale morally permissible or even obligatory is
non-standard within normative ethics.
standards seems to secure a metaphysical tidy art ontology only to fill it

with things that simply cannot be artworks.16
Perhaps the appearance of the paradox of standards is best taken not as
evidence of something rotten in the current state of art ontology but instead
as evidence of the current state of art ontology being more or less on the
right track. That is, presumably an at least implicit assumption behind
the paradox of standards, and any non-trivial philosophical consequences
thereof, is that art and its various relata are in some relevant sense ‘standard’
sorts of things. Should we assume as much, then of course we should
likewise expect ontological enquiry directed at artworks (e.g. paintings,
sculptures, novels, symphonies) to proceed more or less in the same general
lines as ontological enquiry directed at any relevantly similar non-art thing
(e.g. tables, chairs, phone books, ringtones). However, should such
an assumption be absent if not explicitly denied, we ought not be surprised
to find that supporting art-realism in a manner consistent both with standard
art-theorectic commitments as well as with standard artworld practices
and conventions requires artworld onta be sui generis (or at least diverging
to some non-negligible extent from their more ordinary ‘standard’ non-art
kin). In fact, the more sui generis art appears to be the more we should expect
art ontology to be populated not just with categories standardly unavailable
on the ontological menu within contemporary metaphysics but with
sui generis categories by definition off-menu (standardly or otherwise). As
such, from the point of view of independence, the appearance of
the paradox of standards is not the result of art-ontological enquiry
having made some meta-ontological mistake or cultivated some pernicious
methodological insularity from contemporary metaphysics. Instead, the
appearance of the paradox of standards is nothing more than the result
16 Of course, for positions such as Dodd’s sonicism to be consistent with standard art-theoretic
assumptions, one need only qualify the relevant domain at issue. That is, if musical works must be
construed according to the characterization of abstracta standard within contemporary metaphysics, then
musical works cannot be artworks. So, for Dodd’s sonicism to be art-theoretically responsible, he need
only claim its domain to extend no further than musical works simpliciter such that sonicism is a position
within musical ontology rather than a position within the ontology of art. Whether the addition of the
claim that musical works cannot be artworks would somehow now render his view untenable is a matter
best discussed elsewhere.
INTRODUCTION 15
of art-ontological enquiry being correctly adjusted so as to capture the

radical ontic ‘insularity’ of its subject.17
Given the above, how might those inclined toward independence re-
solve the paradox of standards so as to preserve their standard art-ontological
commitments? That is, what might one reasonably expect to follow for our
general ontological commitments given both that artworks must be created
things and that there can be such things as art-abstracta? From the point
of view of independence, should our basic art-ontological aims require for
their satisfaction some degree of departure from the standard characterization
of abstracta, such departures then are not only warranted, but required.
Consequently, the good-making features of any art-ontological position
should principally concern not those traditionally recognized as such within
metaphysics (e.g. internal coherence, ontological parsimony, etc.) but instead
those concerning the theory’s descriptivist pedigree (e.g. the extent to which
it explains, supports, makes sense of, or is at least consistent with, the relevant
surrounding practices and conventions).
Perhaps in order to preserve commitment to art-abstracta, rather than
endorse some radically non-standard metaphysical or art-theoretic claim so
as resolve to the paradox of standards, one might instead more cautiously
split the difference between the two. That is, one might endorse some
moderately non-standard view according to which there are certain sorts of
abstracta (e.g. impure sets, indicated types) that can come into, if not also
go out of, existence so as thereby to preserve abstracta as causally inert
while nevertheless denying that if abstracta exist, they must do so eternally.
One need then simply deny that creation strictly requires causation via
committing to some non-standard (and likely stipulative) sense of create
according to which creation minimally need be neither causal nor causal-
intentional but merely a matter of ontological-dependence (e.g. Moby Dick
names some impure abstractum that Herman Melville created in virtue of his
having created some concrete thing—the manuscript—upon which that
impure abstractum coming into existence ontological depends).
17 Perhaps what the previous analysis has shown with respect to deference and the ontology of art is
that the degree to which art ontology is deferential to the standard characterization of abstracta within
contemporary metaphysics is the degree to which there being artworks so construed either runs afoul of
basic art-theoretic considerations or fails to make adequate sense of the relevant surrounding art practices
and conventions.
Of course, absent some strict causation requirement, this non-stand-

ard (stipulative) sense of creation looks to trade a basic and substantive
art-theoretic necessary condition (artworks must be the products of
intentional actions) for its comparatively trivial cousin (artworks must
be ontologically dependent upon the products of intentional action).
Furthermore, absent some principled, non-arbitrary distinction between
ontologically-dependent things that are creation-compatible (e.g. the
manuscript) and those which are creation-incompatible (e.g. its impure
singleton), appeal to non-causal creation threatens an unchecked,
rampant proliferation of creation, such that, given even a moderately
permissive realism about impure abstracta, from but few acts of artistic
creation would likely flow transfinitely many created things. Addition-
ally, without a strict causal requirement for creation, the precise nature
of how such abstracta could come into existence—whether in number
modest or absurd—would nevertheless conspicuously remain metaphys-
ically mysterious.
Alternatively, one might adopt a more radically non-standard account
of abstracta, specifically one according to which abstracta can themselves,
not just by some causal proxy, enter into causal relations with other things.18
By holding art-abstracta to be causally efficacious, one can then retain the
standard causation sense of artistic creation so as to provide not only
a decidedly non-mysterious explanation as to how such things can come
into and presumably go out of existence but also the means by which
to preclude their unchecked proliferation. Trouble is, however, that the
extent to which this method successfully resolves the paradox of standards
appears to be the extent to which putatively repeatable artworks look to be
construed less and less as abstracta and more and more as concreta. After all,
if Moby Dick names the causal product of certain of Herman Melville’s
successfully executed intention-directed activities, then prima facie Moby
Dick looks name some concrete thing, the ontological particulars of which
would be a matter of debate between competing variants of concreta.19
18 To be sure, there are those within contemporary metaphysics who argue for alternative (non-
standard) views of abstract objects: e.g. see Hale (1987) and Burgess and Rosen (1997).
19 For example, Amie Thomasson—most notably in (1999)—claims fictional characters to be abstract
artifacts. However, the more such accounts employ (or are at least implicitly predicated upon) standard
notions of create and artifact the less fictional characters look traditionally abstract and the more they
instead look to be concreta (if not of the traditional sort then at least some exotic variant thereof).
INTRODUCTION 17
From the point of view of independence, resolution to the paradox of

standards so as to preserve our standard art-ontological commitments
demands rejecting the standard characterization of abstracta in favor of
some non-standard characterization more commensurate with both our
standard art-theoretic commitments as well as those relevant surrounding
art practices and conventions. The degree to which this can be successful
is the degree to which the alternative characterization of abstracta is
non-standard (i.e. abstracta as located causally-efficacious created
things); however, such non-standard sorts of ‘abstracta’ can quickly seem
more and more like concreta (e.g. from standard sorts of physical objects,
events, processes, to the non-standard mereological fusions or four-dimen-
sional space-time compositions thereof). As such, the appeal to non-stand-
ard characterizations of abstracta doesn’t so much preserve realism about
art-abstracta as collapse it into art-ontological nominalism (though likely
one with some comparatively exotic variant of concreta in tow).
Given the options from deference and independence, resolving
the paradox of standards so as to preserve our standard art-ontological
commitments looks entail an art ontology that is either art-theoretically
irresponsible—likely one having a domain unrestricted by even the most
basic of art-theoretic constraints—or radically sui generis metaphysically—
likely predicated (if only partially so) upon some sweeping, general
metaontological view commensurate with a rampant art-ontological
promiscuity.20 Although art ontology ought to be ideally located at the
intersection of aesthetics and metaphysics, the extent to which art ontology,
in its actual current state can be located thusly depends entirely upon the
extent to which the paradox of standards can be coherently resolved.
Regardless of the option one chooses to preserve standard art-ontological
assumptions, the result nevertheless remains the same: An ontology of art
that suspiciously looks to be about neither art nor ontology and to that
20 For example, Amie Thomasson (2007, 2008, 2010) advances a meta-ontological position according
to which ontological enquiry ought to be a combination of conceptual analysis and empirical investi-
gation. My worry here is that philosophers of art already sympathetic to independence may unreflec-
tively consider Thomasson’s work in art ontology (2006, 2010) to license appeal to putatively sui generis
onta within their own work and thereby neglect to consider the fact that doing so entails some pretty
hefty and sweeping meta-ontological consequences that extend well beyond the domain of ontological
enquiry into art.
extent, ‘ontology of art’ in name only. That such results are less than
philosophically palatable I take to suggest that we should view the relation-
ship between philosophy of art and metaphysics in a manner other than that
of deference or independence. Instead, we ought to adopt the view from
which the most art-theoretically and metaphysically responsible ontology
of art can follow.21
IV. An Alternative View

As an alternative to deference and independence, I advance a third option
that I take to be a far more productive and eminently more reasonable view
about the relationship between metaphysics and philosophical aesthetics.
I call this the Reciprocity view.
The Reciprocity View: Any responsible art ontology should be
grounded in a responsible metaphysics, and any responsible metaphysics,
at the end of the day, ought to be able to make sense of art.
That is, aesthetics ought to inherit its art-ontological categories from a
responsible metaphysics, and all things considered, a responsible metaphysics
ought to carve out ontological categories capable of coherently grounding
central issues in philosophical aesthetics (e.g. aesthetic properties, the ontology
of art, etc.).
Consider that most, I assume, hold something like reciprocity with
respect to the relationship between philosophy of science and metaphysics,
such that, presumably whatever one’s metaphysical commitments, if at the
21 A metaphysician might wonder why philosophers of art do not simply abandon the appeal to
abstracta in favor of a neo-Aristotelian account of universals such as D.M. Armstrong’s (1978) view
according to which universals are multiply located but nevertheless concrete, which would allow for
repeatable artworks to be both spatiotemporally located and causally active without thereby collapsing
into art-ontological nominalism (thanks to L.A. Paul for suggesting this). Of course, given that ante rem
realism (i.e. universals wholly exist outside their instantiations) promises no improvement over the
abstracta model, one must be an in re realist about repeatable artworks (i.e. that they wholly exist within
their instantiations) in order to avoid the paradox of standards. However, the in re realist position looks
no less potentially revisionary with respect to the way in which the folk talk about artworks than would
any standard nominalist construal. More precisely, if making sense of the relevant surrounding practices
and conventions explain the appeal to abstracta in the first place, then I suspect that an ontic model
according to which Moby Dick is currently (and simultaneously) located in Paris, Dallas, Angkor Wat,
and Amundsen-Scott South Pole Station would likely prove ill-suited for those with descriptivist
inclinations.
INTRODUCTION 19
end of the day those metaphysical commitments are incompatible with a

coherent account of species, then one ought to seriously rethink those
metaphysical commitments. Likewise, any account of species in philosophy
of biology ought to be compatible with the ontological categories in play for
any responsible metaphysics. So too then I suggest reciprocity as how to
view the relationship between philosophy of art and metaphysics. That is, if
at the end of the day, one’s metaphysical commitments preclude a coherent
account of art, then one should seriously reconsider one’s metaphysical
commitments.22 Likewise, any specification of art-realism in philosophical
aesthetics ought to be compatible with the ontological categories in play for
any responsible metaphysics.23
Of course, reciprocity should not be taken to entail or even to suggest
that the relationship between aesthetics and metaphysics must be one of
equal influence—that there are points of exchange doesn’t suggest that such
exchanges must by any means be equal in both directions. This shouldn’t be
surprising. That is, we really ought to expect exchanges with metaphysics to
be frontloaded for aesthetics and exchanges with aesthetics to be at the
backend for metaphysics (hence the presence of the ‘at the end of the day’
clause for metaphysics). Although according to reciprocity, aesthetics bears
the heavier burden in the relationship (contra independence), it nevertheless
constitutes a non-negligible burden on metaphysics (contra deference)—at
least in this respect, aesthetics appears no different than ethics. Moreover,
adopting reciprocity doesn’t mean that the metaphysician must pause to
consider the consequences of her view for (or whether her view aligns with)
any and all issues central to aesthetics and its cognate areas; nor does it mean
that the philosopher of art must pause to consider whether her view is
22 For example, should standard metaphysical commitments preclude coherent realism about the sorts
of things standardly taken to be paragons of human achievement (e.g. Moby Dick, Hamlet, The Magic
Flute, The Eroica Symphony), those commitments nevertheless being consistent with some banal realism
about tables, chairs, and lumps of clay should provide any metaphysician naught but cold comfort.
Similarly, any responsible philosophy of language conducting enquiry into aesthetic or taste predicates
ought to be consistent with and be informed by the very domain within which such predicates are
standard parts of the critical discourse.
23 Interestingly from this it seems to follow that one method aestheticians could employ to ensure
purchase in a responsible metaphysics is to model ontological categories for art after those responsibly in
play for species which perhaps suggests that any aesthetician holding a dim view of contemporary
metaphysics should consider instead adopting reciprocity with respect to the relationship between
aesthetics and philosophy of science. For related issues in the relationship between metaphysics and
philosophy of science, see Paul (2012).
compatible with all issues central to contemporary metaphysics. Rather,

the principal thrust of reciprocity simply is that there are at least some
substantive and critical issues about which both would do well to pause and
consider the other.24 It is precisely at these substantive and central points of
commonality that productive and informative exchanges can occur, and it is
precisely around such points of commonality that philosophy of art ought
to principally revolve.
Consider the consequences for art ontology on reciprocity. Presumably,
those inclined toward such a view would resolve the paradox of standards
by restricting artworld onta to those standard within contemporary
metaphysics checked against standard art-theoretic assumptions. That is, if
what it is to be an artwork is minimally to be such-and-such a thing, then
art-ontological enquiry ought to be circumscribed according to those ontic
categories standardly available within contemporary metaphysics under
which such-and-such things may coherently fall.
Should one quite reasonably think art theory is prior to (and ought to act
as a constraints upon) art ontology, then in cases of conflict with respect to
the mutual preservation of the standard commitments within, art theory
must win out. From the point of view of deference, there arises a conflict
between our commitment to art-abstracta and our commitment to artwork
creation, and as such, anyone privileging our art-theoretic commitments
must reject art-abstracta. Likewise, those endorsing deference can take no
methodological solace in the fact that the ontology of art may productively
appropriate (in lieu of fashioning their own) such non-standard accounts—
after all, deference demands philosophy of art be deferential to abstracta
within contemporary metaphysics as standardly conceived, such that, the extent
to which adequately answering art-ontological enquiries requires endorsing a
non-standard account of abstracta—and thereby rejecting abstracta as stan-
dardly conceived as an object-kind permissibly admitting art-objects—just is
the extent to which art-ontological enquiry demands a broadly metaphysi-
cally revisionary answer. Although no doubt an in-principle option, absent
24 After all, the ontic categories standardly available within contemporary metaphysics being insuffi-
cient to support anything other than a bankrupt art-realism rather strongly suggests their being likewise
insufficient to support realism about any other product of intention action—if there are no such things as
films, novels, paintings, plays, poems, sculptures, songs, and symphonies, then surely just as equally
unreal must be such things as chairs, commercials, governments, grocery lists, instruction manuals,
nations, office memos, ringtones, and tables.
INTRODUCTION 21
some controversial and sweeping meta-ontological position sufficiently

underwriting independence, solving for art-ontological worries by denying
the standard view of abstracta in contemporary metaphysics should strike us as
a wildly disproportionate, counterproductive, and likely ad hoc method of
preserving realism about artworks.
This suggests that there are but two ways in which to secure an
art-realism that is both art-theoretically and metaphysically responsible.
The first is to endorse a thoroughgoing art-ontological nominalism, such
that, if there are such things as artworks, then artworks must be concrete
things. The second simply is to abandon our commitment to art-realism
(at least insofar as putatively repeatable artworks are concerned) and instead
embrace a radical art-ontological/art-theoretic eliminativism or adopt an
anti-realist fictionalism, such that, for any problematic art onta purged from
the world there is some useful art ficta ready to take its place.
The broad methodological implication of reciprocity for philosophy
of art I take to be this: Prior to declaring abstracta essential fixtures in the
art-ontological fundament, we ought to make sure that the sort of thing an
artwork must minimally be can coherently be the sort of thing that an
abstract object must minimally be, at least as standardly conceived
in contemporary metaphysics. That is, in order to secure a responsible art
ontology, we ought not blithely commit to art-abstracta. Instead,
a responsible ontology of art, in addition to requiring us to ensure our
art-theoretic houses are in order, also demands that we philosophically
reflect on the nature of abstract objects from outside so that we can then
responsibly reflect on the nature of abstracta against some fixed art-theoretic
background.
Perhaps adopting reciprocity will ultimately change little about the way
we think about art ontology other than justifying or increasing our warrant
for thinking it that way in the first place. Alternatively, and far more likely,
perhaps a responsibly constructed art ontology would look quite different.
For example, it might be the case that the principal work done under the
auspices of reciprocity ultimately reveals nominalism to win out.25 Or less
drastically, perhaps we’ll simply find artworks to be no more special an ontic
25 On a purely speculative front, I think that should aesthetics experience anything approximating a
philosophical upheaval in the near future, it most likely will be in the form of nominalism establishing
itself as the dominant art-ontological position.
sort than the relevant works of any putatively more mundane sort (e.g.
poems no more ontologically special a thing than office memos, novels
no stranger sort an object than stereo-instruction manuals, and symphonies
no more metaphysically complex than advertising jingles) and thereby
accordingly expect the ontology of art and the ontology of ordinary objects
to be similarly populated at least with respect to the models their respective
study takes to be available.
Of course, we can further specify the implications of reciprocity by
taking it together with the preference for either revisionary or descriptivist
art ontology. For instance, were we to endorse reciprocity along with a
descriptivist model, it would follow that we ought to count as legitimate
ontic kinds for art ontology all and only those ontic kinds standard within
contemporary metaphysics sufficiently able to adequately capture art
practices and conventions. Notice, however, that on such a view, the likely
consequence of the ontological categories standard within contemporary
metaphysics proving insufficient for adequately capturing basic art practices
and conventions is a bankrupt realism about artworks. Of course, no matter
how deeply held our art-realism may be, it surely must have some threshold
that once crossed (e.g. via recourse to ontological gerrymandering and sui
generis ontic kinds) can no longer be held without revision (i.e. any prima
facie viable realism for artworks must entail nominalism about art-abstracta)
or cannot be held simpliciter (i.e. there can be no viable realism for artworks
and so, there can be no such things that are artworks). By contrast, were we
to endorse reciprocity along with a revisionary model, we ought, all else
being equal, to count as legitimate ontic kinds for art ontology all and only
those ontic kinds standard within contemporary metaphysics sufficiently
able to adequately capture basic art practices and conventions. All else
not being equal, however, we ought to hold purely metaphysical/logical
considerations (e.g. ontological parsimony) as trumping any subsequently
accrued revisions to art practice and convention.
Ultimately, the general methodological implication of reciprocity, at
least with respect to art and abstracta, is that if works of a certain sort, absent
art-theoretic considerations, are best construed as abstracta, then it does not
likewise follow that the putative art status of works of that certain sort
thereby entails or suggests abstracta as a legitimate ontic kind for artworks.
Simply put, our principal methodology should not consist of (i) inquiring as
to the putatively art-relevant sorts of works; (ii) constructing ontic models
INTRODUCTION 23
best suited for works of those sorts (not as artworks but as works of those
sorts simplicter); (iii) declaring the resultant constructed ontic kinds legitimate
kinds for art ontology; then (iv) deflecting charges of practicing capricious
metaphysics via an unreflective, cherry-picked appeal to some controversial
and sweepingly revisionary methodological or meta-ontological indict of
general ontological enquiry itself at least as standardly conducted.26 Rather,
we must first enquire as to the sort of thing an artwork must minimally be
(e.g. the product of intentional action), look to contemporary metaphysics
to find the general ontic kinds consistent with that minimal account, and
then, and only then, can we responsibly ground a legitimate art-ontological
domain (and declare those ontic kinds exhaustive of its limits).
The principal methodological lesson that follows, should we decide to
adopt reciprocity, is that the more we find art ontology to reckon
as legitimate metaphysically queer or sui generis kinds of things, the more
we ought to suspect art ontology of being either blind to basic art-theoretic
considerations or principally motivated by considerations well beyond the
purview of contemporary metaphysics—either way an ‘ontology of art’ in
name only. This suggests that at least insofar as we want to be art-realists,
adopting reciprocity may well require seriously reconsidering, revising,
or perhaps even outright rejecting many of our basic art-ontological
assumptions. Any sustained fruitful exchange between philosophy of art
and contemporary metaphysics must be located at the points within each
where the one takes the other seriously, and as such, the extent to which we
allow these points of exchange to remain absent is the extent to which
the methodologies guiding the relevant enquiries remain philosophically
irresponsible and ipso facto the extent to which the result of any such enquiry
so guided is not itself worth taking seriously.27
26 After all, we should be shocked to find that merely something’s being art requires it to be radically
ontically distinct from its nearest non-art kin.
27 To be sure, the last decade or so has seen a dramatic increase in the number of significant and
exciting exchanges between aesthetics and putatively core philosophical areas, not just in the more
obvious overlap areas (e.g. the nature of fiction and theory of depiction) but also in areas traditionally
considered to be largely in the domain of philosophical aesthetics (e.g. the ontology of music, aesthetic
concepts, predicates, judgments, and testimony). Ideally this recent increase in philosophy of art’s
exchange rate would be the product of some newly fashioned methodology driven by a heretofore
largely absent general, serious, and substantive philosophical concern for all things art and aesthetic.
However, the less than ideal fact of the matter is that often such exchanges seem to progress in a manner
incommensurate, if not outright inconsistent, with their being substantively, let alone principally,
motivated or constrained by basic art-theoretic considerations. As such, this increase is perhaps best
Where deference and independence marginalize, enervate, and insulate

the philosophy of art, reciprocity legitimizes, invigorates, and integrates. It is
through reciprocity that philosophy of art must be taken seriously because it
is from the point of view of reciprocity that it becomes capable of grounding
productive and informative exchanges with other philosophical fields—no
more able to be neglected, ignored, or outright dismissed than philosophy of
science. For our view of the relationship between philosophical aesthetics and
other areas of philosophy to default to anything other than reciprocity is for
the philosophical enquiry we conduct at the relevant intersections to be
ultimately self-undermining. To neglect or ignore the ways in which, or
outright deny the very fact that, issues within aesthetics can substantively and
productively inform issues without requires favoring negligence, ignorance,
and unmitigated arrogance over philosophical progress.
A responsible metaphysics is one for which its standard commitments
in the relevant areas of overlap are themselves informed by and consistent
with standard art-theoretic considerations and commitments. As such, the
extent to which metaphysics is irresponsible is the extent to which its standard
metaphysical enquiry conducted at the relevant areas of overlap ignores,
neglects, or dismisses standard art-theoretic considerations and commitments.
Reciprocity entails that philosophers of art must defer only to a responsible
metaphysics, such that, the extent to which metaphysical enquiry at the
relevant overlap conducts itself irresponsibly is the extent to which philoso-
phers of art needn’t defer to the results thereof. To be sure, reciprocity
entails that philosophy of art must be deferential, but this is not some mealy-
mouthed deference to metaphysics simpliciter but instead a properly earned
deference to a responsible metaphysics. Just as only in the presence of such a
fully responsible metaphysics does reciprocity collapse into deference,
should such a responsible metaphysics be fully absent, does reciprocity
collapse into independence,28 and just as deference seems well earned
when underwritten by a responsible metaphysics, independence looks far
less radical a position when underwritten by the irresponsibility of metaphys-
ics than when underwritten by the deep metaphysical queerness of artworks.
viewed not so much as the product of a philosophical appreciation for aesthetics as a philosophical
appropriation from aesthetics.
28 The difference, of course, being that to move from reciprocity to independence no longer
requires that art be in principle ontologically sui generis.
INTRODUCTION 25
V. Conclusion
In the end, I take the paradox of standards (or at least the appearance
thereof) neither to impugn the general philosophical merits of work
currently being done within the ontology of art nor—despite my own
thoroughgoing art-ontological nominalism—to vindicate any particular
view therein. What the paradox of standards ultimately reveals is that insofar
as philosophers of art and metaphysicians view the relationship between
their respective domains in terms of independence or deference, neither
can responsibly conduct philosophical enquiry at what would otherwise
be eminently productive and informative points of exchange. Absent
the move toward reciprocity, metaphysicians will likely continue to
inexplicably neglect the paragon sorts of human achievement (whilst
fretting over how best to make sense of its utterly banal kin) and philoso-
phers of art will likely respond in similar fashion by defaulting to the
metaphysical queerness of art (thereby further retreating into insularity
and obsolescence).
For any philosophical enquiry at the relevant overlap areas between aes-
thetics and other domains to be responsible and productive comes at a price, a
price all participating sides must pay. To be sure, philosophy of art looks to
bear most of the expense; however, the other side nevertheless incurs a non-
negligible cost in that at the end of the day, it must make sense of art.29
References
Armstrong, D.M. (1978) Universals and Scientific Realism, 2 vols (Cambridge:
Cambridge University Press)
Burgess, John and Rosen, Gideon (1997) A Subject with No Object (Oxford: Oxford
University Press)
Dodd, Julian (2007) Works of Music: An Essay in Ontology (Oxford: Oxford Uni-
versity Press)
29 In this minimal yet substantive sense, for example, ought philosophy of art itself inform contem-
porary metaphysics, and in so doing might philosophy of art then come to stand in the same sort of
relation to metaphysics (with all due respect and apologies to philosophy of science) as applied ethics does
to meta-ethics.
French, Steven and Vickers, Peter (2011) ‘Are There No Things that are Scientific
Theories?’ British Journal for the Philosophy of Science 62(4): 771–804
Hale, Bob (1987) Abstract Objects (Oxford: Oxford University Press)
Mag Uidhir, Christy (2010) ‘Failed-Art and Failed Art-Theory’ Australasian Journal
of Philosophy 88: 381–400
Paul, L.A. (2012) ‘Metaphysics as Modeling: The Handmaiden’s Tale’ Philosophical
Studies (April, 27)
Rosen, Gideon (2001) ‘Abstract Objects’ Stanford Encyclopedia of Philosophy
Thomasson, Amie (1999) Fiction and Metaphysics (Cambridge: Cambridge Univer-
sity Press)
——(2006) ‘Debates about the Ontology of Art: What are We Doing Here?’
Philosophy Compass 1(3): 245–55
——(2007) Ordinary Objects. Oxford University Press
——(2008) ‘Existence Questions.’ Philosophical Studies 141(1): 63–78
——(2010) ‘Ontological Innovation in Art’ Journal of Aesthetics & Art Criticism
68(2): 119–30
I
General Ontological Issues
1
Must Ontological Pragmatism
be Self-Defeating?
GUY ROHRBAUGH
We surely live in a golden age for the ontology of art. What was once a staid
and simple debate has given way to an ever more complex landscape
of views, arguments, and fine distinctions. Part of this is due to a wider
resurgence in metaphysics trickling down into aesthetics, but part of it is also
due to an increasing recognition that works of art present their own
independently interesting challenges. This volume is itself a manifestation
of both tendencies. Perhaps the surest sign of a newly achieved maturity is
the recent, reflexive focus on methodological questions. It has suddenly
seemed worth asking what it is we think we are doing, how we are doing it,
and what the measures of success and failure in this endeavor look like.
I. Pragmatism
When we turn to methodology, it is heartening to find a surprising measure
of agreement on at least one basic point: Answers to methodological
questions are constrained in some way by the practice of art itself. What
this constraint amounts to in detail is, of course, a matter of dispute, but I do
not think that anyone really disagrees with the general idea and it is easy
enough to say why. Whatever works of art are, they manifest themselves in
the world primarily—if not solely in the case of repeatable works—through
our practices. Artworks are, in the first place, the objects of that thought,
action, perception, and judgment, which together make up the apprecia-
tive, productive, art-historical, commercial, and critical practices of the
30 GUY ROHRBAUGH
artworld. Consequently, if one’s methodology were not consistent with and

responsive to these phenomena, one would lose the right to claim that it was
artworks that one was giving an account of in the first place. Theories may
further vie for a title like ‘best explanation of our practices,’ but the more
fundamental role of pragmatic concerns is to secure the very subject matter
of which a theory is a theory.
So, we are all pragmatists of some stripe or other, but what then? We
might draw a simple distinction between what might be called negative and
positive uses of the pragmatic methodology. On the one hand, one might
argue that a certain ontological proposal is inadequate because our practices
demand that works have a certain feature that they do not have according
to the proposal. As much as this is an application of Leibniz’s Law, this
application seems above reproach, and success or failure will often turn on
disputes about what our practices really are and commit us to. On the other
hand, we might try to offer a positive pragmatic argument in favor of a
certain proposal, and this raises different, deeper questions about the relation
of practice to ontology.
Some philosophers take the analogy with scientific theorizing quite
seriously. On this picture, we take the sum total of our practices—perhaps
after rational reflection, enforcement of consistency, and other idealizing—
as identifying what you might call ‘the conceptual role of the work,’
presumably the sort of thing expressible as one of those enormously long
Ramsey sentences. With this in hand, or at least in prospect, we can then
reasonably ask which objects, if any, play this role or play it nearly enough.
This picture comes with two significant philosophical commitments. First,
it makes space for the conceivability of massive error on our part, error
of the sort eliminativists make their stock in trade. If there turn out to be no
such objects, then we will have to seek a different sort of explanation, one
that has no use for the ways in which the practices understand themselves
but instead proceeds in terms of external or alien conceptual resources.
Second, this picture carries with it a commitment to a range of objects that
are candidates for that role, candidates on which we have some kind of grip
independently of that which we have on them through our practices.
Typically, these have been taken to be categories supplied by our more
general philosophical reflections: physical objects; events; sums; and various
abstracta including sets, types, properties, and the like.
MUST ONTOLOGICAL PRAGMATISM BE SELF-DEFEATING? 31
It is fair to say that this approach has faced some difficulties, for none
of these usual suspects have seemed a particularly good match to the
description of the suspect. A lot of ink has been spilled trying to resolve
the tension between creatability and repeatability, with abstractness as
the troublesome middle term. I myself have raised difficulties accounting
for the modal and temporal features of artworks (Rohrbaugh 2003).
Perceptibility and abstractness play poorly together. Proposals that sidestep
many of these issues, like David Davies’ (2004) performance tokens, can
seem incredible on every other score. The discontents are familiar.
Perhaps in reaction to this, a second, alternative motivating picture
sometimes seems to be at work, one that invokes a kind of Copernican
turn of thought. On this picture, it is not the role of our practices to supply
identifying information about the objects of our concern. Rather, our
practices are thought to play some kind of constitutive role, furnishing the
world with their objects through our very inhabiting of them. It is hard not
to avail oneself of dark metaphors here—‘the practices are the very fabric of
the objects of their concern,’ ‘objects that conform to our cognitions,’ and
suchlike—but at the level of motivating pictures, it is clear enough what
broad idea is intended. Both of our prior commitments have been shed, or
traded for new ones. Where the practices are literally productive of their
ontology, there can be no question of massive error nor need we make
out antecedent ontological candidates.
Something like this is surely at work in Peter Lamarque’s (2010) distinc-
tion between work and object, where the existence of a work over and
above some mere object is a matter of on-going appropriate cultural
conditions. Amie Thomasson’s (2004, 2005, 2006) work also seems
informed by this movement between pictures. Although Thomasson
often describes the work of our individuative practices as ‘disambiguating
reference,’ which still suggests a first-picture goal of latching on to one
among several candidates, the heart of her conception is a neo-Carnapian
(contrasting with the overtly Quinean feel of the first picture) view on
which the only meaningful existence and identity questions are ‘internal’
ones, settled by asking whether the conditions set by our own linguistic
usage are met. Where these conditions concern our own activities, as they
do in the case of our artistic practices, their reflexive character leaves no
room for massive error nor need for candidates understood antecedently to
the framework; the existence of (adequate?) practices as they are guarantees
32 GUY ROHRBAUGH
the existence of the works as conceived by the practice. One might also be
tempted to assimilate the contrast between these two pictures to Strawson’s
(1959, pp. 9–12) contrast between descriptive and revisionary metaphysics,
but it is not clear that the two line up in a helpful way. Both pictures
we have sketched might rightly be called ‘descriptive’ in virtue of their
pragmatic commitments and contrasted with views in general metaphysics
that Strawson thought of as revisionary.
II. Deflationism
The first sort of worry concerns ontologically deflationary views in the
ontology of art. My purpose here is not to confront specific arguments
or even positions characterized in any significant detail, but to diagnose
underlying motivations. What I have in mind is manifested clearly in recent
papers by Ross Cameron (2008), who advises a brand of eliminativism about
artworks, and Andrew Kania (2008), who instead urges a increasingly-
familiar fictionalist project. Both writers surely fall within the widest
understanding of our pragmatist methodology, as each is motivated to
preserve the appearances of our practices—in word at least—and each
thinks of his own position as giving as good an account of those practices
as we are likely to get.
In Cameron’s case, this is largely a matter of incoherence within practice,
which leaves no candidate suitable, primarily the conflict between
creatability and abstractness. While Cameron is not an error theorist in
the traditional sense, as his view accords truth to our usual beliefs and
utterances, he holds that the best explanation of their truth involves no
commitment to the objects they purport to be about but instead makes
straight appeal to some more basic ontology, provisionally, arrangements
of simples and abstract structures (Cameron 2008). Given the role of
antecedent ontology and a willingness to countenance massive error, or
some more subtle but related failure, this is a first picture sort of approach.
His engagement with the rest of us is spoiled, I think, by his insistence that
proper ontology can only be understood as a concern with what exists
‘fundamentally,’ and that what does not exist fundamentally does not exist
full stop. Even assuming the availability of a conceptual register in which to
make this claim (cf. Predelli 2009), taking this on board pretty much short
circuits the projects the rest of us are interested in and leaves us with an
unpromising verbal dispute about how we are using terms like ‘exists’ and
‘ontology.’ I do not think anyone actually disagrees with Cameron’s claim
that works of art are not ontologically fundamental, but nor does anyone
think this much matters. As is often the case with skeptical positions, the
import does not lie in any kind of genuine habitability but in the pressure it
exerts in highlighting a philosophical problem, here, how are we to account
for the unity enjoyed by a musical work.
Kania is more interesting, as he starts out less non-negotiably far from
ongoing debates while animated by the same underlying tendencies as
Cameron. In Kania’s story, the ontology of art starts with Wollheim
(1980), Wolterstorff (1980), Kivy (1993), and Goodman (1968) working
only implicitly within our first picture and informed largely by traditional
debates about universals. What they are doing is straight metaphysics, or an
application of it. The aforementioned troubles arise when later figures pay
more attention to the details of our practices and the ways in which they fail
to square with the imported, traditional ontologies. Something very like our
second picture, embodied in the work of David Davies (2004) and myself
(2003), with Jerrold Levinson (1980) cast as a transitional figure, then
emerges in reaction to these pressures. All claim that attention to and respect
for our practices requires some kind of novelty when it comes to
the underlying ontology.
At this point, the structure of a dilemma is said to emerge. We, or at least
Levinson and I (the story for Davies is different), are said to face a problem
squaring our views with the demands of traditional metaphysics. We
describe objects that fit our practices to a tee and then proceed to claim
that there are such objects. Unfortunately, they are not there. Any number
of critics, myself included, have pointed out that the idea of an indicated
type does not really make much sense (e.g. Predelli 2001, pp. 288–9;
Rohrbaugh 2003, p. 199). Despite what Levinson says, one cannot help
but think that he has latched onto the same uncreated types only through a
particular guise, and that saying ‘a new object comes to be’ lacks the
performative oomph it needs. In my own case, Julian Dodd (2004, 2007)
has pressed worries regarding the availability of the very conceptual space
such a view would need, and any number have asked, ‘but what are those
historical individuals you claim to identify?’ Indeed, a sober assessment of
my own paper might find only a negative argument, one that establishes
34 GUY ROHRBAUGH
constraints on ontological proposals without saying what meets them,

or worse, constraints that it is obviously impossible to meet. Allan Hazlett’s
(2012, this volume, Ch. 7) nominalism, for example, is motivated by an
acceptance of my points about the modal flexibility of artwork combined
with frank skepticism about the availability of any ontology of repeatables
that also enjoys such flexibility.
Enter deflationism. Just as we once learned that we need not countenance
orangish-reddish after images in order to give an account of a perceptual
state that has such an after image as its object (Smart 1959), so too we might
rest easier with an account of our various judgments, perceptions, and
actions that dispenses with any ontological commitment to their objects.
On Kania’s story, denizens of the artworld essentially participate in a
collective fiction about ‘artworks’ that ‘have,’ that is ‘have in the fiction
of the artworld,’ exactly those features we attribute to them. We may thus
do away with pesky demands to locate the objects that genuinely play those
roles and confine what would otherwise be contradictory demands to a
fiction in which they remain denatured and harmless. On Cameron’s story,
again, our account of artworld practices would bypass the more obvious
but supposedly doomed project of explaining the truth, veridicality, and
content of these judgments, perceptions, and actions in terms of works in
favor of a direct appeal to the simples and abstract structures which have
good standing even in Cameron’s fundamentalist picture of the world.
Either way, the deflationist offers to preserve what matters to the
pragmatist, at least in name or appearance. At the object-level, our practices
may be recognized as going on just as they do, while our deflationary
attitude at the meta-level need only be known to the philosophical elite
for whom it matters. The deflationist can even claim to do a better job
of respecting our practices. Even apparent inconsistency could be dismissed
as superficial and not in need of revision. It could be said, that is just
how we talk and think and it is clear enough what the conditions are
in which we talk and think that way, so we can leave it be with no
requirement that we try to understand it in its own terms or effect the
repairs such understanding demands. Bring on your finest-grain, most
sensitive observations of our practices; it can only serve as just that much
more grist for the deflationary mill.
This, I think, is the siren lure of deflationism. It offers itself as the natural
endpoint of pragmatic thought about the ontology of art. In particular, it is
the result of moving from the first to the second picture, allowing the
demands of practice to do all the work while shrugging off the demands
of straight, respectable metaphysics. But instead of ending up with a picture
on which our practices give rise to the very objects of their own concern,
we instead end up with, quite literally, nothing.
III. Species
This deflationist narrative loses its grip, of course, if we think there are
perfectly acceptable candidates for the role. Much of the work in my (2003)
paper was negative. It highlighted aspects of the practices surrounding
repeatable works of art—their temporal and modal flexibility, their under-
lying temporal and ontologically dependent natures—that together suggest
the profile, not of a type or universal, but of a historical individual,
something concrete in time with the rest of us, though one dependent on
a causal flow of more basic individuals. I also suggested that we already had a
model for understanding such things in biological species. The parallels
between the cases are striking, especially the need to combine a kind of
repeatability with a real, individual presence in time. My passing, and all too
brief, suggestion is filled out in much greater detail by P.D. Magnus (2012,
this volume, Ch. 5), who reminds us that biologists and philosophers of
biology have routinely worked with a conception of species-as-individuals
for more than four decades. The idea of an individual-with-occurrences
turns out to be, not a novel invention or bit of recherche metaphysics,
but well-credentialed member of our set of antecedently available
ontological options.
There is a risk here, however, that the contagion will simply cross over to
biology. This is not just a matter of some theorists adopting an anti-realist
stance toward species as well as artworks, as Hazlett (2012, this volume,
Ch. 7) does, or of art ontologists claiming species as further examples of their
own theories, as do Dodd (2007, p. 33) and Levinson (1980, p. 81), but of
what look like genuinely unsettled issues within the philosophy of biology
over their status. As Magnus says, ‘there are different ways of filling in
the metaphysical details of the view that musical works are historical
individuals,’ (Magnus, this volume, Ch. 5, p. 119) and the details might
matter.
36 GUY ROHRBAUGH
Orthodox views hold that species membership is a part-whole relation

and that species are scattered individuals, perhaps four-dimensional sums.
Such views bear an obvious relation to art-ontological views like those of
Peter Alward (2004) or Ben Caplan and Paul Matheson (2006, 2008).
I worry that such views succeed in providing a concrete and historical
individual to play the role of the work, but without providing the needed
modal flexibility, for sums are often treated as modally rigid in their
composition. However, the de re modal status of sums is far from clear, so
it hard to say for sure how much of a problem this is (cf. Caplan and
Matheson 2006, p. 67). A deeper worry concerns whether the similarity-
of-parts which such a view offers up to supply the occurrences of a
repeatable work or species can do justice to the kind of generality we are
dealing with. I think someone like Dodd is working with a foundational and
a priori view on which only abstract objects can be ‘generic’ or support
genuine generality through a relation of instantiation, while the unity of
a concrete object is fundamentally different, particular not general.
What might be called ‘big individual’ views accept this picture and leave
themselves open to principled worries that the ‘repeatability’ they supply is
somehow faux or ill conceived.
Magnus turns out to prefer a different way of filling in the details, a
homeostatic-property-cluster account of the sort proposed by Richard Boyd
(1999) for use in a variety of philosophical contexts. The proposal, complex
in its details, bears some resemblance to what I called (Rohrbaugh 2003,
pp. 195–6) a ‘neo-type theory,’ one on which the shared nature of the
instances is supplied primarily though an underlying causal mechanism
responsible for the co-instantiation of properties in individuals. Perhaps
surprisingly, Boyd, Magnus, and others think we have reached a kind
of limit point at which the notions of individual and kind merge, for all
intents and purposes, because any concrete individual might equally well
be conceived as a type of causally-connected stage with a particular causal-
historical origin. While I resisted such a view in (Rohrbaugh 2003),
I confess there is much to be said for it and that it delivers much of what
I was demanding. That such a view offers to resolve the worries surrounding
the first picture and stave off the temptation to deflate is, on its own, a strong
recommendation. The view also naturally takes us to a second worry about
pragmatism, for Boyd takes the legitimacy of HPC-kinds to reside entirely
in their connection to explanation. Such kinds fundamentally provide the
projectable categories undergirding causally sustained generalization in one

or another empirical discipline. Can we say the same for artworks conceived
along the same lines?
IV. Explanation
It is at just this point that Robert Kraut (forthcoming) presses new worries.
While also acknowledging the broad point with which we began,
he worries that the pragmatic methodology, while unavoidable, cannot
perform the work it must do. He too begins by connecting the project of
ontology with explanation:
If such an inquiry is to make any sense, it must be construed as an effort to provide
adequate explanations of customary artworld practices—creation, interpretation,
evaluation, and commodification—by adverting to the sorts of things artworks are;
their ontological status is part of the explanation of why they are treated, or ought
to be treated, in one way rather than another . . . It is because of what Stravinsky’s
Rite of Spring is that it can be performed simultaneously in widely separated concert
halls; it is because of what Picasso’s Guernica is that a forged copy of it lacks the
aesthetic value of the original. (Kraut, forthcoming)
But it is these very demands that Kraut alleges a pragmatic approach to the
ontology of art cannot satisfactorily shoulder. In essence, his worry is that, to
the degree that we are simply codifying regularities we find within our
practices, the less we are able to claim that the ontologies arrived at on that
basis give any genuine explanation of why we do what we do or, worse,
why what we do is legitimated or justified by the kinds of objects with
which they are concerned. Or again, an ontology ultimately driven by a
description of what it is we already do, as it must if it is be to an ontology of
art at all, looks like it will be unable to turn around and informatively
explain or justify any of those doings we described. As he sees it, the
alternative is a bald skeptical position. One might say that there is no such
endeavor as the ontology of art. Rather, when one asks, ‘What sort of thing
is a symphony?’ what one really gets in response is just an expression of the
speaker’s own aesthetic views about what is and is not important about
symphonies, in short, ideology. It would be possible to generalize further
38 GUY ROHRBAUGH
about these attitudes across one or another population, but one would be
doing sociology and not philosophy.
It is pretty clear from his discussion that it is our first picture that Kraut has
in mind, and it is true that many or most ontologists of art would describe
their project in these terms. Their positive applications of the methodology
are meant to have the form of an inference to the best explanation, one that
is modeled on successful applications of this inference in empirical domains.
In the case of, say, the type theory, most authors are self-consciously explicit
that the arguments have this form and that the identification of repeatable
works with types is supposed to be justified by our ability to explain their
repeatability, their distinctness from their occurrences, and so on.
Kraut’s resistance to this picture is not so much motivated by the mis-
match between candidates and data as by what strikes him as a fairly obvious
disanalogy with the scientific case. Over there, ontology understood as
inference to the best explanation makes sense because the nature of the
entities goes beyond the descriptions through which we got onto
them. That is really the point of demanding that the candidate entities be
understood antecedently. If you go in for this kind of talk, the power
of finding that some brain state or other plays the belief role lies in the
way a whole new set of properties and laws, which played no role in
the specification of the role, subsume, connect, and explain the features
that did play an identifying role. In contrast, the objects this method is
supposed to get us onto in the art case do not take us much beyond the
phenomena we started with. A particular type is the sort of thing that has
tokens, those which meet the conditions that give that type its very identity
as an object. His suspicion is that we are just repeating back to ourselves
what we started with using new terminology, and if that is true then we
ought to be more honest about it and adopt an expressivist account.
Such worries might seem to afflict the second-picture conception even
more pointedly. Thought of in Kania’s way, as an ultra-thorough going
descriptivist project that ultimately dispenses with any accountability to
genuine objects and the concerns of metaphysics, all we are left with is a
statement of what we say, and that would seem to have nothing to say to
Kraut’s request for explanations and justifications. So too, it might be
thought to stand as a challenge to Thomasson’s way of approaching matters.
As she summarizes part of her view;
[S]ince facts about the ontology of the work of art are determined by human
conceptions, the resulting facts are, as we might say, ontologically shallow—there
is nothing more to discover about them than what our practices themselves
determine. We can investigate the world to see if it meets those criteria that are
clearly accepted by our practices, and we can try to investigate our practices more
closely to see if they might provide a nonarbitrary way to decide a particular issue
(say, by extending principles accepted elsewhere). Beyond that, any solutions to
these problems must be presented as suggestions or proposals about how we should
(stipulatively) decide such issues, not as discoveries of the real facts. (Thomasson
2005, p. 228)
On the face of it, the kind of ‘shallowness’ Thomasson allows the ontology
of art precludes just the sort explanations and justifications that Kraut thinks
it needs to earn its keep. This is, I suppose, to ignore the further possibility
of purely pragmatic disputes about which kind of terms we should work
with, or why talk one way rather than another, but here again Kraut is likely
to push the thought that we can have no more grounded a conversation
than one in which I push my concerns and you push yours, and this is just
the ontology of ideology offered above.
V. Structures of Thought
There is a moment in the Kania piece that draws my attention each time
I read it. I will quote it for you here:
For if a ‘pragmatic’, descriptivist approach to the ontology of art is the correct one,
then it is not clear what force such appeals to general metaphysical claims or
theories should have. If in proposing a theory of the ontology of art we are really
offering a description of the ‘structure of our thought’ about artworks, then the
existence and nature of such philosophical arcana as types, properties, and so on,
look like they might be beside the point. (Kania 2008, p. 437)
This is, of course, precisely the point at which a thorough-going pragma-

tism is said to jump the rails of genuine metaphysics and come completely
away from first-picture concerns and considerations. What catches my
attention here is the implicit contrast between describing ‘the structure of
our thought’ and, I suppose, describing the actual structure of the world
itself. The latter is taken to be the business of general metaphysics, in
40 GUY ROHRBAUGH
contrast to which the former kind of description must be mere description,

conceptualization, or façon de parler. Now the phrase ‘describing
the structure of thought’ is, of course, Strawson’s, but the contrast isn’t
Strawson’s at all. As is clear from an earlier quotation in Kania’s paper
(p. 434), Strawson contrasts the actual structure of our thought with other,
possible structures, what he would describe as revisionary ones. For someone
like Strawson, no contrast is in the offing between structures of thought and
reality itself, what is there already to be described, but only the contrast
between one structure and another. It is also important to remember that
the structures at issue here are not supposed to be a matter of deploying one
or another empirical concept, ones that we say ‘cut up the world in different
ways,’ or ones that do or don’t pick it apart at the joints. Rather, we are
meant to be on to underlying forms of judgment and not just different
contentful ways of filling them in which differ from thought to thought.
So too, these forms are not regarded as the subject matter of empirical
psychology, but a matter of logic, in the broader sense that figures in the
work of Kant, Frege, and Wittgenstein.
From this point of view, it is not at all clear that the investigation of the
structure of thought is really distinct from the ontological-metaphysical
investigations with which Kania wants to contrast it. I am not at all sure
I know how to divorce my grip on ontological categories from my grip
on the structure of certain kinds of judgment. Ontological categories
like object, property, and proposition are not genuine, maximally general
categories of thing but formal concepts reflecting the structure of judgment
(its abstracted parts and unified whole respectively) and opposed to the
various empirical categories which contentfully fill these slots or play these
roles in thought. Or again, commonplace talk of events as an ontological
category seems a reflection of judgments bearing perfective aspect in their
form of predication. One could put this by saying that these items are really
just the projection or reification of a certain form of thought, but this puts
badly the idea that the two are sides of the same coin. The thought is not
that these forms impose themselves on a world that somehow lacks this sort
of structure, but that the structure of the world and that of thought are one
and the same. The form that displays itself in the merely grammatical
structure of a claim like, ‘That flower is red,’ is in no way alien to what it
correctly characterizes, a flower bearing the property of redness. So too, and
nearer our topic, an apparently ontological claim such as that musical works
are types of sound events might equally be understood as a claim about the
structure of certain judgments, that they are represented as involving a
certain familiar kind of generality.
The suggestion I want to consider is this. In taking oneself to investigate
the structure of thought, understood in this particularly robust way, one
need not abandon the notion that one is doing general metaphysics after all,
for the forms of judgment one discovers and elucidates aren’t really
any different than the arcane objects of concerns to traditional metaphysics.
I have said that the ontology of art calls for innovation in general metaphys-
ics. At the time, I would have been happy to say that we need, indeed
already had need of, new sorts of objects. Putting it this way invites
the question, ‘But which objects?’ But we might put the same thought
in another way here, that talk of kind of objects is, in a way, dispensable
and what is really at issue is the forms of thought that undergird the
phenomena we find in our practices. What we are on the lookout for
here is not another theory pitched at a level like that of Gregory Currie’s
(1989) overall art-historical theory or the ramsification of our collective
artwork platitudes or what have you, for such theories are set apart from,
say, a theory of orbital dynamics or global trade by their content. The
practices that make up the artworld are important to ontologists because it
is through the underlying forms of judgment employed within them that we
find our way to their ontology.
Kania quotes another bit of my text earlier on as evidence of my
commitment to pragmatism, but it is one I think I may understand better
now than when I wrote it:
[A] properly conceived ontology of art is one which provides a metaphysical
framework flexible enough to represent accurately a wide variety of phenomena
and to permit the expression of heterogeneous critical views, views which must
be evaluated in their own terms. Ontologies of art are beholden to our artistic
practices . . . and the critical debates are part of the practices to be captured. It is the
job of the metaphysician to provide the space for further argument, not to cut it
off by fiat. (Kania 2008, pp. 432–3; Rohrbaugh 2003, p. 179)
One point I was registering was that about securing our subject matter, that
with which we began. A second point regards my pretty fervent objection
to the use of ontological or metaphysical claims to bolster particular,
properly aesthetic views. If you are arguing that musical pieces ought to
42 GUY ROHRBAUGH
be played with historically authentic instruments or that they ought not to

be played on a synthesizer, and your argument claims that the very nature of
musical works is what supports your position, then I am pretty sure you are
involved in serious error. So too, whatever objection someone like Aaron
Ridley (2003) might have to a Muzak rendition of ‘Ode to Joy’ can only be
misleadingly expressed by framing it in term of whether or not it counts as a
occurrence, for I am sure that the kinds of objections we want to express
only get a grip if the underlying thought, ‘Here is “Ode to Joy”,’ is shared
by the disputants. What we might do now is recast these points. In general,
the claims on both sides of these disputes almost surely do not differ at the
level of form or structure I am concerned with but only in the content of the
claims and it is this shared form that allows further, aesthetic consideration
to get a grip. What ontology ought to be doing is making such debates
possible by providing the very structure of the claims used within them. In a
slogan: Ontology is in the business, not of supplying aesthetic reasons, but
showing how such reasons possible.
In the end, the deflationist offers to free pragmatists from the metaphys-
ical scruples that stand in their way, but I say there is no stepping away from
genuine metaphysics by retreating to claims about the structure of thought.
The two were the same all along. What is needed is not, or not just,
finding objects that can do the jobs we need them to do, but further
reflection on the kinds of judgments which make up our practices and
their underlying forms.
This brings us to the question whether this understanding of the second
picture has any power to relieve us of Kraut’s worries about explanation.
We might begin by noting that, if anything, the view reinforces
the criticisms Kraut levels at the first picture. It does not seem that the
kinds of objects on offer at the ends of these inferences do differ in any
important way from the identification of one or another fundamental form
of judgment. To claim that works are types is already just to claim to find a
certain form of judgment in our practices. Even so, I do not think that the
type theorist is simply repeating back what he or she finds, and finds
important, in practice. This is shown by the fact that one can give arguments
against such a view, arguments that show the type theorist has misconceived
our actual forms of judgment in important ways.
What I think the view has to offer, however, is a principled reason for
declining the demand that Kraut begins with, the demand for explanation.
As I have tried to indicate, it is not the place of ontology to offer expla-

nations of the kind he has in mind. We do want to know the whys of
creation, interpretation, evaluation, and commodification, but the answers
to these questions lie primarily within further artistic judgments. Ontology,
as I’m trying to envision it here, is a matter of the underlying form of such
judgments, but their ability to answer one or another such question is
primarily a matter of the content of particular such judgments. Most
of what wants explaining or justifying concerns matters on which ontology
is silent, instead occupying itself with making such questions, as well as their
competing answers, possible.
As regards the two specific examples, this seems like the right thing to say
about the question why a forged copy of Guernica lacks the aesthetic value of
the original. Saying, ‘It’s a painting’ does not settle this in any way, but
provides the very terms for setting the question and the term in which
considerations need be framed in order to be brought to bear on it. The
other example, I admit, troubles me more. I feel the pull of wanting to say
that the kind of thing Rite of Spring is explains why it can be performed in
two places at once. But this example strikes me as different than the other
one in an important fashion. It is not, I think we should admit, a question
anyone ever asks within our practices. At best, it seems a question one who
was new to our practices would ask, one that is adequately answered by
saying ‘because it’s a musical work’ and going on to say we do thus and so,
that is, give a description of the practice. Or perhaps it is meant from within
our practices but as a distinctly philosophical question. I overhear your
explanation to the newcomer and ask, as a peer, ‘But why is it so? Why
do we behave in these ways?’ And here it seems to be not uninformative to
begin a discussion about the underlying form of thought we bring to bear in
perceiving, judging, and performing ballets.
We also have the beginnings of a reply to the charge that ontology is
naught but ideology. To get going, that idea already requires participants
with differing ideas about some common subject matter. Kraut discusses
in passing the conflict between someone, like Theodore Gracyk (1996),
who wants to put studio recordings at the center of an aesthetics of rock,
and someone who is moved instead by the phenomena of covers, which
looks to privilege different, conflicting aspects of the same phenomena.
Expressing this dispute in ontological terms seems to Kraut tendentious, as it
does to me. But if it makes sense to talk about the deep forms of these
44 GUY ROHRBAUGH
judgments, then these will be forms shared by the judgments of the dispu-
tants. To lose this is to lose the idea that we have disagreement in place at all.
VI. A Final Suggestion

Our gains here are modest. I have tried to distinguish the original grounds
for pragmatism from its development in two different directions. I have
tried to indicate how the gains of adopting the second picture need not
come at the costs of surrendering any claim to ontological import and
forswearing the obligation to metaphysical respectability. Description of
our practices need not be mere description, for identifying the underlying
forms of the judgments in these practices may be another way of pursuing
the same, genuinely metaphysical, task. And while not standing outside of
metaphysics, I have tried to indicate how this project does stand outside the
project of aesthetic explanation. We should expect few concrete answers to
our pressing questions from ontology, but should recognize that the answers
we do give are wrapped up in and presuppose it; it concerns the form such
explanatory content takes on.
In closing, I would like to broach a suggestion about where this might
lead us. The example of biological species shows how the grip of deflation
might be loosened, but what is the art analogue of a species, something
that might be rightly conceived along second-picture lines (as species are
surely not)? The idea I suspect we need is that of an individual practice.
Surprisingly, this very idea surfaces in the midst of Levinson’s (2012,
Ch. 2) re-examination of indication:
[A]s a composer of a work for performance, or possible future instantiation, he aims in
addition to establish something by that very choice or selection: and what he
establishes is a rule, a norm, a miniature practice. (Levinson 2012, this volume, Ch. 2)
For Levinson, these observations are just a way to get us onto the further
idea of particularized or contextualized structure, one that requires of
its instances both complex auditory properties and a certain history and
means of production. But I think we should stop short of this and focus on
the idea of a miniature practice, for it is this thing which the composer
plausibly brings into existence, and which is both individual and repeatably
manifestable.
It is easy to miss this idea by focusing on the rule that Levinson mentions,
for the content of a rule is such that it can be satisfied or not and thus
suggests a condition of the sort which animates our notion of a type.
But what makes a rule a rule is not this content, but lies in the wider
inferential and practical role (our forms of thought again) such content
plays, here, its embeddedness in practice. Just as we should not confuse
the structure of a work with the work itself, so too we should not confuse
the content of a particular practice with the practice itself, what has that
content. What is philosophically crucial here is that the relation of a practice
to what manifests it (even ‘instantiates’ it, if you like) is entirely different
from the relation between some condition—even the very condition
that correctly describes the rules of that practice—and what satisfies it.
What makes a particular performance of a performance of Chopin’s Piano
Concerto No. 1 is not that it instantiates some particular structure—even a
contextualized or particularized structure—shared with all the others,
but rather that is manifests a particular practice, one that is correctly
described in terms of a structure (though it can be a pure one; the
work that ‘particularlized’ or ’contextualized’ was doing has now already
been done).
I am sure that, on hearing this, some will insist on understanding it simply
as the proposal of a new, causal condition on tokens, like the spatio-
temporally restricted types of the biologist, but this is to give up the
philosophical dividends of recognizing a genuine individual here. Perhaps
individual practices can be classified, sorted into practice types (genres and
art forms), but it is token practices we are interested in here, and along this
dimension they are particulars, not instantiated in sub-practices. They are
also general or repeatable items, for they may be manifested over and over
again in the actions they inform and guide, but the sense of ‘instance’ we are
talking about here has shifted.
Among those dividends are precisely the phenomena I attributed to
artworks in my (2003). An individual practice is the sort of thing that can
come to be and pass away at particular points in history. Levinson is right to
focus on the sort of activity that generates such a thing, and we know
that practices survive only through being lived by their participants.
Such things can die out. They are also the sort of thing that can change
over time, sometimes deliberately and reflectively, sometimes not. The very
conditions, rules, and norms that characterize one can shift without thereby
46 GUY ROHRBAUGH
shifting the identity of the practice which accepts them. We are, I think,
well familiar both with attempts to change such things and with attempts to
preserve them as they are. So too, we can make sense of the possibility of a
practice being somewhat different than it in fact is—its having embodied
somewhat different norms—primarily because we can make sense of the
authors of such practices as having made different creative choices in doing
so or of different changes having been wrought along the way. They are, in
the sense I intended, real things, caught up with us in history with a life story
of their own. Where they exist, they exist in and through the historically
situated agents whose actions they guide and explain.
Further development of this idea must be left for another occasion. What
is important here is its connection to our methodological reflections.
We are contemplating a view on which pragmatic considerations are not
just a mandatory way of getting onto whatever works of art are, but turn out
to already concern the very fabric of those works themselves. Conceived
as miniature, individual practices, repeatable works of art are, in a straight-
forward sense, the product of our wider practices, the traditions in which
artists are working. A fuller investigation of the nature of such individual
practices must involve reflection on the kinds of judgments and their
underlying form that make up such things, through which they are created,
manifested in performance, and, for lack of a better word, lived. I hope here
to have taken a small, further step toward an understanding of works of art as
historical individuals while clearing way some methodological worries
about how we find our way to such a proposal.
References
Alward, Peter (2004) ‘The Spoken Work’ The Journal of Aesthetics and Art Criticism
62(4): 331–7
Boyd, Richard (1999) ‘Homeostasis, Species, and Higher Taxa’ in Species: New
Interdisciplinary Essays, by Robert A. Wilson (ed) (Cambridge, MA: The MIT
Press), pp. 141–85
Cameron, Ross (2008) ‘There Are No Things That Are Musical Works’ British
Journal of Aesthetics 48(3): 295–314
Caplan, Ben and Matheson, Carl (2006) ‘Defending Musical Perdurantism’ British
——(2008) ‘Defending “Defending Musical Perdurantism” ’ British Journal of

Aesthetics 48(1): 80–5
Currie, Gregory (1989) An Ontology of Art (New York: St Martin’s Press)
Davies, David (2004) Art as Performance (Oxford: Blackwell)
Dodd, Julian (2004) ‘Types, Continuants, and the Ontology of Music’ British
——(2007) Works of Music (Oxford: Oxford University Press)
Goodman, Nelson (1968) Languages of Art (Indianapolis: Bobs-Merrill)
Gracyk, Theodore (1996) Rhythm and Noise: An Aesthetics of Rock (Durham, NC:
Duke University Press)
Hazlett, Allan (2012) ‘Against Repeatable Artworks,’ this volume, Ch. 7
Kania, Andrew (2008) ‘The Methodology of Musical Ontology: Descriptivism and
its Discontents’ British Journal of Aesthetics 48(4): 426–44
Kivy, Peter (1993) The Fine Art of Repetition (Cambridge: Cambridge University
Press)
Kraut, Robert (forthcoming) ‘What Is Artworld Ontology?’ The Monist
Lamarque, Peter (2010) Work & Object (Oxford: Oxford University Press)
Levinson, Jerrold (1980/1990) ‘What a Musical Work of Art Is’ Journal of Philoso-
phy 77: 5–28, reprinted in Music, Art & Metaphysics (Ithaca, NY: Cornell
University Press)
——(2012) ‘Indication, Abstraction, and Individuation,’ this volume, Ch. 2
Magnus, P.D. (2012) ‘Historical Individuals like Anas platyrhynchos and “Classical
Gas”,’ this volume, Ch. 5
Predelli, Stefano (2001) ‘Musical Ontology and the Argument from Creation’
British Journal of Aesthetics 41(3): 279–92
——(2009) ‘Ontologese and Musical Nihilism: A Reply to Cameron’ British
Ridley, Aaron (2003) ‘Against Musical Ontology’ Journal of Philosophy 100(4):
203–20
Rohrbaugh, Guy (2003) ‘Artworks as Historical Individuals’ European Journal of
Philosophy 11(2): 177–205
Smart, J.J.C. (1959) ‘Sensations and Brain Processes’ The Philosophical Review 68(2):
141–56
Strawson, Peter F. (1959) Individuals (London: Routledge)
Thomasson, Amie L. (2004) ‘The Ontology of Art’ in Peter Kivy (ed) Blackwell
Guide to Aesthetics (Oxford: Blackwell)
——(2005) ‘The Ontology of Art and Knowledge in Aesthetics’ Journal of Aesthet-
ics and Art Criticism 63: 221–9
48 GUY ROHRBAUGH
Thomasson, Amie L. (2006) ‘Debates about the Ontology of Art: What Are We
Doing Here?’ Philosophy Compass 1: 245–55
Wollheim, Richard (1980) Art and Its Objects, 2nd edn (Cambridge: Cambridge
University Press)
Wolterstorff, Nicholas (1980) Works and Worlds of Art (Oxford: Clarendon Press)
2
Indication, Abstraction,
and Individuation*
JERROLD LEVINSON
I.
Roughly thirty years ago, as part of an exploration of the ontology of
art, I suggested that musical works were not pure abstract structures, like
geometrical forms, but rather, impure indicated structures.1 But what exactly
does that mean? In this chapter, I propose to revisit that old idea of mine in
the hope of clarifying it, before then using it as a springboard for discussion
of artistic indication as a singular psychological act, of the individuation
of indicated objects that results from such indication, and finally, of the
relation between artistic indication and neighboring sorts of action, what we
might call actions of simple indication.
It is necessary, before we start, to briefly explain why it is that musical
works—and by the same token, literary works—cannot be considered to be
pure tonal or verbal structures. For instance, why it is that Shelley’s famous
poem Ozymandias cannot be reduced to the sequence of words: ‘I’ ‘met’
‘a’ ‘traveler’ ‘from’ ‘an’ ‘antique’ ‘land’ . . . , and why it is that Chopin’s
Mazurka in A minor Op. 17, no. 4 cannot be reduced to a complex
sequence of notes starting with an altered F major triad. Here are the most
important reasons why such works cannot be reduced to pure structures.
First, pure structures of elements, which are similar to mathematical objects,
* This chapter originally appeared in ‘Rethinking Creativity. Between Art and Philosophy,’ special
issue of Tropos: Rivista di ermeneutica e critica filosofica, edited by Alessandro Bertinetto Alberto Martinengo,
2011, 4(2): 121–33.
1 Levinson (1980). But see also Levinson (1990b).
50 JERROLD LEVINSON
cannot be created, since they exist at all times, but works of art, and that
includes musical and literary works, surely are created, by specific artists
working in definite historical contexts.2 Second, works of art have a number
of important aesthetic and artistic properties that they could not have were
they pure structures existing atemporally, with no essential links to creative
artists, preceding artworks, preceding artistic movements, and more generally,
surrounding cultural environments.
All this was shown, quite conclusively I think, in Jorge Luis Borges’ brilliant
short story ‘Pierre Menard, author of the Quixote’ and, in the philosophy of art,
through a number of important writings by Gregory Currie, Arthur Danto,
Ernst Gombrich, Jacques Morizot, Kendall Walton, and Richard Wollheim,
to cite but a few authors.3 A musical or literary work, though it is partly
defined by its tonal or verbal structure, is nonetheless, like a pictorial or
sculptural work, a particular human creation; it came into being at a certain
time; it may be destroyed in the future if the conditions for its existence cease
to obtain; and it gets its meaning and produces its effects on us not simply in
virtue of its abstract perceptible form, but in virtue of its status as a statement,
expression, or utterance arising in or emerging from a singular generative context.
A Beethoven sonata would not say the same things, musically speaking, if we
thought it to be a work by Brahms, a Jane Austen novel would not communi-
cate the same message if we considered it to be a heavily ironic Woody Allen
production, and an expressionistic painting by the young Mondrian would
certainly look much different if we saw it as a work painted sixty years later by
a mature Jackson Pollock.
This is roughly why musical and literary works cannot be pure or eternal
structures, but must rather be considered instead as impure, historically
conditioned, temporally anchored, structures. I suggested that such works
are really what I call indicated structures, which are partly abstract sorts of
objects, the result of the interaction between a person and an entirely
abstract structure, such as a sequence or series of words or notes.
The interaction in question is precisely an act of indicating, and it is this
action that creates the link between the abstract structure and the concrete
2 More cautiously, if such structures are not eternal they at least exist as soon as a musical system or
linguistic system is in place, and thus well before the works composed employing elements of such a
system.
3 See Currie (1989); Danto (1981); Gombrich (1963); Morizot (1999); Walton (1970); Wollheim
(1968).
INDICATION, ABSTRACTION, AND INDIVIDUATION 51
individual human that lies at the heart of such an artwork. A paradigmatic

musical work, for instance, Beethoven’s Fifth Symphony, is therefore
a tonal-structure-as-indicated-by-a-specific-composer-in-a-specific-histor-
ical-context;4 similarly, a paradigmatic literary work, such as Flaubert’s
Madame Bovary, is a verbal-structure-as-indicated-by-a-specific-author-in-
a-specific-historical-context.5
The hyphens employed in the above formulation are not idle. They
are meant to draw attention to the particular sort of entity that comes
into existence through the structure-indicating actions in question. So for
instance, it is not the case that what it is to be Beethoven’s Fifth Symphony
is to simply be a tonal structure (S) that is indicated by a specific composer
(C) in a specific historical context (H). Rather, what it is to be Beethoven’s
Fifth Symphony is to be the indicated structure S-C-H, an object distinct from
any tonal structure simpliciter, though an object into which a specific tonal
structure enters essentially.
I know that to describe the act through which an artist creates a work of art
in music and literature as an act of indication might seem odd. I must therefore
say a few words about the kind of indication I mean. Of course, when
composers or writers are creating works, they probably don’t take themselves
to be indicating, and probably they don’t take their acts of creation to be
fundamentally acts of indication. More likely they would be tempted to
describe what they are doing as making, as expressing, as formulating, as
narrating, and there are no reasons to deny that these do apply here as well, at
least generally. But it remains true that what these symphonists and poets are
doing, at least most saliently, is indicating, from among all those available in a
given language or tonal system, the notes or words that will, arranged in
a certain order, constitute the symphony or poem that is theirs.
Naturally, the kind of indication I am discussing here, which we can label
artistic indication, and through which musical and literary works come to be,
4 I here leave aside, as regards standard musical works, that such a work arguably has as its core not
simply a tonal structure, but rather, a tonal-instrumental structure. I do so because my concern in this
chapter is narrowly the role of indication, in the sense I try to elucidate, in the generation of certain sorts
of artworks, and not the debate between sonicists and instrumentalists as to whether performing means
are essential to standard musical works.
5 A similar complication might also be observed in the case of standard literary works, whether poems
or novels. Their core structure is not simply a verbal one, if that be understood as merely a sequence of
words or sentences, but such a sequence partitioned into lines, stanzas, paragraphs, chapters, and the like.
Such a core structure might perhaps be labeled a verbal-segmental structure.
52 JERROLD LEVINSON
should be distinguished from more ordinary kinds of indication, which we

can label simple indication. For example, if I notice something of interest in
the street while on a bus, I may point to it so my traveling companion will
attend to it. This is quite a common type of indication, but not the sort of
indication that creates artworks. Or, if in the midst of a conversation I make
reference to a passerby, there is a sense in which I indicate him or her, yet
I surely do not create either of the people I indicated, nor do I even create a
partly abstract object of which they would be constituents. Or, if I find
myself facing a waitress in a restaurant and I nod my head at the blackboard
special that I want to order, again, I indicate, but I do not thereby bring
something new into existence merely in virtue of my act of indication.6 The
same can be said for any usual everyday action whose purpose it is to draw
someone’s attention to some thing or other. In all such cases, the indication
in play is not plausibly one that thereby brings into existence some thing that
didn’t exist before.
So the specific nature of the indication that lies at the heart of the creating
act, whether in music or in literature, needs to be identified. Let us, then,
compare the indication in operation in the creating of a musical work, say
Chopin’s Mazurka in A minor, Op. 17, no. 4, and a perfectly ordinary act of
indication, say that of drawing my friend’s attention to a strange passerby on
the street. We’ll start with this second indication: I notice the strange person
walking along; I then point my finger toward him so my friend will turn her
gaze in his direction. In other words, I indicate, or point to, this unknown
individual, singling him out as something worthy of attention. If my friend
does indeed look at him as a result of my pointing gesture, then my act
of indication has succeeded; if my friend doesn’t look, then my act of
indication has failed, which is not to say that it never occurred. The action
is purely transitory; it is tightly bound up with a fleeting situation, one that
dissolves almost as soon as it comes to be. The action exhausts itself in
the moment, and doesn’t aim at anything permanent. My goal is simply to
draw my friend’s attention, here and now, to the passing phenomenon,
and there is no goal beyond that. I am not engaged in an activity that is
future-oriented, I have no intention to establish or to build anything, I have
6 The qualification ‘merely in virtue of my act of indication’ is important, since an act of ordinary
indication can clearly bring another event into existence in a straightforwardly causal way. For instance,
my indicating something of interest to my companion can bring about the event of her being aware of
the thing in question.
no desire to leave any traces, however small, on the sands of time. At any
rate, such unconcern about what might follow or issue from my act of
indication is characteristic of indication in the ordinary, non-artistic sense.
Something quite different goes on in artistic indication. What is Chopin
doing, by contrast, when he composes the short but magnificent and heart-
rending Mazurka in A minor? There is a sense in which he too—using
his fingers, whether he inks the notes or plays them on the piano—is
engaged in indication. And what he indicates, in a particular order,
are certain individuals of the tonal realm that existed before the act
of composition. In this case, there is indeed an act, or rather many acts, of
simple indication by which Chopin draws our attention ultimately to a
certain tonal configuration that was there before, hidden in the tonal
domain as a field of possible sounds. But Chopin does more than that
when he is composing, because his intention is indeed to leave a mark
of some sort on the world, to insert something new into the musical culture
that precedes and surrounds him.
So what exactly does Chopin do above and beyond simple indication
when he creates his Mazurka? We can start by noticing that he chooses or
selects notes—here including pitches, rhythms, timbres, and dynamics, and
from both vertical and horizontal perspectives—he doesn’t merely draw
attention to them. That is to say, Chopin has a certain attitude—in part
approval, in part appropriation—toward those particular notes. He doesn’t
in effect merely say: ‘here are some sounds’ but rather, ‘here are some
sounds, they are now specifically mine, I embrace them, and in this exact
sequence.’ When we simply indicate, say by physically pointing or by
referring in conversation, we do not take that perspective with regard to
the object targeted; we don’t choose it, we don’t select it, we don’t
designate it as something that henceforth has an enduring relation
to ourselves.
But in creating his Mazurka Chopin doesn’t just choose or select
a sequence of notes, which he thus puts in a special relationship with
himself. If he were simply improvising, that might suffice as a description
of his activity.7 However, as a composer of a work for performance, or
7 Though it might not. Musical improvisation of some sorts can involve indication as selection, the
difference being that in the improvisational case the musician’s actions of selection have rather broad
targets, such as a given standard or a given style of playing, rather than narrow ones, such as a tonal
structure defined in detail.
54 JERROLD LEVINSON
possible future instantiation, he aims in addition to establish something

by that very choice or selection: And what he establishes is a rule, a norm,
a miniature practice, whereby pianists play a piece by Chopin and not just any
piece of music when they play that sequence of notes chosen by Chopin,
and do so precisely because that sequence was chosen by Chopin. That’s
what makes the Mazurka in A minor exist as a work by Chopin. To
indicate, as a composer, a particular sequence of notes consists precisely in
establishing a rule to reproduce the sounds in a certain way following the
indications of a particular, historically-situated musical mind. And it is as
such an indicated structure that we can identify a classical musical work.
This idea of compositional indication as an action of establishing a
rule was well expressed by Nicholas Wolterstorff in a 1987 essay:
A work of music, then, involves a complex interplay among three sorts of
entities: a performance-kind, a set of correctness and completeness rules, and a
set of sounds and ways of making them such that the rules specify those as
the ones to be exemplified . . . Once we see the contribution of rules to
the constitution ofmusic, it becomes apparent that the three-phase model of
composition . . . [invention, evaluation, selection] . . . is deficient when it comes to
music. Invention is of course still involved, as is evaluation. So too is selection.
But the process of selection is now ancillary to the distinct action of ordaining rules
for correctness and completeness. In light of his evaluation, the composer
selects a set of sounds and ways of making those sounds; but he does this in the
course of ordaining a rule to the effect that exemplifying those sounds and those
actions is necessary for a complete and correct performance.8
The key idea in the above formulation is that of ordaining a rule to be

followed by performers wishing to instantiate or perform a given work, but
clearly this is just a more forceful, more sacerdotal variant of the idea, already
mentioned, of establishing such a rule. And this is an element essential to
artistic indication, one that serves to distinguish it from indication of the
mundane sort.
Let us take stock: Artistic indication, unlike simple indication, normally
involves a deliberate choice, an act of appropriation, an attitude of approval, and
the establishment of a rule or norm. But it is now time for us to explore the
way in which artistic indication creates a link between the abstract tonal or
8 Wolterstorff (1994, p. 120).

verbal world and a concrete individual human being, in other words, to

explore how this act serves to individuate the tones or words that it recruits
for its creative and expressive ends.
We often say, pre-theoretically, that a sonata is composed of tones, and
that a poem is composed of words, albeit ordered in a certain way, and
indeed there is some rough truth to these claims. But to be more accurate,
the altered F major triad that opens the Mazurka Op. 17, no. 4, or the word
‘traveler’ in the first line of Ozymandias, aren’t components of these works
that serve to identify them as such, components that would distinguish them
from possible works that resemble them, at least superficially, to the point of
being perceptually indiscernible from them. The story of Pierre Menard and
‘his’ Quixote clearly shows that works that are perceptually indiscernible
are not necessarily identical; in fact, such works can be dramatically different
in meaning, significance, or content. Therefore, even a given series or
configuration of notes or words, however complex it may be, is not
sufficient to fix or uniquely individuate the musical or literary work in
question. This individuation must rather rest on the unique identity of
the artist who, in the interests of self-expression, combines these brute
elements—abstract notes or words—in a specific creative context, and
who then ends up combined with them, so to speak, in the resulting work.
It is for this reason that the true identifying elements of such works aren’t
really the purely abstract tones and words that the artist uses or appropriates
in fashioning his or her work but, as it were, those tones and words as-
indicated-by-this-artist-in-his-or-her-singular-artistic-situation. The elements of
the work so understood in effect guarantee that the work really is the
work of that artist, and it is by indicating, in the sense I have been trying
to make clear, the abstract elements of a given language or system, that the
artist brings into being these half-abstract and half-concrete entities that
I call indicated structures. The indicated structures are more individual, more
personalized, we might say, than the pure structures that can, unlike indicated
structures, be part of any work of any artist. Again, only the indicated
structure, not the pure structure, can be created by the artist. Only that
structure, and not the purely abstract one whose existence predates that of
the artist himself or herself, can have the aesthetic, expressive, and semantic
properties proper to the work of art as such.
Having somewhat clarified, at least in a contrastive manner, the nature
of the indication involved in the creation of indicated structures, I must
56 JERROLD LEVINSON
now take note of a difficulty concerning the type status of such entities. Is an
indicated structure, that is, a structural-type-as-indicated-in-a-context, itself
strictly speaking a type? Well, odd as it may seem, perhaps not, and for the
following reason. If, as many philosophers maintain, types are wholly
defined in terms of essential properties, ones that must be possessed by any
token of the type, and if, in addition, such properties, even when relational,
are held to be eternal, and so not subject to creation, then types will also be
eternal, and equally not subject to creation.9 But that is precisely the
opposite of what indicated structures and initiated types are supposed
to be, and so insisting that they are types would undermine one of the
main motivations for introducing them. Thus, perhaps the act of artistic
indication that operates on a preexisting structural type and yields an
indicated structure or initiated type should not be conceived as having, as
output, a type tout court.
But if initiated types are not, sensu strictu, types after all, then what are
they? One possibility would be to assimilate them to qua objects, items such
as Obama-as-President, or Venus-as-seen-from-Earth.10 However, that is
not a happy suggestion, for at least two reasons. First, is that it would render
initiated types too insubstantial, too aspectual, and thus poor candidates
for creation in a robust sense. Second, is that the qua object model seems
inadequate to capture the intuition that in creating a musical work of the
standard sort one is constituting it from or making it out of some preexistent
sound structure.11 Be that as it may, if initiated types are then neither
types as classically defined, nor qua objects, they are nonetheless recognizably
what Wollheim called generic entities, that is, things that can have instances
and that can be exemplified in a concrete manner. In the case of musical
9 An argument of this sort has been offered in Dodd (2000, 2002). For discussion and partial defusing
of the argument, see Howell (2002). What I call indicated types Howell there calls types-in-use, and the
reservations aired here as to whether indicated types are finally properly thought of as types would apply
equally well to Howell’s types-in-use.
10 A theory of qua objects is worked out in detail in Fine (1982).
11 This is convincingly argued in Evnine (2009). Evnine also articulates an attractive positive
conception of what the work of making a musical work involves, a conception not far removed from
that of artistic indication as elaborated in the present chapter: ‘The labour, in the case of composition, is
not transformative of the sound structure out of which the work is made. But in some looser sense it is
work on that sound structure. It is the work of locating it within the saturated sound space and
distinguishing it from other sound structures’ (Evnine 2009, p. 215).
initiated types, such instantiation-exemplification occurs through musical

performance.12
That there is perhaps, at bottom, not that much of an issue here is nicely
formulated in a recent essay by Robert Stecker on the methodology of the
ontology of art:
Some believe that types are by definition eternal and unchanging (Rohrbaugh,
Dodd), and others think that at least some types are created and are subject
to change and can cease to exist (Levinson, Howell, Thomasson). On this latter
issue it seems to me that there is no real dispute other than a dispute over which
entities should be called ‘types’. As long as one can give a consistent, intelligible
description of a kind of entity, it is not important what we call it.13
II.
In the rest of this chapter I address three further questions, ones concerning
the sorts of indication effected by works of art, as opposed to the sort of
indication with which we have so far been concerned, the sort that operates
to create works of art, at least in certain art forms.
First question: How does a work indicate that it is a work of art, and is thus
to be appreciated as such? A brief answer is this. On the view of arthood that
I have proposed and defended over the years, a work of art is such—that is,
is a work of art—because it, or the object or event or structure it contains,
has been projected by an artist with a specific intention, that of making it the
case that the work be taken or treated, or regarded, or engaged with more or
less as some works of art have been taken or treated, or regarded, or engaged
12 One difficulty, though, surely remains, namely, saying what it is to instantiate the historical-cultural
aspect of an indicated type. The instantiation of the structural aspect—the sound/performance means
structure or performance type at the core of the indicated type—by a concrete performance is a familiar
and relatively unproblematic idea, but instantiation of the other aspect is arguably not. Perhaps all one
can say is that to produce a performance that complies with the structural aspect of a musical work in a
way that is mindful of and conformant with the historical-cultural aspect is thus to instantiate that aspect as
well.
13 Stecker (2009, p. 385). One might still insist that a reasonable methodological concern remains,
namely this: That the extent to which the kind of entity one posits resists being classified according to the
standard metaphysical taxonomy is the extent to which the kind of entity posited can seem sui generis or
ad hoc. But in response to that I would point out that it would be surprising if new metaphysical insights
or proposals did not often require such posits.
For a judicious review of the pros and cons of conceiving musical works as types of some sort, see
Davies (2011), ch. 2.
58 JERROLD LEVINSON
with in the past.14 It follows immediately that anything can be, or at least can
become, a work of art, if it is simply sincerely and seriously projected in such
a manner, and thus that there are no foolproof external signs that something
presented for our consideration is a work of art.
That said, some features serve as reliable, if not infallible, signs that what
we are presented with is a work of art. For instance, and confining ourselves
here for simplicity to the visual arts, features such as extraordinary beauty,
striking form, enigmatic appearance, ostensible reference to earlier art-
works; employment of a traditional artistic medium, a tag or label attaching
to the object, an autograph or signature on the object itself, or finally,
location in an art gallery or exhibition space. In many instances, these are
some of the ways that a visual artwork indicates or signals to us that it
is indeed a work of art.
Second question: What does a work of art indicate in the world beyond it?
A brief answer is that it indicates as much as can be indicated by any other
symbolic vehicle and, in some respects, rather more. Works of art can
indicate worldly objects in many ways: by representing them; by expressing
them; by exemplifying them; by evoking them; by alluding to them; by
serving as metaphors for them, and so on. Works of art sometimes go
beyond these general ways of referring in ways that are more evaluatively
loaded, as with homage, pastiche, parody, satire, burlesque, or caricature.
Of course, the question of what exactly works of art indicate about the
world outside them, in how many registers or modes, and with what import
for their overall meaning, cannot adequately be answered short of a study in
effect summarizing the whole of art criticism and meta-criticism.
Third question: What things do works of art enable us to glimpse or
discern without explicitly indicating them? In other words, what do artworks
indicate indirectly, without being meant to, without expressly drawing our
attention to them, and so to speak, in spite of themselves? A brief answer is
that they so indicate many things, and possibly more than they directly
indicate. In some sense, artworks indicate indirectly whatever one can
reasonably conjecture, given the work and the artistic context in which it
was created, about the creator and the process of creation. That is to say,
works generally reveal or betray, without aiming to do so, a range of things
about their creators and the processes by which they were created.
14 See the relevant essays in Levinson (1990a, 1996b, 2006).

For instance, the way in which a novel was written can tell us much about
the neuroses of its author even if the novel doesn’t depict a neurotic narrator
and doesn’t directly address neurosis. Something of that sort might be true
of Kafka’s The Trial. A symphony, merely by being excessively long, can
convey the difficulty faced by the composer when trying to finish
the composition, even if the symphony does not directly exemplify this
difficulty. Bruckner’s Fifth Symphony is perhaps an example of such a
work. The violence of a painting’s facture and the savage character of its
forms can suggest a feeling of self-loathing that the artist himself might not
be aware of, without the painting actually expressing, representing, or
symbolizing that feeling. I am thinking here of some of Williem’s De
Kooning’s paintings from his Woman series.
In short, works of art do not only indicate what their creators meant to
convey or communicate, and which in successful cases we understand above
all in engaging with them; they can also indicate much about those creators
and their creative processes that was never meant to be conveyed or
communicated.15
If we label the sort of indication I have been concerned with in this
section work indication, the question can be raised of how work indication
relates to the sort of indication of principal concern in this chapter, namely,
artistic indication, that involved in the creation of literary and musical works
of art. We might ask, in particular, if work indication is figurative indication,
while artistic indication is literal indication. I think not. My view is that
both are literal, though of course there are salient differences between them.
First, their agents are different, artists in the one case, and artworks in
the other case. Second, the agency involved is different, being immediate
in one case, and mediated in the other. That is to say, the author or
composer directly performs actions that constitute the selectings, fixings,
and ordainings identified earlier as central to literary and musical art making,
the artist’s agency not being dependent on that of other agents, whereas
their artworks perform (or ‘perform’) actions only in virtue of having been
constituted as artifacts of a certain sort by their creators. Finally, the action
involved in work indication, if not quite figurative, is admittedly action in a
weaker sense than that involved in artistic indication, one that does not imply
15 For related reflections on what artworks indicate directly and indirectly, see Levinson (1996b).
60 JERROLD LEVINSON
will, intentionality, or goal-directedness, and amounting, more or less, to

bringing about an effect or result.16
In conclusion: The work of art, and more specifically, the musical
or literary work of art, can be considered a site where several kinds of
indication intersect. There is, on the one hand, the kind of indication
performed by the artist, which makes the work what it is, through
transforming an ensemble of abstract tonal or verbal elements into an
individualized such ensemble, one tied essentially to the artist himself or
herself. On the other hand, there is the kind of indication performed by
the work, which encompasses all of what the work signifies, in a wide
sense, about the world and the artist, directly or indirectly, for better or
for worse.
References
Currie, Gregory (1989) An Ontology of Art (London: Macmillan)
Danto, Arthur (1981) The Transfiguration of the Commonplace (Cambridge, MA:
Harvard University Press)
Davies, David (2011) Philosophy of the Performing Arts (Oxford: Blackwell)
Dodd, Julian (2000) ‘Musical Works As Eternal Types’ British Journal of Aesthetics
40: 424–40
——(2002) ‘Defending Musical Platonism’ British Journal of Aesthetics 42: 380–402
Evnine, Simon (2009) ‘Constitution and Qua Objects in the Ontology of Music’
British Journal of Aesthetics 49: 203–17
Fine, Kit (1982) ‘Acts, Events, and Things’ in Sprache und Ontologie (Vienna:
Holder-Pichler-Tempsky)
Gombrich, Ernst (1963) Meditations on a Hobby Horse (London: Phaidon Press)
Howell, Robert (2002) ‘Types, Indicated and Initiated’ British Journal of Aesthetics
42: 105–27
Levinson, Jerrold (1980) ‘What a Musical Work Is’ Journal of Philosophy 77: 5–28
——(1990a) Music, Art, and Metaphysics (Ithaca, NY: Cornell University Press)
——(1990b) ‘What a Musical Work Is, Again’ in Levinson (1990a)
——(1996a) The Pleasures of Aesthetics (Ithaca, NY: Cornell University Press)
——(1996b) ‘Messages in Art’ in Levinson (1996a)
16 It is in that respect similar to the sort of action invoked when speaking of the gravitational action of
one mass on another or of sulfuric acid on susceptible metals.
——(2006) Contemplating Art (Oxford: Oxford University Press)

Morizot, Jacques (1999) Sur le Probleme de Borges (Paris: Kime)
Stecker, Robert (2009) ‘Methodological Questions about the Ontology of Music’
Journal of Aesthetics and Art Criticism 67: 375–86
Walton, Kendall (1970) ‘Categories of Art’ Philosophical Review 79: 334–67
Wollheim, Richard (1968) Art and Its Objects (Cambridge: Cambridge University
Press)
Wolterstorff, Nicholas (1994) ‘The Work of Making a Work of Music’ in P. Alperson
(ed) What Is Music?, 2nd edn (University Park: Penn State University Press)
3
Destroying Artworks
MARCUS ROSSBERG*
I. Introduction
Since the work of Nelson Goodman (1968) and Richard Wollheim (1968)
it should be clear that not all works of art belong to the same ontological
category. Marble sculptures are unique concrete objects, while works
of literature can exist in many copies and symphonies can be performed
over and over again. Ontological questions are thus considered to be tied
to an art form, rather than to art in general. Works of art that are unique, like
marble sculptures, are plausibly taken to be concrete objects. On the other
hand, it seems that works of other art forms—in particular, arguably repeat-
able works like perhaps poems, novels, string quartets, or symphonies—
cannot be construed in this way. The repeatability of such works suggests
that they are universals rather than particulars, so the argument goes, and thus
perhaps they should be construed to be some kind of abstract object. No less
plausible is the suspicion that artworks are created by artists, and that what is
created can be destroyed. The creatability of a work of art is often held to be a
non-negotiable feature of any successful account of artworks (see e.g. Levin-
son 1980). But the contingency immanent in the creatability and destructibil-
ity of artworks seems to be at odds with the atemporal and necessary character
usually attributed to abstract objects.
This chapter investigates the question whether the following two
claims are consistent: (i) that artworks can be abstract objects; and (ii) that
* I am indebted to Matt Clemens, Daniel Cohnitz, Roy Cook, Brandon Cooke, Charlotte Geniez,
Michael Lynch, and in particular Christy Mag Uidhir for helpful comments and discussions.
DESTROYING ARTWORKS 63
all art is created and can be destroyed. To this end, we investigate two ‘case
studies’—music and computer art—to try and determine whether abstract
ontologies are plausible for either and what such ontologies would be like.
In particular, we investigate what kind of abstract object artworks might be.
We then turn to the possibility of the destruction of artworks that are
construed as abstracta of the identified kinds. The outcome of our discussion
will be that (i) and (ii) are inconsistent, given only general methodological
assumptions.
One might wonder whether a discussion of any length is required at all:
How could abstract objects possibly be destroyed? Abstract objects
are usually thought to be non-causal, but ‘destruction’ seems to indicate
causality. The inconsistency of (i) and (ii) would hence be a special case of
the general principle that abstractness and destructibility are mutually ex-
clusive. Not so; certain kinds of abstract objects arguably come into being at
a certain point in time. By a prima facie plausible symmetry, we would
expect them also to be able to go out of existence at a certain later point.
Consider, for instance, abstracta of impeccable mathematical pedigree: sets.
Sets are tied ontologically to their members. So-called impure sets are sets
that have objects as members (or members of members, and so on) that are
not sets. Such members could be abstracta that are not sets or they could be
concrete objects. What is of interest for our discussion are those impure sets
that have concreta in their so-called transitive closure (members, or
members of members, or . . . ), so let it henceforth be understood that the
relevant ‘impurities’ are concreta. Since sets are ontologically tied to their
members, impure sets may come into existence when their members do,
and they may likewise perish together. The (impure) singleton set of Igor
Stravinsky thus arguably went out of existence with Stravinsky’s death in
1971; if this is so, then maybe Johann Nelböck destroyed Moritz Schlick’s
singleton set on June 22, 1936, by shooting Schlick in the chest.
Types too have been considered to be creatable abstracta. If it seems
implausible to think that there is a fixed collection of atemporally existing
types to choose from, then perhaps a type is created together with its first
token. If types are thus created, maybe they can also be destroyed in
a fashion that mirrors their creation. One might think it odd that the type
‘A’ should continue to exist after, say, our sun has blown up and consumed
the Earth (assuming humans or ‘A’ tokens will not travel to different solar
systems). A type perhaps may perish with its last token.
64 MARCUS ROSSBERG
With these two examples, there is a case to be made that abstractness

might not preclude destruction. In order to determine what it would take to
destroy, say, Mahler’s Symphony no. 5 if it were an abstract entity, we will
now investigate what the different categories of abstracta are that might
plausibly serve as an ontology of artworks. Once this is achieved, we turn, as
mentioned above, to the question of their destruction.
II. Music
For well-known reasons, chiefly the repeatability of the works mentioned
in the beginning, various kinds of universals have been suggested as an
ontology for music. Abstract, platonic universals might fit the bill: The
possibility to instantiate works of music at different times and places can be
accounted for in this way (Dodd 2000, 2002; Kivy 1983, 1987, 1988). Such
universals are traditionally conceived of as atemporal, or at best eternal, and
hence they are not good candidates for an ontology that allows for the
creation and destruction of a work (Levinson 1980). Composers, it is
sometimes argued in response, discover works and bring them to our
attention rather than create them (Wolterstorff 1975 and 1980, Section 2.
VI), pointing out that discovery can be as difficult and commendable as
creation: Think of Alexander Fleming’s discovery of penicillin or Kurt
Gödel’s discovery of the incompleteness of arithmetic (Kivy 1983,
pp. 112–13; compare Currie 1989, ch. 3; Davis 2003, pp. 31–2 (without
endorsement)). Others are dissatisfied with the prospect that pieces of music
could thus only be found and forgotten, but not created and destroyed.
Indeed, Jerrold Levinson argues that creatability is a requirement on an
adequate account of musical works (Levinson 1980, p. 9).
To overcome this problem, is has been attempted to employ an aristo-
telian1 conception of universals instead. Universals are here taken to be
ontologically dependent on their instances; they are where and when their
instances are. For the case of music, the relevant instances would presum-
ably be performances. Stephen Davies suggests this conception, holding that
1 It is customary to distinguish ‘Platonic’ (pertaining to Plato) from the non-capitalized ‘platonic’ (for
some position broadly inspired by themes of Platonic philosophy). I follow this convention also for
‘aristotelian.’
such musical universals come into existence with their first instance and
cease to exist with their last (Davis 2003, p. 32). This conception is
tempting, but one might wonder how the survival of the works between
performances is accounted for, or whether we can still hold that Beetho-
ven’s Ninth Symphony came into being when Beethoven composed it,
rather than with its first performance. Works that are never performed pose
an additional problem: They appear to await creation in some kind of
eternal limbo. Maybe these problems can be overcome, possibly by
allowing scores and the like as instances of the work. More important for
our context is that this conception of universals might as well go the whole
hog and side with those who argue that universals are concrete rather than
abstract (cf. Armstrong 1978, 1989); musical works would be nothing over
and above all of their (concrete) performances. Thus construed, the proposal
falls outside the scope of our investigation.
Advocates of platonic universals for the ontology of music often specify
them to be kinds, in particular, norm kinds (Davies 2003, Section 2; Dodd
2000; Wolterstorff 1975, 1980). The distinction between properties and
kinds may be useful in general and instructive for the case of music, but
the fine details are tangential to the discussion here: Whatever features
are peculiar to kinds, an investigation of their fundamental ontology as
abstract universals will suffice for our purpose. Let us thus neglect the
finer distinction.
Another specific proposal for suitably abstract universals is to construe
works of music as structures (Kivy 1988; Levinson 1980). Structures are
well-respected entities in mathematics and prominent items in the philoso-
phy of mathematics (see e.g. Shapiro 2000). For the case of music, sonic
structures are invoked, which are instantiated when a work is performed.2
Structures are usually taken to be necessary, atemporal, abstract universals,
which again means that all a composer can do is draw attention to specific
independently existing sonic structures. Invoking patterns as contingent
counterparts of structures seems little promising; creatable patterns are
arguably concrete, and thus offer no significant advantage over adopting
an ontology of performances without the detour. Employing that what is
2 There are differences in the accounts proposed by musical structuralists regarding, for instance, the
question to what extent such structures determine performances, in particular, regarding timbre and
the instrumentation of the work. Again, for the basic ontological question these details can and will be
omitted.
66 MARCUS ROSSBERG
common to two concrete pattern instances—the Pattern, as it were—seems

no different from appealing to abstract structures (or types, see below).
It is interesting to note that at least in the initial introduction of the idea
that works of music might be much like mathematical objects sometimes
analogies are drawn between the repeatability of works of music in different
performances of the work and the alleged repeatability of numbers (see e.g.
Kivy 2002, pp. 211, 213). A ‘2’ can and does occur over and over again, as
indeed it does in this very passage; yet we talk about the number two, which
is a unique abstract object. While ultimately the proposals do not rest on
this, it should be noted that it would be a mistake to confuse numbers with
numerals. Frege (1903) warned against this error with great force—and great
stamina (pp. 80–138). The numeral ‘2’ can and does recur any number of
times, i.e. the sign ‘2’ has many inscriptions in this passage and elsewhere;
but the number two—which is what all the numerals ‘2’ denote—is an
abstract particular according to the prevalent platonist ontology. That
there is exactly one (natural) number two is a theorem of arithmetic:
There is exactly one immediate successor of the number one. The number
two is not a universal and not repeatable. So, if works of music are
to be repeatable, then they are not like numbers, they are not abstract
particulars (although they might be like mathematical structures, abstract
universals). Maybe, however, they could be like numerals. The numeral ‘2’
is repeatable. It is usually taken to be a type, and all its inscriptions in this
passage and elsewhere are tokens of the type. Our third at least prima
facie plausible option for an abstract ontology for music is thus to construe
them as abstract types.
There is a question whether types are sui generis abstract objects. Types have
been construed as equivalence classes of their tokens. The type of the numeral
‘2’ would thus be the class of all its inscriptions. This would render types
abstract particulars (classes are objects) rather than universals, and the apparent
repeatability of the type would come down to the particular inscription’s
being a member of the relevant equivalence class. Rather than trying to
decide whether types are sui generis universals or equivalence classes, let us
put down equivalence classes as our fourth option for an abstract ontology for
works of music. The members of these classes are the performances of the
work. This however raises a problem regarding unperformed works again.
Classes are identical if, and only if, they have exactly the same members. The
equivalence class of an unperformed work is empty, however, and hence any
two unperformed works would be identical: hardly a welcome consequence

(Davis 2003, p. 32; Wolterstorff 1980, p. 44).
This failure of equivalence classes to account for unperformed works
is sometimes put forward as a criticism of Goodman’s (1968) account of
the ontology of music (e.g. Davis 2001, p. 40; Price 1982, p. 326; with
qualifications, Wolterstorff 1980, Section 2.X). Not to put too fine a
point on it, but Goodman’s proposal does not actually involve classes. He
sometimes uses talk of compliance classes—classes of performances that
comply with a score—as convenient shorthand, but since Goodman is a
nominalist, his ontology does not, strictly speaking, include classes.3 The full
and correct version of Goodman’s proposal does not mention classes of
performances; rather, this official version only refers to scores and to
performances that comply with these scores (see Goodman 1968,
Section IV.4, and in particular, p. 131 fn. 3; compare also Cohnitz and
Rossberg 2006, ch. 6). Goodman’s ontology of music thus only includes
concrete objects: Inscriptions of scores, and actual performances, which
are physical events. A correct performance of a work of music is one
that complies with (a perfect copy of) the definitive score. Together
with classes, the problem of the empty class of performances vanishes for
Goodman too; if the work remains unperformed, there is still the score
and vice versa. Moreover, other notations, or recordings, or maybe
even memories may be sufficient, provided they fulfill Goodman’s strict
requirements regarding notation and compliance.
Let us speculate that the problem of the empty class of performances
can be overcome by an account that appeals to equivalence classes of
performances, perhaps, inspired by Goodman, by including scores
or other suitable records in some way. Equivalence classes of performances
are impure sets and thus prime candidates for destruction.
III. Computer Art

The starting point for the discussion of computer art must be Dominic
Lopes’s seminal investigation of this subject (Lopes 2009). Lopes discusses
a range of computer artworks, most of which include more than just a
3 Wolterstorff is aware of this: compare Wolterstorff 1980, p. 100.

68 MARCUS ROSSBERG
display on a screen. Scott Snibbe’s Boundary Functions (1998, cf. Lopes 2009,
p. 25), for instance, consists of a platform that its users4 are to step on, a
camera and projector overhead pointing down to the platform, and a
(hidden) computer that operates the camera and projector. If two or more
people—registered by the computer by means of the camera—step on the
platform, lines will be projected onto the platform, so that each user is
enclosed by a boundary. The lines move and change according to the
movements of the users so that users cannot step outside of their boundary
lines—unless they leave the platform. For simplicity’s sake, we will here
neglect ontologically complex works like Boundary Functions. The issue of
an ontology of abstracta is likely to be obscured rather than furthered by the
discussion of works that essentially incorporate concrete items. We will thus
focus on works of what might be called ‘pure computer art,’ ‘software art,’
or ‘internet art’ to further our discussion. It should be straightforward to
adapt our findings for ontologically more complicated works. We will
simply stick to the label ‘computer art’ in what follows: while the works
we focus on are indeed typically accessible via the internet, this seems
inessential to the art form; ‘pure’ has a misleading evaluative ring; ‘software
art’ just is not entrenched in the literature.
Typical works of this species of computer art that we consider here are,
for instance, Angelo Plessas’s Towers and Powers (2009, <http://www.
towersandpowers.com>) or Damian Lopes’s Project X (1997, cf. Lopes
2009, p. 22). Towers and Powers present several squares with different
patterns and a circle with a spiral pattern on a black background. The
objects can be moved around, using a mouse or similar input device and
will interact when they touch each other; items are displaced, and will ‘fall’
as if governed by gravity. The circle will start spinning if it is held by the
mouse pointer. The pattern of the squares are black and white, but will
change to color when they are touched by the circle and usually, but not
always, change back to black and white when the contact to the circle is lost.
Lopes’s Project X is ‘a poetry-multimedia installation exploring discovery,
technology, colonialism through Vasco da Gama’s first voyage from Portu-
gal to Africa and South America’ (<http://projectx.damianlopes.com>).
4 I will follow Lopes’s convention (Lopes 2009, p. 76) to speak of the users of a work of computer art,
alluding to computer users, but also to emphasize the interactive character of works of computer art, to
which we will come shortly.
The work displays a text in which each word is hyperlinked. Each click on a
word will take the user to a different screen, either displaying verse or
entries in the ship log, or they may call on a picture or encyclopedia-style
definition, which is displayed besides the text. The resulting new screen is in
some way connected to the word that was clicked on and will in some way
describe part of da Gama’s journey: It may elaborate on the previous theme,
but it may also show things in a different light. Each use of the work has a
random starting point. The journey that users take from there is determined
by what words they click on. Through the wealth of material that is linked,
each use of the work is very likely to be in a unique series.
Works like these can be used on different computers and at different
times; the work is multiply instantiated. Moreover, works of computer
art are essentially interactive (we will not count digital pictures on their
own as computer art in this sense). The input of the user, in whichever way
it may proceed, determines what the particular instance of the work is like,
much like players of a game (be it chess or World of Warcraft) determine by
their actions what the particular playing of the game will be like. The
instances of the works obviously are in some way created by computer
programs and the inputs that the user makes and that are processed by the
program. As Lopes (2009, pp. 63–4) points out, however, it will not do
to identify the work with its program. Two programs that produce the
same result could be written in two different programming languages.
Furthermore, the language that the program is actually written in will be
either compiled, or run on a so-called interpreter, in order to translate the
relatively readable (C++, HTML, Perl, etc.) code to the proverbial zeros
and ones a computer handles. Moreover, since everything is just zeros
and ones in a computer,5 deviant processing of code may display the code
for a sound as a picture, the code for a picture as a long string of text
characters, or play text as a sound (although, to make some of these actually
work will involve some competent hacking). The layers of complexity do
not stop there, however. In order to produce the same result, the codes that
result after compiling or interpreting will be vastly different depending on
the operating system the computer uses (Linux, Mac OS, Windows, etc.)
and this will cause vastly different states of electrical potential in the network
5 This, of course, is not strictly speaking true; even the zeros and ones are mere representations of
different voltage levels in the transistor circuits.
70 MARCUS ROSSBERG
of microchips, depending on the architecture of the computer. At none

of these levels can ‘the program’ be uniquely identified as the work. All of
it will have to fit together so that work is instantiated correctly.
Computer scientists, if they want to abstract from these technicalities,
usually talk of the algorithm that is implemented by the programmer to create
a code that can be run on one or many different systems. The abstract
algorithm is taken to be independent of any of the concrete programs
or computer states that are its implementations. Since we too might be
inclined to disregard the technicalities of a work of computer art, it seems
that algorithms have brought us a step forward and led to an attractive
candidate for an abstract ontology for computer art. Indeed, online galleries
(e.g. <http://www.rhizome.org>) that host computer art ask their artists to
complete a questionnaire regarding technical requirements to display the
work correctly (operating systems, web browsers, scripting languages, etc.)
and for the artists’ views on migrating the work to a different platform or
reinterpreting the work (if, for instance, HTML should go out of use). The
question whether a work of computer art survives its translation to another
platform is, admittedly, akin to the problem of restoration for paintings,
sculptures, and other works of more traditional art forms. I do not wish to
suggest that the existence of the above mentioned artist questionnaires
demonstrate that the ontology of a work of computer art depends on
the decision an artists makes when completing such a form—to no greater
extent, at any rate, than a painter’s view on restoration has an influence
on the ontology of painting. However, if works of computer art were
identified with specific computer programs, migration and reinterpretation
would in principle be impossible.
Lopes endorses an ontology of algorithms for computer art (Lopes 2009,
ch. 4; more explicitly in Lopes 2001, p. 77; see also Tavinor 2011), but
he points out that the correct display of the work must be guaranteed. He
suggests that the algorithm alone is not enough; playing Project X through
an MP3-player will not count as a correct instance of the work. It may be
possible to construe the identity conditions of an algorithm in such a way
that only implementations that produce the correct display count as
belonging to the algorithm in question. For our purpose here, we could
go either way.
There is, however, a problem with identifying algorithms as (part of) the
ontology of computer art. Characterizations of the concept of an algorithm
usually are similar to this: ‘Algorithms are simply . . . step by step instruc-
tions, to be carried out quite mechanically, so as to achieve some desired
result’ (Chabert and Weeks 1999, p. 1). There are even quasi-definitions like
this fairly typical one: A procedure is algorithmic, if and only if, ‘at each
step it is clear what we have to do, and it can be done “mechanically”
by finite manipulations’ (van Dalen 2001, p. 246). Standard textbooks
(e.g. Cutland 1980, ch. 1) identify algorithms with effective procedures,
mechanical methods, or set of rules to compute a function.6 So, algorithms
are certain kinds of procedures, and Lopes’s proposal is thus to specify
procedures as (part of) the ontology of computer artworks. A proced-
ure—according to textbook definitions again—is a specific prescription
for carrying out a task or solving a problem: it is a bunch of rules. Rules
or instructions do not appear to be the right ontological category for
artworks. Pretending for a moment we have even the slightest idea what
the ontology of rules might be, it just seems wrong, or even incompre-
hensible, to describe, say, Plessas’s Towers and Powers as a rule, akin to
modus ponens, or to an instruction, such as, ‘Pick up the red ball.’
Nonetheless, there is something undeniably right about Lopes’s observa-
tion that algorithms are of importance for computer art. I want to suggest,
however, they are only indirectly connected to the ontology. As mentioned
above, Lopes emphasizes the significance of the prescribed displays of
a work of computer art: The abstract algorithm alone might not do.
6 The problem that there is no proper, universally agree upon definition of an algorithm is not usually
discussed in the literature on computer art. An algorithm is meant to be an effective procedure for the
calculation of a function, but the boundaries of effective calculability elude definition. According to the
so-called Church-Turing Thesis (Church 1936 Turing 1936), effective calculability is identified with what
can be calculated by a Turing machine. (See Blass and Gurevich 2003 on the question whether it is
indeed true that Church and Turing were aiming for the same notion, as it is usually assumed.) Despite
the fact that a wide range of computability notions—including what can be calculated with an abacus—
have been shown to be equivalent to Turing computability, there is no strict proof that there are no
effective methods of calculation that go beyond what a Turing machine can calculate. There are two
lines of response to this possible problem; we can leave it to the mathematicians to eventually figure out
exactly what an algorithm is—whatever the outcome, this is the ontological category of computer art; or
one could observe that whether or not the Church-Turing Thesis is true, our computers are, in fact,
Turing machines, and given that this sufficient condition is fulfilled, every work of computer art is in fact
algorithmic.
As a final note on the technicalities: Since the process of a work of computer art depends on non-
predetermined inputs of the user, the works cannot be pure Turing machines. This poses no problem,
however. A device (in this case: the user) that supplies inputs while the program is running is standardly
called an ‘oracle’ by computer scientists. The theory of Turing machines with oracles is well studied and
does not present any theoretical difficulties (cf. Cutland 1980, Section 9.4, or any other standard
textbook on computability).
72 MARCUS ROSSBERG
He observes that being produced by the same algorithm is a necessary7

(but perhaps not sufficient) condition for the sameness of the work: Two
displays are instances of the same work only if they are produced by the
same algorithm (Lopes 2001, p. 76; 2009, p. 63). We can distill an abstract
ontology from this that essentially involves algorithms, but without identi-
fying the work with the algorithm. The algorithm as a criterion of identity
for an artwork can be turned into the defining feature of the equivalence
class of the relevant displays. Being produced by the same algorithm is an
equivalence relation. Let the first option thus be to identify the work with
the defined equivalence class of displays. (Should being produced by the
same algorithm not be sufficient for work identity, the relevant further
conditions—perhaps socio-historical context or authorship—will have to
be added.) Such an equivalence class is an impure set, and thus a promising
candidate for destruction.
Note that this proposal can distinguish works of computer art in
the following potentially troubling case. Imagine two works that are
based on different algorithms but that, given certain inputs by the users,
coincidentally produce the same displays. Despite the same appearance, the
displays will get identified in the correct way, since the identity criterion is
the algorithm that the work is based on and not perceptual indiscernibility.
The displays will be members of different equivalence classes—even
if perchance the works only gave rise to displays that could have
been produced by the respective other algorithm. Since the displays at
their actual space-time locations are actually produced by two different
algorithms, they will get sorted into the appropriate and distinct equivalence
classes (despite the fact that a given user might not be able to tell the
difference). Since the actual, concrete displays will be distinct, the two
equivalence classes will share no members.
We would, however, re-encounter the problem of artworks that have
no instances. Given programming practices, it is highly unlikely that an
artist writes a program for a computer artwork and never runs the program
which would constitute an instance of the work; the standard way to make
7 A further complication here may be that, for any algorithm, there are always several (indeed,
infinitely many) functionally indistinguishable different algorithms. For computer artworks that might
mean: Different algorithms that determinately generate the same series of displays given the same inputs
by the user. We cannot go into this problem here, but appealing to equivalence classes again might be a
promising route.
sure that the program runs without ‘bugs’ is to try and run it—repeatedly.
However, it is entirely possible that two works are written and would run
‘bug-free’ but are actually never executed. The relevant equivalence classes
of both works would be empty, and thus the works would be identical.
Appealing to possible instances would, in addition to encountering
the ontological problems of bare possibilia, render the work indestructible.
We will thus forego such attempts.
A solution would be to argue for an alternative option, namely to
identify the work with the equivalence class of triples consisting of
computer programs implementing the algorithm, the suitable interpret-
ing/compiling/operating system, and a suitable computer architecture.
This could include programs written in different programming languages
(together with different ‘home environments’). A computer program
itself is repeatable, of course; it can run on different computers and at
different times. In order not to jeopardize destructibility, we can follow
our now familiar method and opt for the plausible account of programs (and
operating systems) as equivalence classes of inscriptions. The inscriptions
will typically not be ink on paper but electronic and on some computer
storage device such as a hard drive, memory card, or old-school floppy
disk. Either way, such inscriptions will be concrete, physical objects. Let
us now, with a number of different plausible options for abstract ontologies
for artworks, turn to their destruction.
IV. Destruction
It is clear that if artworks are construed as necessary, atemporal,
abstract universals, be they platonic universals or abstract structures akin to
mathematical structures, then they cannot be destroyed—this is pretty much
true by definition. Stravinsky’s 1919 Firebird Suite would have existed even
if Stravinsky had never lived. What about the alternatives?
Levinson advocates an ontology of structures for music, but, as
mentioned above, insists that the creatability of a musical work is a
necessary requirement for an adequate account of music (Levinson 1980,
p. 9). He argues that these structures are initiated types, as he calls them.
They have a beginning in time, when the work is composed, and
are ‘contextually qualified, person-and-time-tethered’ abstract objects
74 MARCUS ROSSBERG
(p. 216). The indicated structure (indicated by a person at a time in a socio-

historical context) that is the work of music is an initiated type that is
created by its composer (at a time, in a context). With this construal,
Levinson not only accounts for the creatability of musical works, but also
allows for two composers to create distinct works that happen to have the
same sonic structure.
Levinson also addresses the question of the destruction of works of music
explicitly, albeit briefly (Levinson 1990, pp. 261–3). Works of music are
created on his account, but how would one go about destroying one?
He considers various options but as the most plausible one he identifies
the ‘permanent elimination of all records and memories’ (p. 262). (We will
refine this below.) Levinson cannot quite bring himself fully to embrace this
option, however, and instead entertains the idea that musical works linger
on eternally, once created. He is criticized for this by Peter Kivy. Kivy
deems the notion that we could create something that can never be
destroyed implausible, but finds entertaining this idea instructive, since, as
he contends, it shows that Levinson’s proposal for the creation of types
already is incoherent: An indicated structure is ‘after all, an abstract object
that occupies a spaceless, timeless realm with which we cannot causally interact;
that is to say, a world of objects that neither we nor anything else in the
physical world of time and space can do anything to’ (Kivy 2002, p. 222,
emphasis in the original).
Levinson does not argue for the indestructibility of works of music but
merely entertains the ‘comforting thought’ that our music, once composed,
will endure until the end of time. Even if he did argue for their indestruc-
tibility, however, we would at best have reached a standoff. Kivy does little
more than assert the orthodox view of platonic abstracta: That they can
neither be created nor destroyed. Levinson, however, does not intend
indicated types to be platonic universals. Kivy puts his finger on a point,
however, that demands further investigation. Levinson (1980, p. 21; 2012)
holds that indicated structures arise from an indication of a pure, abstract,
and atemporal structure. He is reluctant, however, to identify indicated
structures with concrete instances of pure structures. Indicated structures
cannot be identified with the pure structures either, however, for if
they were identical, indicated structures would be eternal abstracta after all
(cf. Levinson 2012). These initiated types would thus be no different from
types conceived as platonic universals, and initiating them would be no
different from discovering them, or drawing our attention to them, as

the musical platonist would have it. Levinson straddles the divide between
orthodox abstract universals and concrete objects, and may best be
understood as postulating a new category of objects that are neither, but
somehow in between: creatable and abstract (cf. Levinson 2012).
To evaluate this proposal, a more general discussion of types might help.
A type has been proposed to be an abstract universal that is created with its
first token and ceases to be when there are and can be no more of its tokens.
As suggested by Levinson, all records and memories of the type or its tokens
must perish for this to be the case, but this might not be enough. Consider a
time when all ‘A’ tokens have been destroyed and everyone who ever had
a memory of ‘A’ has died. Now imagine a machine still in existence that was
used to produce ‘A’ tokens at the push of a button. The machine itself has
no ‘A’ tokens anywhere embedded in its innards; maybe it works like a laser
printer that has instructions for printing an ‘A’ stored. There is also a
less high-tech version of this: Written instruction (in a language not using
an alphabet that contains ‘A’) that, without mentioning ‘A,’ tells you how to
draw such an inscription. Arguably, the type then still exists, ‘un-tokened,’
and independently of whether the button on the tokening machine is ever
pushed or the written instructions are ever followed. As a third condition
for the destruction of a type I thus want to suggest the utter destruction of
any means to produce a new token of the type.
Wollheim (1968) is usually credited with being the first to take Peirce’s
type/token distinction into service for the ontology of artworks. It
might merit mentioning, however, that Goodman (1968), simultaneously,
considers types as an ontology for repeatable artworks, and dismisses them
(Goodman 1968, in particular, p. 131 fn. 3). Indeed, Goodman had argued
for the superfluidity of types already in his Structure of Appearance (1951,
Section XI.2, following the strategy of Goodman and Quine 1947).
Any ontology that appeals to types of the creatable and destructible fashion
can do without them, since it is in fact only concrete tokens that are ever
essentially appealed to. Inessential uses of types, like ‘the (inscription-
type) “Paris” consists of five letters,’ can be eliminated from the theoretical
discourse in favor of constructions like ‘Every “Paris” inscription consists of
five letter inscriptions.’ For this method, it does not matter whether the
types in question are inscription-types, event-types, action-types, or any
other kind of type. It thus seems that the very feature that allows us to
76 MARCUS ROSSBERG
construe types as creatable and destructible, makes them utterly gratuitous

in our ontology, and thus subject to Ockham’s Razor. Translating, for
instance, Levinson’s highly worked out theory of musical works as initiated
types into an equivalent type-less version, following the recipe gestured at
above, would of course not only be laborious but also lead to some awkward
and difficult to comprehend phrasings. Note, however, that this would be
an artifact of translating a theory that mentions types into one that does not,
and not a general feature of theories that dispense with types. The recipe
merely suggests a systematic way of eradicating types from any theory, thus
demonstrating their dispensability in a perfectly general way—it is not a
suggestion that theories should indeed be constructed in this roundabout
way. Formulations that only refer to concrete scores and performances from
the outset—like Goodman’s proposal, for instance—will presumably be
much more straightforward. Either way, it appears that on purely technical
grounds we must conclude that initiated types are either ontological surfeit,
or non-creatable after all.
Equivalence classes of e.g. musical performances or computer art displays
are impure sets, as observed above, which arguably makes their creation
and destruction a consequence of the creation and destruction of
their members.8 Like creatable types, however, the contingency of these
equivalence classes is tied to their dispensability. Goodman and Quine
(1947) showed how a simple form of impure class talk can be eliminated
in favor of class-free discourse, following recipes similar to the one described
above. This simple form of impure class talk is all that is required to employ
equivalence classes of concrete objects or events, however. Alternatively,
if the nominalist austerity seems too much to bear, the talk of impure
classes relevant here can be construed as plural talk without any loss, as
demonstrated by George Boolos (1984). It is worth noting that in set theory
proper (like ZFU) impure sets are not eliminable, since here the set-forming
operations are iterated. {a, {b, c}} is importantly distinct from {a, b, c}; the
former has two members, the latter three. Simple, non-iterated classes,
8 How this could actually be worked out for e.g. musical performances is a non-trivial matter;
performances are events, and it seems that, in particular, a past event cannot strictly speaking be
destroyed. Presentists might hold that a particular event ceases to be as soon as no part of it is still
present—in our case, arguably, when the last note of a musical performance has faded. Perhaps musical
works, if identified with equivalence classes of performances, should be considered to perish as creatable
types do: When there are and can be no more performances.
however, like algorithm-determined equivalence classes of displays or

computer programs (etc.), or classes of musical performances, have no
such theoretical pressure to require their employment. Destructible classes,
thus, like destructible types, have no place in our ontology of artworks,
unless we want to clutter our ontology praeter necessitatem. They will neither
enable us to do any theoretical work that cannot be done by appeal to their
concrete members in lieu of these classes, nor, therefore, furnish us with
a plausible abstract ontology for artworks.
A different tack is chosen by Amie Thomasson. Inspired by Roman
Ingarden’s ideas (see e.g. Ingarden 1962), Thomasson argues that it is
a mistake to try to identify one of the familiar ontological categories as the
one that artworks of certain art forms belong to (Thomasson 2004, p. 88).
Thomasson argues that we should respect the common opinion that works
of music do not exist necessarily; rather, their existence is dependent on
human intentionality, viz. a creative act of a composer, despite their being
abstract objects. Thomasson contends that we have to investigate
new ontological categories since artworks (and other social objects, like
marriages or scientific theories) ‘fall between’ the traditional categories.
What Thomasson requests are ontological categories that incorporate
the dependence of their objects on human intentionality, and that means
for works of art forms that appear to demand an abstract ontology: abstract
artifacts.9
Most previous ontological investigations have tried to identify an
established category of abstracta that would be a plausible candidate for
the ontological category of works of certain art forms and then asked
whether it is plausible that works can be created and destroyed if they
belong to this category. Thomasson, on the contrary, suggests that ‘rather
than trying to make works of art fit into the off-the-rack categories
of familiar metaphysical systems’ we should attempt ‘to determine the
categories that would really be suitable for works of art as we know them
through our ordinary beliefs and practices’ (2004, p. 90).
There is no space here for a detailed discussion of Thomasson’s
proposal,10 but we can observe three points. First, Thomasson advocates a
9 The theory of abstract artifacts was first devised as an account of fictional characters (Thomasson
1999), but Thomasson more recently adopted it to provide answers to questions regarding the ontology
of artworks (Thomasson 2004, 2005, 2006; in fact the idea was already floated in Thomasson 1999).
10 But see Chapter 1 and Chapter 8 in this volume.
78 MARCUS ROSSBERG
general meta-ontological position according to which ontological catego-

ries are determined by our social and linguistic practices. Not only repeatable
artworks but all artworks, including paintings and sculptures (Thomasson
2004, p. 89), and in fact all intentionally created objects are subject to an
ontological reassignment. The category ‘abstract artifact’ is not available on its
own; it cannot just be inserted into traditional ontologies, but it is part of a
revisionary meta-ontology that can only be adopted, or rejected, as a whole.11
Unless Thomasson’s entire framework is accepted, little more is achieved than
giving a name to a problem.
Second, Thomasson’s arguments for the new ontological framework
draw on analogies between artworks and other socially constructed ‘things’
that seem to be created by pronouncement or social convention: marriages,
for instance; or scientific theories; or money (Thomasson 1999, pp. 12–13).
Human intentionality, Thomasson argues, is essential for the relevant
conventions to obtain and thus for these objects to come into being. Let
us grant for the sake of our discussion that marriages etc. are, indeed, bona
fide objects. Thomasson appeals to marriages and money to argue for the
coherence and plausibility of ontological categories that essentially involve
appeal to intentionality. It seems that Thomasson argues that the coherence
and plausibility of ontological categories in tandem with the apparent
demand of our social and linguistic practices for such categories is sufficient
for their existence. Granting also this, the question arises whether sympho-
nies are sufficiently similar to marriages in the relevant respects to warrant
the assignment of a similar ontological status. Further, there is the question
whether the fact that intentionality is required for a work of art to come
about should find its way into its ontology. One might plausibly argue that
the proper place to account for this intentionality is in the theory of art
instead (see e.g. Mag Uidhir 2010; Mag Uidhir and Magnus 2011). To try an
analogy, an important part of the theory of meteorites is that they are objects
that were formed in space, fell to Earth, and survived the impact. Nothing is
a meteorite that did not come about in this way. However, the ontological
category of meteorites arguably is not concerned with any of these features.
Ontologically, meteorites are plainly rocks. No doubt, they are interesting
rocks, due to their provenance, but that is a feature that pertains to the
theory of meteorites, or the definition of ‘meteorite’, and not to ontology.
11 This point has been pressed by Christy Mag Uidhir (this volume, Introduction).
Similarly, the required intentionality in the creation of an artwork is an

interesting feature that belongs to the theory of art, or the definition of ‘art,’
and not to ontology.
Finally, and perhaps most instructive for our purpose, abstract artifacts are
apparently intended to be abstract objects, i.e. particulars. Accordingly,
Stravinsky’s 1919 Firebird Suite would be a particular, like the marriage
between Sandi and Debbie—since it is this particular marriage that came
into existence with the relevant pronouncement—or like some scientific
theory, or Sherlock Holmes,12 or the number 2. However, the push towards
an abstract ontology for e.g. works of music comes chiefly from the
repeatability of works of the relevant art form. As observed above, if it is
the repeatability of certain artworks that is to be accounted for by appeal to
abstracta, then the employment of abstract particulars does not appear
promising: Particulars are not repeatable, whether they are abstract or not.
Interestingly, however, abstract artifacts share significant features with
types—a likening that Thomasson appears to accept (Thomasson 2004,
p. 91 fn. 2). Thomasson suggests, for example, that works of music
can (and presumably will one day) go out of existence, ‘if all copies,
performances and memories regarding them cease to be’ (Thomasson
2004, p. 89; see also Thomasson 1999, p. 149, and cf. p. 7), which is strongly
reminiscent of creatable types, as discussed above. Are, in fact, those abstract
artifacts that are artworks not particulars, but universals? In this case,
they seem to share a lot in common with Levinson’s initiated type.
Moreover, the question raised above for the relevant similarities between
artworks, now construed as universals, and e.g. (particular) marriages
becomes pressing.
It appears that Thomasson is faced with a dilemma. On the one hand, the
theory of abstract artifacts appears to be designed with abstract particulars in
mind and is perhaps most plausible if understood in this way. For the
ontology of art, however, this means that the proposed solution, while
at least offering a hope for the creatability and destructibility of artworks,
lacks the envisioned benefit of accounting for the repeatability of works.
Thus the reason for the appeal to abstracta, lapses. If, on the other hand, we
interpret the relevant abstract artifacts to be types or similar universals,
we not only appear to forfeit the argument from the analogy to other social
12 See footnote 9 above.

80 MARCUS ROSSBERG
objects, but the proposal is also likely to be subject to criticisms analogous

to those raised for types above.
V. Conclusion
The candidates for abstracta that might serve as an ontology of art which we
identified are: platonic universals; structures; classes; types; and perhaps
abstract artifacts. Leaving aside the general question whether any of these
are too suspect a kind of purported entity to employ, the following are our
verdicts for each of the five candidates. (i) Platonic universals, preexisting
and eternal, can neither be created nor destroyed, but only discovered
and forgotten. (ii) The most tempting account of pure abstract structures
groups these with the former, and hence they are also indestructible.
(iii) Classes—more precisely: equivalence classes of concreta—can be
construed as coming into and going out of existence with their members.
Thus, we can arguably destroy a class by destroying its members. This very
nature, however, is precisely what makes classes dispensable: It seems that
any ontology of art that can be formulated in recourse to destructible
equivalence classes can be formulated without appeal to classes. (iv) Types
suffer a similar fate: Whether or not they are identified with equivalence
classes of their tokens, they appear ontologically superfluous if construed in
a destructible way. (v) Abstract artifacts appear to face a fate similar to that
of types—if construed as universals—or unable to solve the problems that
abstracta were brought in to address in the first place—if construed
as particulars.
Generalizing from our findings, the conclusion appears to be that there is
no room for an abstract ontology of art that accounts for the creatability and
destructibility of artworks. The more plausible the proposals become,
the more they emphasize qualities usually attributed to concreta, rather
than abstracta. Genuine abstracta appear to be pushed so far to the fringe
in this type of account that they do not play any significant role. This is what
makes abstracta theoretically superfluous in these proposals. An ontology
that employs abstracta that are not eliminable in the ways suggested above
would have to include features that are of ontological importance or make
sense of phenomena pertaining to artworks that cannot be explained by
the substituted concrete ontology. In other words: The features that block
the elimination must themselves be of theoretical importance or be entailed

by features that are. Barring such a conception, it seems that we cannot
conclude that an abstract ontology of artworks has been found that allows
for the creation and destruction of works that belong to this category. It is
striking that the only ontologies of art that employ abstracta essentially are
platonist in nature and, in keeping with platonism, insist that artworks
are discovered and forgotten, rather than created and destroyed. Should
it turn out that abstracta are strictly indispensable to give a satisfactory
ontology for works of some art form, then we have to accept that works
of this art form are indestructible. Or, contraposing: If we do not want to
give up on the thought that all art can perish, then there does not appear
to be any way that artworks can be abstract objects that is not ontologically
excessive.
References
Armstrong, D.M. (1978) Universals and Scientific Realism, 2 vols (Cambridge:
——(1989) Universals: An Opinionated Introduction (Boulder, CO: Westview Press)
Blass, Andreas and Gurevich, Yuri (2003) ‘Algorithms: A Quest for Absolute
Definitions’ Bulletin of European Association for Theoretical Computer Science 81:
195–225
Boolos, George (1984) ‘To Be is to Be a Value of a Variable (or to Be Some Values
of Some Variables)’ Journal of Philosophy 81: 430–49
Chabert, Jean-Luc and Weeks, C. (1999) A History of Algorithms: From the Pebble to
the Microchip (Berlin: Springer)
Church, Alonzo (1936) ‘An Unsolvable Problem of Elementary Number Theory’
American Journal of Mathematics 58: 345–63
Cohnitz, Daniel and Rossberg, Marcus (2006) Nelson Goodman (Montreal: McGill-
Queen’s University Press)
Currie, Gregory P. (1989) An Ontology of Art (Basingstoke: Macmillan)
Cutland, Nigel (1980) Computability: An Introduction To Recursive Function Theory
(Cambridge: Cambridge University Press)
Davies, Stephen (2001) Musical Works and Performances (Oxford: Oxford University
Press)
——(2003) Themes in the Philosophy of Music (Oxford: Oxford University Press)
82 MARCUS ROSSBERG
Dodd, Julian (2000) ‘Musical Works as Eternal Types’ British Journal of Aesthetics 40:
424–40
——(2002) ‘Defending Musical Platonism’ British Journal of Aesthetics 42: 380–402
Frege, Gottlob (1893 and 1903) Grundgesetze der Arithmetik, 2 vols (Jena: Pohle);
English translation: Basic Laws of Arithmetic, by Philip A. Ebert and Marcus
Rossberg (Oxford: Oxford University Press, forthcoming)
Gabbay, Dov M. and Guenthner, Franz (2001) Handbook of Philosophical Logic, 2nd
edn, Vol 1 (Dordrecht: Kluwer)
Goodman, Nelson (1951/1977) The Structure of Appearance, 3rd edn (Boston:
Reidel)
——(1968/1976) Languages of Art: An Approach to a Theory of Symbols, 2nd edn
(Indianapolis, IN: Hackett)
——and Quine, W.V. (1947) ‘Steps Toward a Constructive Nominalism’ Journal
of Symbolic Logic 12: 105–22
Ingarden, Roman (1962/1989) Untersuchungen zur Ontologie der Kunst: Musikwerk.
Bild. Architektur. Film (Tübingen: Max Niemeyer Verlag); English translation:
Ontology of the Work of Art, by Raymond Meyer with Jon T. Goldthwait (Athens:
Ohio University Press)
Kivy, Peter (1983) ‘Platonism in Music’ Grazer Philosophische Studien 19: 109–29
——(1987) ‘Platonism in Music: Another Kind of Defense’ American Philosophical
Quarterly 24: 245–52
——(1988) ‘Orchestrating Platonism’ in T. Andersberg, T. Nilsun, and I. Persson
(eds), Aesthetic Distinction (Lund: Lund University Press), pp. 42–55
——(2002) Introduction to a Philosophy of Music (Oxford: Oxford University Press)
——(1990) ‘What a Musical Work Is, Again’ in his Music, Art, and Metaphysics
(Ithaca, NY: Cornell University Press) pp. 215–63
——(2012) ‘Indication, Abstraction, and Individuation,’ this volume, Ch. 2
Lopes, Dominic McIver (2001) ‘The Ontology of Interactive Art’ Journal of
Aesthetic Education 35: 65–81
——(2009) A Philosophy of Computer Art (London: Routledge)
Mag Uidhir, Christy (2010) ‘Failed-Art and Failed Art-Theory’ Australasian Journal
of Philosophy 88: 381–400
——and Magnus, P.D. (2011) ‘Art Concept Pluralism’ Metaphilosophy 42: 83–97
Price, Kingsley (1982) ‘What is a Piece of Music?’ British Journal of Aesthetics 22:
322–36
Shapiro, Stewart (2000) Philosophy of Mathematics: Structure and Ontology (Oxford:
Oxford University Press)
Tavinor, Grant (2011) ‘Videogames, Interactivity, and Art’ in David Goldblatt and
Lee B. Brown (eds) Aesthetics, 3rd edn (Upper Saddle River: Prentice Hall),
pp. 382–6
Thomasson, Amie L. (1999) Fiction and Metaphysics (Cambridge: Cambridge
University Press)
——(2004) ‘The Ontology of Art’ in Peter Kivy (ed) The Blackwell Guide to
Aesthetics (Oxford: Blackwell), pp. 78–94
——(2005) ‘The Ontology of Art and Knowledge in Aesthetics’ Journal of Aesthetics
and Art Criticism 63: 221–9
——(2006) ‘Debates about the Ontology of Art: What are We Doing Here?’
Philosophy Compass 1(3): 245–55
Turing, Alan M. (1936) ‘On Computable Numbers, With An Application to
the Entscheidungsproblem’ Proceedings of London Mathematical Society, series 2,
42: 230–65
van Dalen, Dirk (2001) ‘Algorithms and Decision Problems’ in Gabbay and
Guenther (2001), pp. 245–312
Wollheim, Richard (1968/1980) Art and Its Objects, 2nd edn (Cambridge:
Wolterstorff, Nicholas (1975) ‘Toward an Ontology of Art Works’ Noûs 9: 115–42
——(1980) Works and Worlds of Art (Oxford: Clarendon Press)
II
Informative Comparisons
4
Art, Open-Endedness, and
Indefinite Extensibility
ROY T. COOK
I. Introduction
It is widely (although by no means universally) accepted that there can never
be a set of necessary and sufficient conditions specifying exactly which
objects are, or can be, works of art. This idea—that art is, in some sense,
open-ended—is not new. On the contrary, many influential philosophical
accounts of the nature of art, including the views sketched in Danto (1964),
Gaut (2000), Levinson (1979), and Weitz (1956), can plausibly be thought to
entail some version of this thesis. The purpose of this chapter is first, to
become a bit clearer about what this open-endedness amounts to and
second, to examine the connections, or lack thereof, between the sort of
open-endedness found in art and another kind of open-endedness: The
indefinite extensibility often thought to be characteristic of certain prob-
lematic areas of mathematics, such as set theory, the theory of ordinal
numbers, and the theory of (transfinite) cardinal numbers.
We shall carry out this examination in three distinct stages. First, in
Section II, we shall examine some extant accounts of the nature of art,
focusing on the arguments for open-endedness that these accounts entail, in
order to clarify exactly what the open-endedness of art amounts to. Next, in
Section III, we shall compare the open-endedness of art to another philo-
sophically important and widely discussed sort of open-endedness: The
indefinite extensibility of mathematical concepts such as set, ordinal, and
cardinal. The main focus naturally will be determining whether the class of
artworks (or of possible artworks) is indefinitely extensible in the same sense
88 ROY T. COOK
as these mathematical structures, and our conclusions in this section will be

for the most part negative. In Section IV, however, we shall examine a
second kind of open-endedness within mathematics—Dummett’s idea that
the natural numbers are indefinitely extensible. As we shall see, there are
reasons for thinking that the open-endedness of the concept art is more like
the open-endedness of the concept natural number than it is like the
open-endedness of the concept set. This, in turn, has profound conse-
quences for the sort of conclusions that can be drawn based on the open-
endedness of art.
Before moving on, it is worth making explicit exactly how this chapter
does, and does not, connect to the larger theme of this volume. In what
follows we shall make no assumptions regarding whether some, or all, works
of arts are abstract objects. Instead, the justification for including this chapter
in a volume on art and abstract objects is the careful comparison in Sections
III and IV between the open-endedness of art and the open-endedness of
another class of objects that are often taken to be abstract—mathematical
objects. Of course, the nominalist reader has every right to object that this is
not a legitimate venue for the present examination, since from that perspec-
tive the discussion below has absolutely nothing to do with abstract objects
at all. Fortunately, I am a platonist,1 so I have no qualms about including the
arguments given below in the present volume! It is hoped, however, that
the issues discussed below will be of interest to nominalists and platonists
alike.
II. The Open-Endedness of Art

One of the earliest arguments for the claim that the concept art is open-
ended, as well as one of the earliest clear formulations of what it is for a
concept to be open-ended, is due to Morris Weitz. In ‘The Role of Theory
in Aesthetics’ he characterizes the relevant notion of open-endedness as
follows:
A concept is open if its conditions of application are emendable and corrigible, i.e.
if a situation or case can be imagined or secured which would call for some sort of
1 A logicist even! See e.g. Cook (2009a).

ART, OPEN-ENDEDNESS, AND INDEFINITE EXTENSIBILITY 89
decision on our part to extend the use of the concept to cover this, or to close the
concept and invent a new one to deal with the new case . . . (1956, p. 31)
A few things are worth noting from the outset. First, a concept is open-
ended if and only if there is always a possible situation for which our
previous commitments and decisions are insufficient to mandate2 a decision
vis-à-vis that concept in the novel possible situation. Second, although
Weitz speaks here of an ‘imagined or secured’ situation or case, it is clear
from the context that he does not mean for the open-endedness of a concept
to be hostage to our imaginative abilities. On the contrary, it seems clear
that Weitz would count a concept as open-ended if there were always a
possible situation that required a decision on our part, even if the limitations
of our imaginative capacities prevent us, in some cases, from recognizing
these possibilities. Note that we can recognize that in every case there is a
possibility of the relevant sort without being able to recognize the relevant
possibility itself in every case.
Weitz goes on to argue that the concept art is open-ended in this sense,
on roughly Wittgensteinian grounds:
The problem of the nature of art is like that of the nature of games, at least in these
respects: If we actually look and see what it is that we call ‘art’, we will also find no
common properties—only strands of similarities. Knowing what art is is not
apprehending some manifest or latent essence but being able to recognize, describe,
and explain those things we call art in virtue of these similarities. (1956, p. 31)
In short, according to Weitz there is no list of individually necessarily and

jointly sufficient conditions for an object to be a work of art. Instead, works
of arts are tied together by nothing more than Wittgensteinian family
resemblance, and correct judgments involving the concept art depend on
noting similarities between novel candidates and previous instances of the
concept, as well as on freely chosen decisions in novel cases where similarity
to, or difference from, previous instances does not mandate one judgment
2 There are two ways of understanding the failure of mandate in Weitz’s novel possible situation:
(a) Neither application of ‘art’ nor application of ‘non-art’ is mandated, so either judgment (but
presumably not both at once) are permissible.
(b) Neither application of ‘art’ nor application of ‘non-art’ is mandated, so neither judgment is
permissible.
It seems clear from the text, and from his importation of the Wittgensteinian framework, that Weitz has
something like (a) in mind, however.
90 ROY T. COOK
or the other. As a result, for Weitz the extension of the predicate ‘is art’ has
vague or indeterminate boundaries in exactly the sense suggested by Witt-
genstein’s (1953) analysis of the concept game.
It will be useful to nail down the general structure of this argument a bit
more carefully. Arguments for the open-endedness of art, including Weitz’
argument as sketched above, typically proceed in three stages. First, an
argument is given for the open-endedness of the concept in question,
typically stemming from some particular account of the nature of the
concept itself. Second, the open-endedness of the concept art is then
taken to entail that there can be no explicit, non-question-begging defini-
tion of this concept—that is, there is no finite set of individually necessary
and jointly sufficient conditions for an object to be a work of art. Third, the
open-endedness, and undefinability, of the concept art is taken to imply
that the extension of the predicate ‘is art’ is hazy, fuzzy, vague, indetermi-
nate, indefinite, or otherwise imprecise.
These last two steps are often run together (both in the literature on
aesthetics, and in literature on open-endedness more generally). However,
there are good reasons for keeping the thesis about the undefinability of the
concept art distinct from the thesis about the indeterminacy or indefinite-
ness of the extension of the predicate ‘is art.’ The fact that set theory
contains transfinitely many completely determinate, yet undefinable, sets
already suggests that the latter conclusion does not follow from the former.
And it turns out that the open-endedness of art only supports, at best, the
former conclusion.
As already suggested above, the open-endedness of art amounts roughly
to the lack of any finite set of individually necessary and jointly sufficient
conditions for an object to be a work of art. In fleshing this out more fully,
there are two approaches that suggest themselves. The first is to concentrate
on the idea that any sufficient condition for being a work of art fails to be
necessary. The second is to concentrate on the contrapositive of the first—
that any necessary condition for being a work of art fails to be sufficient. For
the sake of convenience we shall focus on the former approach, since this
formulation will facilitate our comparison with indefinite extensibility
below. Thus, we begin with the thought that the open-endedness of art
amounts to the in-principle failure of any sufficient condition for being a
work of art also being a necessary condition.
This simple gloss, while a useful starting point, is inadequate. A natural

formalization of the thought that there exists no sufficient condition for
being a work of art that is also necessary is the:
Simple Open-Endedness Schema:
Given any predicate not containing occurrences of the predicate ‘is art’:
(8y)((y) ! Art(y)) ! (9y)(~ (y) ^ Art(y))
The Simple Open-Endedness Schema does capture something important:
The restriction on the sort of predicate that can be substituted into the
schema. After all, presumably the predicate ‘is art’ expresses a precise
necessary and sufficient condition for being art. Nevertheless, the Simple
Open-Endedness Schema cannot be the right way to express the open-
endedness of art, regardless of the details of the argument given for the
open-endedness claim. The reason is simple: If we assume (i) that every
actual work of art can be picked out (either using a proper name or via a
definite description—note that we are not requiring that we can distinguish
art from non-art, but only requiring that the works of art are amongst the
larger class of objects that can be picked out); and (ii) that the number of
actual works of art is finite,3 it follows that the Simple Open-Endedness
Schema is false. Just let F be a disjunction of identities of the form:
x = Mona Lisa ∨ x = Appetite for Destruction ∨ x = Maus ∨. . .
where each actual work of art (and no non-artwork) is listed. This disjunc-
tion provides a counterexample to the Simple Open-Endedness Schema. Of
course, without some informative, substantial account of the nature of art,
we might not be able to determine which disjunction is, in fact, the one
listing all and only the works of art. This does not change the fact that there
will be such a disjunction, and that this disjunction (whichever one it is)
provides a counterexample to the Simple Open-Endedness Schema.
The fix is simple, however. Instead of claiming that there exists no
condition holding of only actually existing works of art that also holds of
all actual works of art, we can instead formulate the open-endedness thesis as
the claim that there is no sufficient condition that holds of all actually
existing and possible (future) works of art. More formally:
3 Note that this is much weaker, and more plausible, than the claim that the class of possible artworks
is finite.
92 ROY T. COOK
Weak Modal Open-Endedness Schema:

(8y)((y) ! Art(y)) ! ◊(9y)(~ (y) ^ Art(y))4
This understanding of open-endedness already marks a distinction between
the open-endedness found in art and any similar open-endedness found in
mathematics. Unless one is a quite literal, Brouwerian constructivist,5 there
is no significant distinction between those mathematical objects that actually
exist and those that only possibly exist. The distinction between actual and
possible works of art is crucial to properly understanding the open-ended-
ness of art, however.
Depending on the details of the particular argument given for the open-
endedness of art, we might conclude that the open-endedness of art is not
just a contingent fact about the actual world, but a necessary consequence of
the nature of the concept art itself. As a result, the concept art might not
only be open-ended in the sense sketched above, but necessarily so,
prompting us to adopt the:
Strong Modal Open-Endedness Schema:
□[(8y)((y) ! Art(y)) ! ◊(9y)(~ (y) ^ Art(y))]
None of the arguments below hinge on which version of the thesis we
adopt.
The Modal Open-Endedness Schema (in either its weak or strong form)
seems to capture an important part of Weitz’s conception of the openness of
art: Given any purported necessary condition for being a work of art—that
is, in Weitz’s terminology, any condition that holds of all objects that we
have chosen to call art up to now—there exists a possible situation involving
some object where the necessary condition fails to hold of that object. In
addition, in this novel situation we can legitimately decide to treat the
object in question as a work of art, thus making it one. If there were no
such possible situation—that is, if necessarily every work of art has the
4 Alternatively, we could have retained the formulation of the original open-endedness schema and
reinterpreted the quantifiers as ranging over possible or future as well as actual objects. The formulation
given above better accords with standard logical practice, however.
5 L.E.J. Brouwer believed that mathematical objects were actual mental constructions, and as a result,
the only mathematical objects that existed at a particular time were those that had been explicitly
constructed. See Brouwer (1949).
condition in question—then this would constitute a ‘common property’ of

exactly the sort Weitz rejects. Of course, we can also choose not to treat that
object as a work of art, but the mere possibility of its being a genuine
instance of the concept art (in virtue of our freedom to legitimately treat it
as such) is sufficient to support the Modal Open-Endedness Schema.6
Other views that support this understanding of the open-endedness of art
are not hard to come by. For example, Arthur Danto sketches an account of
the development of the concept art at the end of ‘The Artworld’ (1964) that
implies a version of the open-endedness schema. Loosely put, on Danto’s
account works of art are taxonomized in terms of pairs of predicates that are
‘opposites,’ where two predicates and Y are opposites (relative to some
sortal ) if and only if:
(8x)((x) ! ((x) $ ~ (x)))
Letting the relevant sortal be ‘art created up to and including time t’, it
follows that, at time t, there is a fixed set of opposites that define possible
artistic styles, and all works of art at that time can be categorized in terms of
the relevant stock of ‘opposites’ in operation at time t. Danto then provides
the following description of the evolution of these characterizing opposites:
Suppose an artist determines that H shall henceforth be artistically relevant for his
paintings. Then, in fact, both H and non-H become artistically relevant for all
painting, and . . . the entire community of painting is enriched . . . The greater the
variety of artistically relevant predicates, the more complex the individual members
of the artworld becomes; and the more one knows of the entire population of the
artworld, the richer one’s experience with any of its members. (1964, pp. 583–4)
Of particular interest is Danto’s description of an ‘artistic breakthrough’:

An artistic breakthrough consists . . . in adding a column [a new pair of opposites] to
the matrix. Artists then, with greater or less alacrity, occupy the positions thus
opened up: This is a remarkable feature of contemporary art, and for those
unfamiliar with the matrix [of currently salient opposites] it is hard, and perhaps
impossible, to recognize certain positions as occupied by artworks. Nor would
6 Weitz’s view actually seems to support a stronger thesis: That there are no non-trivial necessary
conditions for being a work of art at all. That is, given any description not containing occurrences of
the predicate ‘Art’:
(9y)( ~ (y)) ! ◊(9y)(~ (y) ^ Art(y))
94 ROY T. COOK
these things be artworks without the theories and the histories of the Artworld.
(1964, p. 584)
Note that opposites, for Danto, need not be either exclusive or exhaustive
in general—they need only be exclusive and exhaustive with respect to the
class of works of art created as of time t. As a result, on Danto’s view, there
could be an artistic breakthrough where an artist decides that property H is
henceforth artistically relevant, and that his works shall have (and be
understood in terms of ) property H, but where no works of art prior to
this decision have H. The artist in question could then produce works that
have property H but which lack all of the opposites relevant to art prior to
this imagined artistic breakthrough. Note that this process requires not only
the artist’s introduction of works that have property H, but also requires the
audience and artworld’s acceptance of H as artistically relevant.
If we make the further assumption that any significant necessary condi-
tion for an object being a work of art will be formulated in terms of the
properties defining various artistic styles, then Danto’s account of artistic
progress entails something much like the Modal Open-Endedness Thesis:
Given any sufficient condition formulated in terms of the ‘opposites’
relevant to the concept Art at time t, there will be a possible work of
art—the result of an artistic breakthrough of the sort described above—that
will fail to have this condition (and further, it will be a work of art, in part, in
terms of its making a new pair of opposites salient for evaluating and
categorizing art).
While Danto’s account of artistic breakthrough does entail something like
the open-endedness of art, on closer examination it appears that something
fishy is going on here. Unlike Weitz, Danto does believe that there is a
relatively simple necessary and sufficient condition for being a work of art:
An object is a work of art for Danto if and only if it expresses some sort of
content about a subject in such a way as to require a certain sort of audience
participation, where that participation requires, in turn, specialized art-
historical knowledge or context (Danto 1964). Note, however, that this
condition is not a condition holding merely of the object in question; rather,
it is a relational condition expressing the existence of complex interactions
between artist, artwork, audience, and (most notoriously) the artworld.
As a result, we need to further restrict the conditions that can be
legitimately plugged into our Open-Endedness Schema if we wish to
claim that Danto’s view entails the open-endedness of art in any substantial,
interesting sense. On Danto’s view, there is no substantial, non-trivial
necessary condition for being a work of art that can be expressed solely in
terms of properties of the object and that does not involve intentional states
of the audience, artist, or artworld. Note that the non-relational properties
of a purported object, understood in this way, need not be identical to the
aesthetically relevant ‘opposite’ pairs discussed above—on the contrary,
many pairs of ‘opposites’ relevant to the categorization of, and appreciation
of, art will involve properties tied up with the intentional states of artists,
audience, and others—that is, they will be relational in the relevant sense.
Nevertheless, any non-relational property could be introduced as one ele-
ment in a pair of relevant ‘opposites,’ and this is enough to carry out a
version of the argument above for the open-endedness of art. Thus, Danto’s
view involves something like the:
Weak Modal Non-Relational Open-Endedness Schema:7
Given any non-relational predicate8 not containing occurrences of ‘is art’:
(8y)((y) ! Art(y)) ! ◊(9y)(~ (y) ^ Art(y))
There is one final observation that needs to be made regarding Danto’s
characterization of the open-endedness of art in terms of opposite pairs.
Unlike Weitz’s view, which merely implies that there is no description that
precisely and determinately distinguishes the class of (possible and actual)
works of art from the class of (possible and actual) non-artworks, Danto’s
view implies a substantial further claim: That any object whatsoever—or at
least any object that we can name, describe, or otherwise pick out
uniquely—could, in the right circumstances, be a work of art. After all,
there seem to be no a priori constraints on the sort of properties and
that can serve as the basis of an opposite pair (other than that such
properties must be contradictories relative to an appropriate sortal). As a
result, any object can be made a work of art, so long as the object is
accompanied by an artistic breakthrough of the sort described above—one
7 There is, of course, an analogous strong version resulting from prefixing a necessity operator.
8 Of course, a full explication of Danto’s view would require us to determine in some detail exactly
what does, and does not, count as a relational predicate. This problem has vexed philosophers as far back
as Leibniz, but since all we need for present purposes is the claim that Danto accepts some substantial
version of the Open-Endedness Schema, we need not enter into these difficult matters here.
96 ROY T. COOK
that introduces at least one new opposite pair in virtue of which the new
artwork (and, as a result, all old artwork) is evaluated.9
This conclusion should not be taken to imply that random objects can
become works of art willy-nilly. After all, we cannot make a urinal a work
of art merely by stipulating that it is a work of art. On Danto’s account,
however, we can make (and Duchamp has made) a urinal a work of art by
specifying (at least implicitly) the relevant properties that are relevant to our
appraisal of the urinal as an artwork (in other word, by specifying the newly
relevant opposite pair or pairs). We should not be fooled by this into
thinking that such artistic breakthroughs are trivial or easy, however, since
on Danto’s view the successful introduction of a novel pair of opposites
requires (eventual) acquiescence on the part of audience and artworld.
Nevertheless, it seems plausible that, on Danto’s account, any object what-
soever can be a work of art so long as (i) it can be introduced along with a
relevant opposite pair (or pairs); and (ii) the artistic relevance of these pairs is
(or at least could be) accepted as artistically relevant by the audience and
artworld.
Other popular accounts of the nature of art support some version of the
open-endedness thesis. For example, cluster accounts such as those explored
in Gaut (2000), where an object is a work of art if, loosely speaking, it has
‘enough’ of the properties from some relevant list supports a version of the
Weak Modal Non-Relational Open-Endedness Schema (and possibly the
earlier, non-relational version, depending on the exact details of the list).
Likewise, on Levinson’s (1979) historical definition of art, any object
whatsoever can presumably be art so long as the artist has proprietary rights
to the object and has the right intentions towards the object, and this again
entails something much like the Modal Non-Relational Open-Endedness
Schema. Thus, many, if not most, views on the nature of art entail some sort
of open-endedness, and as a result it is worth examining more closely the
consequences that do, and do not, follow from the Modal Non-Relational
Open-Endedness Schema. In particular, we shall examine two claims
that might be taken to follow from the open-endedness of art (i) that the
open-endedness/indefinite extensibility of art entails that there can be no
definition of the concept art; and (ii) the open-endedness/indefinite
9 Arguably, Danto’s After the End of Art (1998) can be read, at least in part, as pushing this line of
reasoning in its most extreme form.
extensibility of art, combined with the indefinability of the concept art,

entails that the extension of the predicate ‘is art’ is hazy, vague, indefinite, or
indeterminate.
III. Art, Open-Endedness and Indefinite Extensibility

In this section and the next we shall attempt to flesh out the idea that the
concept art is open-ended by comparing the open-endedness of art to
another sort of open-endedness that has received a good bit of philosophical
attention—the indefinite extensibility of certain mathematical concepts. In
this section we shall compare and contrast the open-endedness of art and the
indefinite extensibility of mathematical concepts associated with the late-
19th century set-theoretic paradoxes: the concepts cardinal number;
ordinal number; and set.10 In Section IV we shall then compare the
open-endedness of art to the indefinite extensibility of a less paradox-prone
mathematical concept: natural number.
Before examining the open-endedness of cardinal number, ordinal
number, and set, it is worth noting that Weitz, in his argument for the
open-endedness of art, explicitly contrasts art with mathematical concepts,
which he takes to be paradigm instances of closed concepts:
If necessary and sufficient conditions for the application of a concept can be stated,
the concept is a closed one. But this can only happen in logic or mathematics where
concepts are constructed and completely defined. (1956, p. 31)11
The idea that at least some mathematical concepts—in particular, the ones
leading to the Russell, Cantor, and Burali-Forti paradoxes—are open-
ended in a sense at least similar to the sense in which Weitz believes that
10 One might wonder, of course, whether the concepts set, ordinal number, and cardinal number
as grasped by 19th century mathematicians mobilizing the (inconsistent) naı̈ve versions of these notions
are the same concepts as those grasped by 20th and 21st century mathematicians (and whether pre-
paradox mathematicians working with the naı̈ve version of these notions were grasping a genuine
concept at all). For our purposes, however, we need only note that the paradoxes generated by the naı̈ve
conceptions of these notions led to a new (and, one hopes, consistent) understanding of these concepts as
indefinitely extensible.
11 To be absolutely fair, Weitz only says that closed concepts only occur within mathematics and logic,
and not that all mathematical and logical concepts are closed. It is hard not to read the latter claim into his
comment about mathematical concepts being ‘constructed and completely defined’ however.
98 ROY T. COOK
art is open-ended is well established, however, and can be traced back to

Russell’s early writings on the paradoxes half a century earlier:
. . . the contradictions seem to result from the fact that . . . there are what we may
call self-reproductive processes and classes. That is, there are some properties such
that, given any class of terms all having such a property, we can always define a new
term also having the property in question. Hence, we can never collect all of the
terms having the said property into a whole; because, whenever we hope we have
them all, the collection which we have immediately proceeds to generate a new
term also having the said property. (1906, p. 144)
Loosely speaking, self-reproductive classes such as the class of sets, the class of
cardinal numbers, and the class of ordinal numbers, cannot be collected
together (despite the somewhat disingenuous use of the term ‘class,’ which
explains our intuitions to the contrary). Any attempt to do so allows us to
identify additional members of the class that are not contained in the collec-
tion. The paradoxes, on Russell’s view, are thus the result of attempting to
collect together all objects falling into a class that cannot be so collected.
Michael Dummett provides a clearer formulation of this idea, and he also
introduces the now standard terminology ‘indefinite extensibility’:
The paradoxes—both the set theoretic and the semantic paradoxes—result from
our possessing indefinitely extensible concepts . . . An indefinitely extensible con-
cept is one for which, together with some determinate range or ranges of objects
falling under it, we are given an intuitive principle whereby, if we have a suffi-
ciently definite grasp of any one such range of objects, we can form, in terms of it, a
conception of a more inclusive such range . . . By the nature of the case, we can
form no clear conception of the extension of an indefinitely extensible concept; any
attempt to do so is liable to lead us into contradiction. (1993, p. 454)
The following definition seems in line with Dummett’s understanding of

indefinite extensibility:
A concept C is indefinitely extensible if and only if there is a function or
operation ƒ mapping collections of objects to objects such that, given any
definite collection ˜ of objects falling under C:
(a) ƒ(˜) falls under C
and:
(b) ƒ(˜) 2
=˜
For our purposes, we can understand the term ‘definite collection’ to be co-
extensive with set.12 The idea that paradox-prone mathematical concepts
such as set, cardinal number, and ordinal number are indefinitely exten-
sible is now a widely accepted part of the mathematical folklore13 (although
there is more disagreement regarding what consequences follow from this
claim and what measures should be taken to insure that indefinitely exten-
sible concepts do not give rise to paradoxes—see Shapiro and Wright 2006
for a more detailed discussion). In addition, I have argued elsewhere that the
concept semantic status (or, a bit more loosely, the concept truth value)
is indefinitely extensible (see Cook 2008, 2009b). Our purpose here is to
determine whether there open-endedness of the concept art can be viewed
along the same lines.
Recall that one of the reasons for interest in the open-endedness of the
concept art is the thought that this open-endedness might entail (i) that
there can be no precise definition of art; and (ii) that the extension of the
predicate ‘is art’ is vague, imprecise, or indeterminate. Thus, if the concept
art turns out to be indefinitely extensible in the same sense as the concepts
set, ordinal number, and cardinal number, then this would provide
substantial evidence for the claim that the concept is indefinable and that
its extension is indeterminate (or, at least, there would be as much reason to
think so as there is reason to think that the mathematical concepts in
question are indefinable, and that their extensions are indeterminate).
There are three (somewhat interconnected) reasons for thinking that the
concept art is not indefinitely extensible in the same sense as are the
concepts set, cardinal number, and ordinal number, however.
First, there is the simple fact that there do not seem to be enough works
of art, or even enough possible works of art. One paradigmatic feature of
indefinitely extensible concepts such as set is that the indefinite nature of
sets entails that the sets form a proper class. In other words, loosely speaking,
there are more sets than can be ‘fit’ into a single set, and this continues to
12 This move would, of course, be viciously circular if our main goal was to gain a deeper understand-
ing of the indefinite extensibility of the concept set. Since our goal is merely to compare and contrast the
indefinite extensibility of set (and related notions) to the open-endedness of the concept art, however,
the assumption is harmless.
13 For example, given any set of sets X, the power-set of the transitive closure of X is not in X, and
given any set of ordinals numbers Y, the ordinal number corresponding to the order type of all ordinals
less than or equal to some ordinal in Y (on the standard ordering) is not in Y.
100 ROY T. COOK
hold no matter how we strengthen our set theory and no matter how large
the sets are that we eventually accept into our domain.14 The collection of
possible works of art is, however, almost certainly set-sized: Presumably the
identity conditions for works of art are determined by some combination of
physical characteristics of the work (or an instance of, token of, or template
for it), its connections (physical, cultural, historical, etc.) to other relevant
objects, and the psychological states of relevant agents. While the class of
works of art that can be obtained by varying one or more of these factors is
no doubt massive, it is still much smaller than the first inaccessible cardinal
number (the size of the smallest standard model of ZFC set theory).
It might appear as if we have pulled a fast one at this point, however.
After all, didn’t we just argue in the previous section that, at least on some
conceptions of the open-endedness of art (such as Danto’s), any object can
be a work of art? If so, then if sets are objects and any set can be a work of
art, then the class of possible works of art contains all sets, and is thus at least
as ‘large’ as the proper class of all sets! Recall, however, that the claim made
in the discussion of Danto’s opposite pairs account was more carefully
worded than this. What was argued for is the claim that any object that
we can name or uniquely describe can be a work of art. This does entail that
many abstract objects, including infinitely many sets, can be works of art, but
so long as our language is countable (or even uncountable but set-sized),
there will be many more sets that we cannot so name or uniquely describe,
and these cannot be art.15 In other words, if a is a set that is neither nameable
nor uniquely describable in our (set-sized) language of set theory, then not
even a latter-day mathematical Duchamp can bring it about that a is a work
of art. Thus, the class of possible works of art remains too small, in the
relevant sense, to be indefinitely extensible in the same sense as are sets,
ordinal numbers, and cardinal numbers.
14 This objection depends on accepting something akin to either the iterative conception of set or the
limitation-of-size conception of set as the underlying rationale for the set theory we adopt (which will as
a result be something along the lines of first- or second-order ZFC, depending on one’s views on logic).
For those readers who prefer some other conception of set (and as a result adopt an axiomatization of set
theory substantially different that ZFC), the second and third reasons for thinking that the concept art is
not indefinitely extensible in this sense will, it is hoped, suffice.
15 Of course, we are being somewhat generous here. There remains the rather interesting, but likely
quite difficult, task of specifying in detail exactly how—even on Danto’s account—one might go about
making it the case that the empty set (or any pure set) is a work of art.
Second, indefinitely extensible concepts are one of the main motivations

for rejecting absolutely general quantification. More often than not, when
we assert statements involving universal quantification, the universal quan-
tifier is restricted to some relatively small domain. For example, when I say
‘Everyone has arrived’ I presumably mean that all the invited guests have
arrived, not that every human being on the planet has arrived. Reflecting on
this, it makes sense to ask whether we can ever use ‘all’ (or any other form of
universal quantification) to quantify over absolutely all objects, instead of
over one or another restricted domain—that is, to ask whether absolutely
general quantification is possible. Opponents of absolutely general quantifi-
cation point to the existence of indefinitely extensible concepts: If quantifi-
cation over a class of objects requires that the objects be collected together
into a definite domain (read: set), then we cannot quantify over all sets, or all
ordinal numbers, or all cardinal numbers (and hence we cannot quantify
over all objects whatsoever), since we cannot collect these objects into a
single, determinate domain.
The problem, for our purposes, is this: Although there is something at
least plausible about the claim that we cannot quantify over all sets, or all
ordinal numbers, or all cardinal numbers, there seems to be no good
reason—no inclination at all, even—for rejecting the idea that we can
quantify over all works of art, or even over all possible works of art. If the
indefinite extensibility of a concept entails that we cannot quantify over all
instances of that concept, then this strongly suggests that the concept art is
not indefinitely extensible (at least, not in this sense).
Third, both the weak and strong versions of the Modal Open-Endedness
Schema are consistent.16 The Russell/Dummett account of indefinite ex-
tensibility, however, is intended to be a diagnosis of the problematic aspects
16 The following is a simple model showing the deductive consistency of the Strong Modal Open-
Endedness Schema. Let our language be the language of Peano Arithmetic supplemented with a
necessity operator and a primitive non-logical predicate ‘Art.’ Then:
M ¼ <W; R; I>
W ¼ fwn : n [ øg
R ¼ f<wn ; wm > : n mg
Each world a standard model of arithmetic where:
IðArt; wn Þ ¼ fm : m ng
Needless to say, this model is not meant to provide any insight into the actual ‘behavior’ of the concept
art, but merely to provide a formal proof of the consistency of the schema in question.
102 ROY T. COOK
of concepts such as set, cardinal number, and ordinal number that, when
left untreated, lead to paradox. There are no pre-theoretic paradoxes
involving the concept art that the indefinite extensibility account might
help us to understand however, so there seems to be little reason to think
that the concept art is indefinitely extensible in this sense. To put things
bluntly, if the concept art doesn’t have the symptoms, then there seems
little reason to apply the diagnosis.
Thus, the (relatively speaking) limited number of possible works of art,
our apparent ability to quantify over all possible works of art, and the lack of
any pre-theoretic paradoxes associated with the concept art all tell against
the conclusion that this concept is indefinitely extensible in a manner similar
to the indefinite extensibility of cardinal number, ordinal number,
and set.
IV. Art, Open-Endedness, and Arithmetic

Although the concept art is not indefinitely extensible in the same sense as
are the concepts set, ordinal number, and cardinal number, we should
not be too quick to reject the more general idea that the concept art is
indefinitely extensible in some sense. It turns out that Dummett intended
his account of indefinite extensibility to apply more widely than the set-
theory-centric discussion of the notion above would indicate. In addition to
arguing that set theoretic concepts were indefinitely extensible, Dummett
also proposed that there was a substantial and important sense in which
apparently simpler mathematical concepts (with far fewer instances) such as
natural number are indefinitely extensible:
While it is impossible to give a precise and coherent characterization of the totality
of all objects that might be called ‘ordinal numbers’, it is not impossible to give such
a characterization to the totality of natural numbers . . . the concept . . . [natural
number] . . . itself is perfectly definite, but our language prohibits us from giving
explicit expression to it; the concept guides us, however, in approaching ever more
closely to this unattainable ideal. (1963, p. 197)
Dummett’s idea here is simple: Unlike the case with set and similar set-
theoretic notions, it is not the class of natural numbers itself (i.e. the class of
objects) that is indefinitely extensible. Instead, it is the class of truths about
natural numbers that is indefinitely extensible. I shall call this notion Gödelian
indefinite extensibility (for reasons that shall soon become obvious). We can
formalize this alternative understanding of indefinite extensibility as follows:
A concept C is Gödelian indefinitely extensible if and only if there is a

definable function or operation ƒ mapping collections of objects to objects
such that, given any definable collection ˜ of objects falling under C:
(a) ƒ(˜) is in C
and:
(b) ƒ(˜) 2
=˜
The key difference between Gödelian indefinite extensibility and the earlier
notion is that the Gödelian version requires not only definiteness but
definability.
If we understand the term ‘definable’ to be equivalent to ‘recursive,’ then
Gödel’s first incompleteness theorem entails (roughly, with various standard
caveats regarding consistency in place) that, given any mathematical theory
T containing basic arithmetic, the concept truth-of-T is Gödelian indefin-
itely extensible (the function ƒ just maps each consistent set of statements
D—i.e. each theory—onto the Gödel sentence for D). Further, the Göd-
elian indefinite extensibility of the concept truth-of-T entails that the
following condition holds:17
Arithmetic Open-Endedness Schema:

Given any arithmetic predicate not containing occurrences of the
predicate ‘True,’ where ‘True’ is a primitive predicate expressing ‘truth
of T’:
(8)((<>) ! True(<>)) ! (9)( ~ (<>) ^ True(<>))18
The similarity between this version of mathematical open-endedness and
our earlier formulation of the open-endedness of the concept art is obvi-
ous. This suggests that art is indefinitely extensible in something much like
17 Strengthening this to a biconditional requires an additional choice-like assumption of the form:

(8)(9)(R(, ) ! (9ƒ)(8)(R(,(ƒ()))
18 <> is the Gödel code of formula F. The principled and systematic ambiguity introduced into the
discussion through the use of natural numbers as names for the truths (and falsehoods) or arithmetic
strengthens Dummett’s implicit claim that the (Gödelian) indefinite extensibility of the concept truth
of arithmetic amounts, in the end, to the indefinite extensibility of the concept natural number itself.
104 ROY T. COOK
the weaker, Gödelian sense: The only difference between the indefinite
extensibility of the concept arithmetic truth and the indefinite extensi-
bility of the concept art is that, given a particular sufficient condition for
arithmetic truth, a truth not falling under this condition can be found.
Given a sufficient condition for art, however, a work not meeting this
condition might need to be created (hence the possibility diamond in the
formulation for art). Unlike the indefinite extensibility of concepts like set,
however, this sense of indefinite extensibility does not entail anything about
the size or paradoxicality of the class of instances of the concept in question,
nor anything about our ability to quantify over all of them at once.
Now that we have a better grasp of the connections between the open-
endedness of art and indefinite extensibility, it is time to examine what
conclusions we can draw. As already noted, the two conclusions at issue are
(i) that the open-endedness/indefinite extensibility of art entails that there
can be no definition of the concept art; and (ii) the open-endedness/
indefinite extensibility of art (perhaps combined with the indefinability of
the concept art) entails that the extension of the predicate ‘is art’ is hazy,
vague, indefinite, or indeterminate.
We can handle the first of these simply and decisively. As we have already
seen, the open-endedness, and hence the indefinite extensibility, of the
concept art is consistent with explicit, precise definitions of art such as
Danto’s definition or Levinson’s definition—at least, this is the case if we
understand the open-endedness along the lines of the Modal Non-
Relational Open-Endedness Schema. Thus, the Gödelian indefinite exten-
sibility of the concept art does not by itself entail that a precise set of
necessary and sufficient conditions for being an instance of ‘is art’ is impos-
sible. What is entailed, however, by the open-endedness and indefinite
extensibility of art as understood above is the non-existence of a definition
of the concept art that can be formulated purely in non-relational terms. As
a result, any adequate definition of art must include relational notions (such
as those included in both Danto’s and Levinson’s accounts).
Of course, it is plausible that the stronger version of the thesis lacking the
restriction to non-relational predicates—that is, the version of the Modal
Open-Endedness Schema attributed to Weitz above—is incompatible with
the existence of any precise definition of art. Of course, this is in one sense
unsurprising, since one of Weitz’s explicit goals is to argue against any such
definition (following Wittgenstein). The critical point for our purpose,
however, is this: There are substantial and highly non-trivial versions of the
thesis that the concept art is open-ended that are compatible with the
existence of a precise definition of the concept art.
The second issue—whether the extension of the predicate ‘is art’ is hazy,
vague, or indefinite, is of more interest. Dummett himself argues that the
correct logic for reasoning about indefinitely extensible concepts of both
the original and the Gödelian variety, such as set and truth of arithmetic,
is intuitionistic logic (see e.g. Dummett 1963, 1993). If this were right, then
it would provide good reasons for thinking that the extension of the
concept art (like the extensions of set and truth of arithmetic, on this
account) is indeterminate. After all, the rejection of bivalence on the part of
intuitionistic logicians can be fruitfully understood as the rejection of a
sharp, determinate boundary separating the truths (and hence, in this
particular case, true predications of ‘is art’) from falsehoods (see e.g. Tennant
1996). Thus, if the logic for reasoning about art, and, in particular, for
reasoning about the applicability conditions for the predicate ‘is art’ were
intuitionistic, then this would provide good reasons for thinking that the
extension of the predicate is indeterminate, and indeterminate in a particu-
larly well-understood way.
There is a general consensus that Dummett is wrong about the logic of
indefinitely extensible concepts, however. Even if intuitionistic logic is
attractive for other reasons,19 and even if intuitionistic logic should as a
result be applied when reasoning about any concepts, including indefinitely
extensible concepts, the indefinite extensibility of these concepts does not,
on its own, provide an argument for the correctness of intuitionistic logic.
The paradoxes and puzzles that plague indefinitely extensible concepts such
as set and truth of arithmetic are not blocked merely by adopting
intuitionstic logic and rejecting bivalence (see Williamson 1998 for useful
discussion). Thus, the indefinite extensibility of the concept art gives us no
more reason to accept that the extension of ‘is art’ is vague or indeterminate
than the indefinite extensibility of truth of arithmetic gives us any reason
to think that there is no sharp distinction between those statements of
arithmetic true on the standard model and those statements that are false
on the standard model.
19 It is! See Cook (forthcoming)

106 ROY T. COOK
Reflecting on the arguments for open-endedness and indefinite extensi-

bility given above, however, this is as it should be. After all, the open-
endedness and indefinite extensibility of art seems to have nothing to do
with any (metaphysical) indeterminacy regarding the extension of the
predicate ‘is art’ (recall the point made above regarding the fact that, at
any time t, there will be a finite, albeit long, predicate holding of exactly the
works of art that exist up to time t). Rather, it has to do with limitations in
our ability to delineate the boundaries of this precise, determinate exten-
sion. In short: We are never in an epistemologically privileged position
where we can list or exactly (non-relationally) describe all objects that will or
could be works of art, just as we are never in a position where we can list or
exactly describe all statements that are true of the natural numbers. This is
consistent with there being a precise, determinate set of truths of arithmetic,
and thus consistent with there being a precise, determinate set of works of
art (and a precise, determinate set of objects that will be works of art, and a
precise and determinate set of objects that could be works of art). The open-
endedness, and indefinite extensibility, of these concepts does not reflect
any metaphysical haziness or indefiniteness in their extensions, but instead
reflects an epistemological and linguistic limitation on our ability to ‘get at’
their extensions in a precise way.
V. Conclusion
As we have seen, if the concept art is open-ended, then it is open-ended in
a sense similar to the (Gödelian) indefinitely extensibility of the concept
natural number. This open-endedness, while of immense philosophical
interest, entails neither the undefinability of the concept art nor the inde-
terminacy or indefiniteness of the extension of the predicate ‘is art.’ Instead,
the openness of art is not a metaphysical thesis regarding any indeterminate-
ness in the class of works of art, but instead reflects an epistemological
limitation, our in-principle inability to precisely (non-relationally) specify
exactly which objects are, will be, or can be works of art. These conclusions,
hopefully, provide some novel insights into the nature of both individual
works of art and the concept art. Perhaps more importantly, however,
the arguments and issues examined above provide a starting point for
more substantial interaction and collaboration between two areas of
philosophy that have traditionally resided at opposite, unconnected ends

of the philosophical spectrum: the philosophy of mathematics and the
philosophy of art.
References
Brouwer, L.E.J. (1949) ‘Consciousness, Philosophy, and Mathematics’ in Proceed-
ings of the 10th International Congress of Philosophy, Amsterdam 1948 3: 1235–49
Cook, Roy T. (2008) ‘Embracing Revenge: On the Indefinite Extensibility of
Language’ in J.C. Beall (ed) Revenge of the Liar: New Essays on the Paradox
(Oxford: Oxford University Press), pp. 31–52
——(2009a) ‘New Waves on an Old Beach: Fregean Philosophy of Mathematics
Today’ in Oystein Linnebo and Otavio Bueno (eds) New Waves in Philosophy of
Mathematics (Aldershot: Ashgate), pp. 13–34
——(2009b) ‘What is a Truth Value, and How Many Are There?’ Studia Logica 92:
183–201
——(forthcoming) In D. Cohnitz, P. Pagin, and M. Rossberg, Erkenntnis (Special
Issue): Monism, Pluralism, Relativism: New Essays on the Status of Logic
Danto, Arthur (1964) ‘The Artworld’ Journal of Philosophy 61: 571–84
——(1998) After the End of Art (Princeton: Princeton University Press)
Dummett, Michael (1963) ‘The Philosophical Significance of Gödel’s Theorem’
Ratio 5: 140–55
——(1993) The Seas of Language (Oxford: Oxford University Press)
Gaut, Berys (2000) ‘Art as a Cluster Concept’ Theories of Art Today. Edited by Noel
Carroll (Madison, WI: University of Wisconsin Press: 25–44)
Levinson, Jerrold (1979) ‘Defining Art Historically’ British Journal of Aesthetics 19:
232–50
Russell, Bertrand (1906/1973) ‘On Some Difficulties in the Theory of Transfinite
Numbers and Order Types’ in Bertrand Russell, Essays in Analysis (New York:
Braziller), pp. 135–64
Shapiro, Stewart and Wright, Crispin (2006) ‘All Things Indefinitely Extensible’ in
Agustin Rayo and Gabriel Uzquiano (eds) Absolute Generality (Oxford: Oxford
University Press), pp. 255–304
Tennant, Neil (1996) ‘The Law of Excluded Middle is Synthetic A Priori, if Valid’
Philosophical Topics 24: 205–29
Weitz, Morris (1956) ‘The Role of Theory in Aesthetic’ Journal of Aesthetics and Art
Criticism 15: 27–35
Williamson, Timothy (1998) ‘Indefinite Extensibility’ Grazer Philosophische Studien
55: 1–24
Wittgenstein, Ludwig (1953) Philosophical Investigations (London: Blackwell)
5
Historical Individuals Like Anas
platyrhynchos and ‘Classical Gas’
P.D. MAGNUS
In thinking about art ontology, we can distinguish two diametrical

approaches: One—which we might call revisionist—abandons any intuitions
that we have about art in order to make the account metaphysically
respectable. The second—call it reactionary—accepts wild and extravagant
metaphysics in order to preserve our presumptions about art. If these were
the only two options, then there would be no hope of giving an art
ontology which respects both art and ontology. What a disaster that
would be! In this chapter, I consider the idea that musical works are historical
individuals. Guy Rohrbaugh (2003) has proposed this for artworks in
general, and Julian Dodd (2007) objects that the proposal is outré metaphys-
ics, too far beyond the pale to take seriously. Their disagreement could
easily be seen as a skirmish in the broader war between revisionists
and reactionaries. Which of metaphysics and art should trump the other
when there is a conflict? It is a morass of philosophical methodology.
Fortunately, the ontology of works as individuals need not devolve
into that—or so I argue. My primary strategy is to show, contra Dodd’s
accusation, that historical individuals are familiar parts of the world.
Although the ontological details are open to debate, it is the standard
opinion of biologists is that biological species are historical individuals. So
there is no conflict here between fidelity to art and respectable metaphysics.
What suits species will fit musical work as well.
In Section I, I introduce the opposition between revisionists and
reactionaries. In Section II, I review Rohrbaugh’s argument and Dodd’s
objections. In Section III, I consider whether pluralism could dissolve rather
HISTORICAL INDIVIDUALS LIKE ANAS PLATYRHYNCHOS 109
than resolve the disagreement. No, I argue, because even for the pluralist a
core disagreement remains; just whether the view of musical works as
historical individuals is even minimally sensible. In Section IV, I review
the notion of species-as-individuals. Although this general idea is the majority
view in biology, there is no consensus about the precise metaphysics of
historical individuals. Nevertheless, it gives us a sketch that we can deploy to
give Rohrbaugh’s suggestion some ontological sophistication. Once we
have done so, it becomes clear that Dodd’s arguments against musical
works as individuals simply miss the mark. In Section V, I conclude with
a discussion of the homeostatic-property-cluster (HPC) conception
of species. An HPC is a regularity in the co-occurrence of properties,
maintained by an underlying causal nexus. Although it is often taken to
be an alternative to the view that species are individuals, it is better to think
of individuals as a variety of HPC. With this in mind, I suggest that we can
think of musical works as HPCs.
I. Revision and Reaction

Philosophical enquiry into the ontology of art engages at once with
the practice and institutions of art (on one side) and the methodology and
results of philosophy (on the other). One possible approach is to look to
metaphysics for the menu of possible ontologies. The ontology of art is then
just a matter of selecting a menu item. If none of them make sense of the
way we usually think about art, then our thinking about art has to change.
Call this the revisionist approach.
A danger of the revisionist approach is that it may end up describing
something that would not count as ‘art’ in any usual sense of the word.
This difficulty leads Amie L. Thomasson (2006) to begin from the other
direction, privileging art practice. The way we use words like ‘song’ and
‘sculpture’ determine the kinds of things that we are talking about. If those
kinds of things do not appear on the metaphysicians’ list of options, then it is
the list of options rather than our understanding of art that needs to change.
She insists that ‘traditional ontological bifurcations . . . are not exhaustive,
and properly handling the ontology of . . . works of art requires accepting a
broader system of ontological categories’ (p. 247). To strain the menu
metaphor, art ontology must pack its own lunch. Thomasson’s specific
110 P.D. MAGNUS
proposal is that repeatable works like books and songs are abstract artifacts—
a suggestion which traditional metaphysics would treat as a category
mistake. Thomasson’s work thus exemplifies a reactionary approach.
The labels ‘revisionist’ and ‘reactionary’ distinguish the two approaches
with respect to how they treat art practice. From the direction of how they
treat traditional metaphysics, Thomasson is the one advocating revolution.
So I do not intend the labels as praise for one side or the other. Rather, it
would be an unhappy end if we had to pick between these two options.
Andrew Kania (2008) offers a way to reconcile the two: If the reactionary
ontology fits what we say and do, then that is our picture of art. A practice-
first analysis would thus be in the business of unpacking our assumptions,
not of showing that they are tenable or true. It would be descriptive
metaphysics, in the fashion of Strawson (1959). The revisionist explications
would, in contrast, describe what there actually is in the world. This
broaches fictionalism, according to which ‘art’ talk is merely as if there
were artworks. Although fictionalism is one way out of the staring contest
between revisionists and reactionaries, it should not be our first resort for
resolving the dispute. We should only think that ‘art’ is a fiction of practice
if it really is impossible to resolve the tension between art and sensible
metaphysics. If there is a way to make sense of art which both respects
practice and could possibly be true, then fictionalism would be otiose.
In this chapter, I will just be considering musical works. If works of
art have a common ontology, then my conclusions will generalize. If they
do not, then my conclusions might still apply to other repeatable works
like books and movies. Regardless, such extensions are beyond the scope of
this chapter.
The philosophical debate about the ontology of music is often framed as
being about pure music, also called abstract or absolute music. A work of pure
music does not include semantic content (e.g. lyrics) or performative
elements (e.g. gesticulation). This restriction comports with one standard
view about music: Sonicism, according to which a musical work simply is a
sound structure. Because refusing the restriction to pure music would
beg the question against sonicism, I accept the restriction for the sake
of argument.
II. Is a Song Forever?

Works of art (Guy Rohrbaugh argues) are modally flexible, temporally
flexible, and temporal. Moreover, these features are best captured by
understanding artworks to be historical individuals.
First, to say that artworks are modally flexible means that a particular work
of art could have had different properties than it actually did. Take the
example of Mason Williams’ ‘Classical Gas.’ This 1968 piece for solo guitar
won two Grammy Awards for Williams (best instrumental theme and
best instrumental performance) and one for Mike Post (best instrumental
arrangement). Williams’ handwritten lead sheet for the piece simply
had ‘Thing for Guitar’ written at the top. He writes, ‘I didn’t really have
any big plans for it, other than maybe to have a piece to play at parties when
they passed the guitar around. I envisioned it as simply repertoire or “fuel”
for the classical guitar, so I called it Classical Gasoline.’1 Williams was later
signed and decided to record the song for his first album. In a recording
session, the copyist wrote ‘Gasoline’ simply as ‘Gas.’ That is the title that
stuck. Given a slightly different course of events, the piece could have had
a different title. Since we are treating it as a work of pure music, however,
the title does not enter into it. Yet, just as the title change resulted from a
copyist’s error, we can imagine a similar small difference in Mike Post’s
arrangement. It is natural to say that a single note higher or lower, shorter or
longer, would have made ‘Classical Gas’ somewhat different than it actually
is, rather than to say that it would have made for an utterly different work
than the one that has been played so many times in the actual world. In
jargon, the intuition is that ‘Classical Gas’ is modally flexible.
Second, artworks are temporally flexible, meaning that they can change
over time. Rohrbaugh focuses his attention on literary works and musical
scores which have appeared in multiple editions. It would be wrong, he
argues, to treat the editions as utterly separate works. It would also be wrong
to treat one of the editions as canonical, treating copies of the other editions
as deficient instances of the canonical type. As he notes, the intuition that a
song might change over time is even stronger for ‘folk music . . . passed
down through an oral tradition, where there is no written ur-text on which
1 The Story of Classical Gas, <http://www.classicalgas.com/gasstory.html> (last accessed May 31, 2011).
112 P.D. MAGNUS
to fixate’ (2003, p. 188). Indeed, for music, fixation on the score is a red
herring. If the musical work is an abstract type of sound structure, then the
same musical work would exist even if the score for it had never been
written down.
Consider ‘Classical Gas’ again. There is an online video of Sungha Jung
playing an arrangement by Ulli Bögershausen. One commenter on YouTube
says of the Jung video that they do not really like this arrangement. Another
describes Jung’s performance as a really nice rendition. The thing to note is
that both commenters take it to be an arrangement and a rendition of
‘Classical Gas,’ not some separate work. With multiple editions by a single
composer, one might imagine that the composer’s authorial power connects
revised scores to earlier ones. The historical separation here is a step further,
because Bögershausen rather than Williams’ did the arranging. And even
the original recording of the work was itself arranged by someone other
than the composer. Nevertheless, it is natural to say that Williams playing
Post’s arrangement was a performance of the same ‘Classical Gas’ that was
performed when Jung played Bögershausen’s arrangement. This means that
Bögerhausen’s arrangement extends the range of things that can count as
performances of ‘Classical Gas.’ In jargon, the work is temporally flexible.
Third, works are temporal, meaning that they come into existence when
composed and could go out of existence. There was no ‘Classical Gas’ in the
Jurassic period, in the Renaissance, or during the Second World War. It
simply did not exist until Mason Williams invented it. If Williams were
annihilated along with all of the recordings, sheet music, guitars, and
guitarists, then there would be no more ‘Classical Gas’—it would be
gone. Note that the claim of temporality is not that the musical work only
exists when actually being played or performed. ‘Classical Gas’ does not
snuff out of existence when guitarists go to sleep and return to existence
when the first note is strummed again.2 Rather, it continues to exist because
it is part of the available guitar repertoire. It is maintained by guitarists,
recordings, and all the rest.
These features are incompatible with the view that musical works are
abstract sound structures. Instead, Rohrbaugh suggests that we should think
2 ‘Classical Gas’ has been played on radio and television literally millions of times, perhaps more than
any other musical work. So it is possible that it has been continuously occurring somewhere for most of its
history. This, of course, is beside the point.
of musical works as historical individuals. ‘Classical Gas’ is embodied in

recordings, sheet music, guitars, musicians, and performances. Some but
not all of these embodiments count as occurrences of the work; the
work occurs when and where it is performed, but does not occur when
and where the sheet music just sits in a drawer. (Note that ‘embodiment’
and ‘occurrence’ take on a somewhat technical meaning here.)
Julian Dodd objects to Rohrbaugh’s account, insisting that musical works
are neither modally nor temporally flexible. I think, given appropriate
rhetoric, it is possible to summon up intuitions for both sides. With some
concentration, I can summon up an attitude from which it seems OK to say
that Post’s arrangement of ‘Classical Gas’ and Bögerhausen’s are two distinct
musical works, similar in many respects and derivative of William’s original
theme. As Dodd acknowledges, however, his Platonist type/token theory
forces him to say that even the slightest difference makes for a different
musical work. It is an affront to intuitions—even to Dodd’s—to say that one
note higher or lower, shorter or longer, makes for a different musical work.
He insists, however, that this is a price we must pay in order to have a
coherent ontology. He writes, ‘[W]hat is a relatively minor conflict be-
tween the type/token theory and our ordinary thought and talk about
works is a small price to pay for avoiding the problems that Rohrbaugh
faces . . . ’ (2007, p. 90).
Although Dodd tries to finesse the point by calling it a ‘minor’ affront and
a ‘small price,’ this is still biting a bullet. He accepts Rohrbaugh’s insistence
that if musical works were modally and temporally flexible, then the
metaphysical view that musical works are types would have to be jettisoned.
Dodd insists that it should be saved because (he says) Rohrbaugh’s alternative
metaphysical view is a disaster.
Dodd complains that musical works as historical individuals would be
‘cross-categorial entities’ that violate the usual metaphysical sensibilities,
that they would be ‘ontologically novel’ and ‘highly revisionary’ (2007,
pp. 145-6). Rohrbaugh himself describes his approach as ‘innovation at the
level of metaphysics, the identification of a new ontological category.’
He cautions that this would be ‘overkill and bad methodology to boot’
unless the novel category can be made to do systematic philosophical work
(2003, p. 197). Even if novel ontological machinery could make sense of art,
however, it might still seem like overkill and bad methodology. It would be
114 P.D. MAGNUS
odd, after all, if art had a sui generis ontology. Call this the argument from
bizarro ontology.
Dodd also objects that he cannot make sense of how a musical work qua
historical individual could be repeatable.3 Since an individual does not have
instances, the separate performances cannot instantiate the piece. As an
historical individual, the musical work has embodiments—the things in the
world, which either constitute it or upon which it ontologically depends.
Yet, Dodd asks, how can embodiments count as occurrences of the work? As
he notes, there are many kinds of ontological dependence which do not
amount to occurrence. For example, the impure set {Mason Williams}
ontologically depends on Mason Williams. Although Williams meets the
conditions for being the embodiment of the set, he is certainly not an
occurrence of the set. Call this the argument from repeatability.
Dodd offers some further objections that only apply to specific develop-
ments of Rohrbaugh’s proposal. For example, historical individuals may be
understood as four-dimensional objects extended through time. On this
view, the individual is identical to an historical process. Dodd objects to
this in two ways.
First, it is natural to say that someone who listens to the three minutes and
six seconds of ‘Classical Gas’ has heard the whole thing. This would not be
true if the work is a long process that stretches from the late 1960s to the far
future. He argues that ‘it is plainly intuitive to think that an audience hears
entire works in a performance’ and that the four-dimensionalist construal
cannot accommodate this (2007, p. 157). My intutions differ on this, and I am
perfectly comfortable saying that the audience hears an entire performance of
the work but does not hear the entire work simpliciter. In any case, from
Dodd’s handling of modal and temporal flexibility, it is clear that one
intuition one way or another will not settle this matter. The core of the
issue is whether Rohrbaugh’s suggestion could possibly be respectable
metaphysics.
Second, Dodd insists that the construal of works as four-dimensional
objects fails because they would be ontologically multifarious (2007,
pp. 160–2). The worry is this: A particular occurrence of the work would
be a temporal part of the overall work; e.g. the part from 8:00 p.m. to 8:03
3 Dodd says ‘continuant’ instead of ‘historical individual,’ but he is clear that he intends to mean by
the former exactly what Rohrbaugh does by the latter; see Dodd (2007), p. 144, fn. 1.
on a particular day. Yet when two guitarists play ‘Classical Gas’ at just that
time in different places, the temporal part would contain two performances.
So some occurrences would be spatially unified (when there was only
one performance of the work during that time slice) and others spatially
scattered (when there was more than one performance at that time).
Occurrences altogether would lack any sensible ontological unity. Call
this the argument from multifariousness.
Apparently, Rohrbaugh himself does not favour the four-dimensionalist
construal of his proposal.4 Rather, he prefers to think of historical individuals
as abstract objects which ontologically depend on but are not constituted
by their embodiments. A work would thus be wholly present every time
and place it is performed. It would endure, rather than perdure. Note,
however, that there is nothing in Rohrbaugh’s original argument which
necessitates this. On face, an historical individual seems to have none of the
markers of an abstract object. Abstract entities are not perceptible, do not
enter into the causal order of things, and exist outside space and time.5 In
contrast, historical individuals may be perceptible. They necessarily enter into
the causal order, coming into being and going out of being. They are spatially
and temporally local. I will not pursue this point, since most of Dodd’s
arguments do not depend on whether a musical work is a thing constituted
by its embodiments or an abstract object dependent on its embodiments.
To sum up, I have distilled Dodd’s resistance to Rohrbaugh into
arguments from bizarro ontology, repeatability, and metaphysical multifari-
ousness. At the core of these is an insistence that the ontology of musical
works as historical individuals just does not make sense.
III. Pluralism and Fictionalism

Taking the reactionary line, the argument from bizarro ontology shows
not that Rohrbaugh’s account of art is unsatisfactory but rather that the
sensibilities of metaphysicians are too timid. If the account fits what we say
and do, then that is our picture of art. Andrew Kania, considering Dodd’s
attack on Rohrbaugh’s ‘new category,’ notes that Rohrbaugh’s argument
4 Dodd says that Rohrbaugh has said as much in personal communication (2007, p. 148).
5 I take this list from from Zoltán Gendler Szabó (2003), but it is entirely typical.
116 P.D. MAGNUS
relies crucially on the way that we actually think and talk about art (2008,
p. 433). As a descriptive or anthropological account, it does not need a
‘further metaphysical defence’ (p. 434). As we saw earlier, he recommends
fictionalism—an option that we should not embrace too quickly.
In a paper coauthored with Christy Mag Uidhir, I argue for art concept
pluralism (Mag Uidhir and Magnus 2011). Our claim is that there
are multiple legitimate ways of thinking about art. Applying different art
concepts, we argue, yields different answers to questions about which
objects are occurrences of which works.6 Applying the aesthetic art concept
to music, two identical sound structures will both count as instances of the
same work. Applying the historical art concept, two sound structures must
have the same provenance in order to count as instances of the same work.7
For ordinary purposes, it is unnecessary to tease these apart. Since they will
typically yield the same judgments, there is an advantage in the looseness of
our usual talk about ‘art.’ Specifying one or the other is only required
to make sense of extraordinary cases and philosophical disputes.
Art concept pluralism explains why there are intuitions that line up on
both sides: The competing intuitions are honed by different art concepts.
We can see the different intuitions as guiding when and how each concept
is applicable.
It is tempting to say that there are simply different ontologies appropriate
to the different concepts: Musical works qua aesthetic art are types. Musical
works qua historical art are individuals. This would be too quick, however.
As Mag Uidhir and I are careful to point out, pluralism should not be used as
a universal solvent to dissolve any philosophical dispute. If Dodd is correct
that works-as-individuals is an unsatisfactory and incoherent ontology, then
it will not do even as the ontology for works qua historical art.
In order to be a non-fictionalist pluralist and insist that musical works
even as considered under specific art concepts really are historical individuals,
I must answer Dodd’s allegation that the ontology of historical individuals is
a category-violating disaster. The easiest way to do that is to show that it is
not novel—contra claims by Rohrbaugh, Dodd, and Kania. Quite the
contrary, it has been familiar to philosophers of biology for decades. It is
6 In the paper, we talk about ‘instances’ of works. I say ‘occurrence’ here in order to be neutral
between Rohrbaugh’s and Dodd’s ontological proposals.
7 The conventional art concept would also require common provenance.
now almost a consensus among biologists that biological species are histor-
ical individuals. In the next section, I explain this view and argue that
musical works would be historical individuals in much the same sense.
IV. The Species-as-Individuals Consensus

The claim that species are individuals was originally proposed in the
1960s by Michael Ghiselin (1966, 1974). It was later popularized by David
Hull (1976, 1978). It is not hollow bravado when Ghiselin calls it the
‘philosophical concensus’ (2002, p. 153; 2007, p. 283; 2009, p. 253). Even
Richard Boyd, opposing individualism, acknowledges it as the default
position (Keller et al. 2003). Members of a species are part of an historical
lineage. Take the species Anas platyrhynchos, the common mallard. All of
the mallards alive today share common ancestors and figure in the same
historical narrative. Biological explanations of why ducks are like they are
include evolutionary explanations in terms of this shared history. As such, an
arrangement of molecules which came together randomly in a distant galaxy
with the same intrinsic properties as a female mallard would not be a
member of the species. The explanation for its having duck-features
would be entirely different than the explanation for why female mallards
have duck-features.
Ghiselin and Hull express individualism as an opposition to the view that
species are classes. They explain this in terms of the relationship between a
species like A. platyrhynchos and an organism like a particular duck. If the
species were a class or set, then the duck would be a member of it—and
instance of the type. If the species is an individual, though, the duck is part of
the species. Ghiselin explains, ‘Organisms, which are also individuals, are
not instances of species, but parts of them, just as cells are parts of organisms’
(2009, p. 254). This is naturally construed along four-dimensionalist lines,
such that the species is the sum of its embodiments.8 So if the argument
from metaphysical multifariousness discredits the view that musical works
are individuals, then it should also work against the view that species are
individuals.
8 As Reydon (2003) notes, however, some authors take the species-as-individual to be merely the
sum of the organisms that exist at a particular time.
118 P.D. MAGNUS
The argument, recall, was that a three-minute temporal part of the work-
as-individual could include one performance or many; so the temporal
part might be spatially local or spatially disjoint. Apply this to a species-
as-individual. A temporal part of A. platyrhynchos typically includes
a great many organisms and is spatially disjoint. Yet we can imagine a
near-genocide, which kills all but a single duck. A temporal part after the
duckpocalypse would include just one organism and would be spatially
local. This does not show any problem with the species-as-individuals
view. Rather, it shows that a particular organism is not merely a temporal
part of a species. An organism is a spatiotemporal part. There is not one
spatially disjoint mallard alive at the present time, but instead a great many
individually local mallards.
The same answer can be given to Dodd’s argument about musical works.
An occurrence of a work, as a part of the historical individual, is not a
mere temporal part. Rather, a performance is a spatiotemporal part. When
there are two performances of ‘Classical Gas’ at the same time but in
different places, there are two occurrences of the work. The occurrences
are distinct parts of the work-as-individual, just as two ducks on a pond are
distinct parts of the species-as-individual. So much, then, for the argument
from metaphysical multifariousness.
Consider next the argument from repeatability. If a musical work were
a type, then each performance would be an instance of the type. Dodd’s
worry was that there was nothing akin to instantiation for historical individuals.
Yet species provide a model here. Each individual duck is a member of
the species. Even though it is odd to say that each performance is a ‘member’
of the work, this is just a terminological difference. The performance may
relate to the work as the duck relates to the species.
As Dodd notes, there are plenty of things that have embodiments (stuff in
the world upon which they ontologically depend) but which do not have
occurrences. Embodiments will not always count as occurrences. Mallards
are repeatable in the straightforward sense that there are a great many
of them now and that they will produce future generations that include
even more. The objection that embodiments do not always underwrite
repeatability is beside the point. Species-as-individuals provide a model of
how some embodiments can count as occurrences. Of course, not all
embodiments of a musical work will count as occurrences of the work.
A printed score for ‘Classical Gas’ is an embodiment of it and can contribute
causally to the production of further performances, but it is not itself

a performance or occurrence of the work. Although it is not usually
mentioned in accounts of species individualism, the parallel claim is true
about species. Mallard sperm and a mallard egg might be combined in vitro
and used to produce a new mallard. Yet the cells in a test tube are not
themselves mallards. So much for the argument from repeatability.
The mere fact that individualism is a familiar and comfortable option
in philosophy of biology suffices to answer the argument from bizarro
ontology. So Dodd’s concerns have all been answered. Since historical
individuals are respectable onta, there is no reason to bite the bullet and
deny that musical works are modally and temporally flexible. Considered as
historical art, musical works are historical individuals.
V. Causes in the Wings

We saw earlier that there are different ways of filling in the metaphysical
details of the view that musical works are historical individuals. The analogy
with individualism about species does not settle the issue. The general
picture of species-as-individuals—just like the suggestion that musical
works are individuals—might be ontologically filled out in various ways.9
Talk of individuals and parts, as opposed to classes and instances, should
not be taken as a brief for mereology over set theory. What is centrally
important to the idea is just the core insistence that a species has continuity.
In an early paper advocating individualism, David Hull is especially clear
about this:
By ‘individuals’ I mean spatiotemporally localized cohesive and continuous entities
(historical entities). By ‘classes’ I intend spatiotemporal unrestricted classes . . . The
terms used to mark this distinction are not important, the distinction is. For
example, one might distinguish two sorts of sets: those that are defined in terms
of a spatiotemporal relation to a spatiotemporally localized focus and those that are
not. (1978, p. 336)
9 Ghiselin himself sees individualism about species as ‘the inspiration for a new ontology with
profound implications for knowledge in general’ (2009, p. 253). Rohrbaugh’s treatment of musical
works as individuals would thus be a specific development in that program. Engagement with Ghiselin’s
more detailed metaphysics is beyond the scope of this chapter, since the approach is decidedly reaction-
ary (in the sense of Section I).
120 P.D. MAGNUS
Similarly, the arguments in defense of works-as-individuals (Section II)

are not at the level fundamental ontology. Historical individuals might, at
the fundamental level, be localized and causally-specified types. This still
contrasts with the usual view of works as types, and it is certainly not the
view that Dodd defends. So when Rohrbaugh considers such a proposal,
he observes that, for works-as-types theorists, the requirement of causal
connectedness would be ‘already to concede defeat’ (2003, p. 196).
Whatever exact metaphysical account is to be given, the continuity of an
historical individual is causal continuity. As Rohrbaugh says, causation
‘clearly waits in the wings’ (2003, p. 200). In the remainder of the chapter,
I turn my attention to the account of species as homeostatic-property-clusters
(HPCs). For those familiar with the HPC view, this may seem like an odd
choice; it is often seen as a rival to species-as-individuals.10 I argue that the
HPC view is better seen as a specification of what it is to be an individual:
Individuals are HPCs united by a single token causal history. This way of
thinking fits both species and musical works.
The HPC account was originally devised by Richard Boyd as an account
of kinds like rationality, reference, and moral goodness. Yet the stock
example an HPC is a biological species, and the account has been most
widely accepted by philosophers of science.11 Consider mallards again.
There are intrinsic properties typical of ducks. Although none of these are
necessary or even together sufficient for species membership, they do occur
together quite often. Duck properties come clustered together in many
spatially-separated packages: distinct organisms that look like ducks, walk
like ducks, and quack like ducks. The central idea behind the HPC account
is that the kind is not characterized just by these typical properties. Instead, it
is unified by the causal processes which produce and maintain packages of
these properties.
As I argue elsewhere (2011), we should not treat the property cluster as
merely one list of typical properties. It is better to treat it as a structured
complex of properties: The duck is characterized by different properties at
different points during her life cycle, and female and male mallards differ in
10 See, e.g. the discussion of HPCs in Richards (2010).

11 Richard Samuels and Michael Ferreira write that ‘philosophers of science have, in recent years,
reached a consensus—or as close to consensus as philosophers ever get—according to which natural
kinds are Homeostatic Property Clusters’ (2010, p. 222).
systematic ways. This property complex is produced and sustained by causes

at several levels. When we look at a specific duck, the explanation for
why the duck properties remain together can be given in terms of their
physiological processes. The explanation for why the properties play out
over time as they do (as a cluster of duckling properties before, a cluster of
duck properties now, and a cluster of geriatric-duck properties later) can be
given in terms of their developmental processes. The explanation for why
there are several similar clusters of duck processes on the lake, that there
were last year, and that there will be next year can be given in terms of
the reproductive processes of mallards. All these processes are themselves
situated in the evolutionary history of the species. The evolutionary origin
of all ducks is the same token event.
Imagine again that there were intrinsically-identical, duck-like organisms
here and in a different galaxy. The explanation for why they maintain bodily
integrity would be the same. The immediate physiological processes would
be of the same type. Yet the explanation for why there were duck-like
creatures there would be different than the explanation for why there are
ducks here. The former would have to be given in terms of whatever
produced the organism in that distant galaxy, the latter could be given in
terms of the evolutionary history of the species here on Earth. Even if the
histories were of a similar type, they would be different token processes.
Evolution is a madly contingent business, and it is the shared evolutionary
history that explains how members of a species share what they do.
Recall Hull’s minimal conception of historical individuals as ‘localized
cohesive and continuous entities’ (1978, p. 336 cited above). The cohesion
and continuity are importantly causal cohesion and causal continuity. The
different parts are produced as part of the same historical narrative,
the same token causal story. An HPC kind that is unified by a single
token causal history will count as an ‘individual’ in this sense. Indeed,
Boyd urges that ‘the distinction between natural kinds and (natural)
individuals is . . . merely pragmatic’—‘almost just one of syntax’ (1999a,
pp. 163, 164).
Now let’s return to musical works as individuals: There are patterns
of sounds that recur when a work is played. None of these precise sounds
are necessary in order for this to be an occurrence of the work, because
the work might be interpreted or arranged differently. For each perfor-
mance of the work, there is a skilled musician who produces such a sound
122 P.D. MAGNUS
structure.12 Yet the patterns of sound are not sufficient for an occurrence of
the work. Performances of the work all share a single, distal cause: The
original composition of the work. All the occurrences of the work as similar
sound structures are part of the same token causal story, just as all the mallards
as similar organisms are part of the same token history. That is to say the
musical work is an HPC unified by a token history. It is an individual.
One might object that Boyd himself thinks of species-as-HPCs but not as
individuals. The complication is that Boyd thinks that a single species might
originate multiple times in different places. However, he is not concerned
with philosophical examples like molecular duck dopplegangers in a distant
galaxy. Instead, he is concerned about species that result from hybridization
(see Boyd 1999b). Imagine that two parent species can be crossed to produce
viable offspring and that different specimens from the parent species are crossed
on several occasions. Boyd insists that the hybrids from each cross and their
offspring, although part of separate causal histories, would count as members of
the same species. So he insists that a species-as-HPC might be unified simply
by a type of unifying cause. A proper individual, in contrast, must be unified in
the broadest scope by a single token cause. If species must be individuals, then
the separate hybrid lineages—even if they were indistinguishably similar—
would be distinct species. So, because he allows the possibility of separate
hybrid origins for the same species, Boyd denies that species are individuals.
Boyd’s worries about hybrids do not obviously arise for musical works.
Imagine note-for-note identical compositions being written by two
composers. These would be separate works, at least considered under the
rubric of historical art or conventional art rather than aesthetic art. We might
instead imagine a case in which a single musician composes a tune, writes it
down, forgets about it, and later composes a note-for-note identical tune.
Structurally, this would be like the cases of repeated hybridization that
worry Boyd. The important difference is that Boyd’s case is biologically
plausible; it actually occurs among plants. The parallel case of repeated
musical composition seems more fantastic. I am unsure of what to say
about it. Suppose one said that the two tunes would be distinct works; the
two works would each be historical individuals. Suppose instead that one
12 Recorded and synthesized occurrences of the work complicate the story somewhat. Each time
a computer plays a MIDI file of ‘Classical Gas,’ there is a different kind of local cause than when there is a
live performance of it. Nevertheless, there are still local causal explanations to be given in both cases, and
the two share a distal cause (the original composition).
said that the two tunes would be the same musical work; the works would
not be individuals, strictly speaking, but property complexes unified by
causal origins that were sufficiently similar. In this latter case, both causal
origins would still crucially share the same individual actor—the selfsame
composer. So even if we allow that the same work could be composed
twice in this way, it would still be an HPC, and there would be significant
constraints on what could count as a composition of the work. This odd
kind of counterexample to works-as-individuals still puts the work in
the same category (HPCs). It would provide no comfort for traditional
works-as-types approaches, such as Dodd’s Platonist sonicism.
VI. Conclusion
Rohrbaugh argues from our understanding of musical works, their flexibility
and temporality, to the conclusion that they are historical individuals. Dodd
objects that this is an unacceptable ontology. Seen as a struggle between a
revisionist (Dodd) and a reactionary (Rohrbaugh), this could only be settled at
the level of philosophical methodology. Yet Dodd’s objections fail, as the
analogy with individualism about species makes clear. The analogy does not
fully resolve things, but instead leaves us the task of elaborating what ‘histor-
ical individual’ must mean. I have argued that ‘individuals’ are best under-
stood as homeostatic-property-clusters (HPCs), which are unified by a
common history rather than merely by a type of process. So we can see
musical works, under the rubric of historical art, as HPCs.
Since HPCs are important for a general account of the world, seeing
works of music as HPCs is not a reactionary position at all. If the account fits
with art practice, then the view is not revisionary either.
References
Boyd, Richard N. (1999a) ‘Homeostasis, Species, and Higher Taxa’ in Robert
A. Wilson (ed) Species: New Interdisciplinary Essays (Cambridge, MA: The MIT
Press), pp. 141–85
——(1999b) ‘Kinds, Complexity and Multiple Realization’ Philosophical Studies
95(1/2): 67–98
124 P.D. MAGNUS
Dodd, Julian (2007) Works of Music: An Essay in Ontology (Oxford: Oxford

University Press)
Gendler Szabó, Zoltán (2003) ‘Nominalism’ in Michael J. Loux and Dean
W. Zimmerman (eds) The Oxford Handbook of Metaphysics (Oxford: Oxford
Ghiselin, Michael T. (1966) ‘On Psychologism in the Logic of Taxonomic
Controversies’ Systematic Zoology 15(3): 207–15
——(1974) ‘A Radical Solution to the Species Problem’ Systematic Zoology
23(4): 536–44
——(2002) ‘Species Concepts: The Basis For Controversy and Reconciliation’
Fish and Fisheries 3(3): 151–60
——(2007) ‘Is the Pope A Catholic?’ Biology and Philosophy 22(2): 283–91
——(2009) ‘Metaphysics and Classification: Update and Overview’ Biological
Theory 4(3): 253–259
Hull, David L. (1976) ‘Are Species Really Individuals?’ Systematic Zoology
25(2): 174–91
——(1978) ‘A Matter of Individuality’ Philosophy of Science 45(3): 335–60
Kania, Andrew (2008) ‘The Methodology of Musical Ontology: Descriptivism and
Its Implications’ British Journal of Aesthetics 48(4): 426–44
Keller, Roberto A., Boyd, Richard N., and Wheeler, Quentin D. (2003) ‘The
Illogical Basis of Phylogenetic Nomenclature’ The Botanical Review 69(1): 93–110
Mag Uidhir, Christy and Magnus, P.D. (2011) ‘Art Concept Pluralism’ Metaphilo-
sophy 42(1/2): 83–97
Magnus, P.D. (2011) ‘Drakes, Seadevils, and Similarity Fetishism’ Biology and
Reydon, Thomas A.C. (2003) ‘Species Are Individuals: Or Are They?’ Philosophy
of Science 70(1): 49–56
Richards, Richard A. (2010) The Species Problem: A Philosophical Analysis
Rohrbaugh, Guy (2003) ‘Artworks As Historical Individuals’ European Journal of
Samuels, Richard and Ferreira, Michael (2010) ‘Why Don’t Concepts Constitute
A Natural Kind?’ Behavioral and Brain Sciences 33: 222–3
Strawson, P.F. (1959) Individuals: An Essay in Descriptive Metaphysics (London:
Routledge)
Thomasson, Amie L. (2006) ‘Debates About the Ontology of Art: What Are We
Doing Here?’ Philosophy Compass 1(3): 245–55
6
Repeatable Artwork Sentences
and Generics
SHIEVA KLEINSCHMIDT AND JACOB ROSS*
We seem to talk about repeatable artworks, such as symphonies, plays, films,

dances, and so on, all the time. We say things like, ‘The Moonlight Sonata has
three movements’ and ‘Duck Soup makes me laugh.’ But how are these
sentences to be understood?
On a simple treatment of these sentences, they have ordinary, subject-
predicate form. The subject1 refers to a repeatable artwork (e.g. The Moonlight
Sonata, or Duck Soup) and the predicate ascribes some property to this
artwork. The sentences are true just in case the individual referred to by the
subject has the property ascribed to it by the predicate. Accepting this simple
interpretation of the semantics of these sentences will have immediate conse-
quences for our theory of repeatable artworks: The truth of the sentences will
require that things like The Moonlight Sonata exist, that they are singular
entities, and that they have the properties that the predicates in our sentences
about them pick out.
But perhaps the simple treatment of these sentences isn’t the correct one.
The surface form of a sentence isn’t always a sure guide to its logical form.
We are familiar with this phenomenon with respect to sentences like the
following:
* We would like to thank Alexi Burgess, Ben Caplan, Stephen Finlay, John Hawthorne, James
Higginbotham, Christy Mag Uidhir, Mark Schroeder, and Gabriel Uzquiano for very helpful comments
and discussions. Our greatest debt is to Karen Lewis, Barry Schein, and Joshua Spencer for very helpful
discussions on numerous occasions.
1 Here and throughout the chapter, when we say ‘subject’ we will be speaking of the grammatical
subject.
126 SHIEVA KLEINSCHMIDT AND JACOB ROSS
1. The average man has two and a half children.

2. Nothing is inside of the room.
Neither of these sentences is such that the subject refers to some individual
entity and the predicate tells us something about that entity. We will argue
that the same is true of the following sentences:
3. The polar bear has four paws.
4. The Moonlight Sonata is roughly fifteen minutes long.
We will argue, in Section I, that sentences such as (3) and (4) do not have
simple subject-predicate structure, wherein the subject refers to some indi-
vidual, the predicate picks out a property that is ascribed to that individual,
and the sentence is true if and only if, the referent of the subject has the
property picked out by the predicate. Then, in the remainder of the chapter,
we will consider alternatives to this simple account. In Section II we will
consider, and argue against, one view according to which these sentences are
not subject-predicate sentences at all. In Section III we will consider, and
argue against, a view on which these sentences are subject-predicate sen-
tences, but on which the subject phrase has a special kind of plural reference,
and on which the predicate phrase picks out a higher-order property. Finally,
in Section IV, we will propose a view that combines features of the second
and third views, and which avoids the problems facing each of them.
If we are correct, then unlike the Simple Semantics that is commonly
assumed to apply to sentences about repeatable artworks, the correct se-
mantics of these sentences will be ontologically undemanding. It will not
require, for instance, that we expand our ontology to include an individual
that is the referent of The Moonlight Sonata, in order to account for the truth
of (4). In fact, it may even give us reason to prefer a more sparse ontology, at
least with respect to repeatable artworks.
I. Simple Semantics
In what follows we will be interested in sentences such as (3) and (4), which
we take to fall into a single class. To understand which class of sentences this
is, consider how (3) differs from:
5. The polar bear is eating a fish.
REPEATABLE ARTWORK SENTENCES AND GENERICS 127
The noun-phrase ‘the polar bear’ is used very differently in (3) and (5). In
(5), it picks out some particular, contextually salient polar bear (say, a polar
bear the speaker is watching), whereas in (3) it does not. Instead, (3) seems to
be saying something about how polar bears generally, typically, or charac-
teristically are. That is, it seems to be a kind of generic statement, relevantly
like ‘dogs bark’ and ‘ducks lay eggs’: it is a generalization about entities of a
given kind, though in order to be true the relevant property needn’t be had
by all or (in some cases) even most of the entities of that kind.2
There is a similar divide between sentences about repeatable artworks.3
Contrast (4) with:
6. The Moonlight Sonata will begin any minute now.
The noun-phrase ‘The Moonlight Sonata’ is used very differently in (4) and
(6). In (6) it picks out some particular, contextually salient performance,
whereas in (4) it does not. Instead, in (4) we seem to be making a claim
about how performances of The Moonlight Sonata generally, typically, or
characteristically are. And this claim will be true even if some, or even most,
of the performances are not roughly fifteen minutes long.
We will be interested in sentences about repeatable artworks, like (4), and
generic sentences, like (3). Sentences of both kinds are such that, though their
surface structure suggests they have subject-predicate form with the subject
referring to an individual, the subject does not seem to pick out some
particular, contextually salient manifestation or instance of the relevant re-
peatable artwork or kind. We will call these sentences the target sentences, and
we will be investigating how these sentences are to be understood.
One fairly natural view about the target sentence is this.
Simple Semantics (SS): Each of the target sentences has simple subject-
predicate form, and is true, if and only if the individual entity referred
to by the subject has the property picked out by the predicate.
Theorists who endorse Simple Semantics have two options. They can take
the predicates in the target sentences to pick out exactly the properties they
2 For an introduction to generics, see Carlson and Pelletier (1995) and Leslie (forthcoming).
3 The similarities between repeatable artwork sentences and generics has been noted elsewhere. Most
recently by Stefano Predelli (2011), but he points to discussions of this similarity by Wollheim (1968 and
1980), Wolterstorff (1975), and Levinson (1980), among others.
Q: What’s the difference between ‘Polar bears have four paws,’ and ‘The Moonlight Sonata has four
movements?’ A: The first contains a bear plural.
seem to, or they can take the predicates in question to pick out some other,
less obvious properties. That is, they must accept or reject the following.
The Straightforward Predication View (SPV): In the target sentences, the
predicate picks out the very same property that it appears to pick out.
That is, it picks out the very same property it picks out in ordinary
subject-predicate sentences where the subject refers to an ordinary
individual.
We will call the denial of the Straightforward Predication View the Obscure
Predication View (OPV). We will argue against Simple Semantics by present-
ing a dilemma. If the proponent of Simple Semantics accepts the Straight-
forward Predication View, then her view will both overgenerate and
undergenerate: It will overgenerate by predicting that there are true readings
of some target sentences that do not seem to have true readings, and it will
undergenerate by predicting that there are not true readings of some target
sentences that do have true readings. We will then show that if the propo-
nent of Simple Semantics accepts the Obscure Predication View, then
although she can avoid both of these problems, she will face a third
problem, the familiar problem of anaphoric predication.
The basic problem with the combination of Simple Semantics and the
Straightforward View is this. If we adopts such a view, then we will need to
say that there is some entity that is the referent of ‘the polar bear’ in target
sentences of the form ‘the polar bear is F’ and some entity that is the referent
of ‘The Moonlight Sonata’ in target sentences of the form ‘the Moonlight Sonata
is G,’ and so on for other target sentences. But for any entity that we may
identify as the referent of ‘the polar bear,’ it will have features that polar
bears don’t typically have, and it will lack features that polar bears do
typically have—and similarly for ‘The Moonlight Sonata.’ Consequently,
this view will predict that seemingly false target sentences are true, and
that seemingly true target sentences are false.
If SS is true, the subject of any given true target sentence refers to some
entity. Thus, in (3), there is some entity referred to by ‘the polar bear,’ and
in (4), there is some entity referred to by ‘The Moonlight Sonata.’ Suppose we
were to endorse the dominant view, on which entities like these are aspatial
and atemporal.4 We might, for instance, identify them with abstract prop-
4 The view that repeatable artworks are abstract has been defended by Currie (1989), Dodd (2004),
Kivy (1983), Levinson (1980), and Wolterstorff (1980). For a more complete list of defenses of this view,
erties. Perhaps ‘the polar bear’ refers to the property being a polar bear, and
perhaps ‘The Moonlight Sonata’ refers to the property being a Moonlight Sonata
performance. Unfortunately, this kind of view undergenerates. Consider the
following:
7. The polar bear is roughly eight feet long.
Both of these sentences have true readings. But, given our current assump-
tions, we should predict that they do not. For we are now assuming that (7)
ascribes the property of being roughly eight feet long to an aspatial entity, and,
similarly, that (4) ascribes the property of being roughly fifteen minutes long to
an atemporal entity. It is reasonable to assume, however, that only objects
present in space can have size, and only objects present in time can have
temporal duration. (Later, we will discharge this assumption.) It would
seem, therefore, that on our current suppositions, sentences (7) and (4)
should be unambiguously false. This view, therefore, undergenerates. It
also overgenerates, since this view predicts that the negations of (7) and (4)
are unambiguously true.
What if, instead, we combine SS and SPV with the view that the subjects
of our target sentences (e.g. ‘The Moonlight Sonata,’ ‘the polar bear’) refer to
entities that are present in space and time? There are then three alternatives
that are fairly natural:
(i) The referent, x, of the subject in question is some privileged individ-
ual among what we would ordinarily take to be the instances or
manifestations of x. Thus, even in sentences such as (3) and (7) in
which there is no contextually salient polar bear that we are picking
out, there is still a unique polar bear that is the referent of ‘the polar
bear.’ Similarly, in sentences such as (4), ‘The Moonlight Sonata’ refers
to some particular Moonlight Sonata performance, or to some other
particular manifestation of The Moonlight Sonata, such as particular
written score, recording, or mental tokening.5
see Caplan and Matheson (2006, p. 59). However, one may claim repeatable artworks are kinds while
claiming that they are present where their instances are: Wollheim (1980, p. 76) is an example of a
proponent of such a view. Views on which these entities are present in space and time will be discussed
shortly.
5 We wish to remain neutral on which sorts of entities count as manifestations of The Moonlight Sonata.
(ii) The referent, x, of the subject in question is located at the fusion of

all of the regions occupied by its instances or manifestations (and
may be identical to the fusion of all of those instances or manifes-
tations). Thus, the referent of ‘the polar bear’ is spread across every
region at which there is a polar bear, being just partly present in each
region. Similarly, ‘The Moonlight Sonata’ is spread across every
region at which there is a Moonlight Sonata performance—or perhaps
every region at which there is a performance, written score,
recording, or mental tokening.6
(iii) The referent of the term in question is a multilocated entity that is
wholly present at each of its complete instances or manifestations.
Thus, just as one might hold that a time traveller can be wholly
located at each of two distinct locations at the same time, so the
proponent of the present view maintains that the referent of ‘the
polar bear’ is an entity that is wholly at each of the regions where
there is a complete polar bear, and similarly, that the referent of ‘The
Moonlight Sonata’ is wholly located at each of the regions where there
is a complete Moonlight Sonata performance—or perhaps at each of
the locations where there is a performance, a score, a recording, or a
mental tokening.7
Suppose we endorse SS, SPV, and (i). Then for any target sentence with ‘the
polar bear’ as its subject, the referent is some individual polar bear or other.
Suppose for the sake of simplicity that every polar bear is either completely
inside or completely outside of Alaska. Since each of the following two
sentences fall into our group of target sentences, this combination of views
predicts a true reading of one of:
8. The polar bear is present only in Alaska.
9. The polar bear is present only outside of Alaska.
But, though we can truly say, for instance, that the polar bear is only present
on planet Earth, we cannot, absent some contextually salient polar bear,
truly utter either of (8) or (9). So this combination of views overgenerates.
And, because the negations of (8) and (9) both seem true, but this combina-
6 Recently, a version of this view has been proposed and defended by Caplan and Matheson (2004,
2006, and 2008).
7 This view was proposed by Tillman (2011), and further defended in Spencer and Tillman (2012).
tion of views predicts that at least one of them will lack any true readings,
this combination of views undergenerates.
One might try to avoid these consequences by claiming that which polar
bear is the referent of the subject of any given target sentence will depend on
the sentence. However, it is not clear what the grounds would be for
claiming that the subject of (8) happens to have a referent that is located
completely outside of Alaska, whereas the subject of (9) happens to have a
referent that is located completely inside Alaska. Further, they will need to
claim that in these two sentences,
10. The polar bear has four paws and is present only in Alaska.
11. The polar bear has four paws and is present only outside of Alaska.
what the referent of ‘the polar bear’ is in the first conjunct of each sentence
depends on what comes after that conjunct. But this is implausible.
Suppose instead that we endorse (ii). If we accept SS and SPV, and we
want to allow that sentences such as (7) and (4) have true readings, then this
will impose serious constraints on what sorts of things can be the referents of
the terms in question. For in this case the referent of ‘the polar bear’ in (7)
will have to be roughly eight feet long, and the referent of ‘The Moonlight
Sonata’ in (4) will have to be roughly fifteen minutes long. But according to
(ii), the referent of the subject of (7) is the size of the fusion of all polar bears,
which is much longer than eight feet. And the referent of (4) will have the
duration of the fusion of all Moonlight Sonata performances, which will be
much longer than fifteen minutes. So this view will undergenerate. It will
also overgenerate, in producing true readings of the negations of (7) and (4).8
It seems, then, that the only remaining natural alternative for the pro-
ponent of SS and SPV is to opt for (iii), and claim that the referent of ‘the
polar bear,’ and similarly the referent of ‘The Moonlight Sonata,’ is a multi-
located entity that is wholly located wherever there is a complete polar bear
8 This objection isn’t new. An objection similar to this one was presented by Dodd (2004, p. 353), and
responded to by Caplan and Matheson (2006, pp. 61–3). That discussion leads us to believe that Caplan
and Matheson would simply reject (4), and accept sentences like ‘The second half of The Moonlight Sonata
was missing from that complete performance.’ But sentences like that one seem particularly bad.
Similar worries face temporal parts theorists: e.g. we might say of a persisting point-sized object, x, ‘x
has exactly one part.’ If accept Simple Semantics, and take the referent of the subject to be x rather than,
say, its current temporal part, then we must either reject the Straightforward Predication View, reject
four-dimensionalism, or deny the truth of the sentence. It is not uncommon to think that such sentences
are, strictly speaking, false.
or a complete Moonlight Sonata performance. Unfortunately, there is also a

problem with this view.9 (We’ll focus here on ‘The Moonlight Sonata,’
though the same problem arises for ‘the polar bear,’ and for the subjects of
our target sentences more generally.)
There are two claims we need to make to generate the problem. Locative
Claim 1: Each part of The Moonlight Sonata is located where and only where
it is performed.10 This is supported by the same reasons that motivated us to
accept the multilocation view about The Moonlight Sonata. E.g. assuming SS
and SPV, we will have trouble explaining the truth of ‘the first movement
of The Moonlight Sonata is roughly six minutes long’ if we claim the first
movement is either non-located or that it is located at the fusion of its
performances. And taking the referent of ‘the first movement’ to be some
particular performance of the first movement is unhelpful for reasons like
those given above.
Locative Claim 2: If some but not all of a wholly material object, o, is
present within some region, r1, then there must be some other, disjoint
region, r2, such that some of o is present in r2, and the parts of o present in r1
and the parts of o present in r2 together make up o. For example, if some, but
not all, of the sailboat is located in the garage, there must be somewhere else,
perhaps the driveway, where the rest of it is at, such that the parts in the
garage and the parts in this other place together make up the whole sailboat.
Now we face a serious problem: the problem of incomplete instances. Con-
sider a world in which there are exactly two Moonlight Sonata performances:
P1 (performed in region R1) and P2 (performed in region R2), such that P2
occurs one hundred years after P1. And suppose that, in the concert hall
where P2 occurs, a bomb is detonated right before the Recapitulation is
performed. Thus, P1 is a complete performance of The Moonlight Sonata, in
which every part of The Moonlight Sonata is performed, whereas P2 is
9 In fact, there are several problems. Here’s one: (iii) requires either claming that identity is intransitive
(because, e.g., The Polar Bear will be identical with each of several distinct polar bear manifestations), or
claiming that there’s colocation (of The Polar Bear and particular polar bears). If we posit colocation,
barring same non-standard views, we must either posit genuine, ground-level colocation, or deny the
(perhaps region-relativized) uniquencess of composition, which will also cause a grounding problem.
10 One could instead claim that whenever there is a merely partial performance of The Moonlight
Sonata, the entire Moonlight Sonata is present. However, for any merely partial performance of The
Moonlight Sonata, there is some part of The Moonlight Sonata that is unperformed. So, with respect to
where the performance is, we can truly say: ‘Part of The Moonlight Sonata is absent.’ However, on this
view, we can also truly say: ‘No part of The Moonlight Sonata is absent.’ But these two sentences cannot
even possibly both be true.
an incomplete performance of The Moonlight Sonata, in which all but the

Recapitulation is performed.11 Suppose further that, apart from P1 and P2,
no part of The Moonlight Sonata is performed on any other occasion.
It follows from Locative Claim 1 that every part of The Moonlight Sonata
except the Recapitulation is located in R2. And it follows from this and
Locative Claim 2 that the Recapitulation is present somewhere else, and
that the Recapitulation at that region and the parts of The Moonlight Sonata
at R2 together make up The Moonlight Sonata. Since, in our case, the
Recapitulation is performed only once, namely in P1, Locative Claim 1
entails that the Recapitulation is located at just one place: a subregion of R1.
Thus, the Recapitulation within R1 and the parts of The Moonlight Sonata in
R2 together make up The Moonlight Sonata.
Recall, however, that P1 precedes P2 by one hundred years. This, then,
has two problematic consequences. First, it means that The Moonlight Sonata
has a scattered location, with one part present in one performance, and the
rest of it present in a distinct performance a century later. This is exactly the
kind of consequence that those endorsing the multilocation option are
hoping to avoid by claiming that entities like symphonies are wholly present
in each of their performances. Second, since the Recapitulation in R1 and
the parts of the Sonata in R2, together, make up The Moonlight Sonata, it
seems that in at least one manifestation of The Moonlight Sonata, the Recap-
itulation precedes the rest of the Sonata. But in addition to being inherently
implausible, this has implausible semantic implications when combined
with the SS and SPV. For this sentence seems clearly true:
12. The Moonlight Sonata is such that its Recapitulation always comes
after its Exposition and its Development.
Assuming SS and SPV, this sentence attributes to referent of ‘The Moonlight
Sonata’ the property of being such that its Recapitulation always comes after its
Exposition and its Development, and according to the multilocation view, in
the case we’ve described the referent of ‘The Moonlight Sonata’ does not have
11 One could instead say that for any part of The Moonlight Sonata to be present in a performance, all of
The Sonata must be performed. But this would mean that, no matter how close to being complete a merely
partial performance of The Moonlight Sonata is, not a single part of the symphony would be present in it,
though the whole Sonata would be present in a complete performance that was ever so slightly different.
Further, anyone at a performance of The Moonlight Sonata could not know whether they were in the
presence of The Sonata until they knew whether the performance would be a complete one.
that property. So this combination of views will predict that there are no
true readings of (12), and will thereby undergenerate. (Similarly, it will
overgenerate due to predicting a true reading of the negation of (12).)
Thus, it seems that if the proponent of SS accepts SPV, then her view will
have unacceptable consequences regardless of what she takes the subjects of
the target sentences to refer to. For, regardless of what we take to be the
referent of ‘the polar bear’, or the referent of ‘The Moonlight Sonata,’ either
there will be some feature had by that referent that will, via SS and SPV,
guarantee true readings of sentences that we take to be false, or there will be
some feature that the referent lacks that will, via SS and SPV, guarantee false
readings of some sentences we take to be true. We’ve focused on just some
of these features, but there are many ways one can run an argument of this
form against the combination of SS and SPV.
There is, however, an objection to our argument that deserves some
comment. In presenting this argument, we assumed that if ‘the polar bear’
and ‘The Moonlight Sonata’ are aspatial and atemporal, then these entities do
not have the sorts of features that appear to be ascribed to them in sentences
(7) and (4). For while the sentence ‘polar bears are roughly eight feet long’ is
true, aspatial entities don’t have size. And while the sentence ‘The Moonlight
Sonata is roughly fifteen minutes long’ is true, atemporal entities don’t have
duration. There are, however, some who would deny this, maintaining that
non-located entities can have such ordinary features as sizes and durations.
They could do so by appealing to the following claim, defended by David
Liebesman (2011):
Inheritance Hypothesis: Whenever k is a kind whose members generi-
cally have some property, F, k itself is F.
Suppose that the Inheritance Hypothesis is true, and suppose that ‘the polar
bear’ refers to a kind, specifically, the kind the polar bear. Suppose further that,
generically, polar bears are roughly eight feet long. In this case, it will follow
from the Inheritance Hypothesis that the kind referred to by ‘the polar bear’
is itself roughly eight feet long. Hence it will follow, from SS and SPV, that
sentence (7), ‘the polar bear is roughly eight feet long,’ is true. Similarly, if
‘The Moonlight Sonata’ refers to a kind to which all and only the Moonlight
Sonata performances belong, and if, generically, Moonlight Sonata perform-
ances are roughly fifteen minutes long, then it will follow from the Inherit-
ance Hypothesis that The Moonlight Sonata is roughly fifteen minutes long.
Hence it will follow from SS and SPV that sentence (4), ‘The Moonlight
Sonata is roughly fifteen minutes long,’ is true. These are exactly the results
we want.
Unfortunately, this explanation of the truth of (7) and (4) is unacceptable,
since it leads to paradox. To see how, let K be the kind to which all and only
performances of The Moonlight Sonata belong. Note that, necessarily, Moon-
light Sonata performances are not kinds. But if something is true necessarily
of all Moonlight Sonata performances, then (perhaps excepting any necessar-
ily finkish dispositions) it is true generically of Moonlight Sonata perfor-
mances. And so it follows that K is a kind whose members are, generically,
not kinds. Hence it follows from the Inheritance Hypothesis that K is not a
kind. And so we have deduced that the kind to which all and only
Moonlight Sonata performances belong is not a kind, which is a contradic-
tion. And so the Inheritance Hypothesis is unacceptable. We conclude,
therefore, that one cannot plausibly object to our argument via appeal to
the Inheritance Hypothesis.
One might respond by claiming that the Inheritance Hypothesis is more
general than what is needed to explain the truth of (7) and (4), and that we
should instead appeal to a restricted Inheritance Hypothesis instead. How-
ever, not only is it unclear how we can do this in a way that would avoid the
sort of worry just raised (for we can raise it with respect to other properties
such as being spatial, lacking members, etc.), but any restriction would leave us
without a unified explanation of the truth conditions of this sort of generic
sentence. The sentence ‘Moonlight Sonata performances aren’t kinds’ seems
relevantly similar to the sentence ‘Dogs bark.’ So if one is true in virtue of
the relevant kind having the predicated property (and it is generic in virtue
of it having inherited that property from its instances, which generically
have it), the other should be as well.
So far we have seen that problems result from combining SS with
SPV. Let us therefore turn to the view that results from combining SS
with OPV. That is, let us consider the view that when a predicate like ‘has
four paws’ or ‘is roughly fifteen minutes long’ occurs in a target sentences, it
can thereby pick out some property that differs from the property it would
normally pick out. If we adopt such a view, we can claim that in the
sentence
‘has four paws’ picks out not the property having four paws but rather the
property being such that its typical instances have four paws. Similarly, one can
maintain that in
‘is roughly fifteen minutes long’ picks out the property being such that its
typical performances are roughly fifteen minutes long. By adopting such a view,
the proponent of SS could endorse (3) and (4). For she could claim that the
referent of the subject of (3) is a kind, and that this kind has the property
picked out by the predicate of (3), since it is such that its typical instances
have four paws. And she could say something similar about (4).12
Such a view, however, faces a well-known problem—the problem of
anaphoric predication. Consider the following dialogue:
13. Yes, and so does this puppy, Chompy.
Because of the anaphoric reference in (13), it’s reasonable to assume that the
two sentences ascribe the same property to the entity referred to by their
respective subjects. But on the view we are now considering, in (3) ‘has four
paws’ picks out the property being such that its typical instances have four paws.
And so, if the two sentences ascribe the same property, then (13) must
ascribe to Chompy the property being such that his typical instances have four
paws. But Chompy, presumably, isn’t the sort of thing that has instances.
And so the view under consideration seems to predict, falsely, that (13) can’t
be true. A similar problem will arise if we claim that in (4), ‘is roughly fifteen
minutes long’ picks out the property being such that one’s performances are
typically roughly fifteen minutes long. For such a view falsely predicts that, in the
following dialogue, the second sentence cannot be true.
14. Yes, and so is the time-out I just received.
12 Wolterstorff endorses this. He follows Wollheim in claiming that, for any kind, k, and predicate,
P, if P can apply to k, and P picks out (at least in some instances of use in natural language) a property that
must be had by an entity in order for it to be an ideal member of k, then p can be truly predicated of k
(Wollheim 1968, pp. 64–5), (Wolterstorff 1975, p. 328). However, he then claims that, in many cases of
this, the kinds and their members do not share the relevant properties. He says that instead, there is
‘analogical predication’; the predicate picks out one property when applied to the member, and a distinct
property when applied to the kind. (Wolterstorff 1975, pp. 326–8)
For the time-out referred to in (14) presumably doesn’t have instances, and
so it isn’t such that its typical instances are roughly fifteen minutes long.
To recap: If the proponent of Simple Semantics adopts the Straightfor-
ward Predication View, then her view will overgenerate (predicting that
sentences have true readings when they don’t) and it will also undergenerate
(predicting that sentences don’t have true readings when they do). If, on the
other hand, she rejects the Straightforward Prediction View, then she will
face the problem of anaphoric predication just described.
II. The GEN Operator View

Maybe we should take seriously the idea that ‘the polar bear has four paws’ is
not a claim about some singular entity referred to by ‘the polar bear,’ but is
rather a general claim about polar bears, and similarly, that ‘The Moonlight
Sonata is roughly fifteen minutes long’ is really a general claim about
Moonlight Sonata performances.13
Clearly this view would be implausible if it implied that the target
sentences were universal generalizations, since they do not require full gener-
ality. Instead, these sentences seem to be about how polar bears and
Moonlight Sonata performances typically or generally are. To capture this, we
can appeal to an unvoiced, two-place operator, the GEN operator, which
functions as an adverb of quantification (like ‘usually,’ ‘typically,’ ‘gener-
ally,’ etc.).14 Thus, instead of having simple subject-predicate form, sen-
tences containing the GEN operator have three components: the two-place
operator, and its two arguments.15 Using this operator, for any predicates F
and G, we can state that things that are F are usually, typically or character-
istically G, as follows:
15. GEN(x)[F(x),G(x)]
One might then propose that the logical forms of (3) and (4) are as follows:
3a. GEN(x)[Is-a-polar-bear(x), Has-four-paws(x)]
4a. GEN(x)[Is-a-Moonlight-Sonata-performance(x), Is-roughtly-15-
minutes-long(x)]
13 Perhaps including some merely possible Moonlight Sonata performances.

14 For an introduction to the GEN operator, see Carlson and Pelletier (1995), and Leslie (forthcom-
ing).
15 Leslie (forthcoming).
On this view, in (3), ‘the polar bear’ is not a name referring to some
individual. Rather, it picks out a property, namely the property being a
polar bear, which occurs within the scope of the GEN operator. Similarly, in
(4), ‘The Moonlight Sonata’ isn’t a name referring to an individual, but rather
picks out the property being a Moonlight Sonata performance. And this
account seems to explain nicely why (3) and (4) are true: (3) is true because
polar bears typically have four paws, and (4) is true because Moonlight Sonata
performances are typically roughly fifteen minutes long.
However, not all sentences whose subject phrases are ‘the polar bear’ or
‘The Moonlight Sonata’ can be interpreted in this way. For many such
sentences do not seem to make generic claims about polar bears or perfor-
mances. Consider:
16. The polar bear is widespread.
17. The polar bear is nearly extinct.
18. The Moonlight Sonata is popular.
19. The Moonlight Sonata is two hundred years old.
Clearly (16) and (17) don’t say that, typically, individual polar bears are wide-
spread or extinct, and (18) and (19) don’t say that, typically, individual perfor-
mances of The Moonlight Sonata are popular or over two hundred years old.
In order to explain the difference between (3) and (4) on the one hand,
and (16) through (19) on the other, one might hold that ‘the polar bear’ and
‘The Moonlight Sonata’ don’t refer to individuals in (3) and (4), but that they
are names referring to individuals in (16) through (19). Thus, one might
hold that the logical forms of (16) and (17) are as follows:
16a. Widespread(The Polar Bear)
17a. Popular(The Moonlight Sonata)
However, cases involving anaphoric reference raise problems for this view.
Consider the following sentence:
20. The Moonlight Sonata is popular and is roughly fifteen minutes long.
On the view under consideration, the logical form of this sentence would
have to be as follows:
20a. Popular (The Moonlight Sonata) & GEN(x)[Is-a-Moonlight-Sonata-
performance(x), Is-roughly-15-minutes-long(x)]
But this can’t be the form of the sentence. For in (20), the reference in the
second conjunct is anaphoric on the first, and so the two conjuncts must
have some element in common. But in (20a), this is not the case.
This problem can be solved, however, via appeal to Gregory Carlson’s
theory of generic statements (applied to repeatable artwork sentences by
Stefano Predelli (2011)). Carlson (1982) posits a ‘realization’ relation holding
between the relevant manifestations and the entity being manifested.16 On
this treatment of generic sentences, (3) has the form (3b):
3b. GEN(x)[M(The Polar Bear, x), Has-four-paws(x)]
Here, M(y, x) indicates that x is a member of y. We can thus take ‘the polar
bear’ to refer to a single entity, like a kind.
Similarly, on Predelli’s view of repeatable artwork sentences, ‘The Moon-
light Sonata’ is always a name referring to a single entity. Sometimes, as in the
standard reading of (18), this name serves as the subject of a simple subject-
predicate sentence. And in other cases, such as the standard reading of (4),
this name lies within the scope of a GEN operator. Thus, on Predelli’s view,
the logical form of (4) is (4b):
4b. GEN(x)[R(The Moonlight Sonata, x), Is-under-15-min-long(x)]
Here R(y, x) indicates that x is a performance of y. More generally,
according to Predelli, where ‘m’ is the name of a musical work, sentences
with the surface form ‘m is F’ can have readings with either of the following
two logical forms:
21. F(m)
22. GEN(x)[R(m, x), F(x)]
Similarly, for Carlson, where ‘k’ refers to some natural or artificial kind,
such as the polar bear or the Honda Fit, sentences with the surface form ‘k is F ’
can have readings with either of the following two logical forms:
16 For Carlson, this relation can be stood in by instances and the kinds they belong to, or by stages and
the individuals they belong to. The latter is helpful in capturing the meaning of ‘characterising
sentences,’ where a property is claimed to be generically had by an individual (e.g. ‘Marzette garden’).
23. F(k)
24 GEN(x)[M(k, x), F(x)]
And in the case of a mixed sentence with the surface form ‘m is F and m is
G,’ the two conjuncts may differ in their logical form: One may have a
simple subject-predicate form, while the other may involve the GEN
operator. And this, according to Predelli, will also be true in the case of
sentences like (20). According to Predelli, the standard reading of (20) has
the following logical form:
20b. Popular (The Moonlight Sonata) & GEN(x)[R(The Moonlight
Sonata, x), Is-roughly-15-min-long(x)]
On this view, we can make sense of the reference in the second conjunct
being anaphoric on the first, since the two conjuncts share a common
element, as they both refer to The Moonlight Sonata.
The Carlson/Predelli View also solves the undergeneration problem
faced by SS. Recall from the last section that, if we assume SS and SPV,
then regardless of what we take ‘The Moonlight Sonata’ to refer to, we will
predict of some sentences that they lack true readings when in fact they have
true readings. If, for example, we take ‘The Moonlight Sonata’ to refer to
some atemporal entity, then our view will predict that,
has no true reading, since atemporal entities don’t have durations. The
Carlson/Predelli View solves this problem, for according to Predelli’s
view, (4b) is a possible reading of (4), and on this reading, (4) is true even
if the referent of ‘The Moonlight Sonata’ is atemporal. Similarly, according to
Carlson’s view, one possible reading of (3) is (3b), and on this reading it is
true. More generally, since the Carlson/Predelli View implies that the target
sentences have readings on which they make generic claims about the
members of the kinds to which the subject refers, rather than claims about
the referent of the subject, this view implies that such sentences can be true
even when this referent lacks the property picked out by the predicate.
Unfortunately, though the Carlson/Predelli View solves the problem of
undergeneration, it doesn’t solve the problem of overgeneration. That is,
the Carlson/Predelli View, like the view that arises from combining SS with
SPV, implies that sentences have true readings when they don’t. The reason
is this: In order to explain the existence of true readings of sentences such as

‘the polar bear is widespread’ and ‘The Moonlight Sonata is popular,’ which
don’t seem to predicate being widespread or being popular generically of polar
bears or of Moonlight Sonata performances, the view under consideration
claims that the target sentences have a reading on which they have simple
subject-predicate form. Hence, this view implies that such sentences have a
reading on which they have precisely the Simple Semantics discussed in the
previous section. And, in order to avoid the problem of anaphoric predica-
tion, the view under consideration will be paired with the SPV. But we saw
in the last section that if we combine SS with SPV, we will predict that some
sentences have true readings that they lack. Hence, the Carlson/Predelli
View will likewise imply that some sentences have true readings that they
lack. Suppose, for example, that we take ‘The Moonlight Sonata’ to refer to an
atemporal entity, the sort of entity that has no duration. In this case the
Carlson/Predelli View will predict that,
25. The Moonlight Sonata is not roughly fifteen minutes long.
has two readings, namely:
25a. :(Is-roughly-15-minutes-long (The Moonlight Sonata))
25b. ::GEN(x)[R(The Moonlight Sonata, x), Is-roughly-fifiteen-
minutes-long(x)]
While on the second of these readings the sentence will be false, on the first
reading it will be true. Hence, the Carlson/Predelli View will incorrectly
predicts that (25) has a reading on which it is true. Similarly, if we take ‘The
Moonlight Sonata’ to refer to a concrete entity that has some location or
other, then this view will predict, for some unambiguously false sentence
concerning the location of The Moonlight Sonata, that it has a true reading (in
the manner indicated in Section I).
This same problem of overgeneration extends straightforwardly to the
corresponding view of generic sentences. Moreover, since Carlson’s view of
generics applies not only to generic sentences involving definite singulars
(e.g. ‘the polar bear has four paws’) but also to generic sentences involving
bare plurals (e.g. ‘polar bears have four paws’), it faces an additional prob-
lem. Consider the following generic sentence involving a bare plural:
26. Kinds that are not members of themselves are lovely.
On Carlson’s view, this statement would be interpreted as follows:
26a. GEN(k)[M(K, k), Lovely(k)]

Where k ranges over kinds and K is the kind consisting of all and only those
kinds that are not members of themselves. But there is a problem. There is
no kind consisting of all and only those kinds that are not members of
themselves; the supposition that there is such a kind leads to Russell’s
paradox. Hence, on the view of generics we are now considering, (2)
involves reference to a non-existing kind, and so it cannot be true. But
clearly (2) (or some sentence relevantly similar to it) could be true. And so
the view of generics we are considering doesn’t provide an adequate,
general account of generics.
It seems, then, that we should look for an alternative account of the target
sentences.
III. The Higher-Order Predication View

Let’s take stock. We want to make sense of sentences such as the following:
20. The Moonlight Sonata is popular and it is roughly fifteen minutes long.
27. The polar bear is nearing extinction and lives on a diet of marine
animals.
In each of these sentences, the anaphoric reference suggests that the two
conjuncts are about the same thing. And yet, the conjuncts seem to be about
different things: The second conjunct of (20) appears to say something
about individual Moonlight Sonata performances, while the first does not;
and similarly, the second conjunct of (27) appears to say something about
individual polar bears, while the first does not. One response, which we
considered in the previous section, is to interpret each first conjunct as
referring to something other than polar bears or individual Moonlight Sonata
performances, and to interpret each sentence’s second conjunct in light of its
first, as involving reference to that same entity. This, response, as we have
seen, faces serious problems. But there is another option: Perhaps we should
take the second conjunct at face value, and interpret the first conjunct in
light of the second. In other words, perhaps we should claim that both
conjuncts of (20) are about Moonlight Sonata performances, and that
both conjuncts of (27) are about individual polar bears, and that these two
conjuncts differ only in terms of what sort of thing they say about these
entities. On this view, the first conjunct of (20) says of Moonlight Sonata
performances that collectively they are popular, and the first conjunct of (27)
says of polar bears that collectively they are nearing extinction. By contrast, the
second conjunct of (20) says of Moonlight Sonata performances that individu-
ally they are roughly fifteen minutes long, and the second conjunct of (27) says of
polar bears that individually they live on a diet of marine animals.
An analogy may be helpful. Consider the following:
28. Outside our tent there were three lions: Leo, Lex, and Lionel. The
lions were gathered around the water hole and they had shaggy
manes.
It is very plausible that both conjuncts of (28) are simply about Leo, Lex, and
Lionel. These conjuncts differ in that while the first says something about
them collectively, the second conjunct says something about them distribu-
tively.
A view of generics along these lines has been proposed by Kathrin
Koslicki (1999), who develops an account of generics that borrows from
the view of plural quantification developed by George Boolos, James
Higginbotham, and Barry Schein. On Koslicki’s view, the logical form of
(3) can be represented as follows:
3c. Have-four-paws((ØX)(8y)[X(y)$Is-a-polar-bear(y)])
Here, the expression within the outer parentheses is to be read as follows:
‘the things such that, for all y, y is one of them just in case y is a polar bear.’
Thus, (3c) says, of the things such that, for all y, y is one of them just in case y
is a polar bear, that, generically speaking, they have four paws. In
other words, (3c) says of the polar bears that they generally, usually, or
characteristically have four paws. Similarly, on Koslicki’s view, the logical
form of,
is as follows:
16b. Widespread((iX)(8y)[X(y)$Is-a-polar-bear(y)])
Thus, (16b) says of the polar bears that they are widespread. And so (3) and
(16) are about the same things, namely, the polar bears. They differ only in
the sort of thing they say about the polar bears: The predicate ascribed in
(16b) (denoted by ‘Widespread’) is collective, whereas the predicate
ascribed in (3c) (denoted by ‘Have-four-paws’) is distributive (though
not strictly universal). Since such a view implies that (3) and (16) have the
same subject, this view can make sense of the anaphoric reference in
sentences such as ‘the polar bear has four paws and is widespread’ or ‘the
polar bear is nearing extinction and lives on a diet of marine animals.’
Such a view can straightforwardly be extended to repeatable artwork
sentences. On the resulting view, the logical forms of (4) and (29) will be as
follows:
4c. Are-15-min-long((ØX)(8y)[X(y)$Is-a-Moonlight-Sonata
-performance(y)])
17b. Popular((ØX))(8y)[X(y)$Is-a-Moonlight-Sonata-performance(y)])
Once again, because this view implies that (4) and (17) have the same
subject, this view can explain the anaphoric reference in sentences such
as (20).
This view also solves the Russell’s paradox problem we discussed at the
end of the last section. For according to the present view, the logical form of,
26. Kinds that are not members of themselves are lovely.
is as follows (where k ranges over kinds):
26b. Are-lovely((ØX)(8k)[X(k)$Is-not-a-member-of-itself(k)])
The expression within the outer parentheses is to be read as follows: ‘the
kinds such that, for any kind, k, k is one of them just in case k is not a member
of itself .’ This sentence thus says, of those kinds, that, generically speaking,
they are lovely. Hence, on the view we are now considering, sentence (26)
makes no reference to any kind consisting of all those kinds that are not
members of themselves. And so this view avoids Russell’s paradox.17 Note
that on Koslicki’s view, the predicates that appear in the target sentences are
not ordinary, first-order predicates, but are rather second-order predicates
(we are using the bold font to represent second-order predicates). Thus, on
17 The use of second-order quantification to avoid problems raised by Russell’s paradox goes back to
Boolos (1984). Q: How much is understood about truth conditions of generics? A: Not one iota,
Koslicki’s view, there is an important difference between sentences (3) and

(29), on the one hand, and sentence (30) on the other:
29. Polar bears have four paws
30. Chompy has four paws.
According to Koslicki, (3) and (29) are two ways of saying the same thing. In
each case, the logical form is as follows:
3c. Have-four-paws((ØX)(8y)[X(y)$Is-a-polar-bear(y)]).
By contrast, on Koslicki’s view, the logical form of 30 is this:
30a. Has-four-paws(Chompy)
Sentence (30a) involves an ordinary, first-order predicate (denoted by Has-
four-paws), a predicate that applies directly to an individual. By contrast, (3c)
involves a higher-order predicate (denoted by Have-four-paws), a predicate
that applies plurally: It is the sort of thing that can be true not, e.g. of some
particular polar bear, but rather of polar bears. Koslicki acknowledges that
there is an important connection between these two predicates. According
to Koslicki, in order to understand the predicate Have-four-paws, one must
understand that this predicate is true of the Xs just in case, generically, ‘Has-
four-paws’ is true of the individual members of the Xs. Or symbolically:
31. Have-four-paws(X)$GEN(x)[X(x), Has-four-paws(x)]
Similarly, the predicate ‘Are-15-min-long’ that appears in (4c) is a higher-
order predicate that differs from, but is importantly related to, the ordinary
predicate ‘Is-15-min-long’ that applies to individual performances or other
events. Their connection can be indicated as follows.
32. Are-15-min-long(X)$GEN(x)[X(x), Is-15-min-long(x)].
According to Koslicki, collective predicates, such as Widespread in (16b)
and Popular in (17b), are likewise higher-order predicates that can be
elucidated in terms of first-order predicates that apply directly to individ-
uals. But in the case of such collective predicates, their elucidations in terms
of first-order predicates will generally be significantly more complicated
than those given in (31) and (32).
While Koslicki holds that second-order predicates are importantly related

to first-order predicates, and that the truth conditions of the former can be
elucidated in terms of the latter, she is very clear that these first-order
predicates do not figure in the logical form of the sentences that they
elucidate. In her words, logical form (LF) is:
that level of linguistic representation at which all grammatical structure relevant to
semantic interpretation is made explicit. That is, LF represents the contributions of
grammar to meaning . . . [Consider the following elucidation]
‘cut’ is a verb that applies truly to events e, involving a patient y, and an agent
x who, by means of some instrument z, effects in e a linear separation in the
material integrity of y.
. . . The paraphrases presented above [as in (31) and (32)] should be viewed as similar
in nature to the elucidation of the meaning of ‘cut’. They reflect part of the lexical
knowledge a speaker possesses concerning verbs . . . and are, for this reason, not
represented as part of the logical form of the sentence.18
But now we have a problem. For consider sentences such as the following:
33. The polar bear has four paws, and so does Chompy.
34. The Moonlight Sonata is roughly fifteen minutes long, and so is the
time-out I just received.
On Koslicki’s view, in the first conjuct of (33), ‘has four paws’ picks out a
higher-order predicate. While understanding this conjunct might require
appreciating the connection between this higher-order predicate and
the first-order predicate Has-four-paws that applies to individuals, it is only
the higher-order predicate, not the first-order predicate, that figures in the
structure of the first conjunct. However, what the second conjunct ascribes
to Chompy cannot be a higher-order predicate. It must instead be the first-
order predicate Has-four-paws. And since the latter doesn’t figure in the first
conjunct, it is very hard to makes sense of the anaphoric predication in (33).
And the same applies to (34). If Koslicki’s view is extended to repeatable
artwork sentences, then it will imply that in (34) ‘is roughly fifteen minutes
long’ picks out a higher-order predicate, which is not the sort of thing that
can apply to directly to the referent of ‘the time-out I just received.’ Hence,
on this view, we cannot make sense of the anaphoric predication in (34).
18 Koslicki (1999, pp. 461–3).

IV. The Multi-Operator View

In the first two sections, we considered some views according to which ‘the
polar bear’ and ‘The Moonlight Sonata’ are singular terms. We saw that these
views face a problem of overgeneration; regardless of what entity is taken to
be the referent of ‘the polar bear’ or ‘The Moonlight Sonata,’ the view in
question will predict that some target sentences have true readings which, in
fact, they lack. Then in Section III we considered a view that seems to avoid
this problem. On Koslicki’s view, ‘the polar bear’ is not a singular term
referring to some special entity, but rather, it refers plurally to the polar
bears. Since this view avoids the implication that ‘the polar bear’ refers to
some entity distinct from polar bears, it avoids the implication that there are
true sentences of the form ‘the polar bear is F’ where F cannot be truly
predicated of polar bears (either individually or collectively). The corres-
ponding view of repeatable artwork sentences has the same advantage in
avoiding overgeneration. However, as we have seen, this view comes with a
cost: It implies that the predicates that figure in the target sentences are
higher-order predicates, and it implies, further, that at the level of logical
form, there is nothing in common between the target sentences and sen-
tences about ordinary objects. Thus, this view implies that the sentences ‘the
polar bear has four paws’ and ‘Chompy has four paws’ have nothing in
common at the level of logical form, since the logical form of former is:
3c. Have-four-paws((ØX)(8y)[X(y)$Is-a-polar-bear(y)])
While the logical form of the latter is:
30a. Has-four-paws(Chompy)
And so this view has trouble making sense of anaphoric predication, as in
(33).
One might think that the problem facing Koslicki’s view could be solved
with a simple modification. Perhaps the logical form of (3) is not (3c) but
rather:
3d. Has-four-paws((ØX)(8y)[X(y)$Is-a-polar-bear(y)])
Thus, perhaps, while the sentence involves plural reference, the predicate it
involves is simply an ordinary, first-order predicate. Such a view would
make sense of the anaphoric predication in (33), but it faces serious difficul-
ties. The problem is that it is unclear how an ordinary first-order predicate
could be applied plurally to the polar bears. To see the force of this problem,
consider the following sentences:
35. Polar bears consume a ton of fish per day. And so does Moby Dick.
36. Coral reefs cover a huge area. And so does Alaska.
On the view we are now considering, the logical forms of (35) and (36) are
as follows:
35a. Consume-a-ton-of-fish-per-day((ØX)(8y)[X(y)$Is-a-polar-bear
(y)]) & Consume-a-ton-of-fish-per-day(Moby Dick)
36a. Cover-a-huge-area((ØX)(8y)[X(y)$Is-a-coral-reef(y)]) & Cover-
a-huge-area(Alaska)
But there’s a problem with this view. For sentences (35) and (36) are
ambiguous. There is one reading of (35) on which the first sentence
means that polar bears usually, typically, or characteristically consume a
ton of fish per day, and there is another reading on which it means that,
collectively, polar bears consume a ton of fish per day. Similarly, there is a
reading of (36) on which the first sentence means that coral reefs usually,
typically or characteristically cover a huge area, and there is another reading
on which it means that, collectively, coral reefs cover a huge area. What this
shows is that there is more than one way in which a given predicate can
apply to the polar bears, or to coral reefs. The very same predicate, ‘con-
sumes a ton of fish per day’ can be said to apply to the polar bears because it
applies to them individually, or because it applies to them collectively. By
analyzing the logical form of (35) simply as (35a), the view we are now
considering obscures this distinction, and hence it fails to account for the
ambiguity of (35). Similarly, in analyzing (36) simply as (36a), this view
obscures the distinction between the two ways in which the predicate
picked out by ‘covers a huge area’ can be said to apply to the coral reefs.
Hence, this view fails to account for the ambiguity of (36).
So how are we to account for the ambiguity of (35) and (36)? We don’t
want the two readings of (35) simply to involve two distinct predicates
ascribing two distinct properties. For, on each of the two readings of (35),
we must explain the anaphoric predication in ‘so does Moby Dick.’ But what
is said of Moby Dick does not vary between the two readings of (35). And so
it seems to follow that the same property must be picked out by ‘eats a ton of
fish’ on the two readings of (35). Nor do we want to say that the two
readings of (35) differ because the subjects have different referents. We don’t
want to say, for example, that the difference between the individual and
collective readings of (35) is that on the individual reading, ‘polar bears’
refers to individual polar bears, whereas on the collective reading ‘polar
bears’ refers to the collection of polar bears. For consider the following pairs
of sentences:
37. Polar bears consume a ton of fish per day. And they have four paws.
38. Polar bears consume a ton of fish per day. And they are widespread.
Note that in each of (37) and (38), there are two possible readings of the first
sentence. But the explanation of this ambiguity can’t be that, in the case of
the individual reading, the subject refers to something individual, whereas
in the case of the collective reading, the subject refers to something collec-
tive. For if this were the explanation, then on the reading of (37) where the
first sentence has the collective reading, the anaphoric reference in the
second sentence would make no sense, and on the reading of (38) where
the first sentence has the individual reading, the second sentence would
make no sense.
And so we are in the following predicament. We want to explain the
two readings of ‘polar bears consume a ton of fish per day.’ We don’t want
to explain this difference by positing an ambiguity in ‘polar bears.’ More-
over, we don’t want to say that the bare plural ‘polar bears’ is a singular
term referring to some special entity (e.g. an abstract object, or the fusion,
or polar bears, etc.) for then we will run into the same kind of over-
generation problem we discussed in Section I. Instead, we want a view on
which the bare plural ‘polar bears,’ like the definite singular ‘the polar
bear,’ refers plurally to polar bears. Thus, we want it to be the case that, on
both interpretations of (35), ‘polar bears’ refers to the things such that, for
all y, y is one of them just in case y is a polar bear. And so, on either
reading of,
39. Polar bears consume a ton of fish per day.
we want the subject to be:
40. (ØX)(8y)[X(y)$Is-a-polar-bear (y)]
Moreover, in order to allow for the anaphoric predication in (33) (‘the

polar bear has four paws, and so does Chompy’), we want it to be the
case that, on either reading of (39), the logical form includes the ordinary,
first-order predicate ‘consumes-a-ton-of-fish-per-day,’ the very same
first-order predicate that applies to Moby Dick. And yet, in order to allow
for the two possible readings of (39), there must be some additional
element, over and above (40) and the property picked out by the predicate
‘consumes-a-ton-of-fish-per-day.’ There must be something that tells us
how this property applies to the polar bears. In particular, there must be
something that tells us whether this property applies to the polar bears
individually or collectively.
So here is our suggestion: In addition to an element that picks out the
polar bears and an element that picks out the property consumes a ton of fish
per day, each of the two readings of (39) contains a third element
that indicates how this property applies to the polar bears. On one of
the readings of (39), it contains a generic operator which indicates that the
property applies generically to the individual polar bears. And on the other
reading, it contains a collective operator that indicates that the property
applies collectively to polar bears as a whole. Let us denote these two operators
with a superscript G and a superscript C, respectively. Thus, on one
reading of (39), it contains the generic (G) operator, and it has the following
form,
G
39a. Consumes-a-ton-of-fish-per-day((ØX)(8y)[X(y)$Is-a-polar-
bear (y)])
and on the other reading of (39), it contains the collective (C) operator, and
it has the following form:
C
39b. Consumes-a-ton-of-fish-per-day((ØX)(8y)[X(y)$Is-a-polar-
bear (y)])
On the first of these readings, (39a), the sentence means that usually, typically,
or characteristically, individual polar bears consume a ton of fish per day.
Hence, on this reading, (39) is true if and only if the following is true:
41. GEN(x)[Is-a-polar-bear(x), Consumes-a-ton-of-fish-per-day(x)]
On the second reading, (39b), the sentence means that together polar bears
consume a ton of fish per day.
Thus, on the view we are proposing, the target sentences refer plurally to
many individuals. And if a target sentence involves an ordinary, first-order
predicate (that is, the kind of predicate that can apply directly to an
individual), then the sentence involves an unvoiced operator that indicates
how the property picked out by the predicate applies to the many individ-
uals referred to by the subject.
Note, however, that some target sentences do not involve ordinary, first-
order predicates. Rather, they involve higher-order predicates that cannot
apply directly to individuals. Consider, for example:
42. Polar bears are nearly extinct.
Unlike the property picked out by the predicate in (39), the property picked
out in (42) cannot be applied directly to individuals: No individual polar
bear, bald eagle, or hump back whale can be nearly extinct—it’s polar bears,
bald eagles, or hump back whales that can be nearly extinct. Thus, while
the subject that the predicate ‘consumes a ton of fish per day’ applies to
is individual, the subject that the predicate ‘are nearly extinct’ applies to is
plural. We don’t, therefore, need an operator that indicates how this
property applies to the polar bears. Hence, on our view, the logical form
of (42) is as follows:
42a. Nearly-Extinct((ØX)(8y)[X(y)$Is-a-polar-bear(y)])
Thus, when it comes to target sentences involving higher-order predi-
cates—predicates that pick out properties that cannot be applied directly
to individuals, such as ‘nearly extinct’ and ‘widespread’—our view coin-
cides with Koslicki’s: At the level of logical form, these sentences consist in a
higher-order predicate applied to an expression that refers plurally to many
things. Our view differs from Koslicki’s view only with respect to sentences
that involve what appear to be ordinary predicates, such as:
And it makes sense that (3) should differ in its logical form from (42) and
from (16) (‘the polar bear is widespread’). For we can say, ‘the polar bear has
four paws, and so does Chompy,’ but we can’t say, ‘the polar bear is nearly
extinct, and so is Chompy,’ nor can we say ‘the polar bear is widespread,
and so is Chompy.’ And so we should want an account according to which
(3) involves an ordinary, first-order predicate whereas (42) and (16) do not.
Another advantage of our view is that it can explain some prima facie
puzzling differences between different generic sentences. We noted earlier
that sentences like the following are ambiguous, as they allow for an
individual and a collective reading:
39. Polar bears consume a ton of fish per day.
43. Coral reefs cover a huge area.
Note, however, that the following sentences are not similarly ambiguous:
44. A polar bear consumes a ton of fish per day.
45. A coral reef covers a huge area.
There is no reading of (44) on which it means that polar bears together
consume a ton of fish per day, nor is there a reading of (45) on which it
means that coral reefs together cover a huge area. Note, further, that while
each of the following sentences could be true:
17. The polar bear is nearly extinct.
46. Polar bears are widespread.
42. Polar bears are nearly extinct.
The following sentences could not be true:
47. A polar bear is widespread.
48. A polar bear is nearly extinct.
All of these differences can be straightforwardly explained, on our view, if
we hold that the difference between the definite singular (‘the polar bear’),
the indefinite singular (‘a polar bear’) and the bare plural (‘polar bears’) in
the subject position of the sentences under consideration marks a difference
in the unvoiced operators that the sentence may contain. In particular, the
indefinite singular, when it is used in a sentence in which no particular
contextually salient individual is being referred to, requires the generic operator;
this requirement explains both why there are no collective readings of (44)
and (45) and why there can be no true reading of (47) and (48). Thus, our
view can nicely explain some differences between sentences involving
indefinite singular, definite singular, and bare plural noun phrases in the
subject position—differences which are mysterious on some other views of
these sentences, such as Koslicki’s.
Most of the target sentences we’ve discussed have involved definite

singular noun-phrases, such as ‘the polar bear.’ It is important to note that
frequently, the definite singular precludes application of the collective
operator. Consider, for instance:
49. The polar bear weighs over a thousand pounds.
50. Polar bears weigh over a thousand pounds.
While we can get a collective reading of (50), sentence (49) cannot be used
gramatically to make an assertion in the absence of a contextually salient polar
bear. While this pattern doesn’t hold for all predicates, we do find similar
results with ordinary predicates involving size, shape, mass, possession of
parts, etc.19 Since these are first-order predicates, on our view, they must
combine with an unvoiced operator that indicates how the first-order predi-
cate is to be applied. But in these sentences the operator in question can’t be
the collective operator, since the definite singular excludes the collective
operator from being combined with these predicates. And so the operator in
question must be the generic operator. Hence, on the view we have
proposed, target sentences involving ordinary, first-order predicates, such as,
will contain a generic operator, and will thus have a form like the following:
G
39a. Consumes-a-ton-of-fish-per-day((ØX)(8y)[X(y)$Is-a-polar-
bear(y)])
By contrast, on our view, target sentences with higher-order predicates that
cannot be applied directly to individuals, such as,
will involve no such operator, and will thus have a form like the following.
16b. Widespread((ØX)(8y)[X(y)$Is-a-polar-bear(y)])
19 There are some properties that can be combined with definite singular noun-phrases collectively.
For instance, it seems fine to say ‘the automobile is a big polluter.’ However, such predicates seem to be
the exception to the rule. There is an intuitive, natural class that the collective-operator-excluding
predicates form, though what it is in virtue of that those predicates, rather than these more unusual ones,
exclude the collective operator, is a project that is beyond the scope of this chapter.
Q: Why was the linguist relieved when she learned of the role ‘wide’ plays in ‘widespread?’
A: Because a little morpheme always helps with the pain.
If we extend this view to repeatable artwork sentences, and thus treat

sentences of the form ‘The Moonlight Sonata is F’ in the same way we treat
sentences of the form ‘the polar bear is F,’ then our view will imply that
insofar as such sentences involve ordinary first-order predicates, such as,
they will involve a generic operator, and will thus have a form like the
following:
G
3e. Has-four-paws((ØX)(8y)[X(y)$Is-a-polar-bear(y)])
G
4d. Is-roughly-15-min-long((ØX)(8y)[X(y)$Is-a-Moonlight-Sonata-
performance(y)])
And insofar as a repeatable artwork sentence involves a predicate that cannot
apply directly to particular performances, such as,
51. The Moonlight Sonata is often performed.
it will contain a higher-order predicate and no unvoiced operator, and will
thus have a form like the following:
51a. Often-performed((ØX)(8y)[X(y)$Is-a-Moonlight-Sonata-per-
formance(y)])
Note that this view solves the problem of overgeneration faced by the views
we considered in Sections I and II. Recall that on views on which ‘the polar
bear’ and ‘The Moonlight Sonata’ are singular terms referring to abstract
entities or kinds, sentences (3) and (4) will each have at least one reading
on which they are false. For abstract entities and kinds are not under fifteen-
minutes long, nor do they have four paws. By contrast, on our view,
sentences (3) and (4) are unambiguously true, since (excluding readings
on which some particular, contextually salient polar bear or performance is
being referred to) our view implies that the only readings of (3) and (4) are
(3e) and (4d) respectively, and on these readings, these sentences are true.
Furthermore, our view solves the problem of anaphoric predication that
arose both for the Obscure Predication View we considered in Section I,
and for Kathryn Koslicki’s view, considered in Section III. Neither of these
views can account for the anaphoric predication in sentences such as:
33. The polar bear has four paws, and so does Chompy.
34. The Moonlight Sonata is roughly fifteen minutes long, and so is the
time-out I just received.
But such sentences pose no difficulty for our view. For on our view these
sentences have the following form:
G
33a. Has-four-paws((ØX)(8y)[X(y)$Is-a-polar-bear(y)]) & Has-four-
paws(Chompy)
G
34a. Is-under-15-min-long((ØX)(8y)[X(y)$Is-a-Moonlight-Sonata-
performance(y)]) & Is-under-15-min-long(the time-out I just
received))
In each case, the two conjuncts contain the same predicate, and so the
anaphoric predication is unproblematic.
V. Conclusion
We began this chapter by noting that the semantics we give for repeatable
artwork sentences can have consequences for our ontology of repeatable
artworks. Insofar as we take these sentences to be true, we can learn about
the world by learning about their truth conditions. This is nothing new; it is
not uncommon, both in the literature about repeatable artworks and in
metaphysics more generally, to add to one’s ontology in an attempt to get
the right semantic results. For instance, if something like Simple Semantics
is true, and we think that some repeatable artwork sentences, like ‘The
Moonlight Sonata is under twenty minutes long,’ are true, then we will be
committed to positing individual entities with which we can identify things
like The Moonlight Sonata. But, contrary to what may be otherwise assumed,
we have argued that Simple Semantics is not true, and we have presented an
alternative semantics for our target sentences, which has no steep
ontological commitments. This semantics has motivation that is independ-
ent of ontological considerations, and unlike Simple Semantics, this view is
ontologically undemanding. It requires only that we posit manifestations of
the relevant artworks. That is, to account for the truth of sentences about,
e.g. The Moonlight Sonata, we need merely posit some (at least possible)
Moonlight Sonata performances. We are free to deny the existence of any
controversial entities like abstract kinds, multilocators, or even fusions of

distinct performances.
In fact, not only does the sort of semantics we’ve argued for give us the
option of being eliminativists about repeatable artworks, it seems to give us
a reason for taking that option. For, given this semantics, being realists
about repeatable artworks wouldn’t merely require us to posit some strange
entities in addition to the more familiar ones, it would require us to posit
entities that we never seem to speak of (at least outside of philosophical
contexts). The sentences that were the best candidates for being
about these entities are, on their most plausible interpretations, actually
not about those entities at all. So Semantic considerations for from requir-
ing an expansion of our ontology, seem to push for a restriction of it.
References
Boolos, George (1984) ‘To Be is to Be a Value of a Variable (or to Be Some Values
of Some Variables)’ Journal of Philosophy 81: 430–49
Caplan, Ben and Matheson, Carl (2004) ‘Can a Musical Work Be Created?’ British
Journal of Aesthetics 44: 113–34
—— —— (2006) ‘Defending Musical Perdurantism’ British Journal of Aesthetics
46(1): 59–69
—— —— (2008) ‘Defending “Defending Musical Perdurantism”’ British Journal of
Carlson, Gregory (1982) ‘Generic Terms and Generic Sentences’ Journal of Philo-
sophical Logic 11: 145–81
Carlson, Gregory and Pelletier, Francis (1995) The Generic Book (Chicago: Univer-
sity of Chicago Press)
Dodd, Julian (2004) ‘Types, Continuants, and the Ontology of Music’ British
Kivy, Peter (1983) ‘Platonism in Music: A Kind of Defense’ Grazer philosophische
Studien 19: 109–29
Koslicki, Kathrin (1999) ‘Genericity and Logical Form’ Mind and Language 14(4):
441–67
Leslie, Sarah Jane (forthcoming) ‘Generics’ The Routledge Companion to the Philoso-
phy of Language
Levinson, Jerrold (1980) ‘What a Musical Work Is’ Journal of Philosophy 77(1): 5–28
Liebesman, David (2011) ‘Simple Generics’ Noûs 45(3): 409–42

Predelli, Stefano (2011) ‘Talk about Music: From Wolterstorffian Ambiguity to
Generics’ The Journal of Aesthetics and Art Criticism 63(3): 273–83
Rohrbaugh, Guy (2005) ‘The Ontology of Art’ in Berys Gaut and Dominic Lopes
(eds) The Routledge Companion to Aesthetics, 2nd edn (New York: Routledge)
Spencer, Joshua and Tillman, Chris (2012) ‘Musical Materialism and the Inheri-
tance Problem’ Analysis 72(2): 259–64
Tillman, Chris (2011) ‘Musical Materialism’ British Journal of Aesthetics 51(1): 13–29
Wollheim, Richard (1968) Art and Its Objects: An Introduction to Aesthetics (New
York: Harper & Row)
——(1980) Art and Its Objects: With Six Supplementary Essays, 2nd edn (Cambridge/
New York: Cambridge University Press)
III
Arguments Against and
Alternatives To
7
Against Repeatable Artworks
ALLAN HAZLETT*
There seem to be repeatable artworks. Plays and musical works,

like Shakespeare’s Hamlet or Mussorgsky’s ‘Pictures at an Exhibition’, can be
performed again and again; installations, like Sol LeWitt’s Wall Drawing #260
(1975), can be installed over and over; dishes of food, like Jean-George
Vongerichten’s tuna tartare, can be prepared many times. Many philosophers
have embraced the existence of repeatable artworks; for example, Julian Dodd
(2007) writes that ‘the fact that [musical] works are repeatable’ is a fact that any
‘satisfactory’ theory of the ontology of musical works must explain (p. 20).
Here I’ll argue, on the contrary, that there are no repeatable artworks.
First, in Sections I–IV, I’ll present the best argument against repeatable
artworks: repeatable artworks are abstract objects, abstract objects have
all their intrinsic properties essentially, but artworks don’t have all their
intrinsic properties essentially. Then, in Section V, I’ll consider two alter-
native ways of thinking about would-be repeatable artworks.
The argument I’ll present has been articulated by Guy Rohrbaugh (2003,
pp. 177–95) as an argument against identifying repeatable artworks with
types: artworks are modally flexible (they have some accidental properties),
temporally flexible (their properties can change), and temporal (they come
into and go out of existence). Types, however, are neither modally flexible,
nor temporally flexible, nor temporal. Repeatable artworks, therefore,
aren’t types. Here I’ll focus only on modal flexibility, as it seems to me
that the modal inflexibility of types explains their temporal inflexibility and
atemporality (Section IV). Moreover, I’ll treat the modal inflexibility of
types as an instance of the modal inflexibility of abstract objects in general,
* Thanks to Sherri Irvin and Christy Mag Uidhir for comments on this chapter.
162 ALLAN HAZLETT
and offer an explanation of why abstract objects must be conceived of as

modally inflexible (Section II). Finally, my conclusion will be that there are
no repeatable artworks, pace Rohrbaugh’s view (2003, pp. 197–200) that
repeatable artworks are akin to biological species (Section I).
The argument, as I’ll articulate it, goes like this: repeatable artworks are
abstract objects; abstract objects have all their intrinsic properties essentially;
artworks don’t have all their intrinsic properties essentially; therefore there
are no repeatable artworks. A bit more exactly: if there are repeatable
artworks, they are abstract objects;1 no abstract object has any accidental
intrinsic properties; would-be repeatable artworks have at least one
accidental intrinsic property; therefore, there are no repeatable artworks.
The argument is valid, so it remains to defend its three premises.
I. Repeatable Artworks are Abstract Objects

Artworks seem to divide into two categories. On the one hand, certain
artworks are token concrete physical objects: paintings; sculptures; and
tapestries, for example. On the other hand, certain artworks seem not to
be token concrete physical objects: musical works; plays; and literary works;
and perhaps conceptual works and culinary dishes. The reason these works
seem not to be token physical objects is that they seem to be distinct from
various token objects or events that are in some sense their instances.
Thus ‘Pictures at an Exhibition’ seems distinct from the particular token
performances that are performances of it; the same, mutatis mutandis, for
Hamlet. Wall Drawing #260, at least on one understanding of the piece,
seems distinct from the particular token installations that are installations of
it; in the same way Jean-George’s tuna tartare seems distinct from the
particular token plates of food that are preparations of it. These artworks
seem to live double lives, unlike paintings: they seem to be repeatable, in
virtue of their (being capable of ) having many instances (performances,
installations, preparations).
1 The scrupulous reader may substitute this material conditional below, mutatis mutandis, when I talk
about repeatable artworks, as well as when I talk about abstract objects. Alternatively, you will be fine if
you allow that repeatable-artwork talk and abstract-object talk aren’t ontologically committing (Cam-
eron 2008, 2010).
AGAINST REPEATABLE ARTWORKS 163
If there are repeatable artworks, to what ontological category do they

belong? When it comes to musical works, many argue that musical works
are types, kinds, or universals, with some defending the view that
the relevant types are eternal, non-spatiotemporal, and discovered by
composers (Dodd 2007; Kivy 1993, pp. 35–74), and others defending the
view that the relevant types are non-eternal, temporal, and created by
composers (Davies 2003, pp. 30–46; Levinson 1980; Wolterstorff 1980,
pp. 33–105).2 On these views, musical works are taken to be ‘norm-kinds’
(Wolterstorff ), ‘indicated sound structures’ (Levinson), ‘prescribed, sound-
event kinds’ (Davies), or ‘norm-types of sound-sequence-event’ (Dodd).
These ideas could naturally be extended to include other intuitively
repeatable artworks. The essential idea is that of token objects or events
standing in the relation of ‘being an instance of ’ to the artwork itself. Thus
even Rohrbaugh, who rejects the identification of repeatable artworks with
types, says that repeatable artworks have ‘occurrences’ (2003, pp. 197–9).
This means that, if there are repeatable artworks, they are abstract objects,
since no concrete object has instances. Despite the ambiguity (Lewis 1986,
pp. 81–6) and perhaps vagueness of the distinction between the abstract and the
concrete, things with instances are unambiguously and determinately abstract.
Note well that being an instance of a type (kind, universal, etc.) is
different from being a member of a set. You might want to say that sets
with concrete members are themselves concrete (Lewis 1986, p. 83), or
(more importantly) that sets with temporal members, capable of entering
into causal interactions, are themselves temporal and capable of entering
into causal interactions (cf. Section IV). But repeatable artworks aren’t
sets, and their instances aren’t their members: sets have the members that
they have essentially, but it isn’t an essential property of ‘Pictures at an
Exhibition’ that it was performed in Glasgow, on 27 October, 2010, by the
Royal Scottish Academy Brass (Dodd 2007, p. 18).3
2 Davies says musical works are socially constructed (2003, pp. 30, 40), but it’s not clear what he means
by this. His lists of other socially constructed things include inflation, books, universities, weeds, the
market, parking tickets, general elections, the Open Championship, Chardonnay, and fluffy dice. It’s not
clear what all these things have in common that warrants calling them ‘socially constructed’. Davies does
suggest a characterization of socially constructed things, writing of musical works that ‘[t]hey are subject
to variability and change in their form and substance, depending on behaviors and organizations that
people contingently choose to adopt or revise’, (2003, p. 40) and he goes on to reject four other claims
that might be associated with the claim of social construction (pp. 41–6). The notion remains obscure.
3 This is so, even if some artworks are sets (with essential members), e.g. a series of screenprints.
164 ALLAN HAZLETT
Being an instance of an abstract type is also different from being a member

of a species. Dodd (2007) compares musical works to biological species in
defence of the familiarity of ‘norm-types’, of which there can be ‘improp-
erly formed tokens’, concluding that ‘[n]orm-types are part of the fabric of
the universe’ (p. 33), and Rohrbaugh (2003) sees the connection between
repeatable artworks and biological species as being even tighter: both
repeatable artworks and biological species are ‘abstract historical individuals’,
whose ‘occurrences’ share a common history. On this view:
We are not all human . . . because of our shared structure. Rather, the shared
structure is something to be explained by what makes us all human, a matter of
the historical nature of the species. (Ibid., p. 184)
Man is [an] abstract historical individual, an entity distinct from but dependent
on the historical succession of individual human beings. (Ibid., p. 199)
But this is the wrong way to understand the ontology of biological species.
The theory of evolution by natural selection not only undermined the idea
of biological categories as eternal and unchanging, but also provided us with
resources for a nominalist understanding of biological species. After Darwin,
we are able to abandon the conception of species as universals (whether
Platonic or Aristotelian, whether eternal or temporal) and move towards the
conception of species as ‘groups of interbreeding natural populations that
are reproductively isolated from other such groups’ (Mayr 1988, p. 318).
Biological species aren’t abstract historical individuals; they’re just groups of
organisms, related in certain ways (including historical ways). (In general,
one could use a relational understanding of group membership to avoid
positing abstracta.) With Mayr’s ‘biological species concept’, we can give
up the problematic talk of biological species as entities (whether as universals
or as particulars) and replace it with (relatively) unproblematic talk of
individual organisms related to one another. On this new conception, the
boundaries of ‘species’ (equivalence classes on the relevant relations of
geographic isolation and interbreeding ability) will be fuzzy and subject to
continual change. And that’s what we want, because that’s what Darwin
discovered. Thus, in contemporary biology, talk of ‘species’ is shorthand
for talk of groups of related organisms.4 We cannot analogize repeatable
artworks to biological species—unless we say that talk of ‘repeatable
4 Against this, see P.D. Magnus (this volume, Ch. 5).

artworks’ is shorthand for talk of groups of related physical objects or

events, which suggests the anti-realism about repeatable artworks that
I favour (Section V).5
II. Abstract Objects Have All of Their Intrinsic

Properties Essentially
Essences do two things at once: they individuate and they constrain. On the
side of individuation, the essential properties of x are meant to distinguish x
from other things, i.e. to explain why x 6¼ y, when x 6¼ y. On the side of
constraint, the essential properties of x are meant to tell us the ways in which
x couldn’t be different from how x actually is, i.e. to specify the modal
boundaries of x. (To put all this another way, essences are meant to provide
identity and persistence conditions.) As Stephen Yablo (2002) observes,
abstract objects seem fundamentally different from concrete objects in the
degree to which their essences constrain them:
Not a whole lot is essential to me[. O]f my intrinsic properties, it seems arguable
that none are essential, or at least none specific enough to distinguish me from
others of my kind [ . . . ] I have by contrast huge numbers of accidental properties,
both intrinsic and extrinsic. Almost any property one would ordinary think of is a
property I could have existed without [ . . . ] Abstract objects . . . are a different story.
I do not know what the intrinsic properties of the empty set are, but odds are that
they are mostly essential. Pure sets are not the kind of thing that we expect to go
through intrinsic change between one world and another. Likewise integers, reals,
functions on these, and so on. (p. 220)
It appears that abstract objects have all their intrinsic properties essentially,
and that concrete objects do not have all their intrinsic properties essentially.
Abstract objects have maximally ‘big’ or ‘thick’ essences; concrete objects
never do, and typically have relatively ‘small’ or ‘thin’ essences. This raises a
puzzle for the naturalistically inclined metaphysician. Abstract objects are
noticeably different, in this regard, from the paradigm of a naturalistically
5 As well, the normativity of biological species is rightly controversial (Bedau 1993; Dretske 2000;
Millikan 1984). Very briefly, to derive some normative claim about x from some claim about x’s natural
history seems to be a instance of the ‘naturalistic fallacy’. So we cannot easily explain the normativity of
repeatable artworks by appeal to the supposedly unproblematic normativity of biological species.
166 ALLAN HAZLETT
acceptable entity: concrete physical objects. So we’re faced with the ques-
tion of whether abstract objects are naturalistically acceptable, given their
divergence from our paradigm.
More importantly, abstract objects not only diverge from our paradigm,
but they diverge from it in a surprising way. Again, Yablo writes:
[T]he null set and the number 11 are thought to exist in every possible world. This
is prima facie surprising, for one normally supposes that existence is inversely related
to essence: the bigger x’s essence, the ‘harder’ it is for x to exist, and so the fewer
worlds it inhabits. And yet [(pure) abstract objects are] extremely well endowed
in the essence department, and missing from not even a single world. (Ibid.,
pp. 220–1)
How can there be objects with maximally ‘big’ or ‘thick’ essences that are
eternal and exist at every possible world? Consider the contrast when it
comes to destruction: concrete objects, which have relatively ‘small’ or
‘thin’ essences, are going out of existence all the time, in virtue of losing
one of their few essential properties; (pure) abstract objects, which have
maximally ‘big’ or ‘thick’ essences, never go out of existence, anywhere, ever.
This aspect of the puzzle arises in connection with pure abstract objects,
but this premise—that abstract objects have all of their intrinsic properties
essentially—applies to impure abstract objects: {Eiffel Tower} essentially
has the Eiffel Tower as its only member. Just as pure sets aren’t intrinsically
different in different worlds (or at different times), impure sets aren’t
intrinsically different in different worlds (or at different times). Their
members may well be—the Eiffel Tower could have been intrinsically
different than it actually is, and it undergoes changes at the actual world at
all the time—but impure sets do not inherit the intrinsic properties of their
members. This isn’t obvious with singletons, but it is with any other impure
set: consider the intrinsic properties of {Brutus, Caesar} vis-à-vis the
intrinsic properties of Brutus and the intrinsic properties of Caesar.
So what’s going on here? Why are abstract objects modally inflexible?
A first step in answering this question is recognizing that the existence of
(pure) abstract objects (whatever this amounts to) makes no demands on the
world (in a sense that would eventually need to be articulated). There is
nothing the world must be like for it to be the case that (in whatever sense
this is true) (pure) abstract objects exist. The necessary existence of all
instances of the number ‘2’ (whatever this amounts to) isn’t a matter of
there being a stubborn regularity across the space of possible worlds, with
each of them having a ‘2’ in it.
Something similar needs to be said of impure abstract objects. {Eiffel
Tower} exists only if the Eiffel Tower exists, so {Eiffel Tower} seems
supervenient in a way that all instances of the number ‘2’ isn’t. But the
existence of {Eiffel Tower} also makes no demands on the world, at least in
the sense that the existence of {Eiffel Tower} is nothing over and above the
existence of the Eiffel Tower. There is no additional way the world must be
to ensure that, in addition to the Eiffel Tower, it also includes {Eiffel
Tower}. The existence of {Eiffel Tower} makes no demands of its own on
the world; the only demands made are those made by the existence of the
Eiffel Tower.
That the existence of abstract objects (whatever this amounts to) makes
no demands on the world explains why it can be the case that abstract
objects have all their intrinsic properties essentially. Recall the puzzle about
destruction. We can now see: concrete objects get destroyed so often
because their existence makes demands on the world, and with changes in
the world come changes in their existence. If abstract objects were like that,
their existence would be even more fragile, given their ‘bigger’ or ‘thicker’
essences. Only because their existence makes no demands on the world are
they able to survive in a plurality of diverse and changing worlds. Having
accidental intrinsic properties is the solution to this problem that concrete
objects use; but a solution is only needed if your existence makes demands
on the world in the first place.
What does it mean to say that the existence of abstract objects makes no
demands on the world? Yablo writes that the truth of statements about
abstract objects ‘does not depend on what may be going on in the realm of
concrete objects and their contingent properties and relations’ (p. 228).
Along similar lines, Agustı́n Rayo (2009) and Ross Cameron (2010) argue
that the truths of mathematics are trivial. Cameron writes:
[T]he truth of any mathematical claim does not demand the existence of numbers,
and it does not demand anything else, hence it does not demand anything at all;
mathematical truths are trivially true, in the sense that they have trivial truth
conditions—no conditions have to be met by the world in order for them to be
true. (Cameron 2010, p. 205)
He goes on to argue that, in this sense, mathematical objects exist, but that
they do not really exist—for the truths of mathematics do not require the
168 ALLAN HAZLETT
existence of numbers or any other objects. This view could naturally be

extended to the existence of abstract objects, in general. Something like this
has got to the right, otherwise we’re stuck with Yablo’s puzzle: abstract
objects have the most maximally constraining essences, but what seem like
the most minimally demanding existence conditions.
But it turns out to be their maximally constraining essences that explain
why the existence of abstract objects can be gotten for free in this way. For
suppose that x has some accidental intrinsic property F. In some worlds x is F,
and in others x isn’t F, and there are worlds in which x changes from being F
to not being F. But if x is like that, then it becomes impossible to see how the
existence of x could ‘made no demands on the world’, for if x is like that, then
x is clearly a denizen of the world, not a trivial existent that can be had for free,
as it were, as Yablo and Cameron argue. Note well that we are concerned
with intrinsic properties here. It is not hard to imagine an abstract object with
accidental extrinsic properties: all instances of the number ‘2’ are such that
Obama is President of the United States, but all instance of the number ‘2’ will
lose this property someday. The existence of all instances of the number ‘2’
can make no demands on the world, while being effected by changes in the
world, so long as the changes are changes in extrinsic properties. What we
cannot make sense of is the idea that the existence of all instances of the
number ‘2’ make no demands on the world, whilst all instances of the number
‘2’ undergoe intrinsic changes as a result of changes in the world. So
I conclude that abstracts objects have all their intrinsic properties essentially
(cf. Dodd 2007, pp. 53–6; and pace Rohrbaugh 2003, pp. 197–200).
III. Artworks Don’t Have All of Their Intrinsic

Properties Essentially
Artworks, on the other hand, don’t have all their intrinsic properties
essentially (Rohrbaugh 2003, p. 182). In particular, repeatable artworks
don’t have all their intrinsic properties essentially. ‘Pictures at an Exhibition’
could have not included the reprise of the ‘Promenade’ between the sixth
and seventh ‘picture’, had Mussorgsky not included it. Vongerichten could
have used onions instead of shallots in his tuna tartare. Horatio could have
been killed at the end of Hamlet.
This is compatible with the claim that artworks have some of their
intrinsic properties essentially (and with the claim that some artworks have
all of their intrinsic properties essentially). Hamlet, you might think, would
not have been Hamlet unless it concerned avenging a father’s murder;
Vongerichten’s dish would not have been that dish if it had been made of
strawberry ice cream. But there are intrinsic properties of artworks—at least
of some artworks, and in particular some paradigm would-be repeatable
artworks—that are not essential properties.
The fact that artworks don’t have all of their intrinsic properties essentially
allows us to make some important observations. For example, Stephen Davies
(2003) argues that some musical works have ‘thicker’ essences than others:
A piece that is specified solely as a melody and chord sequence, leaving instrumen-
tation, elaboration, and overall structure up to its performers, is thinner in consti-
tutive properties than one in which those features are also work-determinative.
(p. 39)
Davies observes that ‘the historical trend has been toward the thickening
of musical works’ (ibid.) and offers as explanation the fact that composers
have lost control over their works: having in the 17th and 18th centuries
conducted their own pieces, they came in the 19th and 20th to see their
scores printed and distributed all over the word (ibid., p. 40). The
‘thickening’ of musical works was an attempt to regain some of that lost
control. Works from the 19th and 20th century have ‘thicker’ essences than
17th and 18th century works, however, only if musical works don’t have all
their intrinsic properties essentially. For if they do have all their intrinsic
properties essentially, then their essences are, and always have been, max-
imally ‘thick’.
The premise is neutral on the question of origin essentialism, since the
relevant properties—e.g. being composed by so-and-so—are not intrinsic
properties. For what it’s worth, it seems to me that we should reject origin
essentialism. Some defenders of musical works (cf. Davies 2003, p. 44; Kivy
1993, pp. 70–3; Levinson 1980, pp. 20–1) argue that musical works are
essentially composed by their actual composer. Once upon a time (I
would like to say when I was very young) I confused the early modern
philosopher Francis Bacon, 1st Viscount Saint Alban, with the 20th century
painter Francis Bacon when, knowing nothing of the latter, I encountered
his Figure with Meat (1954) and briefly considered it as one of the most
170 ALLAN HAZLETT
prescient and original and downright incredible works of the 17th century.
Not quite. But had the Viscount made that painting back then, it would have
been an extraordinary achievement, and a very different achievement than
that of its actual creator. The coherence of this idea suggests the falsity of
origin essentialism.
The essentialist might appeal to the fact that we sometimes want to make
the following kind of modal claim:
Anybody with a wig stuffed full of counterpoint could have written the Toccata in
E major. But only Bach could have composed the Toccata, Adagio and Fugue in C,
the Fantasia in G, and the incomparable Passacaglia in C minor.6
This is ambiguous, as between an essentialist reading, on which the Toccata

in E major is essentially composed by Bach, and an alternative reading,
which is neutral on origin essentialism. Compare: you might think that only
Grant could have won at Shiloh; but no one thinks that Shiloh was
essentially won by Grant. This is a causal, rather than an essentialist, reading
of the claim that only so-and-so could have done such-and-such. The causal
interpretation is more charitable, when it comes to repeatable artworks and
their connections with their creators: it would be a strange essentialism that
took Grosses Meringues to be essentially created by Carême (which only he
could have invented) while denying that Canapés Melba were essentially
created by Escoffier (because my kid could have come up with that!). Either
all repeatable artworks have essential origins, or none do.7
The premise is also neutral on the question of the context-relativity of
meaning. You might argue that the identity of artworks is modally fragile
in virtue of such relativity: Menard’s Quixote is not the same work as
Cervantes’ Quixote (Borges 1939), and not just because they have different
authors, but because they have different meanings, given the historical
context of their compositions. (On this view, time of creation will be an
essential property of artworks.) Take this idea a few steps further, and you
6 Gramophone, July 1975, p. 71.

7 Renee Cox writes that ‘Tristan and Isolde could not possible have existed . . . at the time of Josquin,
for its existence is contingent on human constructs developed centuries later’ (Cox 1985, p. 371), but this
suggests a causal, and not an essentialist, interpretation. Kivy (1993) writes that ‘only Picasso could have
made’ Head of a Bull (1943), that ‘a different personality than Beethoven’s could no more give forth with
Beethoven’s 7th Symphony than could two different people have the same handwriting’, and that
‘certain objects are so unique as to be discoverable only by a single individual’ (pp. 72–3). This, too,
suggests a causal interpretation.
might think that the meaning of an artwork is changing all the time, that
every change in the world engenders a change in the meaning of the work,
conceived of as a socially, culturally, or historically constructed individual.8
On any such view, the meaning of an artwork is not among its intrinsic
properties. And essential extrinsic properties are not my concern here.
Dodd (2007) argues that musical works are abstract types, and accepts that
this entails that musical works have all of their intrinsic properties essen-
tially (2007, pp. 54–6, 86–92). Thus he accepts that musical works
could not have been intrinsically different than they actually are, that (as a
consequence) musical works cannot be intrinsically changed, and that:
‘Completing a work’ . . . is not a matter of bringing to an end a process of moulding a
single thing; it consists in arriving at a work by means of successively indicating
other entities on the way. Any development of the score, however minor, amounts
to the indication of a distinct object. (Ibid., p. 91)
Dodd says that this is ‘the tiniest wrinkle’ for his theory, which remains
‘quite coherent’ (ibid., p. 92). We should, on his view, consider this
‘a relatively minor conflict between the type/token theory and our ordinary
thought and talk about works’ (ibid., p. 90). Peter Kivy (1993) also takes the
intrinsic properties of musical works to be essential properties, such that two
works with the same sound structure are a fortiori the same work (pp. 60–6).
So perhaps we should reject the premise that repeatable artworks have
accidental intrinsic properties for theoretical reasons: because the best
theory of repeatable artworks available entails that repeatable artworks
don’t have accidental intrinsic properties. Against someone committed
to such a theory, my appeals above (e.g. to the fact that ‘Pictures at an
Exhibition’ could have been different) would beg the question.
This philosophical problem has a familiar structure: the three premises
cannot be consistently conjoined with realism about repeatable artworks. So
either one of the three premises or realism must go. Whether rejecting
realism—which I favour—is plausible will depend on what sense the
anti-realist can make of what seemed like repeatable artworks. We’ll return
to this briefly below, and sketch two anti-realist accounts of would-be
repeatable artworks.
8 Cf. Davies (2003), who supports ‘ontological contextualism, which acknowledges the socio-
historical embeddedness of some of the features making up the work’ (p. 34).
172 ALLAN HAZLETT
IV. Causation, Creation, and Consumption

There’s a cluster of objections to repeatable artworks that are based on
worries not about the mere existence of repeatable artworks, but on worries
about how they could causally interact with concrete reality, and thus about
how they could be created or consumed. Thus the objection that abstract
objects can’t be perceived (Dodd 2007, pp. 12–16, 92–4), the objection
that abstract objects can’t be created (Caplan and Matheson 2004; Dodd
2007, pp. 53–6, 86–92; Kivy 1993, pp. 38–47, 66–9; Levinson 1980, pp. 7–9),
and the objection that abstract objects can’t be destroyed (Kivy 1993,
pp. 46–7; Rohrbaugh 2003, pp. 193–5, 200). These objections are closely
related to the objection presented here. It seems to me that the modal
inflexibility of abstract objects explains why, to whatever extent this is
true, abstract objects cannot be perceived, created, or destroyed.
You might think that impure abstract objects can causally interact with
concrete reality, even if pure abstract objects can’t (cf. Caplan and Matheson
2004, pp. 117–23). So the defender of repeatable artworks might identify
repeatable artworks with certain impure abstract objects. I argued against
treating repeatable artworks as impure sets above (Section I), but this leaves
open the possibility that repeatable artworks might be identified with some
other impure abstract objects, capable of causal interaction with concrete
reality. The argument from modal inflexibility (Sections I–III), however,
shows why this identification cannot work.
The same point, mutatis mutandis, applies to worries about creation and
destruction stemming from the eternity of abstract objects (Dodd, pp. 53–6,
86–92; Kivy 1993, p. 66–9). To the extent that abstract objects are eternal,
their modal inflexibility explains their eternity.
This argument is thus superior to the argument from the premise
that abstract objects cannot causally interact with concrete reality (or the
argument from the premise that abstract objects are eternal), and for two
reasons. First, you might think that impure abstract objects can causally
interact with concrete reality (and that they are not eternal); if so, their
abstractness per se does not threaten the causal efficacy (and temporality)
of repeatable artworks. Second, the modal inflexibility of abstract objects
explains their acausal and eternal nature, thus providing an account of what
would otherwise be an undefended premise in the anti-realist argument.
V. What to Say about Would-Be Repeatable

Artworks
My conclusion is that there are no repeatable artworks. But we should try
to say something about what is going on, metaphysically, when it comes
to situations that would normally be thought of as involving repeatable
artworks. Here I’ll sketch two anti-realist ways of thinking about would-be
repeatable artworks.9 In my view, the availability of these accounts should tip
the balance in favour of rejecting repeatable artworks.
However, you could rethink the first premise of my argument (Section I),
and adopt a reductionist account of would-be repeatable artworks, on which
they are to be identified with some set of non-abstract objects. So, for
example, you might return to the theory, rejected above, that would-be
repeatable artworks are sets of token events or objects. A possibility we must
now take seriously is that the category of would-be repeatable artworks is
metaphysically heterogeneous. So, for example, you might argue that
conceptual artworks (like Wall Drawing #260), rather than being abstract
objects, are token thoughts in the mind of the artist (Section II), while
conceding that this ‘conceptual theory’ would not be at all plausible if
applied to musical works (Dodd 2007, pp. 26–9).
First anti-realist account: A profligate eliminativist account, on which
would-be instances of would-be repeatable artworks are each distinct
works of art, united only by intrinsic similarity and a common origin.
Consider Vongerichten’s tuna tartare. If we reject the existence of a repeat-
able artwork here, a culinary work, a dish or recipe, then what remains?
A plurality of token plates of tuna tartare, that seem to unified in some
way—we should have liked to say: unified by being instances of the
repeatable culinary work that is Vongerichten’s tuna tartare. What, if
anything, unites these plates? The present proposal says that these distinct
culinary works are united only by intrinsic similarity (similar shape, similar
taste, similar constituents, etc.) and a common origin (preparation in the
kitchen at Jean-George’s, causal connection to Vongerichten himself, etc.).
Culinary works are incapable of being repeated (being concrete tokens); but
they can be grouped together in terms of relations of intrinsic similarity and
9 There is also a semantic task for the anti-realist, vis-à-vis language that appears to refer to repeatable
artworks. For a charitable semantics of this kind, see Cameron 2008 (see also Dodd 2007, pp. 20–5).
174 ALLAN HAZLETT
common origin. Again, one might apply this idea to would-be repeatable
artworks in general, or to specific types of works.
Second anti-realist account: An austere eliminativist account, on which
would-be instances of would-be repeatable artworks are understood as copies
of some original artwork. Each token plate of tuna tartare is not a culinary
work in its own right, but a copy of some original token plate, an original
culinary work created by Vongerichten; this is what unites the token plates.
The original culinary work, on this view, is incapable of being repeated
(being a concrete token), but capable of being copied. As above, one might
apply this idea to would-be repeatable artworks in general, or to specific types
of works.
The profligate and austere views of culinary works differ not in their
ontology, nor even in their understanding of the fundamental metaphysics
of cuisine; they differ only in their individuation of culinary works. Given
one hundred plates of tuna tartare (based on Vongerichten’s recipe), the
profligate view sees one hundred (derivative) culinary works of art, while
the austere view sees one culinary work of art (viz. Vongerichten’s original),
with one hundred copies of that work.
With these options in hand, the anti-realist should be able to make sense
of would-be repeatable artworks, perhaps (as seems plausible) applying
different options to different types of works. The austere account has
some intuitive appeal when applied to literary works: we can identify
Sense and Sensibility with a token manuscript written by Austen herself,
and understand would-be instances of Sense and Sensibility as token copies of
that original token manuscript. And we should not be afraid of applying the
austere view to the performing arts, on the ground that copying is artless or
mechanical. Nor should we assume that in copying one aims only at fidelity.
When I make printed copies of your handwritten novel with movable type,
my concern is not only with fidelity, but with creating copies of your novel
that are more readable than the original. And I can make a copy of
something whilst aiming for something less than perfect fidelity, either for
prudential or for artistic reasons (e.g. a particular intended method of
copying might be essential to a particular original artwork). Copying
might require great skill and involve significant creativity; thus the austere
view applied to the performing arts does not violate the intuition that
performance involves creativity rather than slavish fidelity (Davies 2003,
pp. 62–3; cf. Dodd 2007, p. 33). There may be a fine art of copying; so to say
that performances of Hamlet are copies of some original performance is not

to denigrate the performing arts.
We might alternatively think of performances of Hamlet as copies of some
original play (rather than some original performance). As in the case of the
handwritten novel, in copying one need not aim to create something
intrinsically similar in all respects to the original. Convention may deter-
mine the appropriate standards of similarity for a given type of work. Think
here of the sense in which a printed photograph is a copy of the negative.10
You might object that there is a difference in kind between, on the one
hand, reading a paperback copy of Sense and Sensibility, and, on the other,
looking at a photograph of Guernica (1937). In the former case, you consume
the artwork itself; in the latter case, you fall short of this, because you are not
presented with Guernica itself, but merely with an image of it. On the
present view, reading a paperback copy of Sense and Sensibility would fall
short in just the same way: you would not be presented with Sense and
Sensibility itself, but merely with a copy of it. But, so the objection goes, that
is absurd.11
My reply is that the difference here is not metaphysical but rather a matter
of convention. Indeed, when you read a copy of Sense and Sensibility you are
only presented with a copy of it. But that consumers will be presented only
with copies is a convention of the literary arts that authors and readers
understand perfectly well. For this reason, and in this sense, it would be
absurd to think that your reading of Sense and Sensibility had fallen short vis-
à-vis aesthetic engagement, in a way analogous to the way in which your
‘viewing’ of Guernica via photograph falls short vis-à-vis aesthetic engage-
ment. (Compare Norman Rockwell’s The Problem We All Live With (1964),
which was intended for publication in Look magazine.) Convention here
determines the appropriate way to engage with a given work (cf. Davies
2003, pp. 36–7). Alternatively, you might think the difference here to be
more objective, flowing from the nature of painting and the nature of the
novel, respectively. That is fine; for all that is needed to reply to the objection
is a sense in which a reading of a copy of Sense and Sensibility is aesthetically
fine, whilst a ‘viewing’ of Guernica via photograph is not, consistent with
10 On the ontology of photography, see Rohrbaugh (2003) and Mag Uidhir (2012).
11 The same, mutatis mutandis, when it comes to the austere eliminativist approach applied to
culinary works.
176 ALLAN HAZLETT
there being no on to logical difference between the two cases: in neither case
are you face to face with the artwork itself. Whether by convention or by
nature, this is a problem for engaging with a painting, but not for engaging
with a novel. (To the extent that it is unsatisfying to say that we engage only
with a copy when we engage with some type of work, the anti-realist will
need to adopt a profligate eliminativist account.)
We have considered austere eliminativist approaches to several types of
artworks, but a profligate account may be more appropriate in many cases.
Consider the performing arts: on this view each performance of Hamlet is a
distinct work of art. We might then conceive of the relationship between
performance and play as follows: Hamlet the play is a literary work, created
by Shakespeare himself; token performances of Hamlet are theatrical works,
created by various actors, directors, (etc.) united in virtue of intrinsic
similarity and common relation to Hamlet the play.
This kind of view is attractive when we consider the particularity
of would-be instances of would-be repeatable artworks. Recall Walter
Benjamin’s idea:
Even the most perfect reproduction of a work of art is lacking in one element: its
presence in time and space, its unique existence at that place where it happens to
be . . . In the case of the art object a most sensitive nucleus—namely, its authenti-
city—is interfered with. (1936, pp. 220–1)
Token musical and theatrical performances may seem to have this kind of
‘aura’, and so may token plates of food. This suggests that we should treat
these tokens as distinct works of art—as on the profligate eliminativist
approach.12 But it is now clear that the decision to treat a particular token
as a copy of some original or as a distinct work of art is not settled by the
metaphysics of the situation, but on our intuitions about the individuation
of artworks. As Arthur Danto (1979) observed, a ‘perverse connoisseur’
12 The emphasis on mechanical reproduction is a red herring: two distinct token individuals can’t
share an aura simply because they’re two distinct token individuals, which a fortiori enjoy different
locations in time, space, history, culture, etc. Menard’s Quixote doesn’t share an aura with that of
Cervantes, but the one isn’t a mechanical reproduction of the other. A McDonald’s double cheeseburger
in Tokyo has a different aura (a different location in time and space, but also a different social meaning, a
different culinary context) than a McDonald’s double cheeseburger in Texas. Benjamin’s objection to
mechanical reproduction is a plea for the particularity of token individuals, but it makes trouble for
repeatable artworks whether in or out of the age of mechanical reproduction: if the authentic nucleus of
an artwork is tied to a particular time and place, then artworks must be token individuals (objects,
events), and neither abstract nor repeatable.
might maintain that ‘I missed something marvelous if I do not see Last Tango
in Paris at the Trans-Lux 85th Street on Friday at 8:00 p.m’ (pp. 100–1)
thereby applying a profligate eliminativist approach to token film screenings.
My point is that one’s evaluative intuitions will drive the choice between
an austere and an eliminativist approach: if one sees a particular token as
particularly marvelous, one will tend to see it as a distinct work of art. But this
tendency should be tempered by the point made above that copying need not
be artless and mechanical.
I conclude that the anti-realist has a number of attractive options when
it comes to understanding would-be instances of would-be repeatable
artworks. We should therefore maintain that artworks do not (in general)
have all their intrinsic properties essentially, and conclude that there are no
repeatable artworks.
References
Bedau, Mark (1993) ‘Naturalism and Teleology’ in Stephen J. Wagner and Richard
Warner (eds) Naturalism: A Critical Appraisal (South Bend: University of Notre
Dame Press)
Benjamin, Walter (1936/1968) ‘Das Kunstwerk im Zeitalter seiner technischen
Reproduzierbarkeit’ Zeitschrift für Sozialforschung 5: 1, translated as ‘The Work of
Art in the Age of Mechanical Reproduction’ in Walter Benjamin, Illuminations
(New York: Schocken Books)
Borges, Jorge Luis (1939/1998) ‘Pierre Menard, autor del Quijote’ Sur 56: 7–16,
translated as ‘Pierre Menard, Author of the Quixote’ in Jorge Luis Borges,
Collected Fictions (New York: Penguin)
Cameron, Ross (2008) ‘There Are No Things That are Musical Works’ British
—— (2010) ‘Necessity and Triviality’ Australasian Journal of Philosophy 88: 401–15
Caplan, Ben and Matheson, C. (2004) ‘Can a Musical Work be Created?’ British
Cox, Renee (1985) ‘Are Musical Works Discovered?’ Journal of Aesthetics and Art
Danto, Arthur (1979) ‘Moving Pictures’ Quarterly Review of Film Studies 4(1): 1–21,
reprinted in N. Carroll and J. Choi (eds) Philosophy of Film and Motion Pictures: An
Anthology (Blackwell, 2005) 100–12
Davies, Stephen (2003) Themes in the Philosophy of Music (Oxford: Oxford University
Press)
178 ALLAN HAZLETT
Dodd, Julian (2007) Works of Music: An Essay in Ontology (Oxford: Oxford

University Press)
Dretske, Fred (2000) ‘Norms, History, and the Constitution of the Mental’ in Fred
Dretske, Perception, Knowledge, and Belief (Cambridge: Cambridge University
Press)
Kivy, Peter (1993) The Fine Art of Repetition: Essays in the Philosophy of Music
Lewis, David (1986) On the Plurality of Worlds (Oxford: Blackwell)
Mag Uidhir, Christy (2012) ‘Photographic Art: An Ontology Fit to Print’ Journal of
Aesthetics and Art Criticism
Mayr, Ernst (1988) Toward a New Philosophy of Biology: Observations of an Evolutionist
(Cambridge, MA: Harvard University Press)
Millikan, Ruth (1984) Language, Thought, and Other Biological Categories: New
Foundations for Realism (Cambridge, MA: The MIT Press)
Rayo, Augustı́n (2009) ‘Toward a Trivialist Account of Mathematics’ in Otávio
Bueno and ysterin Linnebo (eds) New Waves in Philosophy of Mathematics
(Basingstoke: Palgrave-Macmillan)
Philosophy 11: 177–205
Yablo, Stephen (2002) ‘Abstract Objects: A Case Study’ Philosophical Issues 12: 220–40
8
How to be a Nominalist and
a Fictional Realist
ROSS P. CAMERON*
Consider the following arguments (call them ‘Something from Nothing

arguments’):
(A1) Postboxes and fire trucks have something in common—they are
red. Therefore,
(A2) There is some thing, some property, the property of redness, that
postboxes and fire trucks have in common. Therefore,
(A3) There are properties.
(B1) There are as many members of the Beatles as there are seasons in
the year: four. Therefore,
(B2) There is some thing, some number, the number four, that
numbers both the Beatles and the seasons in the year. Therefore,
(B3) There are numbers.
(C1) Mozart composed his first symphony when he was eight years old.
Therefore,
(C2) There is some thing, some musical work, Mozart’s first symphony,
which Mozart composed when he was eight years old. Therefore,
(C3) There are musical works.
(D1) The fictional monster Cthulhu was created by H.P. Lovecraft.
Therefore,
(D2) There is some thing, some fictional being, Cthulhu, that was
created by H.P. Lovecraft. Therefore,
(D3) There are fictional beings.
* Thanks to Elizabeth Barnes, Ben Caplan, Anthony Everett, Christy Mag Uidhir, Amie Thomasson,
Jason Turner, Tatjana von Solodkoff, Robbie Williams, and Richard Woodward for helpful discussion.
180 ROSS P. CAMERON
Arguments A–D have a common structure. The first line is uncontroversial,

the second line appears to be a restatement of the first, the third appears to
follow trivially from the second, and yet it says something apparently very
controversial, for it commits us to believing in a kind of entity that poses
various metaphysical problems. Of particular relevance for this chapter
are conclusions (C3) and (D3): the ontology of the aesthetic realm—such as
works of music, fictional beings (etc.)—is particularly problematic, since it
seems on the face of it to be an ontology of created abstracta. Abstract objects
are bad enough for the metaphysician who has a taste for a parsimonious
ontology; but created abstract objects! How is it that we create these objects
that are neither spatially nor causally related to us? I can understand how I can
causally manipulate the concrete world to create a table or a house, but
creation of an abstract object sounds simply miraculous. Of course, not all
ontology in the aesthetic realm is abstract: I can understand how one can
create a statue or a painting. But musical works and fictional beings do not
seem to be amongst the inventory of concreta and yet they do appear to be
amongst the inventory of the created; and hence we have a puzzle.
Of course, there are routes out of the puzzle. Julian Dodd (2007) denies
that these abstracta are created; Ben Caplan and Carl Matheson (2006)
argue that musical works are not in fact abstract. I won’t argue against either
here; I’m going to propose a response to arguments of the form A–D that
deflates the sense of getting something from nothing rather than one which
attempts to reconcile us to the situation by claiming that the something
that we get isn’t so bad after all.
I said that the first lines of these arguments are uncontroversial, but this
requires immediate qualification. Of course, we could discover that each of
them are false. Lovecraft might be revealed as a plagiarist; we could discover
that Paul was always dead,1 and hence that there were only three Beatles after
all; etc. But the first premises being susceptible to non-philosophical empirical
refutation does not ease the puzzle posed by the above arguments, for such
discoveries would only lead us to replace the first lines with something else
which would still lead to the troubling conclusion (e.g. that there are as many
Beatles as there are persons in the Trinity, or that the fictional monster
Cthulhu was created by Lovecraft’s plagiarized neighbour). Nor are the first
lines even immune to strictly philosophical refutation: perhaps philosophy
1 I am the walrus.
HOW TO BE A NOMINALIST AND A FICTIONAL REALIST 181
will finally succeed in establishing that the external world is simply an illusion,
and we will therefore conclude that fire trucks, the Beatles, Lovecraft, etc.
simply do not exist. Even so, some arguments such as those above will
remain: we would merely replace (B1) with ‘There are as many Beatles as
there are round squares’, and (A1) with ‘The external world and desert
mirages have something in common—they are both illusory.’ More import-
antly, the first lines’ being susceptible to this type of philosophical refutation
still does not ease the puzzle. For while we might grudgingly admit that the
external world might be an illusion, our credence that it is so is not high; we
invest an extremely high credence in there being some truth of the form of the
first line of each of arguments A–D. Rational arguments should be credence
preserving, so we should invest extremely high credence in their conclusions,
and that is enough to yield the puzzle.
The sense in which the first lines of arguments A–D are uncontroversial is
that philosophical qualms with the conclusions don’t seem like good reasons to
reject the first lines. It would be perverse to conclude that there aren’t as many
Beatles as there are seasons in the year because there are no numbers, or that
Mozart didn’t compose his first symphony at eight years of age because there are
no musical works. The first lines of the ‘Something from Nothing’ arguments
are on a far stronger epistemic footing than any metaphysical principles that
are in apparent conflict with their conclusions. We need another option.
We need to accept a meta-ontology according to which the conclusions
of arguments A–D can be accepted as true without being ontologically
committing to the metaphysically recalcitrant entities in question. There are
various proposals along these lines in the literature. (See, inter alia, Fine 2009;
Schaffer 2009; Williams 2010, ms.) My own view is that the ontological
commitments of a claim/theory are those entities that must be invoked as
truth makers for the claim/theory if it is to be held true. (See Cameron 2008b,
2010a, 2010c.) And, crucially, I hold that an existential claim can be made true
by something other than what it says exists: in that case, the entities quantified
over are not ontological commitments of the claim in question, and hence
metaphysical qualms about those entities are misplaced if taken as a reason to
deny the claim. As long as the entities we need to invoke as truth makers are
metaphysically acceptable by our own lights, we need not worry.
The challenge regarding aesthetic ontology, then, is to give an account of
what makes true our claims about the aesthetic realm so that the truth makers
we invoke do not involve created abstracta (and, for those with nominalist
182 ROSS P. CAMERON
sympathies, such as myself, no abstracta simpliciter). If it turns out to be true

that there are aesthetic objects which are abstract and which were created,
that is no problem, provided this truth can be made true by a palatable
ontology. In this chapter I will sketch out such a truth-making story for the
case of fictional beings (having tackled musical works in previous work).2
I want to start by looking at Amie L. Thomasson’s work, much of which
I am in sympathy with. Thomasson thinks it is a mistake to think that there
is a further question as to whether a fictional being exists after we have
established that there is a relevant literary practice concerning the fictional
character in question: so, e.g. it is simply wrong-headed to accept that there
is a coherent literary practice involving discussion of Bilbo Baggins but to
think that there is a remaining substantive philosophical question as
to whether Bilbo exists. This is not just a methodological point: Thomasson
is not merely saying that the existence and good standing of the literary
practice is good evidence that Bilbo exists—evidence strong enough
to outweigh any philosophical qualms we might have about granting that
there is such a thing as Bilbo. Rather, she is saying that the existence and
good standing of the literary practice is constitutive of the existence of Bilbo:
that if you think there is a remaining philosophical question concerning
whether Bilbo exists after establishing that there is a literary practice in good
standing then you are confused about what it takes for there to be something
that is Bilbo. The mistake you make in asking the further question is not
letting philosophical qualms trump strong common sense evidence; it is
the mistake of accepting that the constitutive conditions for it to be the case
that p are met but asking whether p is true. Here is Thomasson:
Our literary practices . . . definitively establish the existence conditions for fictional
characters—that is to say, they establish what it takes in a given situation for there to
be a fictional character. According to those criteria, what does it take for an author
to create a fictional character? This much is clearly sufficient: That she write a work
of fiction involving names not referring back to extant people or characters of
other stories, and apparently describing the exploits of individuals named . . . The
sentence ‘Jane Austen wrote a work of fiction pretending to refer to and describe a
young woman named “Emma Woodhouse” (not referring back to an extant
2 I discuss the problem for musical works in Cameron (2008a, 2010a). In those papers I didn’t actually
provide a nominalist friendly metaphysic, as the truth makers I put forward appealed to abstracta, albeit
not created abstracta. I think a nominalist version of that broad proposal is available, though I can’t get
into that here.
individual . . . )’ is, in virtue of the nature of the concepts involved, logically

sufficient to ensure that we can make reference to ‘the fictional character, Emma
Woodhouse’ . . . [O]nce we see that the existence conditions for fictional characters
are determined by practices, and recognize how minimal those conditions are, it
becomes difficult—and unnecessary—to deny that there are fictional characters, so
understood. (2003, pp. 148–51)
There are two thoughts here that I am utterly in sympathy with. The first is
that the existence of the literary practice is sufficient for the existence of
the fictional characters in question: as I would put it, that we appeal to
something to do with the literary practice rather than the fictional character
itself when specifying the truth maker for claims concerning the existence of,
and properties of, the fictional being. The second is that it is ‘unnecessary . . . to
deny that there are fictional characters, so understood’. I take it Thomasson
is saying that once you accept that this is the correct account of the
conditions under which fictional characters exist, metaphysical scruples
about their existence are revealed to be misplaced: as I would put it, what
needs to cohere with our metaphysical scruples are what we ultimately invoke
as truth makers—if ‘A exists’ is made true by B then metaphysical concerns
about the nature of A shouldn’t arise—all that matters is whether we’re happy
to admit B into our ontology.
But while I am in sympathy with the main thrust of what Thomasson says
about fictional characters, there are nonetheless some parts of her overall
theory that I want to distance myself from. Thomasson’s deflationary
attitude (2009) towards questions concerning the existence of fictional
characters is part of a broad deflationism with regard to existence
questions in general. Thomasson thinks that there is no good philosophical
discipline of ontology. When it comes to the existence of Fs, the task of the
philosopher is to do some conceptual analysis and discover—by reflection
on the concept of F-ness—the conditions under which Fs exist. It is
then an empirical—and non-philosophical—question as to whether those
conditions are met. So for example, consider the debate amongst ontologists
as to whether there are tables: Sider (2001) and Markosian (1998) say there
are, van Inwagen (1990) and Merricks (2001) say that there are not. All of
them agree, of course, that there are some things arranged table-wise. In that
case, thinks Thomasson, this debate is a confused one: for some conceptual
analysis of the concept table will reveal that what it takes for there to be a table
184 ROSS P. CAMERON
is for there to be some things arranged table-wise. To agree that there

are things so arranged and ask the further question as to whether there is a
table is, she thinks, to betray a philosophical confusion: you are accepting
that the constitutive conditions for it being the case that p are met but
nonetheless asking whether p is the case.
Now again, I have sympathy with what Thomasson says about tables.
I think that what makes it true that there are tables just are the
things arranged table-wise (Cameron 2008b, 2010c), and so I agree it’s a
philosophical mistake to accept that there are things so arranged and yet still
ask whether there are tables. And while I wouldn’t use the term ‘conceptual
analysis’ myself, I agree that the way to discover that what makes it true
that there are tables are the existence of things arranged table-wise is to
examine empirically our use of language and infer from that what worldly
conditions the correctness (and thus the truth) of our assertions about
tables is ultimately sensitive to.
However, while the truth of table-talk might not require the existence of
anything over and above the things arranged table-wise, it did presuppose
their existence. Might this then not be the point at which there can be a
substantive ontological debate? No, thinks Thomasson, the same story will
be told. What things is it we’re talking about that are arranged table-wise?
We must, thinks Thomasson, fill in some genuine sortal (and ‘thing’ is not
one!) in order to even have a meaningful ontological question in the
first place. Once we do—once we ask, e.g. about the electrons arranged
table-wise—the issue will play out just as before: we will need to do some
conceptual analysis to discover what the conditions are for it to be true that
there are electrons (so arranged), and then it will merely be an empirical
(and non-philosophical) question as to whether those conditions are met.
And Thomasson thinks this process will simply keep going. If what it
takes for there to be electrons is for there to be strings vibrating electron-wise
then we now just do the same thing for strings, and so on. There will never
be a distinctively metaphysical question to ask: merely conceptual analysis
and empirical investigation at every stage. This is where I get off the boat.
The process, I think, has to come to an end. We might think that there
are the Xs, but that this requires simply that there are Ys arranged F-ly, and
that for there to be the Ys arranged F-ly requires simply that there are the
Zs arranged G-ly . . . and so on . . . but at some point we have to stop
postponing the question and say how the world is: that for all those
existence claims, what is ultimately required is that there be the As arranged

H-ly, and that what it takes for there to be the As arranged H-ly is just
that—that there be the As arranged H-ly. This is the level at which we’ve
found the truth makers for all the truths on our list: that there are As
arranged H-ly is what makes it true that there are Zs arranged G-ly and
(hence) that there are Ys arranged F-ly and (hence) that there are the Xs. It’s
the things (and the properties of these things) at this level that we must be
robustly realist about, and it will be a substantive philosophical claim that
these things exist (and have the properties they have), and a substantive
and difficult metaphysical question as to what reality is like at this level:3 not
one that can be answered solely by conceptual analysis and empirical
investigation.
So to return to fictional beings, the task to my mind is to pinpoint the
features of reality that we are happy to be robust metaphysical realists about
that can give an adequate grounding to our talk about the existence of, and
properties of, fictional characters. The nominalist’s task, then, is to provide
such an account that doesn’t invoke abstracta at this ground level. As I said,
I agree with Thomasson that the truth maker should be something to do
with our literary practice, but what exactly are the features of the world that
ultimately make true our talk of fictional characters? I want to approach the
answer via discussion of a puzzle concerning fictional characters.
Anthony Everett argues against any theory that holds that there are
literally true claims about the existence of fictional characters on the
grounds that this will lead to a violation of Evans’ admonitions (1978)
against indeterminate identity. He begins by presenting us with a story
that, he thinks, is plausibly interpreted such that it is indeterminate whether
3 It is, I think, a substantive metaphysical question not just what in particular reality is like at the
fundamental level but what the very structure of the fundamental level is like. I’ve been talking as if the
fundamental truths are ones concerning the existence of things and the properties and relations holding
of/between those things, but it being a substantive issue which things exist, and what properties/relations
hold of them, at this level. But it is also a substantive issue whether fundamental reality is appropriately
described in subject/predicate form. The strictest version of truth maker theory holds that the funda-
mental truths are just about what there is and never about how those things are (see Cameron
(forthcoming)). Or one could even hold that the fundamental truths are not even quantificational—
that at rock bottom, reality doesn’t consist of there being things at all (see Turner (2011)); given that
possibility, Thomasson could be right that existence questions are always resolvable by conceptual
analysis and empirical investigation and that there is no good discipline of ontology; but still we
would be left with the question as to how reality ultimately is such as to make true all these existence
claims.
186 ROSS P. CAMERON
certain characters in the story are identical. Here is his story in full (Everett
2005, p. 629):
Frackworld: No one was absolutely sure whether Frick and Frack were really the
same person or not. Some said that they were definitely two different people. True,
they looked very much alike, but they had been seen in different places at the same
time. Others claimed that such cases were merely an elaborate hoax and that Frick
had been seen changing his clothes and wig to, as it were, become Frack. All that
I can say for certain is that there were some very odd similarities between Frick and
Frack but also some striking differences.
Everett then tries to cause trouble for the fictional realist by arguing as
follows. According to the Frackworld fiction, it is indeterminate whether
Frick is Frack. By fictional realism, Frick and Frack exist. Are Frick and
Frack identical (asking now whether the fictional objects are identical
simpliciter, not whether it is true according to the fiction that they are
identical)? Plausibly, the actual identity or distinctness of fictional characters
should be determined by whether or not it is true according to the relevant
fiction(s) that they are identical: Bilbo is not Gandalf in reality because it is
true according to The Lord of the Rings that Bilbo is not Gandalf; Morgoth
is in reality Melkor because it is true according to The Silmarillion
that Morgoth is Melkor. In that case, it is actually indeterminate whether
Frick is Frack; but Evans has shown that indeterminate identity is impossible.
Reductio of fictional realism.
Before proceeding, a couple of points about Everett’s case. I don’t think
the Frackworld fiction actually does much to sell the idea that, according to
the fiction, Frick and Frack are indeterminately identical. Starting the story
with ‘No one was absolutely sure . . . ’ and ending with ‘All that I can say for
certain . . . ’ makes it overwhelmingly plausible, to my ears, that we’re
dealing with a merely epistemic issue: that in this fictional world there is a
simple fact of the matter as to the identity or distinctness of Frick and Frack,
but that the inhabitants of this world are unsure as to what it is.
Also, as Benjamin Schnieder and Tatjana von Solodkoff (2009, pp. 140–1)
point out, we must be careful to distinguish between a fiction leaving it open
whether a and b are (according to the fiction) identical and a fiction
positively determining that the identity or distinctness of a and b is
(according to the fiction) not determinately settled one way or the other.
Frackworld certainly leaves open whether Frick and Frack are identical, but
in just the same way as it leaves open their height. Nothing about Frack-
world suggests that it is true according to the fiction that Frick, say, is
indeterminate in height. Frackworld is simply silent on Frick’s height:
it does not positively settle that Frick’s height is unsettled. This is important,
for compare the following two claims:
(1) If it’s true according to the fiction that a and b are indeterminately
identical then the fictional characters a and b are indeterminately
identical.
(2) If the fiction leaves it open whether a and b are identical then the
fictional characters a and b are indeterminately identical.
(1) is a lot more plausible than (2). There’s no particularly good reason
to think that a fiction’s leaving something open about a character means
that reality leaves it open how the fictional character is in that respect.
Frackworld leaves it open what height Frick is according to the fiction, but
reality plausibly settles what height the fictional entity Frick actually has:
it has no height, because it’s not the kind of entity to have a height
property—although it is the kind of entity to be such as to have a certain
height according to a fiction, and it’s certainly not true according to the fiction
that it has no height. Likewise, reality might settle the issue of Frick and Frack’s
identity even if the fiction leaves it open whether they are identical in the
fiction. This is what Schnieder and von Solodkoff think: if the fiction leaves it
open whether or not they are identical, reality settles it that the fictional
entities are distinct. The idea here is that the default is distinctness, and that
when a fiction uses two names it has to positively settle the identity (according
to the fiction) of the referents for it to be the case in reality that we have one
fictional entity rather than two. If this is right, and I think it is plausible, then
there is no actual indeterminate identity inherited in reality as a result
of Frackworld leaving the identity according to the fiction open.
But what of fictions where it is not simply left open whether a and b are
identical but where it is part of the fiction that it is indeterminate that they
are identical: i.e. where ‘a and b are indeterminately identical’ is true
according to the fiction, as opposed to there simply being no facts about
the identity or distinctness of a and b that are true according to the fiction?
The fictional realist has a couple of options here. One option would be to
restrict one’s fictional realism to characters inhabiting possible fictions. If
188 ROSS P. CAMERON
Evans’ argument against indeterminate identity is sound (and if it isn’t then

there’s no problem to worry about in the first place) then a fiction that
settles it to be the case in the fictional world that there are entities a and b
such that it is indeterminate whether a is b is an impossible fiction, and it is
not ad hoc to claim that in that case there can ultimately be no coherent
practice of talking about this fiction and, hence, that the conditions are
not met for it to be the case that there are the fictional entities a and
b. I don’t have much to say against this, other than to say that fictions
with indeterminate identities don’t seem to be utterly incoherent, and so it
would be better to give an account that could handle them in the same way
as a bog-standard fiction if possible.
Another option is that the fictional realist could insist that even when the
fiction settles the indeterminacy of the identity of a and b, reality settles the
issue of their identity or distinctness one way or the other, so that while it is
true according to the fiction that ‘a=b’ is neither determinately true
nor determinately false, in reality ‘the fictional character a=the fictional
character b’ is either determinately true or determinately false. Which?
Well, one could think that reality always settles it in a principled manner.
Schnieder and von Solodkoff, for example, suggest that even in this case reality
settles the distinctness of the fictional characters. But I am with Ben Caplan
and Cathleen Muller (ms) in not finding this very satisfying: Why should
reality settle things that way rather than the other? In the case where the fiction
simply leaves open the identity or distinctness of a and b it seems reasonable to
think that reality settles the distinctness of the fictional characters: using
different names creates a presumption that they do not co-refer—one that is
easily defeated, of course, but presumably (since it leaves the issue open, ex
hypothesi) the fiction does not defeat it, and hence it is plausible that in that case
there really are two fictional characters. But to claim that distinctness is
the default even when the fiction positively establishes the indeterminacy of
the issue seems wrong: here the indeterminacy is a part of the fiction itself, and
to hold to a presumption of distinctness in reality for these cases is no longer to
simply have reality settle what was left open in a principled manner, it is for
reality to overrule the fiction; and without a clear explanation for why reality
should overrule in one direction rather than the other, this is unsatisfying.
A better option along these lines, I think, would be to say that reality
settles one way or another the identity or distinctness of the fictional
characters, but that it is a brute fact that it gets settled the way it does, and
that it is in principle unknowable which way it gets settled. This builds on a

general proposal concerning indeterminacy defended in Kearns and Magi-
dor (2008).4 Consider an ordinary sorites series: a man with zero hairs is
bald, a man with one million hairs is not. To avoid absurdity, we seem
forced into accepting that there is a maximum number of hairs one can have
and count as bald: the number of hairs such that you are bald if you have that
many hairs but would not be bald if you were to have just one more. But
accepting this is hard to swallow, for what makes it the case that this be the
magic number, given that our usage of ‘bald’ doesn’t seem so fine-grained as
to pick out some particular upper-limit of hairs one can have whilst
satisfying the predicate? Kearns and Magidor answer: nothing makes that
the case, it is a brute semantic fact that ‘bald’ picks out the property that has
that as the upper-limit rather than some other close-by candidate property
that would have a slightly different upper-limit. But it is in principle
unknowable where the upper-limit of baldness lies, since we have no access
to these brute semantic facts. Similarly, one could hold that when, in the
fiction, it is indeterminate whether a is identical to b, it is a brute but
undetectable fact in reality that the fictional character a and the fictional
character b are identical/distinct: that reality settles it one way or the other,
but in no principled manner (and so it’s simply a mistake to ask why it was
settled that way), and that we cannot know how it was settled. I think this is
the best option for the fictional realist who wants a ‘straight’ solution
to Everett’s problem. But I want to offer an option for the fictional realist
who wants to take the indeterminacy in identity seriously.
The most intuitively satisfying option, I think, is to simply grant that
when it is true according to the fiction that it is indeterminate whether a is
b, it is true in reality that it is indeterminate whether the fictional character
a is the fictional character b, and to show how one can grant this without
absurdity in light of Evans’ argument. It’s widely recognized (since
Lewis (1988)) that the Evans argument is not intended to establish that
there cannot be identity statements that are indeterminate in truth-value,
only that there cannot be indeterminate identities (just as the analogous
Barcan/Kripke argument for the necessity of identity is not intended
to show that there cannot be contingently true identity claims, only that
4 Cf. Breckenridge and Magidor (forthcoming) and Cameron (2010b). This view is very close (to the
point, perhaps, of identity) to the view that Caplan and Muller (ms) recommend to the fictional realist.
190 ROSS P. CAMERON
there cannot be contingent identities). Suppose I introduce the descriptive

name ‘Bob’, whose reference is to be settled as picking out the bald man
nearest me. If the world fails to co-operate, it will be indeterminate what
‘Bob’ refers to. Suppose Jim, who has hair enough to make him
not determinately bald but not hair enough to make him determinately
not bald, is the nearest person to me, and the second nearest person to me is
Jean-Luc, who is most determinately bald. In that case, both ‘Bob = Jim’
and ‘Bob = Jean-Luc’ are indeterminate identity statements; but there is
simply no threat from the Evans argument here, for there is no thing such
that that thing is indeterminately identical to Bob.
The fact that identity is necessary is compatible with the contingency
of the identity statement ‘Barack Obama is identical to the President’,
precisely because we can’t conclude from this that there is some thing, the
President, that is contingently identical to Obama. The contingency of this
statement arises not from the contingent identity of the referents of the
terms flanking the identity symbol but rather from the fact that one of
the terms is modally rigid whilst the other is modally non-rigid. Similarly,
‘Bob’ above is precisificationally non-rigid: on some ways of precisifying its
reference it refers to Jim whilst on others it refers to Jean-Luc. ‘Jim’ and
‘Jean-Luc’, on the other hand, are precisifactionally rigid (let’s assume): it is
perfectly determinate who those terms refer to. And this is the sole source of
the indeterminacy of the identity statements ‘Bob = Jim’ and ‘Bob = Jean-Luc’.
We can only conclude that there are some things a and b which are
indeterminately identical—and so only face trouble from Evans’ argument—
when we have two precisificationally rigid designators flanking the identity
symbol in an indeterminately true identity statement.
And so if the fictional realist wishes to grant the soundness of Evans’
argument5 and yet take seriously the thought that indeterminate identity in
the fiction gives rise to indeterminate identity of fictional characters
in reality, she should hold that at least one of the names of a fictional
character is precisificationally non-rigid: i.e. that it is not determinate
what its referent is. So what’s needed is an account whereby when it is
true according to the fiction that it’s not determinate whether a is b, there
are, in reality, multiple fictional characters such that it is indeter-
minate which one is the referent of the term ‘the fictional character a’
5 And of course it’s open for her to simply resist this. See e.g. Barnes (2009).
(and/or ‘the fictional character b’). So if we suppose that Frackworld really

is a fiction in which it is indeterminate whether Frick is Frack, we want an
account that will yield that there are in reality multiple fictional characters
that are candidates to be the referent of ‘Frick’, such that on some ways
of precisifying the reference the referent is Frack, but on other ways
of precisifying the referent is something other than Frack. And so ‘Frick =
Frack’ is, in reality, indeterminate: but that is not because there is some thing
that is indeterminately identical to Frick, it is because it is indeterminate what
‘Frick’ refers to. We only have the indeterminacy of an identity statement,
not the indeterminacy de re of an identity, and so we are in no conflict with
Evans’ argument.
The idea that there are sometimes multiple candidate referents for
terms purporting to refer to fictional characters helps with another puzzle,
introduced and discussed in McGonigal (ms). In 1977, Johnny is watching
Star Wars: A New Hope, and utters the sentence ‘Vader is Luke’s father’. In
1980, after the release of The Empire Strikes Back, Jimmy has just finished
watching A New Hope, and utters the same sentence. Intuitively, there’s
something better about Jimmy’s utterance over Johnny’s (whether or not
Jimmy has actually seen The Empire Strikes Back). And it’s not just that Jimmy
was in a better epistemic position for making that claim than Johnny;
intuitively, Johnny is saying something that is simply not validated by
the fiction, whilst Jimmy is correctly reporting on the fiction. How can
we account for this?
McGonigal suggests that utterance truth is relative to a context of
assessment and that Johnny’s utterance is false relative to the context
of assessment that is his time of utterance but true relative to the
context of assessment that is Jimmy’s time of utterance, thus explaining
why they seem to ‘say the same thing’, and yet one utterance be defective
and the other not. But I am of a temperament to always prefer a contextual-
ist solution to a relativist one. I like the solution that says that there are
two fictions, one which has A New Hope but not The Empire Strikes
Back as a part, and one which has both as parts: call these Star Wars and
Star Wars+ respectively. And likewise, there are two fictional characters that
are candidates to be referents of ‘Luke Skywalker’ and ‘Darth Vader’. There
are Luke Skywalker and Darth Vader, who exist according to the Star
Wars fiction, and Luke Skywalker+ and Darth Vader+, who exist according to
192 ROSS P. CAMERON
the Star Wars+ fiction.6 Luke Skywalker+ is, according to the Star Wars+
fiction, the son of Darth Vader+. That is true, simpliciter. It is false, simpliciter,
that Luke Skywalker is, according to the Star Wars fiction, the son of
Darth Vader; it is also false, simpliciter, that he is not: the Star Wars fiction
leaves it open if there is any familial relationship between those characters,
in exactly the same way it leaves open what Darth Vader has for breakfast
according to the fiction. Both fictions and both sets of fictional characters
exist. Utterances of the terms ‘Star Wars’, ‘Darth Vader’ etc. get their
referents based on the context of utterance. An utterance of ‘Star Wars’ in
1977 plausibly refers to the fiction Star Wars, whereas an utterance of the
same term after 1980 plausibly refers to Star Wars+. And so Johnny’s
utterance is false, simpliciter, since it expresses the false simpliciter proposition
that, according to the fiction Star Wars, Luke Skywalker is the son of
Darth Vader, whereas Jimmy’s utterance is true, simpliciter, since it
expresses the true simpliciter proposition that, according to the fiction Star
Wars+, Luke Skywalker+ is the son of Darth Vader+.
So to return to the issue: we want an account of the truth maker for claims
concerning fictional entities that will cohere with this picture of there being
multiple candidate fictional entities to be the referents of terms designating
fictional entities. My suggestion is that the truth maker is the interpretative
act. When engaging with a fiction, we are mandated to interpret it: this act is
what makes true our true utterances concerning the existence of that fiction
and the existence of, and properties of, the fictional characters that exist
according to that fiction.
And so the ultimate metaphysic is nominalist. These acts are concrete
events; at least, they are concrete events if, as I assume, mental events in
general are concrete events. It is true that there are fictions, and fictional
characters, and such things are abstract (since if you list all the things you can
kick, they are not on that list); but our ontology is not what there is, it’s
what there is that ultimately makes true all that is true, and so this ontology
6 I assume that fictional characters are individuated in part by the fictions to which they belong: no
fictional character belongs to more than one fiction. The present proposal makes it easier to believe this
convenient claim. On some views there would be difficult questions about the identity of fictional
characters across fictions: is the fictional character Frasier Crane who appears in Cheers the same as the
character who appears in Frasier, for example? On my view, there are no difficult issues here: there are
simply three fictions—Cheers, Frasier, and Cheers+Frasier, each with their own Frasier Crane. Which
fictional character we pick out with our term ‘Frasier Crane’ depends on context, and might sometimes
be indeterminate.
is nominalistic. Nominalist scruples should not be upset by the mere

existence of abstracta, provided their being is grounded in an ontology
of concreta.
So here’s the picture. I have in front of me a bound bunch of printed
pages, the first of which has the string of letters ‘The Hobbit, by J.R.R.
Tolkien’ on it. I engage with this entity by interpreting the strings of letters
printed throughout it in a certain way. My act of interpretation makes it true
that there is a work of fiction; that I interpret the fiction as involving
the character Bilbo Baggins makes it true that Bilbo exists. No more is
needed of the world to ensure that the fictional character Bilbo exists.
Our act of interpretation is what makes it true that the fictional character
is such-and-such a way according to the fiction. In one sense, then,
there can be no misinterpretation: provided at least that I have coherently
interpreted my copy of The Hobbit, what I say about how a fictional
character is according to that fiction goes, since what makes it true that it
is that way according to the fiction precisely is that I have so interpreted it.
But this is no problem, for the sense in which misinterpretation is intuitively
possible is something I can allow for. For when I engage with a work, I do
not do so in a vacuum: I do so with the intention of engaging with a literary
community, including the author; and when I use terms like ‘The Hobbit’
or ‘Bilbo Baggins’ I do so with the intention of referring to the fiction, and
the character belong to that fiction, that was created by J.R.R. Tolkien
and is discussed by critics and fans, etc. If I—as we would intuitively
describe it—misinterpret The Hobbit by, e.g. interpreting Gollum to be
Gandalf in disguise, then while it is true on my account that there thereby
exists a fiction and a fictional character who is, according to that fiction, a
wizard who pretends to be Gollum-like, there is no reason to take my
utterances of ‘The Hobbit’ or ‘Gandalf ’ to refer to that fiction or that
fictional being. I could use the terms so, but if I am not being obtuse then
I am trying to engage with my linguistic community and use the terms to
refer to the fiction and fictional character created by Tolkien: the fictional
entities whose existence is grounded in his act of interpretation.7 Hence my
7 I don’t mean to suggest that the author’s own interpretation is always the important one. It may
sometimes be that an interpretation other than the author’s has come to have dominance, and that it is
the fiction that is created by that act of interpretation that we refer to by using the name the author
originally attached to a different fiction. Part of the attraction of the present proposal, I think, is its
flexibility: all these different fictions exist, and which we end up referring to simply depends on our
194 ROSS P. CAMERON
utterance of ‘In The Hobbit, Gollum is Gandalf ’ is not true, and so we are
correct in describing my act of interpretation as a misinterpretation.
Let’s see how the account plays out in the two puzzle cases above. If
you’re co-operating with your linguistic community, you intend to use the
term ‘Star Wars’ to refer to the canonical fiction created by George Lucas.
But what is canon changes over time for a fiction like this: films (regrettably)
get added, books and comics get added—and sometimes it might even be a
vague matter whether something is canon (and hence indeterminate which
of many fictions ‘Star Wars’ refers to). When A New Hope was just released
and not yet a success (and so the existence of a sequel not ensured),
‘Star Wars’ referred to the fiction whose being is grounded in Lucas’s
interpretation of that one film; but later, the fictional universe expands,
and as it does the reference of ‘Star Wars’ shifts to encompass the bigger
fiction. That is why Johnny says something false and Jimmy says something
true despite uttering the same string of words: Johnny and Jimmy simply
refer to different fictions, since the fictional universe was expanded between
the times of their utterances. There is, and always will be, a fiction that does
not settle issues of Luke’s parentage—hence there is a fictional character that
is not, according to that fiction, Vader’s son; but as a result of Lucas’s
history and our own linguistic intentions, these are not the referents
of ‘Star Wars’ or ‘Luke Skywalker’.
What of the indeterminate identity case? What I want to say is that when
our interpretation suggests an indeterminacy in the identity of fictional
characters, both interpretations are (as a matter of contingent psychological
fact) forced upon us: we interpret things both ways, and hence there is
a fiction that has it one way and a fiction that has it the other, and there
is nothing that singles out one fiction over the other as being a better
candidate referent, and hence our terms purporting to refer to ‘the’ fiction
and the fictional characters belonging to that fiction are indeterminate
in reference, and so the puzzle is solved as suggested above. Now, that
literary practices. J.K. Rowling can intend Dumbledore to be gay all she wants, but that interpretation
really doesn’t seem mandated by the words on the pages of the Harry Potter books, and so it’s not
implausible to think that it’s simply not true that Dumbledore is gay according to Harry Potter, despite the
author’s intent.
In addition, there may be various normative facts about how I ought to interpret a fiction, or about
how I ought to interpret it given that I am aiming to engage with it as a work belonging to a certain genre,
etc.; and so insofar as I fail to meet these norms I can be said to be misinterpreting the work. I believe that
there are many such norms, although the current proposal is simply neutral on that.
might not always happen: perhaps in some cases there will be principled
reasons for one fiction being the referent over the other—and in such cases
we should be happy with the verdict that the characters are identical/
distinct, as determined by that fiction; but in the absence of such reasons,
we will be left with (an unproblematic) indeterminacy in the statement
expressing the identity of those characters.
There’s obviously much more that can be said about this proposal, but it
will have to wait. I hope to have done enough here to make plausible two
claims: (1) that we can be fictional realists, in the sense of taking at face value
claims concerning the existence of fictional characters, without abandoning
a nominalist ontology, provided we are happy to adopt a meta-ontology
whereby what matters is not what exists but what has to be invoked as truth
makers; and (2) that appealing to acts of interpretation as the truth makers for
claims concerning fictional beings not only allows us to uphold a nominalist
ontology but also provides an independently attractive solution to certain
puzzles concerning fictions and fictional characters.
References
Barnes, Elizabeth (2009) ‘Indeterminacy, Identity, and Counterparts: Evans
Reconsidered’ Synthese 168: 81–96
Breckenridge, Wylie and Magidor, Ofra (forthcoming) ‘Arbitrary Reference’
Philosophical Studies
Cameron, Ross (2008a) ‘There are No Things That are Musical Works’ British
——(2008b) ‘Truthmakers and Ontological Commitment: Or, How to Deal with
Complex Objects and Mathematical Ontology Without Getting into Trouble’
Philosophical Studies 140: 1–18
——(2010a) ‘How to Have a Radically Minimal Ontology’ Philosophical Studies
151: 249–64
——(2010b) ‘Vagueness and Naturalness’ Erkenntnis 72: 281–93
——(2010c) ‘Quantification, Naturalness and Ontology’ in Allan Hazlett (ed) New
Waves in Metaphysics (New York: Palgrave-Macmillan)
——(forthcoming) ‘Truthmakers’ in Michael Glanzberg (ed) The Oxford Handbook
of Truth (Oxford: Oxford University Press)
Caplan, Ben and Matheson, C. (2006) ‘Defending Musical Perdurantism’ British
196 ROSS P. CAMERON
Caplan, Ben and Muller, Cathleen (ms) ‘Brutal Counting’

Dodd, Julian (2007) Works of Music: An Essay in Ontology (Oxford: Oxford Uni-
versity Press)
Evans, Gareth (1978) ‘Can There Be Vague Objects?’ Analysis 38: 208
Everett, Anthony (2005) ‘Against Fictional Realism’ The Journal of Philosophy 102:
624–49
Fine, Kit (2009) ‘The Question of Ontology’ in David Chalmers, David Manley,
and Ryan Wasserman (eds) Metametaphysics (Oxford: Oxford University Press)
Kearns, Stephan and Magidor, Ofra (2008) ‘Epistemicism about Vagueness and
Meta-Linguistic Safety’ Philosophical Perspectives 22: 277–304
Lewis, David (1988) ‘Vague Identity: Evans Misunderstood’ Analysis 48: 128–30
Markosian, Ned (1998) ‘Brutal Composition’ Philosophical Studies 92: 211–49
McGonigal, Andrew (ms) ‘Truth, Relativism and Serial Fiction’, presented at the
2007 meeting of the Pacific APA
Merricks, Trenton (2001) Objects and Persons (Oxford: Clarendon Press)
Schaffer, Jonathan (2009) ‘On What Grounds What’ in David Chalmers,
David Manley, and Ryan Wasserman (eds) Metametaphysics (Oxford: Oxford
University Press)
Schnieder, Benjamin and von Solodkoff, Tatjana (2009) ‘In Defence of Fictional
Realism’ The Philosophical Quarterly 59: 138–49
Sider, Theodore (2001) Four-Dimensionalism: An Ontology of Persistence and Time
(Oxford: Oxford University Press)
Thomasson, Amie L. (2003) ‘Fictional Characters and Literary Practices’ The British
——(2009) ‘Answerable and Unanswerable Questions’ in David Chalmers,
David Manley, and Ryan Wasserman (eds) Metametaphysics (Oxford: Oxford
University Press)
Turner, Jason (2011) ‘Ontological Nihilism’ Oxford Studies in Metaphysics 6: 3–54
van Inwagen, Peter (1990) Material Beings (Ithaca, NY: Cornell University Press)
Williams, J.R.G. (2010) ‘Fundamental and Derivative Truths’ Mind 119: 103–41
——(ms) ‘Requirements on Reality’
9
Platonism vs. Nominalism in
Contemporary Musical Ontology
ANDREW KANIA *
I. Introduction
Ontological theories of musical works fall into two broad classes, according
to whether or not they take musical works to be abstract objects of some
sort. I shall use the terms ‘Platonism’ and ‘nominalism’ to refer to these two
kinds of theory.1 In this chapter I first outline contemporary Platonism
about musical works—the theory that musical works are abstract objects.
I then consider reasons to be suspicious of such a view, motivating a
consideration of nominalist theories of musical works. I argue for two
conclusions: first, that there are no compelling reasons to be a nominalist
about musical works in particular, i.e. that nominalism about musical works
rests on arguments for thoroughgoing nominalism; and, second, that if
Platonism fails, fictionalism about musical works is to be preferred
to other nominalist ontologies of musical works. If you think in terms
of realism vs. anti-realism about musical works, then one way of putting
this is to say that realism about musical works stands or falls with Platonism
about musical works.2 That’s because, for methodological reasons I discuss
below, a theory according to which musical works are concrete objects of
some sort is not a realist theory of musical works, properly understood. This
* Thanks to Curtis Brown and Christy Mag Uidhir for helpful discussion of the issues addressed in
this chapter, and to Trinity University for financial support.
1 Nominalists about musical works may be Platonists about other entities, such as numbers (and, in
principle, vice versa). When I discuss a view according to which there are no abstract entities at all, I call
it ‘thoroughgoing nominalism’.
2 Thanks to Christy Mag Uidhir for bringing this perspective to my attention.
198 ANDREW KANIA
chapter is thus a contribution to the debate over the fundamental ontology

of works of Western classical music, broadly construed, though its conclu-
sions could be applied to other musical (or artistic or cultural) practices
that are sufficiently similar, if such there be.3
II. Contemporary Platonism about Musical Works

The basic questions in the fundamental ontology of musical works are the
same as those of any topic in ontology: (1) ‘Are there any?’ And (2) ‘If so, what
kinds of things are they?’ These questions cannot be approached separately for
musical works any more than they can for numbers, ordinary objects, persons,
possible worlds, and so on. On the one hand, the nature of the thing in
question may provide strong reasons for thinking there are not any such
things. For instance, one might argue that the concept of a soul is essentially
incoherent, and thus that there can be no souls. On the other hand, one might
take there to be such compelling reasons for thinking a particular kind of thing
exists that one posits it despite its odd nature, or the problems it creates in
other areas of inquiry. For instance, one might argue that numbers (conceived
of as abstracta) are indispensible to our best theories of the world, and thus that
they must exist, even though it is difficult to understand how there could be
such things, or how we could know anything about them.
A couple of basic features of musical works might lead one to think that
their ontology would be no different from that of something like properties:
(i) musical works are ‘multiple’ or ‘repeatable’; they have ‘instances’ (per-
formances), none of which can intuitively be identified with the work. Yet
(ii) we have ‘access’ to or come into ‘contact’ with the work ‘through’ or
‘in’ any one of these instances. In fact, as I have argued elsewhere, the early
history of musical ontology can be read as a kind of applied debate over the
problem of universals (Kania 2008a, pp. 426–7).4 More recently, Julian
3 For the distinction between ‘fundamental’ and ‘higher-order’ musical ontology, see Kania 2008b.
There is some debate over how broadly we should construe ‘Western classical music’ if our aim is to
include musical practices centered around the performance of works. (Lydia Goehr 2007, especially chs
7–9) is well known for arguing for a narrow construal. For a defense of the more traditional broader
construal, see S. Davies 2001, pp. 86–91. I will not enter that debate here.
4 I have primarily in mind here the work of Nelson Goodman (1976 (first published 1968)); Richard
Wollheim (1980 (first published 1968)); and Nicholas Wolterstorff (1980).
PLATONISM VS. NOMINALISM 199
Dodd (2007) has argued, pretty much exclusively on the basis of these two
features, that musical works are eternal, unstructured, unchanging, modally
inflexible, abstract types.5
But musical works have further, equally basic, features that do not allow
their ontology (or debates about it) to be assimilated to that of properties
so easily. For instance, (iii) musical works are intentionally created by
composers; (iv) they are normative, both in the sense that they specify
how their performances should go, and in the sense that they admit of
better and worse performances; and (v) they possess aesthetic or artistic
properties that seem to depend on the cultural context of their composition.
To hold that these latter features, and others like them, are relevant to the
ontology of music is to subscribe to a methodological principle held by
many ontologists of art over the last thirty years, a principle now widely
known, thanks to the work of David Davies, as ‘the pragmatic constraint’
(2004, p. 18). The principle is so-called not because of any connection with
the philosophical theories of Pierce, James, and Dewey, but because it takes
artistic practices to be the yardstick against which ontologies of art should be
measured. As Davies puts it:
Artworks must be entities that can bear the sorts of properties rightly ascribed
to what are termed ‘works’ in our reflective critical and appreciative practice; that
are individuated in the way such ‘works’ are or would be individuated, and that
have the modal properties that are reasonably ascribed to ‘works’, in that practice.
(2004, p. 18)
The basic rationale of the principle is simple, and familiar from other areas
of metaphysics: we ought to believe that those things exist which are
required by our best theories of how things are. When we ask ontological
questions about numbers we rightly take our best mathematical theories to
be our most important evidence base; when we ask ontological questions
about music, we rightly take our best musical theories to be our most
important evidence base. It is worth noting that ‘musical theories’ here
does not just mean music theory, narrowly construed, or even musicology
in a traditional sense. It is, rather, our best understanding of this entire
cultural sphere, of everything that goes on in the production and reception
5 Dodd explicitly acknowledges his debt to Wolterstorff (Dodd 2007, p. 100).

200 ANDREW KANIA
of music, that is our evidence base for ontological claims about music. This
is because, to quote Davies once more:
[O]ur philosophical interest in ‘art’ and in ‘artworks’ is grounded precisely in
[artistic] practice. It is because certain features of that practice puzzle us, or because
the entities that enter into that practice fascinate us, that we are driven to
philosophical reflection about art in the first place. To offer an ‘ontology of art’
not subject to the pragmatic constraint would be to change the subject, rather than
answer the questions that motivate philosophical aesthetics. (2004, p. 21)
Davies has been criticized for making unjustified exclusions from this
evidence base. I, for one, have argued that Davies is led astray by not taking
seriously enough the ontological implications of our artistic practices,
including our ontological intuitions (Kania 2008a, pp. 429–32). In Art as
Performance, his ontological magnum opus, Davies claims that:
. . . in reflecting upon our artistic practice in this way, the intuitions that
are strongest will be those that relate to practical aspects of that practice . . . —
judgments made, ways in which entities are treated, etc.—rather than intuitions
about what works are, ontologically speaking. (2004, p. 22)
But either I misunderstood Davies or he has taken this criticism to heart, for
in a recent discussion of the pragmatic constraint, he says that he does not
‘deny that there are ontological dimensions to some aspects of our practice’,
though he notes that ‘these judgments, like other features of our artistic
practice, can constrain ontological theorizing only when subject to rational
reflection’ (2009, p. 163).6
Robert Stecker has recently argued for a further broadening of the
evidence base for musical ontology:
Of course we should look at our musical practices and linguistic usage, . . . but that
should only be a starting point. There are many sciences that study music, including
musicology, music theory, psychology, and anthropology. Why shouldn’t these
studies generate data that are just as valuable for the philosopher? (2009, p. 383)
I am sympathetic to this approach in general. For instance, I myself have

suggested that those interested in a definition of music would do well
to consider recent work in the psychology of music (Kania 2011a), and
6 Davies does not comment in this essay on whether this acknowledgement would impact his
arguments for his own ontology of the arts.
philosophers discussing musical understanding frequently make reference

to the work of music theorists (e.g. Huovinen 2011). But I am not sure how
much of a departure this implies from current best practices in the ontology of
art, including musical ontology. On the one hand, it is not clear how some
of these disciplines (psychology and music theory) could contribute to
musical ontology in particular.7 On the other, while musicology and an-
thropology seem more promising in this regard, precisely because they aim
to describe musical practices, it seems to me that (good) musical ontologists
already appeal to such evidence. After all, when Davies appeals to the ways
in which people talk about certain artworks (he is concerned with art in
general, not just music), he appeals to what critics and art historians say.
Such evidence seems to be the equivalent of musicological and anthropo-
logical data in this context. Similarly, when Theodore Gracyk (1996) argues
for the work of art in rock music’s being the recording, rather than the song
or live performance, he appeals to rock criticism and musicology. We might
sometimes hope for better musicology and anthropology—more systematic,
objective, and wide ranging—but in the meantime we must make do with
what we have.
The pragmatic constraint is touted as a methodological principle used
to arrive at the best ontological theory of artworks. But as such it can also
be used critically, to reject theories that do not respect the principle. One
such theory that I have already mentioned is Julian Dodd’s ontology of
musical works. Recall that Dodd argues, on the basis of (i) their repeatability;
and (ii) the fact that they can be heard in performances, that musical works
are eternal, unstructured, unchanging, modally inflexible, abstract types
(Dodd 2007). Dodd thus violates the pragmatic constraint in two related
ways. Most obviously, he ignores vast tracts of musical practice (e.g. taking
composition to be work creation and the attribution of aesthetic and artistic
properties to works) until after he has established his preferred ontological
theory, at which point he explains away such data, either by rejecting
it outright (e.g. composition as creation) or by supplying paraphrases of the
relevant judgements (e.g. as attributing aesthetic properties not to works but
to performances). The second violation of the pragmatic constraint is
more subtle, and occurs as Dodd is establishing his view. Dodd argues very
7 Stecker admits that his new suggestion’s ‘potential for providing better data for an ontology of music
is as yet unknown’ (2009, p. 384).
202 ANDREW KANIA
early on that his view is a (in fact the) ‘simple’ one, the ‘default’ view, given
the repeatability of works and their audibility in performances (2007, p. 1
et passim). But judgements of simplicity (to say nothing of default) are always
relative and contextual. Two prominent aspects of the way in which Dodd
motivates his view show that its simplicity depends on taking what
I have called the ‘metaphysical constraint’ at least as seriously as the
pragmatic constraint (Kania 2008a, pp. 434–8). According to the metaphysical
constraint, our ontological theories of art, as far as possible, ought to appeal
only to entities posited by our best general metaphysical theories.8 The fact
that in motivating his view Dodd considers only the two features of musical
works that make them seem most like simple properties is one sign that he
implicitly endorses the metaphysical constraint. If musical works must belong
to a well-established ontological category, then one promising approach is to
ask: Of those things investigated by general metaphysicians, which are most
like musical works? And ‘properties’ seems a plausible answer to this question.
The other, more explicit sign that Dodd subscribes to the metaphysical
constraint is the way in which he talks about types when he proposes them
as the best candidate for the ontological category to which musical works
belong. He claims that upon recognition that musical works are ‘generic
entities’, that is, repeatable:
[w]e are . . . invited to treat [them] as types because . . . we thereby provide a familiar
and plausible explanation of the nature of the relation holding between a work and
its occurrences . . . Rather than being a queer relation of embodiment, it turns out to
be just one more example of the familiar relation that holds, for instance, between
the word ‘table’ and its token inscriptions and utterances. (Dodd 2007, p. 11)
One response to Dodd’s approach, then, is simply to reiterate the ‘primacy

of practice’, the trumping of the metaphysical constraint by the pragmatic
constraint. So David Davies argues that ‘[s]omething that only admitted of
the sort of appreciation and evaluation permitted by [Dodd’s theory] would
not be a work of art in the sense that interests us as philosophers’ (2009,
p. 163). Here are two more responses.
8 This characterization of the constraint is somewhat rough and ready, in part because those who
subscribe to it rarely do so explicitly. For an attempt at working out more explicitly the proper
relationship between the metaphysics of art and general metaphysics, from a perspective sympathetic
to the metaphysical constraint, see Mag Uidhir (this volume, Introduction).
Note that however one formulates the metaphysical constraint it must

appeal to something like our best general metaphysical theories. One
problem with this is that it is not clear what our best general metaphysical
theories are. The fact that contemporary guidebooks to metaphysics still
take the problem of universals to be a central issue in metaphysics suggests
that even if contemporary metaphysics is not mere footnotes to Plato, large
chunks of it might still be considered appendices to Plato and Aristotle.9
For instance, Dodd takes the existence of abstract types to be relatively
uncontroversial, while nominalists about musical works tend to start from
the premise that we ought not appeal to abstracta if we can avoid it at all
(a point I return to below).10
Another problem is with the very idea of a general metaphysical theory,
as opposed to a specialized theory such as an ontological theory of musical
works. There are two ways one could conceive of this opposition,
inclusively and exclusively, but both cause problems for the metaphysical
constraint. Considered inclusively, our best general metaphysical theory is a
metaphysical theory of everything, including, for instance, musical works
(if such there be). But clearly there cannot be a consensus on such a theory
without a consensus on the ontology of musical works, since the
latter is part of the former. Considered exclusively, one need not wait
for a consensus on the ontology of musical works before achieving
consensus concerning the best general metaphysical theory, because the
latter excludes musical ontology. On this conception our best ‘general’
metaphysical theory is our best basic metaphysical theory—a theory of
individuals, properties, modality, and causation, say. One problem for
this conception would clearly be demarcating what is metaphysically basic
in a non-question-begging way. Another, related problem is the mirror
image of the problem with the inclusive conception: it is plausible that
moving on from these ‘basics’ to more complicated things such as musical
works could introduce considerations that will lead us to add to or alter the
ontology required to cover the basics. An interesting application of this
point can be found in Zoltan Gendler Szabó’s introduction to nominalism
9 e.g. Loux and Zimmerman (2003) and Le Poidevin, Simons, McGonigal, and Cameron (2009). For
the best consideration of Whitehead’s famous aphorism that I am aware of, see Lachs (1995).
10 Dodd spends five pages early in his book dismissing nominalism about musical works, employing
standard moves in the debate over universals. For responses to these moves, see Caplan and Matheson
(2006, 2008) and Tillman (2011).
204 ANDREW KANIA
(Szabó 2003). Discussing the problem of the causal isolation of abstracta, he

notes that ‘[t]his sort of argument is applied all the time across the board
against all sorts of abstracta, but the fact that it was originally presented in the
context of the philosophy of mathematics is of utmost importance’ (p. 29).
He goes on to give the example of Jaroslav Hašek’s novel The Good Soldier
Šweik, which is ‘presumably an abstract entity, but one that is
causally dependent on a host of concrete ones’ such as its author and the
Austro-Hungarian monarchy’s involvement in the First World War, and
upon which many other concreta depend, such as Szabó’s use of it as an
example (pp. 29–30). Szabó’s point is precisely that if we ignore things more
complicated than the metaphysical basics, we run the risk of oversimplifying
our metaphysics.
What kind of ontological theory of musical works do we end up
with, then, if we forget the metaphysical constraint and apply the pragmatic
constraint? We get a non-standard Platonism, that is, a theory according to
which musical works are abstract objects, but with features not traditionally
attributed to abstracta. One locus classicus here is Jerrold Levinson’s ‘What a
Musical Work Is’ (1990a [first published 1980]), though Levinson has
modified his view over time.11 Levinson argues that, as features (i) and
(ii) mentioned above suggest, musical works are abstract objects, something
like abstract, multiply instantiable structures of sounds. But, as features (iii)
and (v) suggest, they are not simply sound structures, for two different
musical works could share a sound structure, and yet differ in their aesthetic
properties as a consequence of who composed them, and when. One might
be a simple, naı̈ve piece, for instance, while the other is simple in the service
of a kind of primitivism or biting irony. Those sympathetic to Levinson’s
approach have suggested modifications of the view in light of other
features of musical works. In particular, the normativity and modal flexibil-
ity of musical works have been discussed. The upshot is that the consensus
among those who subscribe to the pragmatic constraint and reject the
metaphysical constraint is that the best ontological theory of musical
works is that they are something like structures of performed sounds made
11 See, in particular, Levinson (1990b, 1996, and this volume Ch. 2).
normative by the production of a score in a particular creative act.12 The details of

the theory are not important for our purposes. The question is: Are there in
fact any such things? Thoroughgoing nominalists will think not. I now
turn to motivations for such a view.
III. Motivations for Nominalism about

Musical Works
David Davies says that ‘[t]o offer an “ontology of art” not subject to the
pragmatic constraint would be to change the subject, rather than answer the
questions that motivate philosophical aesthetics’ (2004, p. 21). It’s not clear
why we couldn’t extend this principle to other cultural practices, such as
religion. It seems just as plausible to say that to offer an ‘ontology of religion’
not constrained by rational reflection on what religious practices imply
about the nature of God or witches, say, would be to change the subject,
rather than answer the questions that motivate philosophical theology.
(Davies himself draws an analogy with philosophy of science.) The
challenge this suggests to the ontologist of art is that though the pragmatic
constraint will deliver our best concept of a musical work, it will not
guarantee that anything falls under that concept.
One response to this challenge is to point out that though the pragmatic
constraint does not guarantee that anything falls under our concept of a
musical work, this is no reason to think that the concept is in fact empty.
Furthermore, the general ontological principle appealed to above—that we
should believe in the things implied by our best theories—suggests a
relevant difference between musical works and supernatural entities, namely
that our best theories of the world (including the cultural world) imply that
musical works exist, but that those same theories imply that there are no
witches or gods. One problem with supernatural entities, for instance,
unlike musical works, is that they (arguably) conflict with scientific theories.
12 This literature is too extensive to summarize here, but for a recent example see Matheson and
Caplan (2008). They consider challenges to Levinsonian views on the basis of the modal flexibility of
musical works, and end up defending the plausibility of something very close to Levinson’s own view
(though they do not endorse the view).
206 ANDREW KANIA
Nominalists, however, will claim that there is just such a conflict—

perhaps an even deeper one—between our best concept of musical works
and our best overall theories of the world: our best concept of musical
works implies they are abstract objects, and our best theories of the world
make no reference to abstract objects. The Platonist might try to reply to
this argument in the same way I suggested she should reply to a proponent
of the ‘metaphysical constraint’: she might claim that to exclude the evi-
dence for musical works from the evidence base for our best overall
ontological theory is to beg the question against the Platonist. But the
situation is different here. The nominalist is not (or should not be) appealing
to some pre-existing, settled metaphysical theory. He is arguing that, even
when we have taken the evidence of musical practice into account, a theory
without abstracta, and thus without anything like what we take musical
works to be, is preferable to one that posits abstracta. Presumably whatever
the details of the argument here, a major component will be an appeal to
something like Occam’s Razor. We would thus have to attempt to weigh
the ontological savings of rejecting abstracta against the costs in other aspects
of the theory, such as simplicity. Obviously at this level of generality there is
nothing I can say that should sway us one way or the other; to go further
with this debate we would have to turn to particular arguments for
or against thoroughgoing nominalism.
It is noteworthy that musical nominalists do not say much to motivate
their thoroughgoing nominalism. They usually briefly appeal to problems of
causal interaction with abstracta, particularly the creation of musical works,
and then move quickly on to considering nominalist proposals. (See, for
example, Caplan and Matheson 2006 and Cameron 2008.) There are several
responses the Platonist can give to the initial problem of causal interaction or
creatability. I have already discussed the first in connection with the meta-
physical constraint. Our conception of musical works could just as easily
(and perhaps less dogmatically) be taken as evidence that some abstracta are
capable of causal interaction, including creation, as that they cannot be
abstracta. In other words, the nominalist’s dialectic here seems to rely on the
(bankrupt) metaphysical constraint. Second, there are of course resources
available to the Platonist for giving a positive account of the nature of
musical works as abstract and creatable. For instance, Caplan
and Matheson (2004) suggest some promising strategies for defending a
conception of musical works as sets or types that are creatable, and Simon
Evnine has suggested that creation does not require causal interaction, in the
case of either concreta or abstracta (2009, pp. 214–15, esp. fn. 25).
What this suggests is that the nominalist’s motivation resides wholly
in quite general motivations for nominalism which are seldom, if ever,
engaged with. Perhaps it is too much to expect the nominalist about musical
works to provide arguments for nominalism in general. However, this does
mean that the nominalist’s case is built on a conditional: if there are no
abstracta, then musical works must be thus and so. This comes out pretty
explicitly in Chris Tillman’s consideration of various nominalist theories
of musical works: ‘If there is a presumption in favor of the material over
the abstract, and if the main motivation for musical materialism is that
materialism is untenable . . . , then musical abstractionism is unmotivated’
(2011, p. 28, emphasis removed).13 The nominalist might reply that the
Platonist’s case is similarly built on a conditional: if there are abstracta, then
musical works must be thus and so. However, the Platonist has the dialectical
advantage here, because, thanks to the pragmatic constraint, the ‘thus-and-so’
in the case of the Platonist is how we ordinarily conceive of musical works.
This means that (a) other things being equal, there’s a smaller cost to accepting
the antecedent of the Platonist’s conditional than the nominalist’s;14 and (b) it
gives us some (perhaps slight) reason to think there are abstracta. Nonetheless,
I doubt these brief reflections will do much to sway anyone already inclined
to thoroughgoing nominalism. I thus turn now to the ontological options
open to a thoroughgoing nominalist when it comes to musical works.
IV. Contemporary Nominalism about Musical Works

The broad sense in which I am using the term ‘nominalism’ encompasses a
variety of ontological theories of musical works. One group of nominalist
theories is the materialist theories, according to which a musical work is
some kind of concrete entity, such as a collection of performances or the
13 In the ellipsis, Tillman refers to the arguments typically marshalled against materialism, which
he finds wanting. He considers these arguments on pp. 20–8. The details do not affect the point I am
making here.
14 The ‘other things’ in this case are elements of the debate between thoroughgoing nominalists and
Platonists. Of course this debate may well not be equal, but part of my goal here is just to see how
ontologies of music relate to more general ontological theories.
208 ANDREW KANIA
particular creative action of a composer. Chris Tillman (2011) has recently

produced a useful menu of some options for the materialist about musical
works. Call whatever you think the concrete manifestations of musical
works are (scores, performances, recordings, etc.) the work’s atoms.
According to musical perdurantism, a musical work is the fusion of its
atoms, and those atoms are its temporal parts.15 According to musical
endurantism, a musical work is its atoms, but only one at a time, as it were;
it is wholly present in each atom, rather than being identified with the
fusion of its atoms. According to musical spannerism, a musical work
is coextensive with its atoms, but it is not identical to them, nor are they
its parts. (As Tillman says, ‘spanning is weird’ (2011, p. 19 fn. 27).)
Another group of nominalist theories is eliminativist theories, according
to which there are only concrete objects, such as performances and the
creative actions of composers, and none of these can be identified with
musical works; therefore there are no musical works. Few have defended
eliminativism about musical works. Richard Rudner (1950) is the closest we
have to a classic source, though it is possible to interpret his position as
materialist. He argues that the best candidate ontological category for
musical works is that of abstract object, since it does better than any
other candidate (individual performance, set of performances, composer’s
intention, etc.) at filling the role our musical practices carve out for it.
However, Rudner considers the peculiarities of the kind of abstractum
musical works would have to be, and the fact that they would have to be
created by composers, deal-breakers for Platonism about musical works. As
a result, he argues, we should stop speaking (at least strictly, as theorists)
of musical works, and talk instead only of performances, compositional
intentions, and so on.
Judging by the title of a recent article in the British Journal of Aesthetics,
‘There Are No Things that are Musical Works’, Ross P. Cameron (2008) is
also an eliminativist, though, like Rudner’s, his position is not easy to
pigeon hole. Cameron argues that we can have our musical works and
eliminate them too. That is, he thinks that when we ordinarily say things
like ‘there are many musical works’, we say something true, even though
15 Ben Caplan and Carl Matheson have defended musical perdurantism at some length (2004, 2006,
2008), primarily in dialogue with Julian Dodd (2002, 2004, 2007). You could even be forgiven for
thinking they subscribe to the view.
there aren’t really any musical works. The reason this isn’t a contradiction,
according to Cameron, is that the truth conditions of ordinary English
sentences such as ‘there are many musical works’ do not require that there
be any things which are musical works. On the other hand, when we
(truthfully) say things such as ‘there aren’t really any musical works’,
we are speaking ‘Ontologese’, the language of metaphysics or fundamental
reality. The truth conditions of sentences in Ontologese do require there to
be referents for terms like ‘musical work’. I have no space to discuss
Cameron’s view in depth, in part because it is a general ontological position,
out of which this theory of musical works falls.16 I do think it is unstable,
however, and threatens to collapse into Doddian Platonism, eliminativism,
or fictionalism, depending on which elements of the theory one holds most
firmly to.
According to both kinds of nominalist theory I have considered here—
materialism and eliminativism—there are no musical works of the kind
implied by our musical practices, since those practices imply that musical
works are abstract. The major difference between the two theories is that
the materialist sees the denial of the existence of musical works as a greater
theoretical cost than the eliminativist. Consider that, in some sense, the
materialist and eliminativist do not (or need not) disagree about the kinds of
things that exist.17 They both agree that there are no abstract objects (or,
at least, no abstract objects that are musical works). They do disagree about
whether there are any musical works, of course, but that is a disagreement
about whether musical works can plausibly be identified with some kind of
concrete entity, not about whether the kind of concrete entity in question
exists. Moreover, both these kinds of theory count it as a cost to deny the
existence of the kind of musical work implied by our musical practices. This
is evident in the case of the materialist by the use of the paraphrase strategy.
The materialist attempts to show that as many as possible of the claims we
make that at least appear to commit us to the existence of abstract musical
16 For some initial criticisms, see Stecker (2009, pp. 378–80) and Predelli (2009).
17 I supply the qualification because any given materialist and eliminativist may of course disagree
about the kind of things that exist. For instance, one may be a perdurantist and the other an endurantist,
in which case they would (arguably) disagree about the existence of temporal parts (on some construal of
that term). But this kind of disagreement is not relevant to the arguments I am currently considering. It is,
after all, a kind of disagreement two materialists could have. The only relevant disagreements here are
disagreements between the materialist and eliminativist qua materialist and eliminativist.
210 ANDREW KANIA
works can be paraphrased into claims that commit us only to concreta. This
saves us from some kind of error (e.g. the error of failing to refer to anything
when we attempt to refer to musical works), but attributes some other
error to us (e.g. not realizing what we are referring to).18 But it is also
evident in the case of the eliminativist, for the eliminativist does not
(or need not) deny the existence of the concreta with which the materialist
identifies musical works. Why, then, does the eliminativist not subscribe to
materialism? Presumably because the eliminativist thinks that it would
do less violence to musical practice to deny the existence of musical
works altogether than to identify them with the concreta the materialist
believes them to be. After all, eliminativism about musical works would
make no sense if the eliminativist did not believe both (1) that our concept
of a musical work is that of a certain kind of thing; and (2) that there are
no such things.
It is at this point that we see that the dispute between the materialist and
the eliminativist is doubly pragmatic: it is pragmatic in the sense that the
materialist and eliminativist agree about what kinds of things exist, but not
about whether to call one kind of thing a musical work. Choosing between
the theories depends on one’s purposes. The dispute is also pragmatic in
the sense that the pragmatic constraint appears to be implicit common
ground. The question the nominalist faces when choosing between
materialism and eliminativism is whether it would be better to give up
talk of musical works altogether or to transform it into talk of, say, fusions
of performances.19 And the measure of what is better here is clearly closeness
to, or coherence with, existing musical practices.
V. Fictionalism about Musical Works

In this final section I will suggest that a virtually ignored ontological theory
of musical works—fictionalism—should be preferred by a nominlist over
18 There may be disagreement among materialists about exactly which errors we are committing and
saved from.
19 It is the fact that this would be a transformation of musical discourse that leads me to say, as I did in
the introduction, that materialism about musical works is not realism about musical works.
both materialism and eliminativism by on the kind of pragmatic grounds just

considered.20 The basic idea of fictionalism is that, given claims in some
domain that appear to commit us to the existence of things that do not
in fact exist, we should conclude that these claims are fictional, rather
than assertoric.21 This is to be contrasted with the eliminativist strategy
of ceasing to make the given claims, and the materialist strategies of substi-
tuting other claims or reinterpreting the given claims as referring to
something that does exist.22
The basic motivation for fictionalism is that a realm of discourse may have
a value other than truth that justifies its continued use. For example, Hartry
Field (1980) argues that mathematics enables us to make inferences about
empirical matters more easily than we could without it. Taken literally,
argues Field, mathematical statements commit us to the existence of
numbers (conceived as abstracta), but we need not take them literally to
get what we want out of them. Thus, we should take them fictionally. Bas
van Fraassen (1980) is a fictionalist regarding scientific discourse about
unobservable entities. He thinks the point of such discourse is not truth
but rather empirical adequacy, that is, roughly, the ability to predict and
explain the observable.
These brief sketches are enough to distinguish two kinds of fictionalism:
hermeneutic (descriptive) and revolutionary (normative). Van Fraassen
purports to be giving an account of the nature of scientific discourse
about unobservables, a nature it has possessed since long before his theory
of it. He is thus a hermeneutic fictionalist—he offers an interpretation of
what has been going on in the discourse all along. Field, by contrast, argues
that mathematicians have actually been engaging in their discourse at face
20 Lydia Goehr is perhaps the best known fictionalist about music works (e.g. 2007, p. 106), though,
despite the title of the book in which she sets forth that theory, this aspect of it is not often commented
upon. Also, as I’ve just mentioned, Cameron’s arguments could be given a fictionalist spin, though he
would clearly rather you just left them alone.
21 I will not say much about what it is for a claim to be fictional, to adopt a fictional attitude towards a
proposition, and so on (likewise for assertion). For an introduction to these topics in connection with
fictionalism, and the literature on them, see Eklund (2007, Section 2) and Sainsbury (2010).
22 Matti Eklund claims that this makes fictionalism primarily a linguistic rather than an ontological
theory, albeit one that is usually motivated by ontological concerns (2007, Section 2.1). But the same
reasoning would suggest that materialism is primarily a linguistic thesis, when it is generally considered an
ontological theory. On the one hand, I think it would be misleading to think of fictionalism as a
linguistic thesis in contrast to the other kinds of theories I have been considering here. On the other
hand, I do think it is valuable to bring out the interconnection of linguistic and ontological matters, as
I have tried to do already in comparing materialism and eliminativism.
212 ANDREW KANIA
value, saying a lot of things that are false because they are about a realm of
entities that do not actually exist. He argues that enlightened mathemat-
icians ought to stop asserting such claims, and start making them fictionally
instead. He is thus a revolutionary fictionalist.23
What is the relevant discourse in the case of fictionalism about musical
works? It is musical discourse—precisely the discourse that gives rise
to questions about the ontology of musical works, and theories that purport
to answer those questions.24
What kind of fictionalist should you be about musical works?
I recommend revolutionary fictionalism. It seems implausible (to me) that
in ordinary musical discourse people are not committed to the existence of
musical works as distinct entities, that they are already speaking about them
fictionally. Application of the pragmatic constraint gives us our best theory
of the kind of thing people are referring to (or attempting to refer to) in such
discourse. If you don’t think there are any such things, and are tempted by
fictionalism, then you should think that people ought, when speaking
strictly, to adopt a fictional attitude towards them; that is, you should be a
revolutionary fictionalist.25
What is the value of musical discourse about works that justifies retaining
it despite its falsehood? It is the value of those musical practices that are
enmeshed with that discourse—practices (apparently) involving musical
works—whatever that value is. This raises a number of issues. The most
obvious is what the value of musical practices involving works is. I take it
that a large part of the answer to this question will be a general theory of the
value of music. I don’t have one to hand, but I take it as uncontroversial that
music is very valuable.26 One might, of course, argue that although music in
23 I am no great fan of these terms, since it’s not clear to me that revolutionary fictionalism is any more
disturbing to our usual way of thinking about what goes on in a domain of discourse than hermeneutic
fictionalism. But the terms are well established, so I will run with them. Note that materialists can also be
divided into hermeneutic and revolutionary camps, according to whether they claim the paraphrases
they provide for ontologically-committing claims supply what we actually mean by those claims, or what
we ought to mean by those claims.
24 As the pragmatic constraint suggests, I take non-linguistic behaviours to be relevant to our
interpretation of the discourse. I assume this is uncontroversial.
25 That said, whether you plump for hermeneutic or revolutionary fictionalism might depend on
other commitments you have in the philosophy of language. For further distinctions between varieties of
fictionalism, see Eklund (2007, Section 2).
26 For an introduction to theories of the value of music, see Goldman (2011); Gracyk (2011); and
S. Davies (2003).
general is valuable, musical practices involving the work concept are perni-
cious. One might even think that what is valuable about these practices
could be retained, and its perniciousness expunged, by reformulating
the practice (including its discourse) to eliminate the work concept.
I think this is implausible, but it is not too far from some views that have
actually been defended in musicology and philosophy of music.27 For
example, Lydia Goehr concludes that:
[i]n the end, musicians must just ask themselves whether the most satisfactory form
of musical criticism is one that is based on the ideal of Werktreue [faithfulness to
a work]. If it is not, they must seek an alternative. No musician is necessarily bound
to this ideal, however pervasive and persuasive the romantic aesthetic. (2007,
p. 279)
She explicitly leaves such questions open, but to do even this is clearly
far from a ringing endorsement of work-based musical practices. Lee B.
Brown (2011) also bemoans musical ontologists’ obsession with the work
concept, but he is more concerned that the obsession has led ontologists
to mischaracterize certain musical traditions, rather than that practices
involving the work concept are less valuable than they could be.
On the other hand, it is possible to construct an argument for precisely
the opposite conclusion: that musical practices involving works are more
valuable than those without the concept. The basic idea would be that
the works are enduring entities that thus admit of (i) being worked on over
time by their creators; and (ii) being appreciated on multiple occasions of
reception by their audiences. I doubt disagreement over these issues
will have much effect on musical practices, even in combination with
fictionalism. It seems unlikely we’ll reach a philosophical consensus about
the values of musical practices such that entire practices will be given up.
And, when it comes to fictionalism, it is practical matters—the value of
some discourse other than truth—that count. Anyway, if we did reach
a consensus about the values of music and this significantly affected
our musical practices, it would not make sense to bemoan the fact. If we
27 Of course, if we decided to get rid of the work concept altogether, due to its perniciousness, the
problem of the ontology of musical works would disappear, just as we are no longer concerned with the
nature of witches.
214 ANDREW KANIA
were to discover the true end of music, it would behoove us to strive to

achieve it.28
Finally, there is some disagreement about which musical practices really
involve the work concept. For instance, I have argued that much jazz is a
tradition without musical works (Kania 2011b). Goehr argues that the work
concept entered Western classical musical practice much later than most
philosophers suppose (Goehr 2007). But, if correct, these theories do not
affect fictionalism about musical works, which is only about
musical discourses that do, implicitly or explicitly, refer to musical works.
We might compare this with van Fraassen’s ‘constructive empiricism’:
van Fraassen is a fictionalist only about the unobservable entities posited by
scientific theories; he believes that when it comes to observable entities
science aims at the truth.
It is worth noting that the rooting of the musical-works fiction in musical
practice means that fictionalism about musical works is immune to
an objection raised against fictionalism about some other things, such as
possible worlds or moral values. R.M. Sainsbury, for instance, argues that in
both these latter cases the fictionalist faces a serious problem about how to
choose the fiction we ought to subscribe to (2010, pp. 190–2, 200–4). This is
because fictionally subscribing to a possible-worlds story, for instance,
is supposed to help us discover modal truths—what is really possible or
necessary.29 But without knowing such things, we will be at a loss to choose
between different possible-worlds stories (for instance, between a story
according to which one has at most one counterpart in any possible world
and a story according to which one has more than one counterpart in some
possible worlds). Since the value of musical discourse is not epistemic, it
does not face this objection.30 If we grant that our discourse about musical
28 This dialectic might be taken even one step further: one might attempt a transcendental argument
that there must be a diversity of musical values since humanity cannot be wrong in pursuing the diversity
of musical practices it in fact pursues. But we’re now in uncomfortably deep waters.
29 The fictionalist about possible worlds is not (thereby) a fictionalist about modality tout court, just as
the fictionalist about musical works is not (thereby) a fictionalist about music tout court. Hence the
unsuitability of the labels ‘modal fictionalism’ and ‘musical fictionalism’, despite their appealing brevity
(Sainsbury 2010, pp. 179–80).
30 I choose fictionalism about possible worlds as my illustrative example because I am not so
convinced by Sainsbury’s arguments against moral fictionalism. Sainsbury argues that the moral fiction-
alist is also in a quandary about which story about moral values to choose, that she will end up choosing
the story that gives the results she antecedently believe in. But he grants that engaging with the moral
fiction might be useful for non-moral ends such as prudential self-interest. What he seems to reject is that
works is (or ought to be) a fiction, and that the practices it is enmeshed in are
valuable for reasons other than the acquisition of truths, there can be no
question about the legitimacy of the pragmatic constraint. The story we
should tell ourselves about musical works is the one implicit in our musical
practice.
Why should a contemporary nominalist about musical works prefer
fictionalism over materialism or eliminativism? As we saw at the end
of Section IV, what motivates the choice between different nominalist
theories of musical works is how closely each theory hews to existing
musical practices. Fictionalism has certain advantages here. It looks, at first
glance, as if we do not need to alter our musical practices at all. We
can continue to talk about musical works in just the ways we have always
talked about them.31 On closer inspection, however, there are a couple of
changes. First, and most significantly, the fictionalism I have recommended
is ‘revolutionary’ in that it recommends moving from a musical-works
discourse aimed at truth to a musical-works discourse aimed at whatever
the value is of practices involving such discourse. So though the practice,
including the discourse, may look the same on the surface, it will be
operating in a different way. What would in the past have been assertions
about musical works, for instance, ought really, according to the fictionalist,
to be put forward as make-believe. Second, as we have just seen, it is
conceivable that the fictionalist with a complete theory would recommend
that some practices be changed, in light of the value of musical
practices involving discourse about works. But this is not a consequence
of fictionalism in particular, since the change is due to the theory of musical
value, not the fictionalism. A Platonist about musical works with a theory
about the value of practices involving musical-works discourse could just as
easily suggest that certain musical practices ought to be changed.
It seems to me that the best response the materialist can give to this line of
reasoning is to press on the fact that, according to fictionalism, there are no
there will be any way for us (practically? psychologically? theoretically?) to neutrally evaluate moral
fictions for how well they achieve that end. That seems an unjustified assumption. To my mind, the
bigger problem for the moral fictionalist is how to avoid the charge that the end substituted for moral
value (e.g. prudential self-interest) is not being appealed to as exactly the kind of entity the moral
fictionalist was motivated by rejecting in the first place (i.e. an objective value).
31 I pass over the distinction between the messiness of actual musical practice and the cleaner theory
we achieve by a process of aiming at reflective equilibrium.
216 ANDREW KANIA
musical works. The appearance of congruity with our discourse about

musical works is merely a façade, since we take that discourse to be about
musical works. The fictionalist’s response to this is to press on the idea that the
value of our musical-works discourse is not truth, but instead whatever
the value is of the broader musical practice that includes that discourse.32
I am not sure what response the eliminativist can plausibly give to the
fictionalist. In a sense, fictionalism is just an eliminativism that preserves our
discourse about musical works (albeit in fictional rather than assertoric
mode). Perhaps the eliminativist could similarly press on the idea that the
point of our musical discourse is truth, and thus that (given nominalism)
eliminativism is the only theory that faces up to the harsh truth that there are
no musical works as conceived of in our practices. But to my mind this just
throws into sharper relief the fact that our musical discourse would be
valuable even if there were no musical works of the kind it implies.
A useful thought experiment available to fictionalists in this connection is
that of the Oracle of Philosophy. Suppose you humbly submit to the Oracle
the question of whether musical works exist, and the Oracle succinctly
answers in the negative. Would you really conclude that we (for any ‘we’)
should give up talking about musical works? If you’re a nominalist inclined
to answer ‘no’ to this question, you should be inclined to fictionalism about
musical works.33
There is something a bit strange about this way of comparing
the nominalist alternatives. For it seems as though our musical practices,
including our discourse, wouldn’t really change under any of these revisionary
theories. The ontological theory of musical works as abstract objects is already
the theoretical philosophical result of a process of reflective equilibrium—
an abstraction, if you will. If the pragmatic constraint is correct, then, as
philosophers we ought to think of musical works, if there are any, as a certain
kind of abstractum. Similarly, if it turns out there are no such things, it is only
as philosophers that we must choose between the nominalist alternatives. But
that’s just the nature of the ontological game. We are trying to figure out
what ultimately exists, and what we ought ultimately to say about what exists.
If my evaluation of the various nominalist alternatives here is correct, then,
32 The fictionalist must also have some theory of empty names, which will allow them to explain how
the discourse can still be about musical works in some sense, just as we can talk about Zeus or Sherlock
Holmes. But everyone needs a theory of empty names.
33 The Oracle makes its first appearance in the fictionalism literature in Burgess and Rosen (1997, p. 3).
ultimately, if there are no abstracta of the sort we seem committed to in our

musical discourse about works, we ought to be fictionalists about musical
works. For all I’ve said here, however, the nominalist still faces the challenge
of showing that there are no such abstracta.
References
Brown, Lee B. (2011) ‘Do Higher Order Ontologies Rest on a Mistake?’ British
Burgess, John P. and Rosen, Gideon (1997) A Subject with No Object: Strategies for
Nominalistic Interpretation of Mathematics (Oxford: Clarendon Press)
Cameron, Ross P. (2008) ‘There are No Things that are Musical Works’ British
Caplan, Ben and Matheson, Carl (2004) ‘Can a Musical Work be Created?’ British
—— (2006) ‘Defending Musical Perdurantism’ British Journal of Aesthetics 46: 59–69
—— (2008) ‘Defending “Defending Musical Perdurantism” ’ British Journal of
Aesthetics 48: 80–5
Davies, David (2004) Art as Performance (Malden, MA: Blackwell)
—— (2009) ‘The Primacy of Practice in the Ontology of Art’ Journal of Aesthetics
Davies, Stephen (2001) Musical Works and Performances: A Philosophical Exploration
(Oxford: Oxford University Press)
—— (2003) ‘Music’ in Jerrold Levinson (ed) The Oxford Handbook of Aesthetics
(Oxford: Oxford University Press), pp. 489–515
Dodd, Julian (2002) ‘Defending Musical Platonism’ British Journal of Aesthetics 42:
380–402
—— (2004) ‘Types, Continuants, and the Ontology of Music’ British Journal of
—— (2007) Works of Music: An Essay in Ontology (Oxford: Oxford University
Press)
Eklund, Matti (2007) ‘Fictionalism’ in Edward N. Zalta (ed) The Stanford Encyclo-
pedia of Philosophy, Spring edn, <http://plato.stanford.edu/archives/spr2009/
entries/fictionalism/>
Evnine, Simon (2009) ‘Constitution and Qua Objects in the Ontology of Music’
Field, Hartry (1980) Science without Numbers: A Defence of Nominalism (Oxford:
Blackwell)
218 ANDREW KANIA
Goehr, Lydia (2007) The Imaginary Museum of Musical Works: An Essay in the
Philosophy of Music, rev. edn (Oxford: Oxford University Press)
Goldman, Alan (2011) ‘Value’ in Theodore Gracyk and Andrew Kania (eds) The
Routledge Companion to Philosophy and Music (New York: Routledge), pp. 155–64
Goodman, Nelson (1976) Languages of Art: An Approach to a Theory of Symbols
(Indianapolis, IN Hackett)
Gracyk, Theodore (1996) Rhythm and Noise: An Aesthetics of Rock (Durham, NC:
Duke University Press)
—— (2011) ‘Evaluating Music’ in Theodore Gracyk and Andrew Kania (eds) The
Huovinen, Erkki (2011) ‘Understanding Music’ in Theodore Gracyk and Andrew
Kania (eds) The Routledge Companion to Philosophy and Music (New York:
Routledge), pp. 123–33
Kania, Andrew (2008a) ‘The Methodology of Musical Ontology: Descriptivism
and its Implications’ British Journal of Aesthetics 48: 426–44
—— (2008b) ‘New Waves in Musical Ontology’ in Kathleen Stock and Katherine
Thomson-Jones (eds) New Waves in Aesthetics (New York: Palgrave Macmillan),
pp. 20–40
—— (2011a) ‘Definition’ in Theodore Gracyk and Andrew Kania (eds) The
—— (2011b) ‘All Play and No Work: An Ontology of Jazz’ Journal of Aesthetics and
Art Criticism 69: 391–403
Lachs, John (1995) ‘Footnotes to Plato’ in Ted Honderich (ed) The Oxford Com-
panion to Philosophy (Oxford: Oxford University Press), p. 284
Le Poidevin, Robin, Simons, Peter, McGonigal, Andrew, and Cameron, Ross
P. (eds) (2009) The Routledge Companion to Metaphysics (New York: Routledge)
Levinson, Jerrold (1990a) ‘What a Musical Work Is’ in Music, Art, And Metaphysics
(Ithaca, NY: Cornell University Press), pp. 63–88
—— (1990b) ‘What a Musical Work Is, Again’ in Music, Art, And Metaphysics
(Ithaca, NY: Cornell University Press), pp. 215–63
—— (1996) ‘Art as Action’ in Pleasures of Aesthetics (Ithaca, NY: Cornell University
Press), pp. 138–49
Loux, Michael j. and Zimmerman, Dean W. (eds) (2003) The Oxford Handbook of
Metaphysics (Oxford: Oxford University Press)
Matheson, Carl and Caplan, Ben (2008) ‘Modality, Individuation, and the
Ontology of Art’ Canadian Journal of Philosophy 38: 491–518
Predelli, Stefano (2009) ‘Ontologese and Musical Nihilism: A Reply to Cameron’
Rudner, Richard (1950) ‘The Ontological Status of the Esthetic Object’ Philosophy
and Phenomenological Research 10: 380–8
Sainsbury, R.M. (2010) Fiction and Fictionalism (New York: Routledge)
Stecker, Robert (2009) ‘Methodological Questions about the Ontology of Music’
Journal of Aesthetics and Art Criticism 67: 375–86
Szabó, Zoltan Gendler (2003) ‘Nominalism’ in Michael J. Loux and Dean
W. Zimmerman (eds) The Oxford Handbook of Metaphysics (Oxford: Oxford
Tillman, Chris (2011) ‘Musical Materialism’ British Journal of Aesthetics 51: 13–29
van Fraassen, Bas (1980) The Scientific Image (Oxford: Clarendon Press)
Wollheim, Richard (1980) Art and its Objects: With Six Supplementary Essays
IV
Abstracta Across the Arts
10
Reflections on the Metaphysics
of Sculpture
HUD HUDSON*
I. Familiar Characterizations of Sculpture

Consider the two-millennia-year-old, stunning and massive, white-marble
sculpture, Laocoön and His Sons, attributed by Pliny the Elder in his Natural
History to three Rhodian sculptors, Hagesander, Athenadorus, and Polydorus.
The statue, which was rediscovered in the early 16th century, represents both
Laocoön (that unfortunate priest of Poseidon who wisely cautioned his fellow
Trojans ‘to beware of Greeks bearing gifts’) and his sons, Antiphantes and
Thymbraeus, at the moment of their destruction by sea serpents.
Although Pliny may have overstated the case in his description of the
Laocoön as the greatest of all works of art ‘superior to any painting and any
bronze,’ it seems beyond doubt that the Laocoön is a magnificent artwork
and a superb example of what sculpture has to offer as a distinctive art form.1
The Laocoön is three-dimensional, thus inviting visual assessment from
multiple points of view, and it is constructed from several pieces of solid
marble, thus permitting tactile assessment, as well. The artwork represents a
(fictional) event by way of a combination of its three-dimensional shape and
its ambient space, and in so doing manifests striking aesthetic properties, in
* For generous and insightful criticism and comments on an earlier draft, I thank Christy Mag Uidhir.
1 As noted in the opening sentence of the solitary essay devoted to the topic in The Oxford Handbook of
Aesthetics, ‘philosophy has not had a great deal to say about sculpture’ (Hopkins 2002, p. 572). Yet despite
the lack of any widely-endorsed or orthodox analysis, there certainly are a number of commonly shared,
informal preconceptions about sculpture, a handful of which I hope to challenge in the present chapter.
(I recommend Hopkins’s article as a particularly helpful introduction and guide to some of the
philosophical literature on sculpture.)
224 HUD HUDSON
this case, by providing an exquisitely-beautiful representation of a horrific

ugliness—the strangling of a man and his children. Finally, it shows how a
single and unified piece of art can emerge as the product of the joint beliefs,
desires, intentions, and volitions of several artists all working in concert.
I suspect that popular thoughts concerning what is distinctive about and
perhaps even constitutive of sculpture have been significantly over-influ-
enced by reflecting on familiar statues like the Laocoön, and I fear that as a
result we suffer from a tendency to mistake accidental characteristics of our
best-known masterpieces for essential characteristics of sculpture in general.2
In this chapter, I will introduce and defend a series of examples designed to
challenge the commonsense theses that sculpture must be three-dimen-
sional—in Section II, or solid—in Section III, or available to visual and
tactile assessment—in Section IV, or in interaction with an ambient space—
in Section V. In Section VI, I will investigate an intriguing feature of one
special class of sculptures—those carved from a single block of marble such as
Michelangelo’s Pieta. In Section VII, I will investigate an intriguing feature
of another special class of sculptures—those related to multiple castings such
as Rodin’s The Thinker. Finally, in Section VIII, I will offer some concluding
remarks on the lessons of these investigations for the ontology of sculpture in
particular and for the subfield of art ontology in general.3
II. On the Dimensionality of Sculpture

The first commonsense thesis to be challenged concerns the dimensionality of
sculpture. Sculpture, we are told differs from two-dimensional art (such as a
2 The ‘we’ in this passage is primarily intended to pick out us non-specialists, for the tendency here
lamented is presumably rather less prominent in those who are familiar with the surprising developments
in sculpture over the last half-century (e.g. the minimalist sculpture of a Fred Sandback or the combines
of a Robert Rauschenburg).
3 Definitions or analyses are often hard to tell apart from generics in the literature on sculpture, and
rather than attributing to particular theorists the official theses in question, let it suffice to attribute to
them descriptions that have (I conjecture) contributed to the overly-narrow conception of sculpture
popular today. Accordingly, for the emphasis on the three-dimensionality of sculpture, see Carroll
(2008); Hopkins (2002); and Langer (1953). For the discussion of the significance of the visual and tactile
assessment of sculpture, see Carpenter (1960) and Read (1977). For the highlighting of the role of the
ambient space of sculpture, see Hegel (1975); Langer (1953); and Martin (1976). For special focus on the
representational function of sculpture and debates regarding just how sculptural representation works,
see Goodman (1976); Schier (1986); and Walton (1990).
REFLECTIONS ON THE METAPHYSICS OF SCULPTURE 225
Van Gogh painting), by being three-dimensional. Strictly speaking, of course,

The Starry Night has as much claim on being a three-dimensional (albeit very
thin) artwork as does The Winged Victory of Samothrace, but the sentiment is clear
enough. Sculpture has a robust, three-dimensionality—not the canvas-thin
thickness of an oil painting that affords no perceptual accessibility or represen-
tational power to its artistically insignificant depth. In fact, fine distinctions in
this vein account for differences between paradigm instances of sculpture on the
one hand and High-relief, Bas-relief, and Sunk-relief pieces on the other, each
of which is diminished in the third dimension in some characteristic way.
Caution is called for, however. Let a receptacle be a region of space-time
possibly exactly occupied by a material object. One non-trivial question
concerning the metaphysics of material objects, then, is—‘Just which regions
are receptacles?’ A highly-defensible answer to this enquiry is that absolutely
any region is a receptacle and that it is largely the business of empirical science
and philosophy to inform us which receptacles are actually occupied (i.e.
which regions serve as hosts to actual objects). Moreover, neither contempo-
rary science nor philosophy speaks with one voice on this topic. One epistemi-
cally-open option takes the material world to be exhausted by a very large
number of point particles (and their sums) and maintains that every statue is a
fusion of countably-many, properly arranged, point-sized items. If that were
the correct metaphysic, statues would turn out to be scattered, zero-dimen-
sional objects, and yet, owing to their exceedingly dense arrangement, would
nevertheless appear quite solid and could still present the senses with all the
splendor of Michelangelo’s David. Alternatively, perhaps some version of
string theory is correct, according to which every material object decomposes
into a scattered plurality of proper parts, each a topologically-connected,
vibrating string. If that were the correct metaphysic, statues would be scat-
tered, one-dimensional objects. Alternatively, they could be scattered, two-
dimensional objects, should some tile-based theory prove the favored posit of
the best physics, and so on up the hierarchy of dimensions.4
Note, moreover, that the hierarchy certainly need not top out at three.
Although difficult (if not impossible) to imagine, there is no logical barrier
to manifolds with more than three spatial dimensions—any arbitrarily large
n-dimensional space is a logically consistent state of affairs. Of course, there may
4 For more on regions, receptacles, and recommended restrictions on dimensionality of material

objects, see Hudson (2006, ch. 2).
226 HUD HUDSON
be a metaphysical upper-limit on n-dimensional spaces, but currently it seems

once again an epistemically-open option that the region in which we live and
move and have our being will turn out to sport (at least) four spatial dimensions,
the fourth perhaps already proudly containing a cross section of every statue that
ever has been produced. That is, if we were to come to learn that material
objects were already extended in four dimensions (just not very appreciably in
the fourth) and that Myron’s Discobolus had an imperceptible tail, as it were, a
tail wholly located in a fourth spatial dimension, we need not thereby retract
our claim that it was an instance of sculpture, after all.
Perhaps this misses the point, however. Perhaps it’s not that the statues
themselves must be three-dimensional objects, but that they must inhabit a
three-dimensional space. Well, the point about higher-than-three-dimensional
spaces still stands, but couldn’t a two-dimensional, non-orientable space
(shaped, for example, like a Möbius strip) house an appropriately shaped,
two-dimensional statue with just as much distinction as its three-dimensional
counterpart region? Facts about the distances between the proper parts of the
sculpture would change were it to be embedded in one hosting space rather
than another, but it doesn’t seem that it would lose its statue-status, should the
region in which it is located undergo a dramatic change in dimensionality.
(More on this theme in Section V below.)
The lesson, then, is simply this: We are not currently in a position to
determine whether any of the material items we regard as artworks are zero-,
one-, two-, three-, four-, or higher-dimensional objects, and thus appealing to
differences of dimension to sort artworks into major categories seems an
unpromising strategy to champion.5
A quick thought: Time is a dimension as well, and barring multi-dimen-
sional-time worlds, one might think there is room for a theory of purely
temporal statues—that is, of objects with, say, a one-dimensional temporal
shape and no spatial shape or location, at all. Cartesian minds are well
qualified for this job, and if they could be temporally disconnected or
gappy, then manipulating those gaps could generate very interesting tem-
poral shapes, indeed—perhaps enough variation to support a plausible and
intricate aesthetics of temporal sculpture. But let us return to our main theme.
5 This critique leaves open the suggestion that artworks may continue to be sorted in accordance with
some folk conception of dimensionality, but then the resulting categories will have less pull on us when it
comes to more rigorous analysis.
III. On the Solidity of Sculpture

Suppose that, despite no real pressure to do so, we agree to confine our
attention to three-dimensional statues, after all. The second commonsense
thesis to be challenged concerns the solidity of sculpture. There is, of course,
the observation that nothing big enough to have ever been seen by any of us
is a solid item (in one sense of that term)—that is, a material object that
exactly occupies a connected region. Where naı̈ve perception declares there
exists a solid wall we have instead a swarm of scattered objects, densely
arranged, acting in unison, manifesting a variety of forces, and presenting
the mere appearance of solidity to anyone whose perceptual capacities are as
coarse as ours. But ‘perceptual solidity’ is a perfectly serviceable notion;
perhaps the commonplace thesis is that statues must at least appear solid.
However, this refined-and-restrictive thesis does not seem to hold water
as would, say, a statue of confined liquid in motion. Moreover, in addition
to liquid statues, we can entertain statues of imprisoned gas or of directed
light.6 Any such piece, however, may be regarded as conceptual art that
merely plays upon traditional categories of sculpture rather than being an
instance of those categories, and thus may not be taken as a legitimate
counterexample to the solidity thesis. Still, setting aside the challenges
presented by liquid, gas, and light sculpture, I believe there is a much
more metaphysically interesting counterexample to the solidity thesis that
is worth some serious reflection.7
Suppose I have three sons and desire to create a beautiful work of art
celebrating each of them. For my oldest and youngest, I carve handsome
statues of marble that are admired (naturally enough) from without and that
so capture the look of my children at play they seem ready to spring to life at
any moment. My middle son, however, is a challenge. Reserved, quiet, shy,
always caught up in his own thoughts—the boy’s inner life is concealed in
comparison with those of his brothers whose every emotion and thought is
transparent to the most casual observer. The best way to portray him,
6 For examples of liquid sculpture see the brief-lived works of Shinichi Maruyama as well as
innumerable fountains, for an example of gas sculpture see Fujiko Nakaya’s Fog Sculpture, and for
examples of light sculpture see the works of Dan Flavin, James Turrell, and Waltraut Cooper.
7 The case presented in the next paragraph originally appeared in Hudson 2007. Also compare
Michael Heizer’s works, North East South West and Double Negative, as additional examples of the
theme to be explored in the text.
228 HUD HUDSON
I think, is from within. Accordingly, I secure an unusually large block of

marble and set to work to create a masterpiece—The Brooder. Rather than
chipping away at the exterior of the block, however, I hollow out a cavity
leaving almost all its outer surface rough and untouched. When finished, the
marble taken away is exactly the marble that would have remained had
I worked as before and sculpted a statue of my middle son in the manner of
my earlier two pieces. The resulting cavity, therefore, has rather interesting
shape properties, the most striking of which is that it is shaped just like the
body of my introverted child when sitting in quiet reflection. With one
exception: The cavity is larger than life—allowing an admirer to enter the
hollow, to position a thin layer of marble that closes off the outside world,
and to appreciate the artwork by candlelight from within.
Attempting to paraphrase away apparent ontological commitment to
such strange items as holes—whether internal cavities, or simple depres-
sions, or piercing tunnels—is a common reaction to a story such as this one.
But not everyone is satisfied with perforation-paraphrases, and a surprisingly
good case can be made for the position that holes are genuine, material
objects in and of themselves.8 To be sure, they are peculiar, hybrid objects—
for (like regions) they can host other material objects and (like material
objects) they can take a turn as guest in a region that instead hosts them. But
they appear to manifest certain causal powers and capacities, their shapes (as
we have just seen) can be the result of the exercise of significant artistic skill,
and (surprisingly) they even seem to be objects of perception.9 All in all,
such cavities seem reasonable candidates for three-dimensional, non-solid
(non-liquid, non-gaseous, and non-luminal) statues.
IV. On the Visual and Tactile Assessment of Sculpture

Suppose that, despite no real pressure to do so, we agree to confine our
attention to three-dimensional and solid statues, after all. The third com-
monsense thesis to be challenged concerns the availability of sculpture to
visual and tactile assessment.
8 The case has been made satisfactorily (in my judgment) in Casati and Varzi (1994).
9 On the perception of holes see Nelson and Palmer (2001) and Bertamini and Croucher (2003).
But first, a brief aside: The cases for consideration (from here on out)
become increasingly fantastical, and the best justification I can offer for
seriously advancing them is simply the reminder that I am not trying to
characterize historical and contemporary sculptures or to make predictions
about the probable features of the range of future sculptures, but rather to
evaluate the status of certain preconceptions about sculpture as an art
form—in particular, to investigate what appear to be popular candidates
for necessary conditions to be included in the proper philosophical analysis of
‘sculpture.’ Since analyses boldly lay down claims about what is necessary,
even foreign and fantastical thought experiments are appropriate tools of
assessment, so long as they satisfy the modest condition of being metaphys-
ically possible. In each case we may ask—‘If there were such a thing, a thing
conforming to the salient features of the case as presented, would it be a sculpture or
no?’ If our answer is positive, then putative necessary-conditions on sculp-
ture will be exposed as mere commonly shared characteristics of actual
pieces rather than essential features of the art form itself.
That said—allow me to introduce you to three (perhaps merely possible
yet) marvelous pieces: The Infinitesimal Dragon; The Medusa; and The Ecto-
plasm Griffin.
Suppose our ability to manipulate material objects outruns our ability to
perceive the products of those manipulations. That is, suppose we have the
capacity to produce objects that do indeed have their sculptural features as
the direct result of an artist’s successful sculptural attempts but that verifica-
tion that these intention-directed activities have been successfully executed
requires the use of some non-perceptual method. (Perhaps computer pro-
grams assist in making a series of imperceptible cuts in some immobilized
material and subsequent interaction with the environment reveals the
material has taken on the desired shape.) In any number of ways we can
infer relevant facts about the three-dimensional and solid shapes of our
careful manipulations, we can pinpoint their locations, we can trade them
on the open market, we can describe their features and display them with
exaggerated mock-ups—we just can’t see or touch them. The Infinitesimal
Dragon is a statue that may not quite live up to his name (but he’s awfully
close), for despite having several replicas of him large enough to admire and
to study, we simply have no direct perceptual access to him whatsoever; he
is simply too small for perceptual capacities as crude as ours to resolve or
distinguish.
230 HUD HUDSON
A certain material is remarkably dangerous to fragile creatures like us. It is

an especially malleable substance, and when under the supervision of an
artist who is in a position to manipulate the arms of a non-sentient robot
(from the safe remove of her studio many hundreds of miles away) it can
become the stuff of a truly splendid artwork. Our artist’s tools include a
special pair of local gloves with which she measures out and fashions a
portion of harmless clay present in her studio. Whatever the local gloves do,
however, is mimicked exactly by the gloves fitted to the robot a world
away, and the mysterious substance always responds to pinches, and pulls,
and pressure just as would another lump of clay. As she works the clay, then,
the artist simultaneously creates two statues—as it so happens on this
occasion, two Gorgon busts. The Medusa is the name she eventually bestows
on the remote artwork, for like its namesake, none of us could survive a
visual encounter with it. Even worse, none of us could have a perceptual
encounter with it of any kind, despite its perfectly respectable size, it is
simply too lethal to permit approach within the range of possible perception.
Sometimes philosophers ask about the location relation that ties material
objects to regions of space-time. One pair of interesting questions is whether
a single object can be located in more than one region (multilocation) and
whether two or more objects can be located in a single region (co-location).
In defense of the thesis of the possibility of co-location, its friends have
occasionally asked us to consider a very different kind of material object—
‘material’ because it is, after all, located at a space-time region, but ‘very
different’ because it does not interact causally with familiar material objects
like you or me (or our everyday rocks, or trees, or chairs). Instead, these
ectoplasm objects, as they have been sometimes called, simply pass right
through us and the familiar objects of our surroundings, unnoticed and
invisible, briefly sharing our locations (but without sharing any of our proper
parts). The Ectoplasm Griffin recently won the award for best artwork
produced by the artists who inhabit Ectoplasmia. These artists (like many
of our own) delight in intentionally and painstakingly shaping the material of
their world for representative and aesthetic purposes, and are furthermore
especially notable for working their material under the direction of some-
thing analogous to sound, but without the benefit of visual or tactile
sensations. Better yet, this prizewinner is sitting right there, to your left,
but unfortunately, despite its perfectly respectable size and non-lethal prop-
erties, it is simply made of a material invisible and otherwise imperceptible to

the likes of you.10
Again, admittedly fantastical, even bizarre cases, to be sure—but if any of
them is so much as metaphysically possible, the thesis that part of what it
takes to be a sculpture is to be available (even in principle) to visual and
tactile assessment is false. Even stranger cases lie ahead.
V. On the Ambient Space of Sculpture

The fourth commonsense thesis to be challenged concerns the notion that
sculpture is distinctive insofar as it must be presented in an ambient space.
One quick way with the commonsense thesis would be to return to the
metaphysical presuppositions operative in the story of The Ectoplasm Griffin
and entertain a space-time-filling statue made of material that did not in any
way causally interact with (and thus accommodated without impeding) the
ordinary material objects we are identical to and familiar with. That is,
space-time would contain an omnipresent statue that precisely matched its
dimensions, topology, and size and would still have room for the rest of us,
to boot. Perhaps such a state of affairs is possible only if God is (and has a
peculiar artistic bent), but it is an interesting challenge to the thesis that
every statue requires an ambient space, for The Ectoplasm Universe, as we shall
call it, intrudes into every region that might be reserved for that purpose.
As before, however, I believe there is a much more metaphysically
interesting counterexample to the ambient-space thesis—one that also
features a Divine artist—worthy of some serious reflection. Substantivalism
is the view that takes space-time to be a concrete particular with an onto-
logical status not reducible to relations between material objects (or any
other entities). On such a conception, one may be a realist about space-time
points (i.e. treat pointlocations as genuine entities in and of themselves),
one may identify regions with fusions or pluralities of these points, and one
may identify individual times or moments with certain hyperplanes or
10 A possible response: Well, just as it is imperceptible to me, so too, it isn’t sculptural for me—that is,
the basic relation is sculptural for, there is no such thing as sculpture—full stop, and to be sculptural for entails
being perceptible to. Such a relativizing strategy seems extreme to me, but it is worth mentioning just the
same.
232 HUD HUDSON
“slices of the space-time block.”11 Several theories in the philosophy of time

are very much at home with such a conception of space-time, especially a
trio of views known as the Growing Block, the Shrinking Block, and the
Disappearing Branch Theory—all of which endorse a special feature salient
in the discussion to come, namely, a recombination principle for space-time
points according to which these points are not modally bound to one
another but rather appear in different combinations when different times
are present.12
A Morphing Block is a space-time that (like its cousins in the paragraph
above) features different combinations of space-time points when different
times are present, but (unlike its cousins in the paragraph above) does not
always suffer growth or diminishment in the same direction and does not
always grow or diminish in increments of a single hyperplane. Instead, a
Morphing Block may change its shape in radical ways as time passes, simply
by way of creation or annihilation of space-time points whether they be
located at the ends of the block (so to speak) or somewhere deep within its
interior.13 If the creation-annihilation stories that account for this change of
shape are just the uncaused and mysterious comings-to-be and obliterations
of the standard Growing Block, Shrinking Block, and Disappearing Branch
Theories, then we no more have a candidate for sculpture than we would if
a blind and intentionless hurricane reconfigured a beach’s worth of sand
into the shape of Neuschwanstein Castle. But if the chisel (so to speak) is
wielded by a Divine artist, determined to bring an object into a certain
determinate shape for the purpose of displaying certain aesthetic properties,
then we have a rather different affair. Play along with the theism for a
moment, and entertain the metaphysical possibility of God’s creating a
disconnected and uninhabited space-time, distinct from ours, which He
subjects to this gradual process of shaping. Consequently, we would have a
substantival entity—a concrete particular—whose shape is the result of just
the kind of chipping away that we associate with the generation of classical
11 For discussion of Substantivalism see Earman (1989) and Nerlich (1994).

12 For discussion of the Growing Block Theory see Broad (1923) and Tooley (1997). For discussion of
the Shrinking Block Theory see Casati and Torrengo (2011). For discussion of the Disappearing Branch
Theory see McCall (1994).
13 For a much more sustained discussion and critical evaluation of the Morphing Block Theory see
Hudson (2011, 2012).
sculpture. And, interestingly enough, the resulting artwork is not presented

in an ambient space—on the contrary, it is identical to the space.
Now, of course, the atheist will not be moved by this example, nor will
the theist who denies that God would ever turn His creative powers to such
frivolous ends. But wouldn’t it be surprising if the proper analysis of
‘sculpture’ presupposed the non-existence of God or else took a stand on
the possible aesthetic creations of a perfect being? Better, I suspect, to err on
the side of caution and to refrain from insisting on the essentiality of being
presented in an ambient space when it comes to sculpture.
VI. The Menagerie

The Menagerie—that’s its name—is a heavy, marble sphere that occupies the
corner of my office desk. It is so-called because it contains many duplicates
(varying only in size and orientation) of every statue of an exotic animal that
ever has been or ever will be sculpted, and they are all of them confined
within the bounds of one and the same sphere with an eight-inch diameter.14
In the early 20th century, Samuel Alexander remarked that ‘elicited by his
own creative art . . . the sculptor discovers his figure in the block’ (1925,
p. 228). The opposing view—both to Alexander and to my remark about the
current contents of The Menagerie—is quite simply that the block and the
sphere contain no actual figures patiently awaiting release, but rather mere
potential figures which are to be created by the artist’s hand, not discovered by
drawing back the layers of a marble curtain. Who is right? Do there exist
genuine objects in the block and the sphere with the requisite shapes prior to
cutting into the marble? If there are such figures now present, is it correct to
think of them as statues or artworks before they are revealed or—if too small,
lethal, or imperceptible to be revealed—before they are separated from their
neighbors and come to inhabit new environments? Good questions!
First question—‘Are the figured objects in the block prior to the sculpting?’
The metaphysics of composition is highly relevant here. Enquiries
into just when a plurality of objects has a mereological sum have been
among the highlights of recent analytic metaphysics. One popular and
14 The Menagerie contains many duplicates of every other statue, too, but the others weren’t relevant to
its christening.
234 HUD HUDSON
surprisingly-defensible position—which maintains that the correct answer is

‘always’—is known as Universalism.
(Universalism): Necessarily, for any objects, the xs, there exists an object,
y, such that the xs compose y.15
According to this liberal theory of composition, whenever there are some
things those things have a mereological fusion no matter what spatio-
temporal and causal relations they may or may not satisfy. Moreover, to
call attention to this fusion is to call attention to an additional piece of the
world’s furniture and is not to be confused with talking about the objects
collectively, or about a property they share, or about the set that has them as
members. Note that on this formulation of the view, part-hood is treated as
a two-place relation with no mandatory indexing to times, thus permitting
both diachronic as well as synchronic fusions. Consequently, on the as-
sumption that the block is itself a fusion of certain microphysical parts,
any arbitrary subset of those parts also has a mereological sum, and so there
they are—all those objects with the figures of lions, and tigers, and bears—
occupying subregions of the block prior to sculpting.
The metaphysics of decomposition is also highly relevant here. Whereas
Universalism assures us that many will yield one, the Doctrine of Arbitrary
Undetached Parts (DAUP) assures us that (an extended) one will yield
many; that is to say, whereas Universalism provides a liberal theory of
composition, DAUP provides a liberal theory of decomposition.
DAUP Necessarily, for any material object, x, and regions, s and s*, if s is
the region x exactly occupies, and if s* is an exactly occupiable
subregion of s, then there exists a material object, y, such that (i)
y exactly occupies s*; and (ii) y is a part of x.16
According to this very liberal theory of decomposition, whenever there is
an extended object, that object has parts entirely confined to each of the
occupiable subregions of the region it occupies. Moreover, to call attention
to these parts is to call attention to additional pieces of the world’s furniture
and is not to be confused with talking about portions of the object, or about
conceptual parts, or about potential parts, or about what would exist if the
15 For discussion of Universalism, see Lewis (1986) and Hudson (2001, ch. 3, from which the next few
sentences in the text have been borrowed).
16 For discussion of DAUP, see Hudson (2001); van Inwagen (1981); and Zimmerman (1996).
rest of the object were cut away. Once again, then, on the assumption that
the block is itself occupying an extended region, any arbitrary subregion of
that region hosts an object too, and so there they are—all those objects with
the figures of lions, and tigers, and bears—occupying subregions of the
block prior to sculpting.
Second question—‘Are the figured objects in the block prior to sculpting also
statues prior to sculpting?’
It seems clear that the answer is ‘no,’ but there are at least two competing
explanations that disagree about why ‘no’ is the correct response—two
explanations that arise from two popular theories about how material
objects persist across time.
Endurantism is the view that an ordinary material object is not spread out in
time as it is in space and that it is composed of different spatial parts at different
times (rather than being composed of temporal parts across times). Persistence
across an interval of time is thus a matter of being located at the different times
in that interval (i.e. of being multilocated at a series of non-simultaneous three-
dimensional regions).17 On this metaphysics (assuming that there is, in fact, a
figured object in the block prior to sculpting) ‘being a statue’ is a temporary-
property or a phase-sortal of a material entity. That is, no new thing comes into
existence as the sculptor practices her art; instead, an already existing thing
comes to have a property it didn’t have before. Strictly speaking, then, on this
view the sculptor does not create a statue (in the sense of bringing into existence
a thing which is a statue) but she does create a statue (in the sense of bringing it
about that something which was formerly not a statue is now a statue).
Perdurantism is the view that an ordinary material object is spread out in
time just as it is in space and that it is composed of temporal parts as well as of
spatial parts. Persistence across an interval of time is thus a matter of being
partly present at the different times in that interval (i.e. of having different
temporal parts located at the different moments in that interval).18 On this
metaphysics (again assuming that there is a figured object in the block prior
to sculpting) ‘being a statue’ is not merely a phase-sortal of some material
objects, it is also a career-predicate of some material objects. That is, as the
material object in the block prior to sculpting continues to persist, many
new things—new temporal parts—are coming into existence. Moreover,
17 For discussion of Endurantism, see van Inwagen (1990).

18 For discussion of Perdurantism, see Sider (2001).
236 HUD HUDSON
the Perdurantist’s willingness to thus countenance many more objects than

does the Endurantist ends up making a difference to just which items are the
best candidates for being genuine statues. Consider the phase during which
the Endurantist would be inclined to say that our figured object is tempo-
rarily a statue. The Perdurantist instead recognizes a distinct entity which
has a life-span confined exactly to that period and which is a proper,
temporal part of the longer-lived object. It is the shorter-lived, temporal
part, however, that seems to be the only non-arbitrary candidate for the title
‘statue,’ for it is distinctive among the available candidates in virtue of both
being a statue at each moment it is present and not being a temporal part of
any other such thing. Strictly speaking, then, on this view the sculptor
creates a statue in yet another sense, for she brings it about that something
that comes into existence when she reaches a certain point in her artistic
endeavors is not merely another temporal part of a pre existing figured
object, but is also a statue.
A disturbing, parting thought:19 If the foregoing discussion is on the right
track, then a sculptor who sets to work on a single block, displays her artistic
skill by exhibiting some preexisting figure or (to accommodate the cases in
Sections II–V above) by changing its immediate environment in salient
ways. In order to display a certain figure currently secreted away in the
interior of the marble sphere on my desk, a sculptor will have to cut into the
sphere and remove pieces of the marble currently obscuring the proposed
artwork—i.e. whichever figure the artist intends to transform into a statue
by freeing it from its surroundings and exhibiting it in the right sort of way.
The cost—and perhaps you will think it only a minor cost—is the willful
destruction of many aesthetically spectacular objects which will be broken
into pieces in the process (and which will, in their scattered state, lose much
of their aesthetic value)—for remember, The Menagerie contains duplicates
of every statue, and any one of them may be exhibited only by way of the
sacrifice of nearly all of the others. There is, of course, no guarantee that our
sculptor will be successful and will present to view (as her accomplishment)
one of the many beautiful figures within, but there is a guarantee that she
will destroy many beautiful non-artworks in the attempt. The artist who
19 . . . which will presuppose a view according to which naturally occurring objects (such as pieces of
marble) can exemplify aesthetic properties (such as beauty) even if no one ever apprehends them, or
appreciates them, or even has any de re attitude whatsoever towards them.
sculpts from a single block of marble, then, always takes a (sometimes

lamentable) risk.
VII. Concrete or Abstract?

But what of sculpture that is not generated by carving, chipping, chiseling,
cutting or in other ways removing matter? What of cast sculpture? Far from
being a preexisting material object imprisoned in a sea of marble awaiting
emancipation and promotion to statue-status, cast sculpture is often
regarded as an abstractum capable of admitting numerous, concrete, non-
art-objects as its instances. On this view, Rodin’s The Thinker does not
reside in the Musée Rodin or any other museum; instead, it is a non-
material, non-solid, non-spatially located, abstract object which is made of
neither bronze nor marble, yet which by way of some twenty or so original
castings is intimately-related to a number of material objects each of which
presents that masterpiece to its audience.
Of course, such a theory presupposes that there are such things as
abstracta in the first place and thus sets itself against an impressive nominalist
tradition that would have it otherwise, and it takes at least one of two
controversial stands—either that cast sculptures do have or else abstract
objects fail to have all of their features essentially.20 But let us (for the sake
of argument) presuppose a liberal ontology of and a permissive modal
profile for abstracta. Given those concessions, should we then embrace
the thesis that sculpture comes in at least two radically different forms—
the concretes arising from the carvings and the abstracta underlying the
castings?
The question is too substantial to receive adequate treatment in this
chapter, but I would like to offer a critique of one natural temptation
leading to an affirmative answer: Suppose, then, that in the case of The
Thinker we are confronted with several numerically distinct material objects
which (despite obvious differences—e.g. one being of marble another of
bronze) have remarkably similar perceptual features, which share certain
features of their causal histories tracing back to the same intention-directed
20 For discussion on the first point see Cameron (this volume, Ch. 8). For discussion on the second see
Hazlett (this volume, Ch. 7).
238 HUD HUDSON
activities of the same artist, and which are called (in a variety of contexts) by
the same name.21 We acknowledge that none is privileged in the right way
and when pressed to identify a unique bearer for the name that we casually
use to pick out the several objects so unified, we resort to officially
bestowing the name as well as the status ‘artwork’ on the single, abstract
object that we believe is operative in such unification—the art-abstractum
which then becomes the official referent of Rodin’s ‘The Thinker.’
Again, I only intend to engage the reason exhibited in this bit of
speculative psychology above in favor of the abstractum thesis for cast
sculpture. That reason consists (i) of acknowledging pressure to select
exactly one referent for ‘The Thinker’; (ii) of conceding that there is no
non-arbitrary concrete candidate for the job; and (iii) of turning to the
world of abstracta to complete our task. Whereas I admit (ii), I suspect both
that the pressure in (i) is actually quite slight and that the world of abstracta
in (iii) is no more likely to produce a non-arbitrary candidate than is the
concrete world. A few brief words of explanation.
First, there was a rather large series of unnamed, intermediate, material
statues (of clay, of plaster, and—if the argument of Section III above was
correct—of holes) involved in the eventual production of the numerous,
original castings of The Thinker. The fact that these objects were merely
intermediate stages in producing further and more widely-displayed objects
should not interfere with the judgment that they were themselves statues;
after all, if Rodin were to have died shortly after his first thumbing of some
clay into that well-known shape, it is hard to believe that it would have
failed to qualify as a statue. If, however, these objects were in fact christened
(ignoring complications arising from the fact that the original title was The
Poet) the many items in this series presumably bore names that were simply
homonyms. The pressure to select exactly one referent for ‘The Thinker,’
then is a bit like the pressure to select exactly one precise hue for ‘candy-
apple red,’ a pressure that relaxes as soon as we remember that for a variety
of pragmatic reasons we can press a single name into heavy service.
Second, turning our attention to abstracta only makes the arbitrariness
problem worse. Amongst the material candidates we at least have potential
21 But what of the causal histories of the posthumous castings of The Thinker? A hard question, I’d say,
and related to the question of whether Tom Stoppard’s Guildenstern is the same fictional character as
Shakespeare’s Guildenstern.
tie-breaking features like ‘the first cast,’ or ‘the first exhibited cast,’ or ‘the
cast which best realized certain artistic intentions’ (whether anyone knows
which piece manifests that favored feature or not). But which abstract object
is the privileged unifier of the instances—which has any chance of being the
abstractum that uniquely deserves the name over its abstract rivals? Recall
that we granted (against the nominalist) that there was a Platonic heaven of
abstracta to choose from, but the problem is that even if we confine our
attention to some uncontroversial set of material instances, there are still
plenty (even infinitely many) abstract objects which do an equally good job
of unifying exactly the members of that set. Moreover, whereas the details
and particulars of the actual historical artistic process significantly reduce the
range of admissible candidates, there is nothing in that process which selects
from the (infinity of) remaining abstract candidates a solitary winner.
Moreover, we presumably couldn’t select just one if we sat down and
tried; the fine-grained differences between the competitors quickly outrun
our abilities to distinguish them from one another. We could, of course,
regard it as indeterminate just which abstract object does the trick, but then
we would have gained no real advantage of uniqueness by retreating from
the concrete world. Accordingly—barring other and more persuasive con-
siderations—I see no reason to deny the status of artwork to each of the
concrete instances and to honor their common origins and other similarities
with harmless homonyms.
VIII. Concluding Remarks

Despite an occasional confident pronouncement here or there in the
preceding sections, I regard myself as an enthusiastic-yet-utter novice in
the field of aesthetics. Art ontology is a bit closer to my own specialization,
and although unprepared to contribute much to the wide variety of aes-
thetically-sophisticated debates, I do hope to offer some modest advice to
those working in this area.
So—by way of summary—on the strength of the examples in Sections II–V,
I would caution the art ontologist against elevating to the status of necessary
conditions the commonsense notions of sculpture as three-dimensional,
solid, and perceptually-accessible artworks set against an ambient space.
240 HUD HUDSON
Additionally, on the strength of the discussions in Sections VI–VII, I would

advise the art ontologist to maintain that sculptures do not come in two
flavors (concrete and abstract) but are instead one and all concrete and
to acknowledge that many of those items (the carved sculptures) are
preexisting figures which are not so much created as they are engaged,
exhibited, and transformed into artworks.
References
Alexander, Samuel (1925/2000) Art and the Material reprinted in Collected Works of
Samuel Alexander (Bristol: Thoemmes Press)
Bertamini, Marco and Croucher, Camilla (2003) ‘The Shape of Holes’ Cognition
87: 33–54
Broad, C.D. (1923) Scientific Thought (London: Kegan Paul)
Carpenter, Rhys (1960) Greek Sculpture: A Critical Review (Chicago: University of
Chicago Press)
Carroll, Noël (2008) The Philosophy of Motion Pictures (Cambridge: Blackwell)
Casati, Roberto and Torrengo, Giuliano (2011) ‘The Not So Incredible Shrinking
Future’ Analysis 71: 1–5
Casati, Roberto and Varzi, Achille (1994) Holes and Other Superficialities (Cam-
bridge, MA: The MIT Press)
Earman, John (1989) World Enough and Space-Time (Cambridge, MA: The MIT
Press)
Goodman, Nelson (1976) Languages of Art, 2nd edn (Indianapolis, IN: Hackett)
Hegel, Georg Wilhem Freiderich (1975) Hegel’s Aesthetics: Lectures on Fine Art,
vol 2, trans T.M. Knox (Oxford: Oxford University Press)
Hopkins, Robert (2002) ‘Sculpture’ in Jerrold Levinson (ed) The Oxford Handbook
of Aesthetics (Oxford: Oxford University Press), pp. 572–82
Hudson, Hud (2001) A Materialist Metaphysics of the Human Person (Ithaca, NY:
Cornell University Press)
—— (2006) The Metaphysics of Hyperspace (Oxford: Oxford University Press)
—— (2007) ‘Lesser Kinds Quartet’ The Monist 90: 333–48
—— (2011) ‘A Metaphysical Mix: Morphing, Mal, and Mining’ Philosophical
Perspectives, Vol. 25: Metaphysics (Cambridge: Blackwell)
—— (2012) ‘The Morphing Block and Diachronic Personal Identity’ in Personal
Identity: Complex or Simple? (Cambridge: Cambridge University Press)
Langer, Susanne (1953) Feeling and Form: A Theory of Art (New York: Charles
Scribner)
Lewis, David (1986) On the Plurality of Worlds (Cambridge: Blackwell)
Martin, F. David (1976) ‘The Autonomy of Sculpture’ Journal of Aesthetics and Art
McCall, Storrs (1994) A Model of the Universe: Space-Time, Probability, and Decision
(Oxford: Clarendon Press)
Nelson, Rolf and Palmer, Stephen (2001) ‘Of Holes and Wholes: The Perception
of Surrounded Regions’ Perception 30: 1213–26
Nerlich, Graham (1994) What Spacetime Explains (Cambridge: Cambridge University
Press)
Read, Herbert (1977) The Art of Sculpture, 2nd edn (Princeton: Princeton University
Press)
Schier, Flint (1986) Deeper into Pictures: An Essay on Pictorial Representation (Cambridge:
Sider, Ted (2001) Four-Dimensionalism (Oxford: Oxford University Press)
Tooley, Michael (1997) Time, Tense and Causation (Oxford: Oxford University
Press)
van Inwagen, Peter (1981) ‘The Doctrine of Arbitrary Undetached Parts’ Pacific
Philosophical Quarterly 62: 123–37
——(1990) ‘Four-Dimensional Objects’ Noûs 24: 245–55
Walton, Kendall (1990) Mimesis as Make-Believe (Cambridge, MA: Harvard
University Press)
Zimmerman, Dean (1996) ‘Could Extended Objects be Made Out of Simple Parts?
An Argument for “Atomless Gunk” ’ Philosophy and Phenomenological Research 56:
1–29
11
Installation Art and Performance:
A Shared Ontology
SHERRI IRVIN*
It has often been thought that there is a fundamental difference between

visual artworks and works in the performing arts. Typical works for per-
formance are susceptible of having indefinitely many performances, often
thought of as tokens of the work. Visual artworks, on the other hand, are
not susceptible of having indefinitely many tokens. For paintings and carved
sculptures, the particular physical object created by the artist is the only
token of the work; and even for works in multiple forms like etching or cast
sculpture, legitimate tokens of the work are typically only those authorized
by the artist and created by a specific process.
Contemporary works of installation art call this dichotomy into question.
They may have a different appearance for each exhibition, with different
configurations or even entirely different materials. The vortex elements of
Ann Hamilton’s (1988/1996) (the capacity of absorption), in the collection of
the Miami Art Museum, are configured differently each time the work is
exhibited.1 Two displays of one of Felix Gonzalez-Torres’s candy spills may
not have any components in common: the candy may be thrown out and
replaced between exhibitions.
In creating an installation artwork, the artist does not simply create or
specify some physical object. Installation artworks centrally involve the
expression of parameters for the constitution of a display; and depending
* Thanks are due to Allan Hazlett for discussion, and to Martin Montminy and Christy Mag Uidhir
for very helpful feedback on earlier versions. I am grateful to Senior Curator Peter Boswell of the Miami
Art Museum for information about works in the MAM collection.
1 Interview with Miami Art Museum Senior Curator Peter Boswell, July 2010.
INSTALLATION ART AND PERFORMANCE 243
on the work, the displays may (or even must) vary dramatically from one
exhibition to the next. Moreover, people other than the artist often con-
struct the display: Curators, conservators, and assistants may imbue it with
aesthetically relevant features the artist did not choose.
The analogy between installation works and artworks for performance,
such as musical compositions, is thus easy to see: The artist specifies param-
eters for acceptable realizations, there is considerable variation among these
realizations, and aesthetically relevant aspects of the realizations are often
introduced by others.2 Should we, then, see installation artworks as analo-
gous to artworks for performance, applying the same modes of understand-
ing in both cases?
This chapter has three objectives. First, I argue that apprehending an
installation work is, in fact, similar to apprehending an artwork for perfo-
rmance: In each case, audiences must recognize a relationship between the
performance or display one encounters and the parameters expressed in the
underlying work. Second, I consider whether and under what circum-
stances realizations are also artworks in their own right.3 I argue that, in
both installation art and performance, a particular realization is sometimes
an artwork in its own right (even as it realizes another work).4 I offer criteria
for determining when this is the case. Application of these criteria will yield
the verdict that performances are sometimes artworks in their own right,
while displays of installation artworks rarely are. However, this is a contin-
gent matter that arises from the conventions of the respective art forms.
Third, I address ontological concerns about entities that are both abstract
and temporal, as many artworks are on my analysis.
To clarify my terminology: With respect to installation art, what we see
on a given occasion is the display, and many displays may be generated for a
single work. A display is to an installation work, then, as a particular
performance is to a work for performance. The installation artwork or
2 These similarities, as well as a number of differences, between time-based media installations and
musical works are discussed by Laurenson (2006).
3 My treatment of this topic is inspired by James Hamilton’s (2007) argument that every theatrical
performance is an artwork in its own right, and no theatrical performance is a realization of any other
work. I discuss Hamilton’s argument in Irvin (2009).
4 There are also performances, e.g. Keith Jarrett’s 1975 Köln Concert (as discussed by Davies 2011,
pp. 135–6), that are not realizations of any underlying work. On my view such cases, which I do not
discuss here, always come out as artworks in their own right.
244 SHERRI IRVIN
work for performance itself is the underlying work. Both displays and per-
formances are realizations of the underlying work.
I. Production (1980)
Let us begin with a case study. Liz Magor’s installation artwork Production
(1980) is made up of some 2,800 bricks that Magor produced four at a time
out of wet newspaper, using a manual press, which is also part of the work.
This labor-intensive process required weeks of full-time work. The bricks
are not attached to each other or numbered. The work had been exhibited in
several different configurations prior to acquisition by the National Gallery
of Canada, so the curator and Magor swapped diagrams and descriptions by
fax to work out details of the new display. One of Magor’s faxes begins:
Yes, there are a thousand different ways to do it. But there’s a notion or rule of
thumb that eliminates some of them and modifies the others. I like it best when the
bricks are trying to act architecturally—they’re trying to make a wall or a column or
something. The ultimate would be that they totally cover a wall, with no space at
the top, bottom or sides . . . But a partial wall is okay too.
The bricks were finally installed in a long wall about two meters high, with
the press positioned right of center, a few feet in front of the wall. On earlier
occasions the arrangement had been quite different: The bricks were once
used to construct two parallel walls, each eleven feet long and approximately
the same height as the artist, with just enough space between them to
accommodate the press and one brick-producing worker, who might
have been constructing her own prison cell.
The nature of this work not only allows but demands reconfiguration.
The work comments on the relation between the labor of production and
the creative task of construction; the laborer simply produces the units,
which may then be manipulated in a multitude of ways. Always to display
the bricks in the same way would be to obscure this fact and thereby to
undermine an important feature of the work. However, not all possible
configurations are appropriate: The bricks can’t be dumped in a heap.5
5 I discuss this work further in Irvin (2006).

II. Apprehending the Underlying Work

This case reveals a number of things about installation artworks. First,
apprehending an installation work is more than apprehending a particular
display, just as grasping a musical work is more than hearing one performance.
Grasping the underlying work involves understanding the parameters
specified by the artist and the possibilities they create. I have written elsewhere
about the artist’s sanction: The full array of creative activity the artist
undertakes, much of which goes beyond manipulation of physical material
(in the case of visual artworks).6 When artists provide instructions for
the constitution of displays, they sanction particular features of their works:
The range of permissible configurations and materials, for instance. This is
as much a part of the creation of the work as is the making of the physical
object; and for some contemporary artworks it has supplanted the latter partly
or completely. When a museum acquires a text-based conceptual work
by Lawrence Weiner, only a certificate of authenticity, and no physical display
material, changes hands.
Attending to the parameters sanctioned by the artist is crucial. To focus
only on a particular realization, without recognizing that realizations with
different features are permissible, would be to misunderstand the underlying
work. In the case of Production, it would obscure a central interpretative
theme of the work: Namely, the idea of units of production that can be
incorporated into a final construction in very different ways. Similarly,
focusing only on a particular performance, without awareness of the permis-
sibility of other, very different performances, would often mislead us about
the underlying work; we might assume that the musical work mandates a
melodic passage which is in fact the product of permissible improvisation.
Even when realizations are heavily constrained, so that little variability is
permissible, this is an important thing to grasp about the work. Imagine a
work, call it Obsession, where only one configuration of the bricks is
permissible, with every brick numbered to ensure it will always end up in
the same position. The range of interpretations apt to such a work is quite
different from the range of interpretations that is apt to Magor’s work
Production.
6 See Irvin (2005, 2008).

246 SHERRI IRVIN
Grasping the underlying work, then, may require sophisticated under-

standing of how the features of a given realization relate to the features of
the underlying work. This understanding will tend to be facilitated by
exposure to varied realizations, though mere awareness that the artist has
sanctioned the permissibility of such realizations will help the audience to
apprehend the underlying work more fully.
III. The Status of Realizations

With regard to apprehension of the underlying work, then, our accounts of
installation artworks and of artworks for performance should be similar. But
what of the realizations themselves? Are they artworks in their own right?
I deny that a realization should automatically be seen as an artwork in its
own right; to allow otherwise would lead to a proliferation of artworks that
is undesirable and unnecessary. I take my approach to be a reasonable
reconstruction of how people tend to understand things in ordinary think-
ing about art. A performance may fit the artist’s prescriptions well enough to
count as a performance of a given musical work, yet do nothing to tempt us
to say that a new artwork has been created. A performance of ‘Hang on
Sloopy’ by a mediocre high school marching band may, if we’re lucky,
competently exhibit some of the potential of the underlying work, but it
does nothing more than that.7 However, when a performer’s interpretation
of a musical composition is striking, we begin to speak of the interpretation
as a distinct artwork. This is especially clear in cases of jazz compositions
that allow for improvisation, but it can also happen in more standard
musical compositions, as with Glenn Gould’s interpretations of the Goldberg
Variations.
In visual art, there are practices of individuation that clearly tend away
from the proliferation of artworks. These, too, give some support to a policy
of conservatism about when we say that a new artwork has been generated.
Pace Mag Uidhir (2009, 2012), we do not credit an artist with making a new
artwork every time she prints from the same photographic negative, unless
7 I do not hold that, in general, aesthetic merit is required for the constitution of a new artwork.
However, I hold that for a realization to be a work in its own right, it must be aesthetically distinctive.
Further discussion is found below.
there is special manipulation going on in the printing process.8 A fortiori,

we do not suggest that a new artwork is created when someone else makes a
print from the negative to the artist’s usual specifications. And, of course,
when a painting or sculpture is re-exhibited in a way that requires little or
no reconfiguration, there is no temptation to say that an additional work is
generated. So, given that some installation artworks are plausibly regarded as
complex works of sculpture, some criteria must be satisfied to tempt us to
think that a new work has been generated.
Below, I propose criteria to distinguish between realizations that are art-
works and those that are not. Application of these criteria to installation
artworks, on the one hand, and works of performance, on the other, has
divergent results: Whereas performances are sometimes artworks in their own
right, the displays of installation artworks rarely are. I explain this contingent
state of affairs in terms of the conventions operative in the different art forms.
Here are the criteria that determine whether the realization is an artwork
in its own right:
(1) The features of the realization are significantly underdetermined by
parameters sanctioned in the underlying work: That is, the para-
meters allow for readily apparent variation among realizations.
(2) The realization makes an aesthetic contribution that is not merely
perfunctory: It possesses aesthetic properties that, either in degree or
in kind, go significantly beyond what is required by the parameters.
(3) Those constructing the realization appropriately (in accordance with
the artist’s sanction) make aesthetic decisions that are not simply man-
dated by aesthetic values expressed by the artist of the underlying work.
When at least one of these criteria is unsatisfied, the realization is not an
artwork.9 Failures to satisfy the criteria tend to occur in different ways, and
for different reasons, in installation art than in the performing arts. I will defend
the criteria in the course of discussing cases in which they are not satisfied.
Here are two examples of failure that are typical of installation artworks:
8 As I understand Mag Uidhir, his nominalist view of photographic and print ontology stems from the
assumption ‘that a particular print is in fact an artwork’ rather than an instance of an artwork (Mag Uidhir
2009, Section 2). While I believe it is sometimes true that a particular print is an artwork in its own right,
I do not regard the assumption as generally acceptable with regard to prints and, especially, photographs.
9 Satisfaction of the criteria is a matter of degree, with the consequence that some realizations may be
borderline cases of distinct artworks. I see no reason to be troubled by this.
248 SHERRI IRVIN
(A) The aesthetically relevant features are so heavily constrained by the

underlying work that all admissible displays will be very similar.10
In such a case, criterion 1 is not satisfied, and it is thus impossible for 2 or 3 to be
satisfied. The possibility that the realization is an artwork cannot get off the
ground at all. If those installing the work are forced to make aesthetic decisions
(perhaps due to peculiarities of the gallery space), they will aim to be guided by
principles expressed by the artist and features of past realizations.11 The
resulting display is not an artwork in its own right, any more than the hanging
of a painting on a particular occasion is or generates a new artwork. It owes all
of its aesthetically relevant features to the underlying work:
(B) A display may fail to be an artwork even though significant vari-
ability is permitted by the underlying work.
In this case, typified by Liz Magor’s Production, the display’s features are
significantly underdetermined by the work’s parameters (1). The realization
may well surprise us, even if we have seen other realizations in the past; and thus
its aesthetic contribution need not be perfunctory (2). Because variability is
permitted, those constructing the realization may be required to make aesthetic
decisions. Crucially, however, every effort will be made to refer back to
parameters explicitly expressed by the artist or implicit from prior realizations.
As my discussion of Production revealed, it is conventionally the curator’s task to
avoid introducing salient aesthetic features; curatorial decisions should, instead,
reflect the artist’s aesthetic values (3). Discussions of contemporary art curation
and conservation emphasize this point: The artist should always be consulted, if
possible, in matters of display and conservation. Long questionnaires and
elaborate best practices have been developed to document artists’ preferences.12
The failure to satisfy criterion 3 suggests that the display is not an artwork
in its own right, even though the displays differ markedly. Those who
constitute the display are acting as agents of the artist, in accordance with
10 An example is Maria Fernanda Cardoso’s 1992 work Cementerio—jardı́n vertical (Cemetery—Vertical

Garden), in the collection of the Miami Art Museum, which is installed in accordance with rigid
parameters and templates that preclude significant variability.
11 As Laurenson (2006, n.p.) notes, cases of this type are very common among time-based media
installations, which ‘operate more like sculpture in that the artist rarely sees him or herself working
within a context that allows for interpretation in the way that a composer or a playwright might.’
12 Major research networks and projects dedicated to developing best practices for installation
and new media artworks include: DOCAM INCCA Inside Installations Matters in Media Art
PRACTICs and the Variable Media Network.
principles that heavily constrain their choices and prohibit autonomous

aesthetic decisions.
Performances typically do not fail to be artworks for either of the reasons
just indicated. The underlying work tends to vastly underdetermine the
features of the performance (criterion 1), with the result that performances
may make significant aesthetic contributions (criterion 2). Regarding most
works for performance, it is clearly appropriate for performers to make
autonomous aesthetic decisions (criterion 3); indeed, it would typically be
wrong for them to refrain from doing so. The conventions of installation art
thus differ significantly from those of the performing arts.
How, then, might a performance fail to be an artwork?
(C) The performance lacks aesthetic significance (2), although the

underlying work substantially underdetermines the features of ac-
ceptable realizations (1).
This might occur even if 3 is satisfied, and those creating the realization
appropriately see themselves as making autonomous aesthetic decisions. They
might simply fail to generate a performance that is anything other than deriva-
tive; and in such a case, their product has no claim to be a distinct artwork.13
A word about criterion 2 is in order. Most contemporary definitions of art
do not require that an artwork be aesthetically significant; it need only have
been created through the right sort of process, or received uptake into the
institutions of art, or something of that nature. Why, then, should aesthetic
significance be required for a realization to be an artwork? First, the criterion
of aesthetic significance is not a criterion of aesthetic value. The realization
might be aesthetically significant, introducing aesthetic features that go well
beyond what the underlying work requires, even while it fails aesthetically.
The issue is not success, but something more like aesthetic distinctness:
Does the realization possess features that are not merely parasitic on the
underlying work? The reason for demanding aesthetic distinctness is one of
parsimony: When everything aesthetically significant about the realization
can be referred back to the underlying work, or to a previous realization,
nothing is gained by invoking another artwork.
13 When a realization is derivative in the very strong sense I mean here, its structure is derived almost
exclusively from the underlying work itself or from another realization. This is not just a matter of
stylistic similarity.
250 SHERRI IRVIN
Second, aesthetic significance is not a matter of appearances alone.

Usually, a photograph of an artwork is not itself an artwork; it is either a
snapshot or a reproduction. However, under some conditions, something
that looks like a reproduction of another work is an artwork in its own right:
Sherrie Levine’s photographic appropriations of other artists’ works are a
well-known example. The context and manner of presentation, as well as
the discursive framework, supply content that mere reproductions lack, and
this distinguishes Levine’s works aesthetically.
It might be possible, then, to produce a ‘straight’ performance, yet do so
in a way, or in a context, or with supplemental communication that makes
for aesthetic distinctness. This does not, however, show that all, many or
most straight performances are artworks. Not all photographic reproduc-
tions became artworks when Sherrie Levine’s photographs of other artists’
works became artworks. The circumstances in which this happened were
rare and special, and something similar is true regarding the aesthetic
significance of straight performances.14
Here is another way a performance might fail to be an artwork:
(D) Although the work significantly underdetermines the features
of the realizations (1), the performers fail to make autonomous
aesthetic decisions (3).
In most such cases, 2 will also be unsatisfied: Those who make no attempt at
artistry rarely produce anything aesthetically significant. But suppose that, by
luck or chance, the resulting performance is aesthetically distinct (so that 1 and
2 are satisfied, but 3 is not). We should deny that it is an artwork, as we would
deny that an aesthetically valuable object produced by natural phenomena is. It
is thus possible that only one of two indiscernible performances is an artwork,
since it is a product of autonomous aesthetic decisions whereas the other isn’t.
Another way in which criterion 3 could go unsatisfied, with regard to
either installation artworks or musical compositions, is this: The person
creating the realization could make autonomous aesthetic decisions that are
inappropriate, since the underlying work does not allow for such decisions in
the construction of the realization.15 If a curator expresses her own aesthetic
14 For discussion of a similar point in relation to the work of Louise Lawler, see Irvan (2012).
15 Note that mistakes (such as wrong notes in a performance) do not fall into this category, since they
are not the product of autonomous aesthetic decision.
values by constructing a display that deviates markedly from the artist’s

parameters, we probably do not have a realization of that artist’s work at
all. We might have another work constituted of the physical stuff normally
used to realize the original artist’s work; but that is another matter.16 Things
are similar in musical performance: If the performer clearly operates outside
the scope of what is permitted by the composition, we may have a distinct
work, but not one that is both a realization of the original composition and a
new work in its own right.
I have presented three criteria for a realization to be a work in its own
right. There is a well-established practice of treating some distinctive
performances or performative interpretations as works in their own right,
and I believe the criteria do a good job of capturing the central elements that
tend to figure in such decisions. This is not to deny that there may be
outliers that the criteria do not account for. Deriving such criteria is a matter
of identifying the curve the runs among the scattered data points of practice.
But are the criteria relevant to installation art? Does critical practice ever
issue the verdict that a display is an artwork in its own right? A first point is
that as art forms evolve, it makes sense to look for established principles that
can be used as norms in new kinds of case. These criteria, descriptive of
practice in relation to the performing arts, can sensibly be used to warrant
and assess evolving critical practice in relation to installation art.
Second, there are cases in which critical practice does seem to recognize
an underlying installation artwork and a set of displays that are treated as
artworks in their own right. Consider Rachel Harrison’s Indigenous Parts,
whose five displays have differed dramatically. Certain sculptural and video
elements are included in every display of the work; other site-specific
elements are amassed by the installation team at each venue. Iwona Blaz-
wick, who curated the work at Whitechapel, describes the situation thus:
Rachel Harrison’s sculptural constellation Indigenous Parts was first installed in a
temporary gallery space in downtown New York in 1995. Since then it has
migrated to numerous locations . . . absorbing and shedding indigenous fragments
at each venue . . . In this way one work becomes many, as Indigenous Parts is iterated
according to the specifics of each context. (Blazwick 2010, pp. 101–3)
16 This is still another reason to reject the view of installation artworks as four-dimensional concreta in
space-time: not everything that happens to the stuff the artist made (if such there is) belongs to the
artwork proper.
252 SHERRI IRVIN
Blazwick’s description of Indigenous Parts itself as something that has ‘mi-

grated,’ and her observation that ‘one work becomes many,’ lends credence
to the idea that each display realizes an underlying work while counting as a
work in its own right. The nomenclature and dating system that has emerged
for the work supports this view; the displays have come to be known as, for
example, Indigenous Parts V, 1995–2010, and Indigenous Parts IV, 1995–2009.
I do not suggest that critical practice in this area is firmly fixed, that there
is no alternative way to understand Indigenous Parts,17 or that there are no
instances that conflict with the picture I have laid out. I do think, however,
that the criteria I have offered are well established in relation to the
performing arts and can sensibly be—and in some instances already have
been—extended to installation art.
IV. Artworks, Instances, and Ontology

I have suggested that installation artworks and their realizations are distinct,
and that installation artworks, like artworks for performance, centrally
involve parameters for the creation of realizations. I have not pronounced
on the nature of these artworks, or on their relationship, identity or
otherwise, to the parameters. But I see no viable way around the idea that
these works are universal or abstract entities of some sort: Their realizations
are instances, or at least occurrences, of them.
The most compelling objection to the claim that artworks are abstract is
that artworks are temporal and, often, temporally flexible, subject to cre-
ation as well as (in many cases) change over time and destruction, whereas
abstract entities are typically conceived as atemporal and unable to enter into
causal relations.18 In the ontology of music, some theorists have attempted
to avoid this difficulty by adopting perdurantism or endurantism about
musical works. Perdurantism, as defended by Ben Caplan and Carl Math-
eson (2006, 2008), is the view that the musical work is the fusion of its
performances. Endurantism, as defended by Chris Tillman (2011), is the
17 The obvious alternative is to think there are five distinct but interrelated works that, though they
share common elements, do not realize any common underlying work.
18 I am less concerned about modal flexibility, as discussed in (e.g.) Rohrbaugh (2003); I’m willing to
say, if necessary, that if the symphony had had one different note, it would have been a (subtly) different
symphony, rather than the same symphony with (subtly) different features.
view that the work occupies all the space-time regions occupied by its atoms
(which may include performances as well as scores, recordings, and perhaps
even memories), but without being identical either to any particular atom
or to the fusion thereof. Instead, the work is ‘wholly located’ wherever one
of its atoms is, just as I am wholly located wherever my current time slice is.
How should we adjudicate among competing proposals about the nature
of the artwork? As David Davies (2004, ch. 1) and Amie Thomasson (2004)
have argued, our aim in identifying the artwork should be to identify the
entity that is relevant to our critical practice: That is, the entity that, to the
greatest degree possible, warrants or makes true the appropriate claims we
make about artworks in our practices of appreciation. The fusion of a work’s
performances is not the right sort of thing to satisfy this role. When we
critically appreciate a particular musical work, our aim is to assess the entity
that the composer has offered to us and that manifests the composer’s
achievement.19 But the fusion of performances may be deeply misleading
in this regard. If the work is performed only once, then on Caplan and
Matheson’s view, all there is to say about the work is what there is to say
about that performance. But this is incorrect: A single performance, espe-
cially a mediocre one, may reveal little of the underlying work’s potentiality
and brilliance. And given that, pace Nelson Goodman (1968), there can be
incorrect performances, a single performance may have features that actively
distort our understanding of the work.
Tillman’s endurantism fares a bit better, in part because Tillman admits
the possibility that scores, as well as performances, might be atoms of the
work.20 Attention to the score at least provides us access to aspects of the
work that performances might not reveal. But the work is not fully revealed
through any of its performances, so to say that it is wholly located where a
given performance is located is uninformative from a critical perspective. To
appreciate the work, it would be a mistake to focus exclusively on a
particular performance, or even on the collection of all actual performances.
Performances contain elements that do not belong to the work; and there
may be aspects of the work that are never revealed through performance.
Moreover, some explanation of why a particular performance counts as a
19 For argument, see Currie (1989, pp. 36–40) and Davies (2004, pp. 52ff).
20 An atom, in Tillman’s view, is an appropriate object of critical attention and, while not identical to
the work, can be seen as a stand-in for it.
254 SHERRI IRVIN
performance of the work is needed; and this explanation must appeal to the
performance’s satisfaction of the relevant parameters or norms.
Consider, as a simpler case, the philosophical work you are now reading.
In a sense, it is wholly located wherever one of its atoms (printings, digital
files) is: if the notation is correct, nothing is missing. Tillman’s endurantist
proposal thus does better by philosophy papers than it does by musical
works. But even here, there are complications. Even if every atom were
in the same font, the font would not belong to the paper itself.21 We don’t
get confused about this; implicit knowledge of the norms of this form of
discourse informs us that font is incidental. But this information cannot be
had simply by consulting atoms of the work.
Allan Hazlett (this volume, Ch. 7) offers another interesting proposal:
Perhaps a work like Magor’s Production is a token event including Magor’s
creation of the bricks and provision of instructions, and the bricks’ being
installed on various occasions. Hazlett’s proposal differs from perdurantism
and endurantism in that it does not construe the work as having atoms, but
instead treats Magor’s creation of the bricks and instructions, along with the
various displays, as constituting a single event extended over a period of years.
Objections that Tillman (2011) raises against perdurantism seem to apply
to Hazlett’s proposal as well: It makes the work out to be something that
may not be complete until long after the artist dies, and it makes it difficult
or impossible for a particular viewer to perceive the whole work. Perhaps
these consequences are not as unpalatable for installation artworks like
Magor’s as they are for musical works, though: We might think that
Magor has enlisted the museum as an agent in the completion of her
work, and a consequence is that the work evolves over time in a way that
does, in fact, limit the accessibility of its whole four-dimensional structure to
any given viewer. Nonetheless, in my view an account like Hazlett’s is not
consistent with appropriate critical practice regarding the work. The actual
displays may or may not do justice to the work itself; they may or may not
reveal the full potential of this set of objects and this set of instructions; they
may or may not comply fully with the instructions. There is an appropriate
object for critical attention here that is distinct from the particular installations:
21 I am not claiming that the font can never be integral to a work. Mark Danielewski’s 2000 novel
House of Leaves does have font and text configuration as integral to some of its passages. The point is that
the font will be integral in some cases and not in others, and atoms of the work, considered collectively or
separately, are not sufficient to determine which holds for a particular work.
It is the entity that Magor created, which centrally involves a set of instruc-
tions for installation. The particular displays count as such partly by virtue
of the fact that they sufficiently comply with said instructions; the work
thus has a normative element that is not captured by identifying it as
an event.22
The ontological category I prefer for artworks is along the lines of the quasi-
abstract entity discussed by Barry Smith (2008). What kind of thing, Smith
asks, is a game of chess that players play without a board, simply by speaking the
descriptions of moves to each other and holding the state of the imaginary
board in memory? The chess game is not a thought or collection of thoughts;
those are representations of the game rather than the game itself. There doesn’t
seem to be any collection of stuff that could plausibly be identified with the
game; it is, to that extent, abstract. However, it also exists in time: It came into
being at a particular moment and will end at a subsequent moment. Moreover,
it can have realizations or occurrences: Someone listening to the players’ verbal
exchange could move pieces about on an actual chessboard to realize their
game, either during the conversation or later.
Quasi-abstract entities are abstract in two senses: In the sense of not being
concrete—not being physical objects or events—and often also in the sense
of being susceptible of instantiation. These two senses of abstraction are
linked. A particular concrete object or event cannot have (other) instances;
there can only be other concrete objects or events, or representations, that
resemble it in various ways. The normative, and thus non-concrete, ele-
ments of a quasi-abstract entity are precisely what allow it to have instances,
namely those objects or events that satisfy the norms.
Is it ontologically promiscuous to suggest that there is a quasi-abstract
entity, a game, that came into existence when two people interacted under
the right circumstances? Must one who holds such a view think that there
are magical processes by which—poof !—strange new things come into
existence and float around in a mysterious ontological realm through
which they engage in mysterious causal interactions with real (in our
realm) objects and events? I don’t think so. Amie Thomasson describes
22 I object on similar grounds to the view of installation artworks as concrete four-dimensional space-
time worms. In addition, many complex sculptural and installation artworks (e.g. the candy spills of Felix
Gonzalez-Torres and Jana Sterbak’s 1987 Vanitas: Flesh Dress for an Albino Anorectic) are physically
discontinuous; there is little or no physical material that survives from one display to another, though
(according to standard critical practice) the existence of these works is not discontinuous.
256 SHERRI IRVIN
what I take to be a promising deflationary maneuver. The existence of an

abstract social entity, she says, is just a matter of the fulfillment of the
relevant conditions. ‘[I]f one grasps the concept of a recession and knows
that the relevant conditions are sometimes fulfilled, it makes little sense to
ask whether there really are recessions’ (Thomasson 2003, pp. 288–9).23 The
existence of the entity is settled by the fulfillment of the conditions that are
criterial for that entity; no further question remains to be asked.24
Thomasson’s account helps to explain how quasi-abstract entities can
participate in causal relations, as is required for their temporality. A recession
comes into existence when certain conditions are fulfilled and ceases to exist
when those conditions are no longer fulfilled. The game of chess came into
existence when a set of norms was expressed, and it simultaneously became
possible for particular games satisfying these norms to be initiated. The
fulfillment of the relevant conditions for the existence of these entities is a
function of concrete events whose participation in causal relations is un-
controversial.
To apply this picture to artworks, we can say that an artwork comes into
existence when an artist engages in activities of communication and/or
fabrication in the right sort of context (given a background of artworld
conventions). These activities, sometimes along with physical stuff that
eventuates from them, fulfill the conditions for the artwork’s existence.
The artwork can change over time through further acts of communication
and/or fabrication. Finally, given a background of conventions for under-
standing artworks, the work can cause reactions in its audience, including
attempts at appreciation and interpretation.
23 Searle (1995, 2010) offers a related account of social facts. As Searle would acknowledge, the ability
of a specific interaction to give rise to a social fact depends on an extensive background of social
conventions and institutions. These, too, exist by virtue of a complex of actions that can ultimately be
explained in terms of ontologically unremarkable episodes of physical particles moving around this way
and that—or, if they can’t, then we will need a more exotic fundamental ontology to make sense of the
goings on in our world.
24 Ross Cameron (2008), in similar deflationary spirit, points out that the truth makers of claims about
abstract entities are perfectly ordinary: They are commonplace events and states of affairs that belong to
or depend on the ordinary realm of physical particles that move around this way and that. To say that
there is a recession, then, is not to commit oneself—poof!—to the coming into existence of a new
nugget in some special ontological realm. Cameron (this volume, Ch. 8) suggests that this sort of account
is compatible with nominalism: ‘What is important for the nominalist is that a world of concreta suffices
to ground all truths.’ If that is correct, then so much the better.
Opponents of abstract or quasi-abstract entities may claim that the effects

I attribute to artworks are due not to these entities themselves, but to
something more concrete. To respond fully to this suggestion would take
us far afield. For the present, I will simply say that our descriptions of our
social world would be dramatically impoverished, and also greatly compli-
cated, by eliminating talk of recessions and chess games in favor of talk of the
multiplicity of more concrete events that fulfill the conditions for existence
of these entities. Many phenomena of our social world would be much
more difficult to understand and explain if we attempted to appeal only to
concrete events and objects existing at some ontologically fundamental
level. One might propose to explain such phenomena by appealing not to
recessions, chess games, and artworks, but to beliefs about such entities
(with the entities themselves being construed as non-existent). I have
trouble seeing the appeal of this sort of story. As Thomasson asks, if the
conditions for the existence of a recession are fulfilled (and, we might add, if
recessions figure in viable causal explanations and accurate predictions of
how things go), why be a fictionalist about the recession? Why attribute
error to everyone who believes and acts as though recessions, and other
quasi-abstract entities, exist? In any event, I take it that any adequate
account of our social world will require something like an ontological
category of quasi-abstract entities, and that category will be seen to have
many members; appealing to them is not an ad hoc maneuver to solve a
narrow or specific problem related to art.
To say that artworks are quasi-abstract is not yet to say much about their
nature; artworks, chess games, and recessions are, intuitively, quite different
from each other. So, what more can we say about artworks as quasi-abstract
entities? Jerrold Levinson (1980) suggests that a musical work is a structure
of sound and instrumentation that has been indicated by a composer.25
While this account may be correct for the instrumental works in the classical
tradition to which Levinson (quite reasonably) restricts his attention, it will
not quite do for musical works more generally. A composer may or may
not indicate a particular sound structure; in John Cage’s Imaginary Landscape
No. 4, what is indicated is not a sound structure, but a form of activity in
25 For the application of such an account to other art forms, see Levinson (this volume, Ch. 2). It
would be consistent with Levinson’s position to see an installation artwork as a set of indicated
parameters for the construction of displays.
258 SHERRI IRVIN
which twenty-four performers, guided by a conductor, control the dials on

twelve radios. A jazz composition may include passages in which it is
specified that the performers should improvise; and for a performer to
‘improvise’ in the same way each time would be a violation of the norms
for performing most such works. It is thus mandatory that the sound
structures realized in particular performances of the work vary. Some instal-
lation artworks, like Magor’s Production, are analogous to improvisatory jazz
works in this respect.
We have a couple of options in the face of this situation. One is to say that
some musical works are indicated sound-and-instrumentation structures, some
(like Imaginary Landscape No. 4) are indicated activity-and-instrumentation
structures, some are simply indicated activity structures, and some (purely
improvisatory performances) are particular, non-repeatable events. Which is
the case and depends on what the artist has specified.26
Another option is to say something more general about what all of these
have in common. In each case, the artist has sanctioned a set of norms for
the creation of realizations; and a realization will be one that satisfies those
norms, whatever they may be (and, typically, is causally connected to them
in an appropriate way).27 Thus, perhaps the artwork in many cases, or even
in every case, just is a set of norms or instructions.
Prima facie, the artwork-as-instructions account seems more palatable for
some art forms than others: Most musical works may be amenable to analysis
in terms of norms or instructions for producing a sound event, while
paintings and sculptures, though they have normative aspects (including
proper configuration), seem to be chiefly physical entities. Installation
artworks lie at various points in between: Some are more like complex
sculptures, others involve some particular physical stuff but without a fixed
configuration, and still others lack any physical substrate that persists from
one exhibition to another.
It is possible to analyze all of these works as sets of norms: It is just that in
the painting case the norms specify that this very canvas should be displayed in
a particular way, while in installation art cases of the latter sort the norms
26 I offer an analogous account for visual artworks in Irvin (2008).

27 It is important to note that not all aspects of the instructions or norms for creating realizations of a
work are directly determined by the artist; some grow out of the general musical or artistic culture of the
time, and function as defaults even if not expressly invoked by the artist. A painting is to be hung with its
representational content right-side up, unless the artist (Georg Baselitz, say) specifies otherwise.
simply specify that, say, a pile of hard candies in brightly colored wrappers
should be dumped in a heap in a corner of the gallery, or that the words A
WALL BUILT TO FACE THE LAND AND FACE THE WATER AT
THE LEVEL OF THE SEA should be inscribed some way or other so as to
be visible to the viewer.28 What the artist has done in each case is to supply
instructions or norms for creating a display; it’s just that in the painting case,
the act of articulating the instructions is inseparable from the act of creating a
canvas with particular features, since the most important instruction is that
this canvas be displayed.
I’m sympathetic to complaints about this sort of account. The suggestion
that works of painting are really sets of norms is radically at odds with the
commonsense idea that paintings are fundamentally physical entities. Be-
cause the norms associated with most paintings (‘hang it with the painted
surface facing away from the wall and the representational content shown
right-side up’) are so heavily convention bound that we tend to overlook
them, it’s common not to notice that painting has a normative element at
all; and it’s not clear that the appreciation of painting suffers when most
people’s grasp of its normative element remains implicit. Even when the
normative element is noticed, the physical aspect still seems to retain
primacy; the temptation, when we consider painting or sculpture in isol-
ation from other cases, is to think the artwork is a physical entity with some
normative features tacked on as an auxiliary.
For visual artworks, especially paintings, sculptures, and installation art-
works that are bound to particular physical material, an option is to say that
the work is a hybrid of physical and normative elements, with the physical
elements having primacy in some cases and the normative elements in
others. Notice, though, that this will be a difficult row to hoe when it
comes to works that do not involve any particular physical stuff: The display
may be a physical entity, but the underlying work is not bound to any
particular physical material, so it’s hard to understand in what sense it is even
a hybrid physical entity.29 And, of course, cashing musical works out as
physical/normative hybrids does not seem feasible.
28 These examples are drawn, respectively, from the candy spills of Felix Gonzalez-Torres and a 2008
work by Lawrence Weiner that is in the collection of the Miami Art Museum.
29 Consider, for instance, Tino Sehgal’s 2002 performance artwork This is propaganda. Sehgal supplied
only a verbal description of the work, with no supporting documentation, when Tate Modern acquired
the work. Tate employees made no notes about the work during meetings with the artist, though they
260 SHERRI IRVIN
Ultimately, we face a dilemma: We can have either a neat, unified

ontology that treats artworks as sets of norms or instructions, or a heteroge-
neous ontology that better captures our intuitions about art forms that have
a stronger connection to particular material stuff. According to this hetero-
geneous ontology, some works (including many musical works as well as
the most dematerialized installation artworks) are sets of instructions, while
others are physical/normative hybrids. I won’t argue here for one of these
options over the other;30 either way, the notion of quasi-abstract entities, or
something like it, supplies resources that are needed to account for the
work’s normative dimension. And, to reiterate what I’ve said above, the
objection that there are no such things doesn’t have much traction from my
perspective. We need to invoke things like instructions, parameters, norms,
and laws to make sense of many aspects of our world. So, whatever
ontological maneuver is required to allow for the existence of these things,
I’ll avail myself of it for the sake of artworks as well.
V. Conclusion
I have argued that the creation of both installation artworks and artworks for
performance centrally involves the expression of parameters for realizations.
Grasping the underlying works in both art forms, then, is a matter of
grasping the parameters.
I offered an account of the criteria that determine whether a particular
realization is an artwork. These criteria allow that some performances may
be independent artworks, while displays of installation artworks typically are
not. This is largely due to contingent facts about the respective art forms:
Installation artworks are often designed to provide a minimum of latitude;
and even when latitude is permitted, the installers are expected to defer to
the artist’s aesthetic values. In the performing arts, the conventions are quite
different: The underlying work’s parameters typically vastly underdetermine
were permitted to make personal notes afterward; the latter may not be printed for circulation (van
Saaze, 2011). Though critical practice is clear that the work persists between performances, no physical
concretum is plausibly identified with it.
30 In Irvin 2008, I come out for the heterogeneous ontology.
the features of performances, and the performers see themselves, quite

appropriately, as making significant aesthetic decisions.
It is possible, of course, for realizations in either art form to cross these
boundaries. A work for performance could involve rigid parameters, such
that little aesthetic latitude is available to the performers and no independent
artwork results. And an installation artwork could enlist the installers in
creating a new artwork by demanding autonomous aesthetic decision. For
this reason, it is sensible to apply the same criteria to the two art forms,
though the verdicts about actual works within the two forms will diverge
rather systematically.
To appreciate such works appropriately involves seeing them as quasi-
abstract entities, susceptible of having instances but also capable of being
created and, in many cases, changing over time and being destroyed. Given
the availability of a Thomasson-style deflationary maneuver, I don’t see that
such entities should trouble us, ontologically speaking. And since quasi-
abstract entities abound in many domains of social living, there is nothing
ad hoc about invoking them to account for artistic phenomena.
References
Blazwick, Iwona (2010) ‘Mystery, Ancient and Modern: Indigenous Parts’ Exhib-
ition catalog, Rachel Harrison: Museum with Walls (Annandale-on-Hudson: Bard
College Center for Curatorial Studies; London: Whitechapel Gallery; Frankfurt
am Main: Portikus), pp. 100–11
Cameron, Ross (2008) ‘There are No Things That are Musical Works’ British
—— (2012) ‘How to be a Nominalist and a Fictional Realist,’ this volume, Ch. 8
Caplan, Ben and Matheson, Carl (2006) ‘Defending Musical Perdurantism’ British
—— (2011) Philosophy of the Performing Arts (Malden, MA: Wiley-Blackwell)
Goodman, Nelson (1968) Languages of Art (Indianapolis, IN: Bobbs-Merrill)
Hamilton, James (2007) The Art of Theater (Malden, MA: Blackwell)
Hazlett, Allan (2012) ‘Against Repeatable Artworks,’ this volume, Ch. 7
262 SHERRI IRVIN
Irvin, Sherri (2005) ‘The Artist’s Sanction in Contemporary Art’ Journal of Aesthetics
—— (2006) ‘Museums and the Shaping of Contemporary Artworks’ Museum
Management and Curatorship 21: 143–56
—— (2008) ‘The Ontological Diversity of Visual Artworks’ in Kathleen Stock and
Katherine Thomson-Jones (eds) New Waves in Aesthetics (New York: Palgrave
Macmillan), pp. 1–19
—— (2009) ‘Theatrical Performances and the Works Performed’ Journal of Aes-
thetic Education 43: 37–50
—— (2012) ‘Artwork and Document in the Photography of Louise Lawler’ Journal
of Aesthetics and Art Criticism 70:79–90
Laurenson, Pip (2006) ‘Authenticity, Change and Loss in the Conservation of
Time-Based Media Installations’ Tate Papers 6: n.p. <http://www.tate.org.uk/
research/tateresearch/tatepapers/06autumn/laurenson.htm>
—— (2012) ‘Indication, Abstraction, and Individuation,’ this volume, Ch. 2
Mag Uidhir, Christy (2009) ‘Unlimited Additions to Limited Editions’ Contempor-
ary Aesthetics 7: n.p.
—— (2012) ‘Photographic Art: An Ontology Fit to Print’ Journal of Aesthetics and
Art Criticism 70: 31–42
Searle, John (1995) The Construction of Social Reality (New York: Free Press)
—— (2010) Making the Social World: The Structure of Human Civilization (Oxford:
Oxford University Press)
Smith, Barry (2008) ‘Searle and De Soto: The New Ontology of the Social World’
in Barry Smith, David Mark and Isaac Ehrlich (eds) The Mystery of Capital and the
Construction of Social Reality (Chicago: Open Court), pp. 35–51
Thomasson, Amie L. (2003) ‘Foundations for a Social Ontology’ Protosociology
18–19: 269–90
—— (2004) ‘The Ontology of Art’ in Peter Kivy (ed) The Blackwell Guide to
Aesthetics (Malden, MA: Blackwell), pp. 78–92
Tillman, Chris (2011) ‘Musical Materialism’ British Journal of Aesthetics 51: 13–29
van Saaze, Vivian (2011) ‘Image Not Available: What Remains of Tino Sehgal’s
Living Sculptures and Constructed Situations?’ Presented at ‘New Strategies in
the Conservation of Contemporary Art’ workshop, Amsterdam
12
What Type of ‘Type’ is a Film?
DAVID DAVIES*
I.
What kind of thing is a film? What kind of thing is a photograph? Philoso-
phers pondering such matters have generally been moved by the following
sorts of considerations. To appreciate an artwork requires at least an experi-
ential engagement of some kind with an instance of that work, where the
latter is an entity that makes manifest to the receiver some or all of the
properties bearing upon its appreciation.1 In the case of a painting, we
require an experiential encounter with a particular physical object usually
located in a particular gallery or museum. This makes it plausible to identify
the work itself with that object or with some set of properties of that object.
With films and photographs, however, there seem to be many different
locations where, at a given time, we might experientially encounter an
instance of a work in the manner necessary for its appreciation. You may be
watching Citizen Kane, or looking at Stieglitz’s The Steerage, in Chicago at
the same moment that I am watching the same film, or viewing the same
photograph in London. In this respect, films and photographs seem to
* I wish to thank Christy Mag Uidhir for helpful comments on an earlier draft of this chapter. I also
wish to thank the Social Sciences and Humanities Research Council of Canada, whose support
facilitated the writing of this chapter.
1 ‘Some or all’ depending upon one’s views in the epistemology of art. For ‘aesthetic empiricists,’ all
properties bearing upon a work’s appreciation will be manifest in such an experiential engagement
unmediated by knowledge of a work’s provenance, save perhaps knowledge of the artistic category to
which it belongs, whereas for ‘contextualists’ access to the properties manifest in such an engagement is
necessary but not sufficient for proper appreciation of a work. On aesthetic empiricism and contextual-
ism, see Currie (1989, ch. 2) and Davies (2004, chs 2 and 3).
264 DAVID DAVIES
resemble musical and literary works. Appreciating a musical work requires

hearing it performed, and appreciating a literary work requires reading it,
but people in different locations engaging with different objects or events
can simultaneously have the necessary experiential encounter with a given
literary or musical work.
I said that an instance of a work is an entity that ‘makes manifest to the
receiver some or all of the properties bearing upon its appreciation.’ ‘In-
stances,’ so conceived, are defined by the particular role that they play in
appreciation. We may talk here of an entity playing ‘the experiential role’ in
the appreciation of an artwork. But entities can be more or less well fitted to
play this role. It seems natural, for example, to say that some performances of a
musical work fail to contain all and only the notes specified in the score, some
screenings of a film involve damaged prints, and some copies of a novel contain
typographical errors. This suggests that we should talk in a stricter sense not
merely of entities ‘qualified’ to play the experiential role for a work, but of
entities that ‘fully qualify’ in virtue of not being flawed in these sorts of ways.
Something is fully qualified to play the experiential role in the appreciation of a
given artwork X at a time t just in case it possesses at t all those experienceable
properties that are necessary, according to the practices of the art form in
question, to fully play this role. Where something fully qualifies in this sense
at a time t, we may term it a ‘strict instance’ of X at t. To say, then, that different
events or objects may fully qualify at specific times to play the experiential role
for a given artwork is to say that the work admits of a plurality of strict
instances.2 It is in allowing for a plurality of strict instances that so- called
‘multiple’ artworks or art forms differ from so-called ‘singular’ works or forms.
Classical music, film, literature, and cast sculpture, as suggested above, are
treated as multiple in this sense. Painting, on the other hand, is treated as
singular because, even though prints and copies may to some extent play the
experiential role in the appreciation of a painting, we treat only one entity as a
strict instance of the work. Only the original canvas, it is assumed, can possess
all of the properties necessary to fully play the experiential role in the appreci-
ation of the work: Only the original canvas is a strict instance of the work.
2 Obviously, since the experienceable properties of an object usually change over time, something
may be a strict instance of a work X at certain times but fail to be a strict instance of X at other times. To
say that a work X whose instances are objects admits of a plurality of strict instances is thus an abbreviated
way of saying that different objects may, at some time in their existence, fully qualify to play the
experiential role for X.
WHAT TYPE OF ‘TYPE’ IS A FILM? 265
II.
Philosophical accounts of multiple artworks must explain what is often
termed their ‘repeatability,’ the defining fact that they can have distinct strict
instances. As Richard Wollheim (1980, pp. 75ff.) poses this question, in what
ways can an entity be ‘generic’ in the sense that it has a potential plurality of
‘elements’ falling under it? The members of a set, certainly, can be seen as
falling under the set, but sets are individuated extensionally: Any difference
in membership entails a difference in set. The Steerage, Citizen Kane, and
Bleak House, however, could surely have had more or fewer strict instances
than they actually have. Such reflections help to motivate the idea that
multiple artworks are types and their instances are tokens of those types.
The type/token distinction is familiar in everyday contexts. For example, if
I inscribe the letters ‘s-h-e-e-p’ three times on a clean blackboard and ask
how many words are on the blackboard, you can correctly answer both ‘3’
(there are three occurrences of word-tokens) and ‘1’ (there are occurrences of
only one word-type). The identity and nature of a type does not change as the
number of its tokens changes. In this respect, the type-token relation mirrors
the relation between a multiple artwork and its strict instances.
However, so some philosophers have claimed, if multiple artworks stand
to their strict instances in the relation of types to tokens, they are types of a
special sort. For they are generally taken to admit not only of strict instances
but also of other entities intimately related to them that play the ‘experien-
tial role’ with respect to these works while falling short of the requirements
for being strict instances. As noted above, we treat as performances of a
musical work not only musical events that meet all the requirements for
right performance, but also musical events containing at least some incorrect
notes. Similarly, a damaged print of Renoir’s La règle du jeu can still provide
an audience with a screening of the film, albeit a flawed one. Types or kinds
that admit in this way of both correct and incorrect tokens have been
termed ‘norm-kinds’ (Wolterstorff 1975, 1980) or ‘norm-types’ (Dodd
2007). Descriptive types, it is claimed, are individuated by the condition
that must be met by their tokens, norm-types by the condition that must be
met by their correct or properly formed tokens.
I said that imperfect performances of musical works and screenings of
films are nonetheless treated as being intimately related to the works in
266 DAVID DAVIES
question, as no less of the work than their strict counterparts. This contrasts
with how we customarily treat copies of paintings or works of carved
sculpture. With such singular artworks, copies, however faithful, are treated
as extrinsic to the work, the latter being taken to have a single strict instance
and no other entities that are ‘of ’ it in the sense that flawed performances or
screenings are taken to be ‘of ’ particular multiple works. This asymmetry in
treatment is noteworthy because it prescinds from the purely epistemological
considerations in terms of which we introduced the notions of ‘instance’ and
‘strict instance.’ A flawed print of a film may be less well suited than a very
good period copy of a painting to play the experiential role for their
respective works, for example. Yet the former is treated as being ‘of ’ the
work in the same sense as its strict instances, while the latter is not.
This aspect of our practice is symptomatic of what I have elsewhere
described (2010) as an equivocity in the (philosopher’s) notion of ‘instance.’
It is because flawed performances and screenings stand in a particular kind of
relation to the provenance of a work that they are treated as being ‘of ’ that
work in an intimate sense that is denied to copies of paintings because the
latter do not stand in this relation to provenance. Entities that stand in this
kind of provenential relation to a work may be termed its ‘provenential
instances,’ or ‘p-instances.’ Those things that are fully or partly qualified to
play the experiential role in the appreciation of a work, however, qualify as
instances on purely epistemic grounds—they are what we may term a
work’s ‘epistemic instances,’ or ‘e-instances.’ A work’s p-instances will
normally be among its e-instances, strict or flawed.3 But there is no prin-
cipled reason why a work, whether multiple or singular, cannot have
e-instances—and indeed strict e-instances—that lack the necessary causal
history to be among its p-instances.4 A text that is a strict p-instance of a
literary work, for example, may also be a strict e-instance of another work,
generated in a very different art-historical context, with which it shares its
text. We need not pursue here the implications of such considerations for
3 Normally, but not necessarily. A p-instance of a work of cast sculpture which has lost all of its detail
through erosion, for example, may no longer have any capacity to play the experiential role in the
appreciation of that work.
4 See my 2010 for an extended argument for this conclusion. My use of e-instance in this chapter is
broader than in my 2010 paper in that I here allow for both strict and flawed e-instances. This is a
terminological rather than a substantial difference however, and reflects the different concerns of the two
papers.
our customary understanding of the multiple/singular distinction as it

applies to artworks and art forms. But the distinction between p- and
e-instances will play a key role later when we look at how some multiple
works can be repeatable without being types in the standard sense.
While many find attractive the idea that multiple artworks are norm-
types of which the work’s strict and flawed instances are well- and improp-
erly formed tokens, this view is not without its problems. Types, we have
said, are generic entities that can have elements that fall under them, but
what are such things? Most philosophers of art who have written on this
subject take types to be abstract objects that do not have determinate spatial
locations. It is therefore questionable whether they can enter into causal
relationships with things that do. This raises concerns about our ability to
epistemically access and refer to types if we take knowledge and reference to
require some kind of causal engagement with what is known or referred to.
Some philosophers, however, argue either that abstract objects can enter
into causal relations (Burgess and Rosen 1997, pp. 24ff.; Dodd 2007,
pp. 13–14) or that knowledge and reference do not require such relations
with their objects (e.g. Brown 1991).
A more serious problem arises if we accept an account of the metaphysics
of types defended by Julian Dodd (2007, chs 2–3). Types, Dodd argues, are
unstructured, temporally and modally inflexible, entities. They are indi-
viduated according to the conditions that must be satisfied by their tokens,
or, in the case of norm-types, by their well-formed tokens. These condi-
tions generally relate to intrinsic properties of these tokens and can be
characterized as correlated properties associated with the types—the prop-
erty of being instanced by strings of the form ‘s-h-e-e-p’ associated with the
word-type ‘sheep,’ for example (Dodd 2007; Wolterstorff 1980). Further-
more, he maintains, a type exists not, as some have thought (Levinson 1980;
Wolterstorff 1975), just in case it is possible for it to have tokens, but, rather,
just in case its property-associate exists. And, he further maintains, proper-
ties cannot be brought into or go out of existence—they exist eternally5
if they exist at all. It follows that types, too, cannot be brought into or go out
of existence. So, if multiple artworks are types, they cannot be created, but
only discovered.
5 Dodd argues that the alternative option—that properties exist atemporally—raises additional prob-
lems about how things existing in time can possess such properties.
268 DAVID DAVIES
For some (Dodd 2007, ch. 5), this consequence of identifying multiple
artworks with types simply shows that we must revise our intuitions about
artworks in light of sober metaphysical reflection. If, however, we wish to
reconcile the idea that multiple artworks are types with their creatability,
one option is to appeal to the normative element in such works if we
construe them as norm-types. For example, it can be argued that, while
types of sound sequence may themselves exist eternally, musical creation is a
matter of making a particular type of sound-sequence normative for a work,
and this brings into being something that did not exist prior to the com-
poser’s creative activity—namely, the eternal sound-structure type made
normative for a work by an individual at a time. Musical works can then be
thought of as ‘initiated types’ created by their composers (Levinson 1980; for
related proposals, see Lamarque 2010, Wolterstorff 1975, and (in a more
sceptical vein) S. Davies 2003, p. 170). But, if works are types, whether
initiated or not, then, if Dodd is right about the individuation and existence
conditions for types, they must still be individuated in terms of a condition
for (correct) instantiation expressible as a property: The property of being
correctly instantiated only by performances that fully comply with the
performance specifications set out by B in context C and that stand in an
appropriate intentional-historical relation to that act of specification, for
example. But then this property, like every property, must exist eternally if
at all, so we have not shown that multiple works, as norm-types, are
creatable (see Dodd 2007, ch. 5).
But both Dodd’s conception of musical works as eternally existing norm-
types and Levinson’s conception of such works as ‘initiated’ types are highly
counterintuitive if extended to other kinds of multiple artworks, such as
films and photographs. For, so extended, each conception entails that the
audio-visual structure of a film and the visual structure of a photograph
preexists its ‘making’ and is discovered by its makers. To see why this is so
counterintuitive, we need to consider certain significant differences be-
tween multiple art forms, and, of more significance in the present context,
between film and photography, on the one hand, and music and literature
on the other.
I shall use ‘initiate’ as a neutral term that applies across the arts and carries
no commitment as to whether initiation is creation or discovery. As a
number of writers have noted, there are at least three ways in which
multiple artworks can be initiated (S. Davies 2003, pp. 159–63; Wollheim
1980, pp. 78–80; Wolterstorff 1980, pp. 90ff.). First, as in the case of literary
works, an artist may bring into existence an instance of the work which
serves as an exemplar. Further instances are then generated through emu-
lating the exemplar in those respects required by relevant artistic conven-
tions in place. Second, as in the case of cast sculpture, an artist may produce
an artefact that, when employed in prescribed ways, generates instances of
the work. This may be termed a ‘production-artefact’ (Wolterstorff 1980).
As in the case of exemplars, further conventions in place in the relevant art
form determine how this artefact must be used if a strict instance is to be
produced. Third, as in the case of classical musical works, an artist may
provide instructions that, if properly followed by those aware of the relevant
conventions and practices, result in strict instances of the work. In such
cases, the instructions call for interpretation, and instances of the work (strict
or flawed) are performances of it.
Photographs and films clearly fall into the second of these categories. An
analogue camera registers light rays entering through its lens as photochem-
ical changes on a strip of film, while a digital camera records, stores, and
digitizes the values of the light levels. While the trace produced photo-
chemically by a standard analogue camera is a negative, the trace produced
by a digital camera is a bitmap. In each case this trace serves as, or is
manipulated to produce, the production-artefact appropriate employment
of which generates those prints or digital images that are p-instances of the
work. Films, on the other hand, have as their production-artefacts those
entities used, in various cinematic media, to generate screenings of those
films—film prints, videotapes, digital files, etc., which either are, or stand in
a ‘copy’ relation to, master encodings of the film.
The inclination to see initiation as creation is strongest in the case of
multiple artworks whose initiation involves a production-artefact. Con-
sider, for example, Rodin’s The Thinker. Rodin conceived the idea for a
sculptural work having certain perceptible properties. He chose to physi-
cally realize the work not by working directly on a sculptable material
selected as the medium for the piece, but by working first on another
kind of physical material which was manipulated to produce a cast which
then, in its turn, was employed (1902) to produce a p-instance of the work.
A number of other p-instances were cast later. Since the work is multiple,
the sort of reasoning rehearsed above with respect to music seems to
commit us to thinking of the work as an eternally existing abstract object,
270 DAVID DAVIES
a type of which its p-instances are tokens. But, if we reason in this way,
then, surprisingly, the decision by Rodin to employ a more indirect method
for generating instances of his piece entails that, whereas a visually indistin-
guishable work of carved sculpture would presumably have been his cre-
ation,6 The Thinker is his discovery, albeit, as Dodd might insist, one that
involved great creativity. Counterfactual situations pose further difficulties.
Suppose Rodin intended to smash the cast immediately after the first
casting, and indeed did so. Because it was still in principle possible for the
work to have multiple strict p-instances—someone could have frustrated
Rodin’s intentions by preserving the cast and casting a second copy of the
sculpture—should we say that the work is multiple, and a discovery, even
though Rodin merely intended to produce a singular work by a more
roundabout causal process? Again, should we say that, once we move
from the daguerreotype to the multiply instantiable calotype, photographs
are no longer created by their makers but discovered, so that, while
Daguerre and Fox Talbot were producing photographic images at the
same time—images that in each case can be viewed as instantiating a type
of visual manifold—the former was creating his images whereas the latter
was discovering his? There is surely something deeply counterintuitive
about this. How could the decision to employ not one kind of causal-
mechanical process, which permits only a single p-instance of the work,
but another equally causal-mechanical process, which permits multiple
p-instances, carry such consequences for the status of the artist as creator?
There is a further problem (Rohrbaugh 2003). Strict p-instances of
multiple works generally admit of artistically relevant variations. For
example, there is no single way in which a given musical work sounds
when correctly performed. However, because of the role that sets of
instructions play in initiating a musical work, the Doddian-type theorist
can identify a condition that all correct performances must satisfy, a condi-
tion that defines the type that is the work. Photographic works also admit of
such variation, having correct instances (prints) that differ in such artistically
relevant properties as tonal contrast, print medium, and size. There is
6 This presupposes that paintings and works of carved sculpture are properly seen as particulars rather
than as types. As Dodd notes, it is possible to reject this presupposition. Another alternative would be to
view paintings and works of carved sculpture as ‘logically proper types’ that are capable of having only
one token. I am grateful to Juhani Yli-Vakkuri for pointing out these possible options for type-token
theorists, albeit ones that Dodd does not take up.
therefore no single visual array that is common to all a photographic work’s

p-instances. But, since photographs are initiated through a production-
artefact, we cannot appeal to instructions to identify common manifest
properties that define a Doddian type.
III.
One might respond by taking the creatability of multiple artworks such as
photographs as given. If types indeed exist eternally and cannot therefore be
created, this shows that such artworks cannot in fact be types. The challenge
then is to say how an artwork can be repeatable if it isn’t a type. Guy
Rohrbaugh (2003) identifies three features which, so he argues, are
common to multiple artworks like photographs and non-multiple artworks
like paintings: (a) modal flexibility—a work could have had different
intrinsic properties; (b) temporal flexibility—a work’s intrinsic properties
can change over time; and (c) temporality—a work comes into and goes out
of existence. Since types, as characterized by Dodd, lack all three of these
features, Rohrbaugh concludes that multiple artworks cannot be types.
What can possess all three of these features and also be repeatable, Rohr-
baugh maintains, are continuants—essentially historical individuals that
depend for their existence on those concrete entities that are their embodi-
ments: ‘Photographs are non-physical, historical individuals, continuants,
which stand in a relation of ontological dependence to a causally-connected
series of physical (sometimes mental) particulars’ (2003, p. 198). Among a
photograph’s embodiments are its occurrences which are distinguished from
other embodiments, such as the negative, in that they ‘display the qualities
of the work of art and are relevant to appreciation and criticism’ (ibid.). It is
in virtue of its capacity for multiple occurrences that a photograph is
repeatable. A similar account is proposed not only for films but also for
musical works.
Some have questioned Rohrbaugh’s account on the grounds that con-
tinuants themselves are somewhat strange beasts (see Dodd 2007, ch. 6).
More crucially in the present context, however, the argument for the
continuant status of photographs depends upon the assumption that photo-
graphs are modally and temporally flexible, and the further assumption that
272 DAVID DAVIES
these are properties that continuants, as historical individuals, but not types,
can possess. The claim that multiple artworks are creatable, then, is hostage
to the claim that they are modally and temporally flexible. But the latter
claim is far from obvious. Dodd, for example (2007, pp. 87–90), argues that
there are no good reasons to think of musical works as either modally or
temporally flexible. Any locutions that might tempt us to think otherwise
are, Dodd claims, most plausibly understood in terms of suitable para-
phrases. Where a work W* differs in its intrinsic properties, either modally
or through time, from a work W, or from a work W at a time t, then we
should view W* and W as distinct works, albeit distinct works that closely
resemble one another, he maintains. While this doesn’t by itself show that at
least some multiple artworks are modally and temporally inflexible, it raises
questions as to the force of Rohrbaugh’s claims to the contrary.
Indeed, Rohrbaugh’s argument for the modal and temporal flexibility of
photographs can be challenged on independent grounds. He begins with
paintings. The latter, he maintains, are physical objects which are them-
selves continuants. As physical objects, they retain their identity even when
they vary in what might intuitively (but incorrectly, according to Rohr-
baugh) be regard as intrinsic properties of paintings. Picasso’s Les Demoiselles
d’Avignon, for example, could have had a different appearance had Picasso so
chosen, and can change over time because the physical object identical to
the painting can itself change over time through natural processes of
material transformation or through damage. And if paintings are temporally
and modally flexible, theoretical unity, Rohrbaugh maintains, supports the
idea that we should say the same about photographs.
This argument is open to at least two objections, however. First, as Dodd
points out (2007, pp. 151–2), to assume theoretical unity is to beg the
question against the type theorist, for whom paintings and multiple artworks
belong to completely different ontological categories. Second, more seri-
ously, it can be argued that paintings cannot be identical to physical objects,
as Rohrbaugh maintains, precisely because the latter stand in flexible relations
to what are usually regarded as intrinsic properties of paintings. Consider a
strategy used in metaphysical debates about the relation between an artwork
like a statue and the physical material of which it is composed. Some have
argued (e.g. Thompson 1998) that this relation is material constitution, not
identity, on the grounds that the statue and the lump of clay have different
persistence and existence conditions. If a clay finger on the statue is replaced
with a different but visibly indistinguishable piece of clay, the statue, but not
the lump of clay, survives the replacement, it is claimed. Similarly, if we heat
the lump of clay making up the statue at a given time and reform it into a
different shape, the lump of clay survives this process but not the statue.
Such thought experiments purportedly demonstrate that the material of
which the statue is composed at any given time is merely constitutive of it at
that time, whereas the shape of the statue is essential to it.
In claiming that paintings are physical objects which are themselves
continuants, Rohrbaugh is assuming, in a similar vein, that a physical object
is not identical with the material of which it is composed at a given time.
But, even if neither a painting nor the physical object with which it might
be identified can be identical to a given collection of matter, the painting
and the physical object may themselves differ in ways analogous to the statue
and the lump of clay. David Wiggins (1980), for example, maintains that a
painting’s intrinsic ‘pictorial’ properties are essential to it qua painting. But it
is precisely these kinds of properties of the physical canvas that can change
over time as a result of damage or deterioration. If such pictorial properties
are essential to the painting but not to the physical canvas, then paintings
cannot be identical to physical objects. Thus there are strong (to some)
intuitions about the temporal and/or modal inflexibility of paintings that
would, if granted, defeat the proposed identity of paintings with physical
objects. So one cannot assume that paintings are identical to physical objects
in arguing for the temporal and modal flexibility of paintings, or, deriva-
tively, for the modal and temporal flexibility of photographs.
IV.
Given the general constraints on an ontology of multiple artworks like films
and photographs to which I have appealed in the foregoing reflections, the
challenge is to find an account of such works that allows them to be both (1)
repeatable and (2) creatable. We should prefer an account that does not rest
upon contestable claims about the modal and temporal flexibility of works.
Indeed, claims about such matters should arguably be resolved on the basis of
an otherwise adequate account of the nature of multiple artworks, rather
than serve as premises in formulating such an account. We might, however,
274 DAVID DAVIES
be encouraged in our search for such an account by Dodd’s concession that

at least some repeatable artworks—he cites photographs—might not be
types, even if there are conclusive reasons to identify musical works with
types.7 Dodd offers no insight into how this might be possible, or why, if it is
possible for photographs, it would not also be possible for multiple artworks
in general. In the remainder of this chapter, I want to suggest a way in which
cinematic and photographic works might be repeatable without being types
in the metaphysically weighty sense elucidated by Dodd. This also has
implications for Dodd’s claims about musical works and, thereby, for
multiple artworks in general.
I have thus far left unchallenged Dodd’s construal of types as abstract
objects that are individuated in terms of a property-associate specifying
those features required in (well-formed) tokens of the type. Dodd argues
with some plausibility that if musical works are what we may term ‘Dod-
dian’ types so construed then they must be discovered rather than created by
their composers. A similar conclusion will follow for other kinds of multiple
works if they are Doddian types having these properties. But Dodd’s
conception of types seems to contrast strikingly with the account sketched
by Richard Wollheim (1980), the first philosopher, in the tradition to which
Dodd belongs, to invoke types in a theory of multiple artworks. It was in
response to Wollheim that Wolterstorff elaborated his conception of such
works as norm-kinds, a view largely taken over by Dodd. What is striking in
the present context is Wollheim’s claim that what distinguishes types from
other kinds of generic entities is that they are postulated where we have a
piece of human invention. This suggests that ‘Wollheimian types,’ as we may
call them, differ dramatically from Doddian types. Indeed, I shall argue,
works construed as Wollheimian types can have exactly the kinds of
properties that multiple works like films, photographs, and cast sculptures
seem to possess—both repeatability and creatability.
7 Dodd (2007, p. 109, fn 12). Dodd, critically discussing Rohrbaugh’s arguments against the type-
token theory of musical works, entertains the supposition that ‘the repeatability of photographs—that is,
the fact that they admit of multiple copies (in the form of prints) . . . indicate[s] that photographs are
types.’ He comments in the footnote: ‘This is a big supposition, of course, and one to which I do not
wish to commit myself. After all, it is possible that the kinds of arguments that lead us to view works of
music as types do not apply in the case of photographs, or that there are special reasons why the
repeatability of photographs is best explained by a rival to the type-token theory.’
Wollheim (1980, pp. 34–5) introduces the idea of multiple artworks as

types in addressing what he terms the ‘logical’ question, ‘Are works of art
physical objects?’ This asks about the logical category to which works of art
belong, and thus the criteria of identity and individuation applicable to
them. To answer the logical question positively is to say that works are
particulars, rather than being non-particulars such as universals, or classes.
Wollheim contrasts the logical question with the ‘metaphysical’ question,
‘Are works of art physical objects?’ Here we assume that artworks are
particulars, and ask what kinds of particulars they are. Alternative answers
are that works are indeed physical objects, or that they are non-physical
particulars of some kind, either ‘Ideal’ objects, existing in the minds of
artists or receivers, or ‘Presentational’ objects, purely phenomenal entities
having only sensible properties.
Wollheim defends the idea of multiple artworks as types as one possible
negative answer to the logical question. This seems to carry with it the
assumption that such works are not objects or particulars of any kind, and
thus not abstract objects or particulars of any kind. Indeed, it is notable that,
in his discussion of multiple artworks as types, Wollheim nowhere mentions
abstracta. This isn’t to say that types might not be abstract entities, but it is to
say that this is not relevant to answering the logical question as Wollheim
conceives it. Rather, as noted earlier, he classifies types alongside classes and
universals as generic entities. Kinds of generic entities differ in how they
group the elements falling under them and in the logical relationships in
which they stand to these elements. Classes, for Wollheim, are simply
constructs out of the particulars that are their members. Universals are
more intimately related to their instances, being present in them, but types
are even more intimately related to their tokens. They are not only present in
their tokens but are treated much of the time as if they enjoy equal status
with them. This is to be explained in terms of the way in which properties
are transmitted between types and their tokens: We can rightly predicate of
the type all of those properties—other than those only possible in particu-
lars—that a token must have in virtue of being a token of that type. Thus,
for example, the Red Flag is rightly taken to be red because all tokens of the
Red Flag must be red.
Although we treat types like their tokens in this way, Wollheim does not
take this as indicating that types are themselves particulars of some kind
standing apart from their tokens as works stand apart from their physical
276 DAVID DAVIES
manifestations on the ‘Ideal’ and ‘Presentational’ theories. What positively

distinguishes types from classes and universals, he claims, is that we postulate
a type ‘where we can correlate a class of particulars with a piece of human
invention; these particulars may then be regarded as tokens of a certain type’
(1980, p. 78). This claim, Wollheim stresses, is an answer to an ‘entirely
conceptual [question] about the structure of our language.’ Furthermore,
questions about the identity of types, and about their individuation, are
essentially questions about the generative relationships in which particular
groupings of tokens stand to such pieces of human invention. Wollheim
speaks of the problem of ‘how the type is identified or (which is much the
same thing) how the tokens of a given type are generated’ (ibid., p. 79), and
of ‘the variety of ways in which the different types can be identified, or (to
put it another way) in which the tokens can be generated from the initial
piece of invention’ (ibid., p. 80). Types so differ in the ways in which they
can be initiated, as spelled out earlier in the case of multiple artworks.
Tokens of a Wollheimian type, therefore, are grouped not in terms of
shared manifest properties, but in terms of a shared relational property—
their having the manifest properties that they do in virtue of a causal history
that relates them to an act of human invention. This accords with what we
have already seen regarding the role a particular causal history plays in
grouping certain entities as p-instances of a photographic, cinematic, or
cast-sculptural work. The multiple nature of works of photography, film,
and cast sculpture—their ‘repeatablity’—is explicable in terms of this shared
causal history. To be a p-instance of such a multiple work W is just to have a
causal history C involving the work’s production-artefact in a way sanc-
tioned within the relevant practice. To be a strict p-instance of W is just to
have certain specific manifest properties in virtue of C, where those properties—
the ones required in order to fully qualify to play the experiential role in the
appreciation of a work—are again determined by norms operative in a given
artistic practice. Screenings of Citizen Kane, for example, are the end
products of a particular process of making mediated by a particular produc-
tion artefact, and a properly formed screening has certain specific manifest
properties, qua visual-acoustic sequence, in virtue of that causal history. All
such screenings, proper or flawed, are thereby p-instances of Citizen Kane.
Wollheim, as we have seen, offers a logical account of types, and of works
as types, tying their identity and individuation to the different ways in
which we group particulars. But this is not to embrace a simple nominalism
about types of the sort criticized by Dodd (2007, pp. 38–41). Wollheim’s
claim is not that statements that use type-expressions are ‘translatable’ into
statements about those particulars that are tokens of the type. Rather, for
Wollheim, to postulate a type is to postulate a particular kind of principle for
grouping particulars that relates them to an act of human invention taking
place within a practice. This enables us to clarify the ‘of ’ locution whereby
we describe the relation that obtains between a multiple work like a film or
photograph and its p-instances. The latter is to be compared with other
locutions in terms of which we group particulars by reference to the
relations in which they stand to acts of human invention. Consider, for
example, the locution ‘copy of this Sunday’s Observer.’ What I am reading is
such a copy because it has certain properties as a result of being produced by
a causal process using a production artefact that is itself the result of a
collective act of human creation. Those things correctly described as ‘copies
of this Sunday’s Observer’ are grouped under this label not merely in virtue
of their manifest properties but also in terms of the histories of making
whereby they came to have those properties.
V.
It might seem, however, that we have simple postponed the ontological
question about multiple artworks. If Citizen Kane, as a Wollheimian type, is
indeed something postulated in virtue of the way in which we group
particular screenings by reference to a piece of human invention, what
kind of thing is it, ontologically speaking, that we postulate? It is here that
we see the temptation to reify types through the kind of analysis furnished
by Dodd. Wollheimian types, it might seem, must be entities that somehow
stand apart from their tokens, and that explain why those tokens are
rightly grouped together. It is instructive, here, to see how Wollheim’s
logical conception of types mutates into the ontological conception in
Wolterstorff and Dodd. Wolterstorff introduces norm-kinds, as a corrective
to Wollheim’s talk of ‘types,’ by analogy with natural kinds such as ‘The
Polar Bear.’ Natural kinds, he assumes, are associated with a set of manifest
properties possessed by well-formed, but not necessarily by all, members of
the kind. This funds the idea that it is through possessing certain properties,
278 DAVID DAVIES
rather than through standing in a particular causal-intentional relation to a

piece of invention, that something qualifies as a token of a norm-type that is
a work, and further funds Dodd’s idea that musical works, as norm-types,
are individuated in terms of property-associates that specify the manifest
properties required in well-formed tokens. This allows Dodd to defend the
‘timbral sonicist’ view that musical works are individuated purely in terms
of how their well-formed instances sound. This divorces status as a well-
formed instance—a strict e-instance, in our terminology—from provenance
(2007, p. 2).
A preliminary Wollheimian response to Dodd would be that, if we
postulate types when we group things relative to a piece of human inven-
tion, then Doddian types whose property-associates specify only the manifest
properties required in well-formed tokens cannot explain such groupings
precisely because Doddian types prescind from provenance. But this might
suggest that what we are postulating when we speak of Citizen Kane is a
Doddian type whose property-associate specifies not only intrinsic but also
relational properties of its correct tokens, such as having manifest features X
in virtue of standing in relation R to a given act of human invention A. Such
a view might seem attractive because (1) unlike Dodd’s ontological account
of musical works, it integrates the role of human invention into its concep-
tion of the multiple artwork; while (2) it also speaks, like Dodd’s account, to
deeper motivations for including Doddian types in an account of multiple
artworks. These motivations are apparent in Levinson’s distinction between
‘implicit’ and ‘initiated’ types. Implicit types are Doddian types that preexist
their ‘discovery’ and exist through belonging to a certain framework of
possibilities. In chess, for example, all possible combinations of legal moves
exist as implicit types given the rules of the game. A player tries to discover
the best way to continue from a given game situation, a way of continuing
open to discovery by multiple players on multiple occasions. Here it seems
right to say that particular games or particular sequences of moves exist prior
to their discovery, and to treat each such discovery as a tokening of a pre-
existing type. Levinson takes given combinations of sound sequences and
instruments used to produce those sound sequences to be implicit types in
this sense. Thus a composer discovers rather than invents one such combi-
nation. But, as noted earlier, he maintains that the work itself is not the
implicit type but, rather, an initiated type—the implicit type as ‘indicated’
by an individual at a time, or in a given musico-historical context. Thus two
composers who discover the same implicit type in different musico-

historical contexts bring into existence works that have different properties
consequent upon that difference in context, and thus different works.
Musical works, then, as initiated types, ‘begin to exist only when they are
initiated by an intentional human act of some kind’ (1980, p. 81).
Levinson, like Wollheim, wants to accommodate the creatability of
multiple artworks by tying them to particular human actions. But, by
elucidating works as initiated types that incorporate acts of discovering
implicit types, Levinson lays himself open to Dodd’s objection, noted
earlier, that an initiated type whose existence is tied to the existence of its
property-associate is no more creatable than an implicit type. To avoid this
conclusion, we must either dispute Dodd’s criterion for the existence of
implicit types, or hold that implicit types and initiated types have different
kinds of existence conditions, or reject the idea that the link between
multiple artworks and human creativity runs by way of the discovery of
implicit types. I shall return to the third option shortly. But neither of the
other options strikes me as promising.
Consider, for example, the ‘chess’ analogy which funds the idea that
composers discover implicit types. The idea is that the sound/performance
means structures distinctive of musical works are discovered rather than
created by their composers, just in the way that combinations of chess
moves are discovered by players. In problematizing this analogy, we
might contrast the essentially inventive nature of the process whereby artists
initiate both singular and multiple artworks with the creative activities of
chess players. To think of writers and composers as discovering preexisting
possibilities fails to take account of the ways in which artists themselves are
responsible, through their creative activities, for establishing the very frame-
work of possibilities that their works explore. This is most obvious in
literature in the works of writers like Joyce (Ulysses, Finnegan’s Wake),
Burgess (A Clockwork Orange), and Hoban (Riddley Walker) who invent
substantial parts of the very language in which their works are written.
But the possibility for such invention is always present in artistic activity,
even for artists who work within the kinds of prescriptions characteristic of
classical French drama, for example. In music, as in painting, the possibilities
can be changed not only by technological change—for example, in
painting, the possibilities for ‘staining’ canvases made available with the
development of acrylics—but also by an artist’s changing the ‘rules’ of the
280 DAVID DAVIES
game—as in twelve-tone musical compositions. The artist’s ability to ex-

hibit what Kirk Varnedoe (1989) termed a ‘fine disregard’ for the existing
conventions within an art form and still produce something that belongs to
that art form undermines the idea of artists as discoverers of what is pre-
existent in a general framework of possibilities.8
There is a further problem with the idea that we gain explanatory
purchase on, or understanding of, multiple artworks by taking postulated
Wollheimian types to be initiated Doddian types, types whose property-
associate incorporates relational properties that relate a perceptual manifold
to a piece of human invention. Such Doddian types clearly exist, since their
corresponding property-associates exist, and, given suitable care in formu-
lating the relevant property-associate, they will presumably comprise all and
only the strict p-instances of a multiple artwork. But the need for care in
formulating the relevant property-associate gives the game away. Unlike
implicit Doddian types that exist independently of their possible tokens, that
can thus be thought of as discovered by composers, and that can explain our
grouping of certain things as strict e-instances of works if Dodd’s proposed
ontology is accepted, the types we are currently considering can only be
characterized ‘after the fact,’ as it were. For the things that are rightly
grouped as p-instances, or as strict p-instances, of a multiple work—the
tokens (or strict tokens) of the Wollheimian type—depend not merely upon
what the artist does but also upon the relevant conventions in place in the
artistic community that determine what must be done in order to generate
an instance (or a strict instance) of a work—for example, how we must
manipulate a production artefact, if a p-instance (or a strict p-instance) of the
work is to result. Thus initiated Doddian types cannot explain our group-
ings of things as tokens of a work, and do not add to our understanding of
the picture already in place in the Wollheimian account. The problem is not
the absence of the relevant type, therefore, but its bearing on the questions
that interest us in our attempts to understand multiple artworks like films or
photographs.
The foregoing also bears upon another strategy that might seem open to a
defender of the idea that multiple works are Doddian types. It will be
8 Varnedoe takes such a ‘fine disregard’ to be a mark of modernism, although it surely also applies to
much premodern art. If there are cultures of artistic production that are usefully thought of on the ‘chess’
analogy, such as French neoclassical drama, they cannot furnish a general model for thinking about how
artworks are initiated.
recalled that Dodd, following Wolterstorff, insists that multiple works of

music are norm-types individuated in terms of a property-associate that
specifies what is required in correct tokens of the type. Our interest in
appreciating artworks is in the artistic content articulated by their correct
instances, or by what we have termed their strict e-instances, not in the
variety of contents that might be articulated by their p-instances, flawed or
strict. But then, should we not say that to postulate a work reflecting how
we group particulars by reference to a piece of human invention is to
postulate, as the result of that piece of invention, not merely a principle
for grouping a work’s p-instances but also a principle for determining which
of those p-instances are strict e-instances of the work. But to postulate a
principle that determines a work’s strict e-instances—those entities fully
qualified to play the experiential role in the work’s appreciation—is to
postulate something that picks out a class of particulars not in terms of
shared causal histories, but in terms of some set of manifest properties that
they fully possess. And a set of particulars grouped by reference to the
satisfaction of such a condition couched in terms of manifest properties
just is an implicit Doddian type.
It is undeniable that there is an implicit Doddian type corresponding to a
p-multiple artwork, the property-associate of which specifies the properties
required in a strict instance of the work. But the precise nature of this type
depends, again, upon the artistic practices within which the work is initi-
ated, practices which determine—given what is specified by the initiating
artist—which features are required for a strict instance of the work. Thus,
yet again, Doddian types come in ‘after the fact,’ representing the groupings
that our practice achieves rather than standing as an independent measure of
the rightness of those groupings. Again, we do not explain the nature of the
Wollheimian types that we postulate, but only describe them, in our talk of
Doddian types. Furthermore, there will also be an implicit Doddian type for
singular artworks such as daguerreotypes (and indeed paintings), a type
whose property-associate specifies (perhaps indexically) the manifest qual-
ities required in strict e-instances of those works. If the existence of such an
implicit Doddian norm-type entails that a work is discovered rather than
created, then no artworks are created. But the sense in which this is true is,
I think, uninteresting for our understanding of art, for the Doddian types
play no explanatory role in that understanding.
282 DAVID DAVIES
VI.
How, then, if not in terms of Doddian types, might we illuminate the nature
of what is postulated when, in identifying a work, we postulate a Wollhei-
mian type? The answer, I suggest, is that we illuminate what is postulated by
attending to the precise combination of human invention and an enabling
practice that does explain and legitimate9 both the groupings of certain
things as p-instances of a work and the identification of the properties
required in its strict instances. To postulate a work qua Wollheimian type
is to postulate a piece of human invention ‘initiating’ the work that is embedded
in a set of practices that issue in the very groupings of tokens that we seek to
understand. To appreciate works as such Wollheimian types is just to
appreciate what was done in two senses—what manifest properties are pos-
sessed by right tokens having the relevant causal history, and what was done
in establishing the preconditions (for example, the generation of a produc-
tion-artefact) for that casual history to take place.
On such a view, our understanding of works as Wollheimian types is not
illuminated by appeal to types in the metaphysical sense. Rather, the things
that play the particular kind of logical role described by Wollheim might be
viewed as what I have elsewhere (2004) termed ‘performances,’ contextual-
ized actions. On the ‘performance theory,’ Citizen Kane is a collective
generative action that brings into existence a production artefact whose
function is to enable multiple screenings of the work. A correct screening is
one that has those distinctive manifest features required by norms embodied
in the relevant artistic practice in virtue of standing in a causal relation to the
work, qua generative process, mediated by the production artefact. I have
argued for the performance theory elsewhere, and shall not repeat those
arguments here. My claim is only that multiple works, viewed as perfor-
mances, can play precisely the role for which Wollheimian types are
postulated, and can, unlike Doddian types, accommodate both the repeat-
ability and the creatability of multiple works. By rescuing Wollheim’s
neglected insights from the misunderstandings of Doddian-type theorists,
9 While I lack the space to develop this further here, I think my talk of groupings being ‘legitimated’
by our practices can be cashed out in terms of the more general model of the embeddedness of norms in
social practice spelled out by Robert Brandom (1994).
we may hope to demystify multiple artworks and appreciate them as the

created yet repeatable entities that they are.
References
Brandom, Robert (1994) Making It Explicit (Cambridge, MA: Harvard University
Press)
Brown, James Robert (1991) The Laboratory of the Mind: Thought Experiments in the
Natural Sciences (London: Routledge)
Burgess, John and Rosen, Gideon (1997) A Subject With No Object (Oxford:
Clarendon Press)
Carroll, Noël (1998) A Philosophy of Mass Art (Oxford: Clarendon Press)
Currie, Gregory (1989) An Ontology of Art (London: St Martin’s Press)
——(2010) ‘Multiple Instances and Multiple “Instances”’ British Journal of Aesthetics
50: 411–26
Davies, Stephen (2003) ‘Ontology of Art’ in Jerrold Levinson (ed) Oxford Handbook
of Aesthetics (Oxford: Oxford University Press)
Dodd, Julian (2007) Works of Music (Oxford: Oxford University Press)
Lamarque, Peter (2010) Work and Object (Oxford: Oxford University Press)
Thompson, Judith Jarvis (1998) ‘The Statue and the Clay’ Noûs 32: 149–73
Varnedoe, Kirk (1989) A Fine Disregard (London: Thames & Hudson)
Wiggins, David (1980) Sameness and Substance (Oxford: Basil Blackwell)
Wollheim, Richard (1980) Art and its Objects, 2nd edn (Cambridge: Cambridge
University Press)
13
Musical Works: A Mash-Up
JOSEPH G. MOORE*
I. Controversing
What role in the individuation of a musical work is played by its prove-
nance—by the musico-historical and biographical conditions that sur-
round the work’s composition and reception? The fact that Beethoven’s
Hammerklavier Piano Sonata (No. 29 in B flat major, Op. 106, 1818)
occurred in the musico-historical setting that it did, and the fact that it was
composed when it was in the sequence of Beethoven’s oeuvre determines
some of the sonata’s broadly aesthetic features. The Hammerklavier estab-
lished a new precedent for the length of a solo work, for example; and
certain features of Beethoven’s later work, such as situating the fugue
within a classical form, emerge here for the first time. It’s natural to
attribute these novelties to the work itself, since they contribute to the
piece’s critical importance.
On the other hand, even a highly informed listener doesn’t encounter
these features in a narrowly perceptual way when the work is performed,
nor do they guide a proper or even insightful performance of the work. And
if a musical work is individuated by those of its features that can be presented
and experienced in performance then provenance shouldn’t figure in the
work’s identity.
* For help with this chapter, I thank Bradley Armour-Garb, Richard Beaudoin, Thomas Bennigson,
Mark Crimmins, Stephen Davies, James Harold, Christy Mag Uidhir, Stephen Maitzen, Lisa Moore,
Margaret Moore, Robert Pasnau, Dave Robb, Nishi Shah, and Thomas Wartenberg, as well as
audiences at the University at Albany, the Eastern Division Meeting of the American Society of
Aesthetics, and the Philoso-Ski Conference in Boulder, Colorado.
MUSICAL WORKS: A MASH-UP 285
But perhaps this illegitimately privileges those aesthetic features that are
narrowly experiential, and these are sometimes of secondary evaluative
importance (think of chance music, or the use of found sounds). In any
case, it’s not clear that we can cleanly distinguish those aesthetic properties
of a work that can be heard in performance from those that can’t. Who
would deny that The Rite of Spring’s revolutionary character was apparent to
its first audience in Paris? (If you think Nijinsky’s choreography was really
the focus, consider instead the reception given to Ornette Coleman’s The
Shape of Jazz to Come.) Isn’t this striking historical property part of the
work’s very nature?
II. Just Yoking

Controversies such as this about the proper role of provenance have been
hotly debated in recent decades. And I think it is fair to say that ‘contextual-
ists’—those who think work identity is sensitive to provenance—are win-
ning the day. If we agree with Jerrold Levinson and other contextualists1
that at least some historical or even biographical properties should figure in
the identity of a musical work, then it’s plausible to construe a musical work
as a sound-structure yoked somehow to certain art-historical features that
surround its composition. There are interesting and subtle disagreements
about how we should best understand and deploy the notion of a sound-
structure, as well as the notion of art-historical provenance, but these are not
my main concern here.2 My concern instead is with the yoking: How
should those of us moved by contextualist considerations best understand
the way structure and provenance mix in our concept of a musical work?
An obvious way to incorporate provenance in a work’s identity is to
construe a musical work as a fine-grained concatenation of a sound structure
and a particular provenance. I’ll take as representative Stephen Davies’s
1 The debate concerning musical works was notably sharpened by an exchange between Jerrold
Levinson and Peter Kivy begun in the 1980s. See Kivy (1983, 1987) and Levinson (1980, 1990). For more
recent discussion see, for example, Davies (2001) and Dodd (2007).
2 For my purposes, we can take sound structures to be abstract rhythmically articulated sequences of
sound types. However, see Davies (2001, 2008), for a thorough discussion, and reasons why we might
include in sound structures not just timbre, but instrumentation. I’ll say a good deal more about
provenance below.
286 JOSEPH G. MOORE
suggestion that ‘a musical work is a performed sound-structure as made

normative in a musico-historical setting.’3 The proposal may allow that
strict similarity in sound structure and provenance is somewhat loosened
in application, but its logical root is conjunctive: Two works are identical if
and only if they have the same sound structure and the same provenance.
This is the main form of contextualism on offer in the literature, presumably
because it’s the only straightforward way to mix structure and provenance.
I propose a different mix. On my view, our judgments of work identity
are more disjointed. They are variously sensitive to structure, or to prove-
nance, or to both in a way that occasionally pulls us in different directions.
Moreover, our identity judgments can shift with the explanatory purposes,
evaluative foci, and pragmatic considerations that characterize the different
judgmental and conversational settings in which we find ourselves. Some-
times we’re focused primarily on a work’s musical structure, while some-
times we’re concerned more with its art-historical status. Most of the time,
both criteria are implicitly at play. Exactly how our sensitivities shift—and
even that they shift at all—is not readily apparent to us because the two
concerns rarely come apart. If I’m right, there’s a sort of tacit semantic
indecision or individuative indeterminacy built into our work concept.
This indeterminacy isn’t worth resolving through revisionary conceptual
sharpening because not doing so yields cognitive and conversational econ-
omy. And as I’ll eventually explain, it doesn’t rule out the world’s answering
to our shifting concept.
This is the view I will develop in this chapter. I first argue that ‘shiftism’
(sorry) is a superior alternative both to contextualism about musical works
and to contextualism’s main opponent—the ‘structuralist’ view that musical
works are simply sound structures.4 In Section III, I argue that this shifting
3 See 2001, p. 97. Although Davies is not out to define musical works, I invoke his ‘suggestion’
because I think it sensibly incorporates Levinson’s inclusion of performance-means, while relaxing the
inclusion of ‘individual,’ as opposed to less restrictive ‘general,’ bits of musico-historical context. (See
Levinson (1980) for this distinction.) Davies’s proposal also incorporates, as Levinson eventually did (see
1990, p. 260), the normativity from Wolterstroff ’s (1980) influential idea that musical works are best seen
as ‘norm kinds.’
4 A note on labels: I use ‘structuralism’ even though neither Kivy nor Dodd apply this label to their
views. My use is meant to be neutral on the disagreement, alluded to above, over whether sound
structures should incorporate such features as timbre, instrumentation, or performance-means more
generally. As I’ve set things up, shiftism is technically a form of contextualism, since it incorporates
provenance in some manner in its account of work identity. But in the interest of clean contrasts and
clarity I’ll henceforth restrict ‘contextualism’ to fine-grained conjunctive proposals like those of Levin-
view gives a better treatment of some hypothetical and actual cases of work
individuation because it explains some unsettled intuitions. And in
Section IV, I argue that considerations that might sharpen this indetermi-
nacy are either tangential or inconclusive, particularly once shiftism is seen
as a theoretical alternative. I then worry, in Section V, that the individuative
indeterminacy that shiftism exposes challenges the existence of musical
works. However, in Sections VI and VII, I develop a more detailed account
of the indeterminacy that I argue, in Section VIII, is compatible with
realism about musical works.
III. Intuition-Mongering
Hypothetical cases are sometimes treated as grist for a theoretical mill, or
presented along with a favored intuition that the theory goes on to honor.
For the moment, I’d like to effect the opposite: Insofar as it’s possible,
consider the following counterfactual scenarios—my own versions of
some familiar examples—with an open, uncommitted, and pre-theoretic
mind.
Scenario #1: The Hawaiian-Hammer. Suppose Beethoven had lived in
some remote part of Hawaii, as yet untouched by the culture of Captain
Cook. In 1817, a Broadwood piano washes ashore. The Hawaiian genius
soon figures out how to play and compose for it, and to do all of this in
classical style no less. Even more miraculously he does so all on his own. By
the fall of the next year, he has composed a piece that is note-for-note
identical to the actual Hammerklavier. Is the Hawaiian-Hammer the same
as the actual Hammerklavier?5
Scenario #2: The Honey-Hammer. Suppose instead that, back in
Vienna, Beethoven sits down to write his 29th piano sonata during the
son and Davies. My taxonomy also leaves out the art-historicist view of Rohrbaugh (2003) and the
musical perdurantism of Caplan and Matheson (see 2006, 2008). These views also individuate works by
context, perhaps at the expense of structure. I don’t have space adequately to incorporate these views
here.
5 This example could be presented as a case of counterfactual doppelgangers, though I see no
significant difference, and some simplicity of presentation in casting it as counterfactual comparison
with the actual Hammerklavier. This type of example has been widely considered, though my version is
essentially a poor man’s version of Currie’s (1989, p. 62), with Hawaii substituted for Twin Earth.
288 JOSEPH G. MOORE
same time period; and he dedicates it to Archduke Rudolph, as he actually

did. Suppose, though, that the sonata’s third movement turns out to be
more upbeat and somewhat sweeter than the actual Hammerklavier’s fam-
ously sorrowful third movement. Suppose our counterfactual Beethoven
has upped the tempo a bit, made a few themes a touch less searching, and
tossed in a few more major tonalities. The differences are subtle, but
noticeable (at least to those who can shuttle between possible worlds),
although the counterfactual piece goes on to much the same critical acclaim
and influence as the actual Hammerklavier. Is this counterfactual 29th
Sonata none other than our actual Hammerklavier?6
The strict contextualist holds, of course, that neither the Hawaiian-
Hammer nor the Honey-Hammer is the same work as Beethoven’s actual
Hammerklavier, while the pure structuralist agrees that the Honey-
Hammer is a different piece but will allow that the Hammerklavier could,
indeed, have been composed in a remote part of Hawaii.
But what do our untutored intuitions tell us? (Or have I already reloaded
them?) The examples are far-fetched, of course, and there isn’t universal
agreement, but some of us draw a blank; and we do so because we’re pulled
in different directions. If we successfully discharge the presupposition that a
determinate ‘yes’ or ‘no’ is called for, we might well say that the Hawaiian-
Hammer is the same as the actual Hammerklavier ‘in one sense but not
another.’ Ditto for the Honey-Hammer, with the senses converted. I don’t
know for certain that I’m in the intuitional majority here, though I suspect
it. (Confirming this might be a project for experimental philosophy.) In any
case, I want to investigate exactly what wide-spread intuitional indetermi-
nacy of this type might show about our work concept.
In the first scenario, we notice a structural similarity and a ‘provenancial’
(if I may) dissimilarity. Since this pulls us in conflicting directions, our
overall judgment about the identity of the Hawaiian-Hammer remains
indeterminate. This would be nicely explained if our work concept shifts,
as I claim, between a structural strand and an entirely provenancial strand.
But the contextualist might claim that our intuitional uncertainty arises
instead because we recognize a similarity in one of the (as he holds) two
components of the musical work, while we simultaneously disregard a
6 This second type of example is much less common, though a notable predecessor is Rohrbaugh
(2003, p. 182).
dissimilarity in the other. Perhaps we disregard provenance because our

consideration of the case has been too hasty, or we haven’t fully appreciated
the truth of contextualism.
This alternative explanation of the Hawaiian-Hammer is already inferior,
since it makes our intuition half-mistaken, but I think the Honey-Hammer
nails the case against the contextualist. Regarding this second scenario, the
contextualist has to say, I think, that any intuitional indeterminacy results
from a looseness in our enforcement of structural similarity. But this gets
things wrong: We don’t blank on the Honey-Hammer solely because we
can’t gauge whether the structural differences transgress some permitted
degree of looseness; rather, our sense that the structural differences are, or
might be beyond the pale is balanced against our sense that the strong
provenancial similarities should rule the day.
We think of a musical work as a product, but also as a process. And the
process, at least, is almost always cultural. (This is one reason the Hawaiian-
Hammer, with its isolated genius, is a particularly far-fetched example.) To
count as a work, a musical offering must have ‘presentational content’—a
performed sound structure, close enough—but someone must also consider
it to be a work, and indeed a work of a certain type. The composer isn’t
entirely authoritative here; and in this respect, musical works are a bit like
marriages and touchdowns. Even when she’s working solo, the composer’s
work is shaped—both constrained and enabled—by a socio-artistic tradition
and practice, which she might shape in turn. This practice helps determine
what features can count as artistic, and how: It helps determine, for
example, which expressive features are conveyed by which transformations
of the medium. But this social practice also plays a role in determining
whether something is a work at all, and how the work is institutionally
recognized, presented, described, categorized, canonized, and so on. All of
these considerations drive the way we individuate works.
So far, of course, the contextualist will agree. What supports my shifting
view against contextualism is the further observation that many art-insti-
tutional means of work-individuation needn’t have much connection with
musical content. In order to count as a musical work, a creation must have
some musical content or other, to be sure,7 but a work’s art-institutional
7 Whether Cage’s silent piece, 4’33”, has any musical content at all is at the heart of one interesting
argument that it isn’t a musical work. See Davies (1997).
290 JOSEPH G. MOORE
individuation needn’t concern the particularities of this content. This sug-

gests that provenance can have an individuating force all on its own.8
Consider the way we think, at least theoretically, of musical works over time,
both during their composition and after their initial or official completions.
A composer confronts a deadline, and very late in the day she decides to alter
the key signature or tempo of her piece, or to tinker with its harmonies, or to
cut and add measures or entire sections, or even to completely change the
principal theme. We nevertheless seem to regard the work as numerically
unchanged for sound practical reasons of a broadly institutional and intentional
kind—a piece has been commissioned, a title has been chosen, or the composer
has simply stuck to the self-imposed task of composing her next new work. On
the other end, a deadline may have passed—an initial performance has oc-
curred, or a score has been published—yet the composer tinkers with her
composition because it dissatisfies her in some way (true of Bruckner and
Chopin), or because she wants to change things to suit the different musicians,
instruments, or audience involved in an upcoming performance (true of
Mozart, Stravinsky, Ellington, and many others).9 We say that the work has
changed, but we seem to mean this qualitatively and not numerically: We
might distinguish drafts or versions in order to mark the relatively stable
musical, or institutionally noted points in the diachronic stream of structural
variation, but we don’t stray from the pragmatic, provenance-driven reasons
for counting the whole structure-shifting enterprise as of a piece, and not many.
I see no reason to think that our work concept is any less modally flexible
than it is temporally so. And once we’re gripped by the force of art-insti-
tutional considerations then our work concept can, I submit, accommodate
structural variations even greater than those involved in the Honey-Hammer.
Suppose Beethoven set out with the express goal of writing a sonata in which
to reconceive the fugue. (For all I know he actually did this.) And suppose he
accomplished this with music that is very different—that doesn’t move by
descending thirds, for example. Couldn’t this still be the Hammerklavier?
I’ve chosen Beethoven as the source of my examples because it’s precisely
in the era he dominates that Lydia Goehr has controversially claimed that
our work concept came into being in its complete and mature form.10 One
8 This possibility is proposed for photographs by Rohrbaugh (2003).

9 For some of these examples and others, see Davies (2007).
10 See Goehr (1992).
mark of this is that Beethoven and his contemporaries could, and did make
normative heightened levels of performative uniformity, precision, virtuosity,
and musical detail. But this actually makes Beethoven less tractable for my
purposes since we tend to focus so intently on, as it were, his music. Consider
instead pieces that are composed with a particular person or occasion preemi-
nently in mind, or that are strongly driven by some non-musical theme.
Couldn’t Handel’s Fireworks Music, Ellington’s Queen’s Suite, or Arlo Guth-
erie’s Alice’s Restaurant survive even greater variations in musical structure?
What’s going on here? Bound up in our work concept are two distinct
metaphysical criteria of individuation, one structural and the other provenan-
cial. Works themselves seem to vary in how two distinct types of properties
bear on their identity: Some works have greater structural detail than others,
and some are surrounded by more forceful and determinative art-historical
particularities. The Hammerklavier has more structural detail than the ditty
I just wrote (trust me); and this same ditty might have been composed anytime
in the past few decades, whereas Bartok’s Concerto for Orchestra, actually
composed in 1943, could not have been composed before 1942, when Sho-
stakovich penned the Seventh Symphony which Bartok satirizes.11
But I think our work concept is doubly shifty: Our relative sensitivity to these
two variably thick ingredients can itself shift and remix as our thought and talk
focus on different aesthetic issues, and as they make salient different kinds of
aesthetic property. So our individuative judgments shift as we consider different
musical works, and also as this consideration is driven by different aesthetic
concerns. Very occasionally, a perfect semantic storm blows in: We stumble
across an example that pulls our two criteria in different directions, as we
consider the case in a manner that leaves our divergent sensitivities in intu-
ition-numbing equipoise. This happens most readily in theoretical discussions,
like this one, in which an example is purposely constructed, and we regard it
without any particular concern (other, perhaps, than understanding the very
nature of this regard). In these rare, untethered, theoretical settings our work
concept blanks, and breaks down. But thankfully, what might be common in
philosophical discussion is almost unheard of in everyday talk of musical works
and performances of them.
11 . . . unless Shostakovich had composed his symphony earlier; but then Shostakovich’s symphony
wouldn’t have had its characteristic wartime theme, unless the war had happened earlier . . . I riff here on
Levinson’s nice example (1980, 1990, p. 71).
292 JOSEPH G. MOORE
IV. Can We Sharpen Up?

My claim so far is that our received concept of a musical work is individua-
tively vague: In the vast majority of normal cases we unproblematically tell
musical works apart and count them cleanly; but in certain cases, such as in
the thought experiments I’ve discussed, our work concept doesn’t individu-
ate cleanly and determinately because its structural and provenancial strands
pull in different directions (and judgmental setting doesn’t resolve this
conflict). But the case I’ve made for this conceptual indeterminacy rests
largely upon a few thought experiments, and intuitions about these are
notoriously complex and various, as are their explanations. One might
suspect that there are other considerations that will resolve this pre-theoretic
indeterminacy.
What about the published debate over the individuation of works?
Portions of that discussion properly concern, in effect, our separate under-
standings of structure and provenance, whereas my concern, again, is with
the more fundamental question of how the two are mixed in our work
concept.12 Also tangential, I believe, is the discussion of whether works are
created or discovered, and how, if they’re created, they can also be abstract
entities. (I’ll suggest at chapter’s end that the existence conditions of musical
works might also be indeterminate.) What remains seems largely a consider-
ation of actual and hypothetical examples, and I’ve here proposed an
alternative and superior way of treating them.
Moreover, it’s difficult to see how further considerations could decisively
sharpen up our work concept. A discovery model seems entirely implausible
in this realm: Musical works are not natural kinds whose essential features
might be discovered through empirical investigation. On the contrary,
musical works seem constructed entities whose identity conditions are
entirely determined by our discursive practice and the conceptual scheme
that drives it.13
We might hope instead for conceptual resolution through reflective equilib-
rium: Perhaps our work concept will sharpen up once we see how it connects to
12 For example, the debate about the role of instruments and performance-means bears on the proper
understanding of a work’s structural component. And the question of how much of a work’s provenance
is identity-determinative—can different composers from the same musico-historical era compose the
same piece?—applies to the provenancial component separately.
13 Here I’m sympathetic with the view of art ontology defended by Thomasson (2005).
broader aspects of the way we think about music and artistic practice. Here too
I’m sceptical. The problem, it seems to me, is that the obvious issues we might
plumb are, on reflection, either independent of the question of work identity or
themselves unsettled (or both). I’ve already suggested that the issue of existence
conditions won’t help. Where else might we look?
We might try looking outside the realm of Western classical music. Here,
though, we will encounter musical practices without works or, more often,
practices with musical works that might be structurally ‘thinner’ but beset by
the same individuative indeterminacy nevertheless.14 Our thought experi-
ments could be made just as inconclusive if we substituted for Beethoven’s
Hammerklavier sonata John Coltrane’s ‘Giant Steps’ or, I suspect, a piece of
Classical Indian, or Javenese gamelan music. Or perhaps we will find un-
helpfully that the concept of a musical work is elsewhere deployed in a quite
different and heterogeneous fashion, as has been argued for rock music.15
We might look to other entities—symbols, flags, car models, books—that
seem both created and repeatable. But here too, I suspect we will encounter
the same individuative indeterminacy. Indeed, the shifting view might be
extended to cover repeatable non-musical works such as Henry Ford’s
Model T and Saul Kripke’s Naming and Necessity, because work concepts
seem generally to yoke a structural or content-sensitive criterion together
with a provenancial one.
Finally, we might hope that some general theory about the production
and reception of artworks might sharpen our work concept. But I’m still
skeptical. Take the debate at play in our opening controversy. To paint in
very broad brushstrokes, the ‘neo-formalist’ holds that what is to be evalu-
ated in an artwork are the significant features we narrowly perceive in
encountering it in a gallery, in a book, or in performance, for example.
(These are sometimes called a work’s ‘aesthetic properties’ as opposed to its
‘artistic’ ones.) On this view, it might be useful to know a work’s art-
historical setting in order properly to perceive it—in order, that is, to
appreciate which of its many perceivable features are artistically relevant.16
But it is only these perceivable features that properly figure in an evaluation
14 Stephen Davies develops the useful notion of a work’s ‘thinness’ in Davies (2001).
15 See Gracyk (1996) and Kania (2006).
16 This importance of this is brought out by Walton’s famous Guernicas example (Walton 1970).
294 JOSEPH G. MOORE
of the work itself as opposed to the artist or the genre; and so it is only these
features that are built into the work metaphysically.
The neo-expressivist holds, by contrast, that since art is at core a communi-
cative activity, cultural and historical setting is essential to an artwork’s com-
municative force. On this view, a sound structure or a canvas may be the crucial
perceptual vehicle of an act of artistic communication, but it is still only one part
of that act. For this reason, the artwork is more inclusive than the sheer
communicative vehicle (i.e. the canvas or sound structure); the artwork also
includes the vehicle’s relations to cultural setting and to the contingent expres-
sive connections that are an indispensible part of the artistic communication.
Who’s right? Perhaps the neo-formalist and the neo-expressivist capture
divergent but equally useful ways to conceptualize the world’s enormously
diverse variety of artworks and artistic practices. This theoretical pluralism
would support, and be supported by the individuatively mongrel character
of our work concept. But even if only one view is right, the independence
of work-individuation from this general debate seems ensured by significant
theoretical looseness in this domain—that is, by unclarity or outright
freedom about whether one theoretical choice point has any bearing
upon another. For example, the neo-formalist might have reason to hold
that, although only narrowly perceptual properties of the artwork are
properly attended to while evaluating the work itself, other features includ-
ing relations to context are nevertheless part of the work. And the neo-
expressivist might favor a conception of the artwork according to which the
work itself is a narrow expressive vehicle, with the rest of the expressive act
captured in events centering around the artist.17 One can imagine moti-
vation for each of these complex views. And their very possibility shows
that the general debate about the nature of artistic practice privileges neither
the structural nor the provenancial strand of our work concept.
So, if I’m right, there’s a deeply irresolvable indeterminacy in our
concept of a musical work. This would defuse the debate about work-
individuation: Structuralism and contextualism overprivilege, in effect,
distinct and sometimes divergent individuative strands bond up in our
mongrel concept of a musical work. Each view is half-right, but also
17 This possible position resembles (in reverse!) David Davies’ (2004) view that a canvas or sound
structure is the ‘work-product’ while the artwork itself is the creative activity by which the artist
produced this product.
half-wrong. But this nice result is balanced against a dark metaphysical

implication—if our work concept is irresolvably indeterminate, perhaps
there aren’t really musical works at all.
V. Interlude: A Quinean Tension

So far, I’ve argued for the following two claims:
Conceptual Indeterminacy: Our concept of a musical work (and the discourse/
practice in which it’s embedded) doesn’t determinately individuate all cases we
might consider; sometimes different individuative strands within our concept pull
us in different directions.
No Sharpeners: There’s no hidden criterion of individuation for musical works
that can be discovered empirically; and reflecting on broader conceptual connec-
tions won’t favor any particular revisionary sharpening of our concept.
Along with these two claims, I also hold that we have a strong prima facie
reason to be realists about the existence of musical works. We seem
successfully to talk about and quantify over works; indeed, we seem to
perform them. Much of our music making seems to traffic in works; and
talk of musical works seems an ineliminable part of the performative,
critical, and even legal discourse that surrounds this music making.
But this combination of views is in tension. If Conceptual Indeterminacy
and No Sharpeners are both true, then it’s hard to see how realism about
musical works can be squared with Quine’s compelling metaphysical injunc-
tion against ‘entities without identity.’ How can we admit musical works into
our considered ontology if they lack clear identity criteria—if we can’t say,
even in principle, whether the Hawaiian-Hammer (or the Honey-Hammer)
and Beethoven’s actual Hammerklavier are one and the same musical work?
We might question or scrutinize Quine’s principle. Indeed, Quine himself
once archly asked whether certain entities (he had in mind meaning-notions
like propositions) might be accepted ‘as twilight half-entities to which the
identity concept is not to apply? If the disreputability of their origins is
undeniable, still bastardy, to the enlightened mind, is no disgrace.’18 But
despite the appeal of enlightened disrepute, I’m with Quine. At least, I’m
18 Quine (1969, p. 23).

296 JOSEPH G. MOORE
with him in opposing the possibility of ‘in-the-world’ metaphysical indeter-

minacy—the putative possibility that, for example, the Hawaiian-Hammer
and the actual Hammerklavier neither determinately stand in nor fail to
determinately stand in the identity relation to one another. For one thing,
the general possibility of this type of metaphysical indeterminacy faces strong
general challenges to its coherence.19 More importantly, the possibility doesn’t
match up well with the source of the conceptual indeterminacy in this case.
For the conceptual indeterminacy in our work concept arises not from a lack
of intuitions, but from conflicts among them. And this suggests that the
indeterminacy resides not in the world itself, but in our representations of it.
Indeed, it suggests that, if need be, we could unpack our work-talk and
say everything we want to say about our musical practice in terms of
individuatively precise entities that answer to our conflicting criteria. If
this is so, then an individuatively determinate world could answer to our
indeterminate work-talk. It is this view that I will flesh out in the remainder
of the chapter. My positive view consists of a model of our unrevised work
concept (Section VI) and a sketch of a metaphysics that grounds it
(Section VII). With this view in place, I argue (Section VIII) that it
compares favorably to a second way one might respond to the Quinean
tension—that is, to abandon realism about musical works.
VI. A Supervaluating Concept

The individuative shifting I’ve argued for can be nicely modeled as a matter
of our work concept’s supervaluating over independent structural and
provenancial criteria.20 According to the supervaluation model, two
token-performances are determinately considered to be performances of
the same musical work—that is, our work concept counts them as perform-
ances of the same work—if they stand in both the same-structure and the
19 See Evans (1984), Moore (2008), and Salmon (1982, pp. 243–6).
20 For the purposes of the conceptual model, we might think of our criteria as tracking (often
overlapping) equivalence classes of actual and merely possible token performances (and playbacks)
formed under the same-structure and same-provenance relations respectively. To leave things there
would be circular, though, since the relations would be informed by the criteria. So in the next section,
I will replace these equivalence classes with entities that are more metaphysically articulated and
conceptually autonomous.
For early articulations of supervaluationism see, for example, van Frassen (1969) and Field (1973).
same-performance relations to one another. They are determinately not

considered performances of the same work if they fail to stand in both of
these relations. Otherwise (if the performance-tokens stand in one but not
another of these criterial relations to one another), it is indeterminate
according to our concept whether the two performance-tokens count as
performances of the same musical work.
So, according to the model, Beethoven’s own performance of his freshly
composed sonata—on an autumn evening in 1818, suppose—and a perform-
ance by Richard Goode in Amherst, MA in 2008 determinately count as
performances of the same musical work, the Hammerklavier, because the two
performances stand in both the same-structure and same-provenance
relations to one another. (Both are members of the same two equivalence
classes formed under the same-structure and same-provenance relations
respectively.) A David Grisman mandolin performance—‘inspired by the
Grateful Dead’—later the same evening in Northampton, MA determinately
counts as a performance of a distinct musical work because it determinately
does not stand in either relation to Beethoven’s performance (or to Goode’s).
It’s indeterminate whether a performance by the Hawaiian genius counts as a
performance of the same musical work as Beethoven’s performance: The
performances stand in the same-structure relation to one another, but fail to
stand in the same-provenance relation. And it’s also indeterminate whether a
performance of the sweeter sonata by our counterfactual Beethoven on the
same autumn evening counts as a performance of the same work: The
counterfactual performance stands in the same-provenance, but not
the same-structure relation to Beethoven’s actual performance.
A more comprehensive and nuanced model of our work concept would
need to be complicated in a variety of ways. And even then, no model will
perfectly capture our work concept. For one thing, folk-constructs such as
this vary somewhat with person, culture and domain of application. Never-
theless, I think this supervaluationism nicely explains our concept’s individ-
uative indeterminacy. It captures the profile of our intuitions, including
conflicted intuitions, that I elicited earlier in the chapter; and it explains why
our work concept functions perfectly well in the vast majority of cases in
which the two relations don’t come apart.
I will comment briefly on two complications at play in the examples I’ve
discussed. The first complication concerns vagueness in each of the two
criteria considered separately. We might allow that Beethoven and the
298 JOSEPH G. MOORE
Hawaiian genius could have indicated the same sound structure despite
some relatively minor differences in their musical specifications. But how
great can these differences be? Not as great as the significant harmonic
differences that yield the Honey-Hammer, I think, but we might allow
differences in key, for example, and slight differences in specified tempo and
ornamentation. Where’s the cut-off point? The allowable degree of vari-
ation seems vague.
The relation of same-provenance also seems loose. Beethoven could have
indicated his sonata later than he did, perhaps even after he had composed
what is actually counted as his 30th piano sonata (op. 109). But could the
very same Hammerklavier, provenancially considered, have been composed
by one of Beethoven’s contemporaries? Considered in the right setting,
I think it could, but I suspect vagueness here as well.21 In any case, as far as
I can see, the vagueness in these two criterial relations, and the best way to
understand it (perhaps with a second application of supervaluationism?) is
independent of my account of the indeterminate criterial mixing.
A second complication has to do with the way that our judgments of
work identity might vary with judgmental setting. The norms and standards
of different musical genres seem to make for differences in the way we
individuate works within them.22 But as I suggested in Section III, our
identity judgments might also vary, even concerning one and the same case,
with ‘judgmental setting’—that is, with the variable conversational, aes-
thetic and practical concerns that are at play when an identity judgment is
called for.23 It’s not my goal here to establish that our work concept is
21 Matheson and Caplan have recently argued that de re modal claims in this realm could be
reinterpreted as claims that are really about de dicto possibilities (see 2008, pp. 496–8). This may be, but
I don’t see that they give any positive reason in favor of such reinterpretation other than that this is
required by theories of work individuation that I’ve challenged in this chapter. Construing my putative
de re modal intuition to be what it seems is certainly more straightforward. If I’m wrong about this, the
same-provenance relation would simply be less flexible modally. In any case, my view problematizes an
application of the de re/de dicto distinction in this realm, since our work concept traffics in two types of
entities (as I’ll argue below).
22 See S. Davies (2001, ch. 1), where the notion of the relative ‘thickness’ of a work is introduced,
though I think this notion properly applies only to a work’s structural component. And see D. Davies
(2004, ch. 5) where provenancial thickness (my expression) is claimed to be work-relative.
23 In other areas of philosophy, these settings might be called judgmental or conversational ‘contexts’;
and a view that honored this type of variation would be called ‘contextualism.’ (See Moore 1999,
pp. 347ff. for an articulation of contextualism in this sense. And see Lewis (1996) for an influential
application of the view to knowledge reports.) But this usage would be confusing in this debate where
‘contextualism’ is already used to pick out a dependence of musical works upon musico-historical
surrounding.
deployed in this type of setting-variable way, but I strongly suspect it. If so,
this would add another layer of complexity to our model, and raise some
new questions.24 However, as far as I can see, this setting-dependence of our
work concept cuts across its individuative indeterminacy and the super-
valuation model I’ve just proposed.
VII. A Schizoid Metaphysics

The conceptual model I’ve just proposed nicely captures the way our work
concept is deployed in musical practice, but it leaves a crucial metaphysical
question unresolved: What do the two criteria track? What makes it the case
that two token performances are structurally or provenancially similar, or
both? The metaphysics that seems to me best to match the conceptual
model is almost flat-footedly simple: The distinct criteria track respectively
two distinct and largely distinguishable types of entities—‘structural-works’
(S-works) and ‘provenancial-works’ (P-works). But what are these?
Before pursuing this question, I emphasize the relative independence that
it has from the conceptual model. To be sure, there must be something
about the world that these criteria track, unless we are to leave the door
open to eliminativism about musical works. But my conceptual account
might remain entirely accurate even if the metaphysics I pursue turns out to
be misguided.
I think we can take structural-works simply to be sound structures. Sound
structures are not, of course, uncontroversial entities. We might wonder
which particular sequences should play the role of S-works. For example,
do sound structures include performance-means?25 As I noted earlier there
are general metaphysical worries about sound structures as abstracta. For
example, how do abstract sequences enter into the causal relations at work
in our putative reference to musical works, and our knowledge of them?
But my claim about the existence and structure of the individuative
indeterminacy in our work concept is, as far as I can see, compatible with
24 How much of this variability is due to a setting-dependent application of the two criteria separately,
and how much to a setting-dependent mixing of the criteria? (How can we tell?) And how can we best allow
this type of setting-dependence without turning judgmental evaluation into a subjective free-for-all?
25 See, for example, Davies (2008). Questions about the exact nature of sound structures strike me as
reflections of looseness in our structural criterion.
300 JOSEPH G. MOORE
a wide range of possible answers to such questions. In any case, these worries
don’t tell more forcefully against my position than the positions to which
I’ve argued it’s superior. We need confidence only that there is some
workable account of sound structures, or at least of sound structure talk.
The more novel construction is that of a provenancial-work.26 I think of
P-works as causal webs of intentionally linked action-tokens grounded
initially in status-endowing actions of a composer and the musical insti-
tutions that surround her. They spread diachronically from the actions
involved in composition and work-endowment through certain performa-
tive, evaluative, and representational events that those acts of composition
influence. Since the acts of composition, work-establishment, performance,
and evaluation are to be specified and understood relative to a background
of expressive and performative traditions that inform them, we can usefully
regard P-works as ‘tradition-threads’—that is, as individual threads in the
broader causal tapestry of musical, psychological, and social events upon
which a musical culture and tradition supervene.
An example might help. The P-work that corresponds to Beethoven’s
Hammerklavier piano sonata (call it the ‘P-Hammer’) is to be specified and
understood against a broad range of performance traditions, contingent
musico-expressive conventions, and prominent cultural events that would
have been at play in the Viennese musical culture in which Beethoven
composed his sonata. This expressive background is roughly a collection of
musical conventions that Beethoven and his contemporaries would have
implicitly regarded as the relevant musical backdrop against which a new
musical composition was to achieve its expressive effects.
The P-Hammer itself starts with some specific actions of Beethoven’s—
his deciding to compose another sonata, or his being commissioned to do
so. They extend through the process of composition, and the moment
when Beethoven (or a publisher or a deadline) determined that his com-
position was done. This is the period during which the P-Hammer came
into existence.
The P-Hammer continues by including the specific performative and
evaluative tradition that Beethoven’s composition has brought about. It
includes all token performances intended to be of Beethoven’s 29th piano
26 Predecessors here might include Rohrbaugh (2003), as well as Caplan and Matheson (2006, 2008),
though I can’t pursue here exactly how my notion compares to theirs.
sonata, or more broadly, performances (and playbacks) that stand in a chain

of referential intentions reaching back to Beethoven’s composition. It also
includes representations—thoughts and words about the Hammerklavier—
that are grounded in causal chains reaching back to Beethoven’s compo-
sitional actions.27 The P-hammer is, then, a single tread in the still on-going
musical tradition that includes all of Beethoven’s compositions and the
many performative and representational events they’ve spawned, as well as
the vastly more numerous musical events that comprise the on-going
history of Western classical music.
As I’ve sketched them, P-works supervene upon particular actions,
including particular psychological and social events. So, to the extent that
such action-tokens are temporally bounded, then P-works are as well. And
this means that P-works come into existence somewhere in the compo-
sitional process, and go out of existence when the causal effects of these
compositional acts come to an end—when the last performance has been
played, the last score has been burned, and no one thinks about the
Hammerklavier any more.
A number of questions remain for a fleshed out account of tradition-threads,
but this sketch suffices for my purposes here.28 It’s time now to situate
shiftism—to show how it allows us to maintain realism about musical works.
VIII. How to be a Realist about Musical Works

I’ve argued that bound up in our concept of a musical work are two distinct
criteria of individuation, one musico-structural and the other provenancial.
And I’ve also suggested that the relative force of the two criteria can shift
with judgmental setting. Usually the two criteria are harmlessly aligned. But
occasionally—when we confront certain examples or when we ask certain
philosophical questions about musical works—the criteria pull in different
27 My intent here is to capture the intuition that a work, even in its provenancial sense, might be kept
alive even if there were no more performances (or playbacks) of it. This might be so, I think, if a culture
continued to think and write about the work, or the work continued to influence the composition of
other works. A plausible minimal condition for the continued existence of P-works, suggested by
Rohrbaugh’s 2003 account of photographs, might be that there remains at least the potential for
performance.
28 I investigate these in a follow-up manuscript ‘Musical Works as Tradition-Threads.’
302 JOSEPH G. MOORE
directions. When this happens, our concept doesn’t yield a determinate

judgment about work individuation.
I’ve also argued that this conceptual indeterminacy is no reason to aban-
don realism about musical works. According to the metaphysics I’ve
sketched, the world contains individuatively determinate entities (sound
structures and tradition-threads) that our work concept gets at, and responds
to in a fashion nicely modeled by supervaluationism. When the application
of our work concept is unsettled we can, if we wish, fully describe the
situation in terms of these individuatively determinate entities. But a general
conceptual revision or divorce is uncalled for. Not only would sharpening or
abandoning our work concept be inefficient, but something important
would be lost. The distinct criteria arise from a cluster of features that are
bonded together in a unified work concept for artistic and cultural reasons—
by the way music is thought about in practice, appreciation, and evaluation.29
However, even if one agrees with all this, one might still wonder whether
the view really honors realism about musical works. For one thing, there is
not, metaphysically considered, a unified kind of entity that answers to our
work concept. If there are metaphysical categories, then sound structures
and tradition-threads are surely to be sorted differently—perhaps as abstracta
and concreta respectively. This on its own doesn’t undermine realism, of
course. A concept can be disjunctive in application without violating the
broadly realistic condition that it get relatively straightforwardly at entities
in the world. For example, we can competently apply the concept of jade
without realizing that it comprehends entities (jadeite and nephrite) that are
categorically distinct when considered from within a mineralogical frame-
work. And this small misfit between concept and reality isn’t enough to
undermine realism about jade.
But my view posits a bigger misfit than this. It holds that in the normal
case, there are two entities that answer to our talk of one musical work. And
I’ve argued that we can’t construe a musical work, in the contextualist way,
as a concatenation of these two entities. It’s surely surprising, then, to learn
29 This commitment to the existence and utility of a unified work concept rules my view out as a form
of ‘concept pluralism.’ Still, in allowing that our core-concept binds together distinct conceptual strands,
my view is a near neighbor. Indeed, as far as I can see, shiftism about the individuation of individual
musical works is compatible with pluralism about the application conditions for our artwork concept(s).
Christy Mag Uidhir and P.D. Magnus plausibly defend such a view (2011).
that when we count the thirty-two individual musical works that comprise
Beethoven’s complete piano sonatas, we traffic in sixty-four distinct entities!
If I’m right, our work concept involves a tacit semantic presupposition that
is false—namely, that there is a realm of metaphysically unproblematic entities
that answer in a clean, one-one fashion to our talk and counting of musical
works. This presupposition might have arisen from a sort of tacit and collective
linguistic pretense. It’s musically and culturally useful to making certain ‘work-
wise’ distinctions among groups of performances and topics of appreciation
and evaluation. Speakers within a musical practice adopt or simply inherit the
practice of talking as if there are ‘things’ to which these groupings and topics
answer in a one-one fashion. Individual such ‘things’ are given names which
draw revealingly on both musical and provenancial features. And a work
concept and a surrounding discourse are quickly up and running.
The false presupposition is never exposed by the normal conditions in
which the concept is applied. Most people never consider, and never need
to consider whether there is really one metaphysically coherent ‘thing’ that
‘The Hammerklavier Sonata’ picks out. A philosopher might point out that
this musical work can’t comfortably be identified with a score, with a psycho-
logical idea in Beethoven’s head, with a class of performances, or with any
other prima facie plausible candidate; and she might note that musical works
are philosophically strange in other ways besides. But the utility of the practice
is largely untouched, even if conceptually-minded composers might stretch it
by playing off the philosophy. In short, our work concept continues to
function perfectly well despite theorectical problems with the presupposition.
All of this might seem to make my theory a sort of non-eliminative
fictionalism about musical works, since it seeks to support our thought and
talk about musical works against a metaphysical backdrop that only imper-
fectly matches our conceptual scheme. But there is a crucial difference,
I think, between my view and the interesting work-fictionalisms that have
recently been proposed by Andrew Kania (2008) and by Ross Cameron
(2008). The central claim that those theories are fictionalist about—‘Musical
works exist’—is not, on my view, true only when considered one way but
not another, or true only relative to one way of speaking English and
not another.30 On my theory, this existence claim is true simpliciter. It’s
30 Thus, Kania distinguishes between the claim as made in descriptive metaphysics and as made in real
metaphysics (metaphysics of the ‘fundamental level’), while Cameron distinguishes between the truth of
304 JOSEPH G. MOORE
determinately true that musical works exist because it’s true on each
sharpening of our work concept—S-works exist, and so do P-works. This
seems enough for the view to merit the label ‘realism’ and avoid ‘fictionalism.’
In the end, the view defended here is perhaps best characterized not by
any label but by the truth-status it accords various claims about musical
works. While the existence claim comes out determinately true, the mis-
match my view situates between semantic functioning and semantic phe-
nomenology shows up in the indeterminacy it posits elsewhere. I have
argued at length, of course, that it can be indeterminate whether the actual
Hammerklavier is identical to the Hawaiian-Hammer, and also whether it is
identical to the Honey-Hammer. But the view also holds, for example, that
it is indeterminate whether sound structures are musical works.31 And
although I have not argued for it here, the view also allows, more notably,
that it is indeterminate whether a musical work can change musically over
time, and whether musical works can come into and go out of existence.
Tradition-threads satisfy these conditions, but sound structures don’t.
The mismatch is not that such claims seem to us to be clearly true, or to
be clearly false. Indeed, I think conflicted intuitions about these issues help
sustain debate about the nature of musical works. The mismatch is rather
that my theory challenges the assumption, connected to the false presuppo-
sition discussed above, that these claims must have some determinate truth-
value or other—that is, that unless we are to abandon our work concept or
regard it with suspicion, these questions about the nature of musical works
this sentence in English and the falsity of a homophonic sentence in ‘Ontologese,’ English’s metaphysi-
cally more considered counterpart. Despite my sympathy for these views, I worry that musical and
philosophical discourse, especially concerning such existence claims, doesn’t always fall cleanly and
discernibly on just one side of some such linguistic-cum-semantic divide. And my theory has the
advantage of not requiring one. In any case, these fictionalist views are driven by the desire to avoid
the eternal existence of musical works, while (as I’ll observe in a movement) my view partially obviates
this motivation by allowing that the existence conditions for musical works are indeterminate.
31 The sharpening ‘sound structures are sound structures’ is true while the sharpening ‘sound
structures are tradition-threads’ is false. This result—that neither S-works nor P-works are determinately
musical works—accords nicely, I think, with the philosophical view I’ve tried to motivate. I should
acknowledge, however, that not everything runs smoothly in this world of supervaluations. For
example, interpreted in a straightforward fashion ‘Musical works are individuatively determinate’ is
determinately true; and of course this seems to be precisely the claim I’ve been arguing against!
However, it’s not implausible, I think, to give this sentence a meta-linguistic reinterpretation. In fact,
I suggest that the sentence really claims that talk of an individual musical work univocally refers to a
unique and individuatively determinate entity. And the supervaluation account of our work concept
allows us to deny this claim.
must have clear and determinate answers. I hope I’ve shown that this
needn’t be so: We can hold that musical works exist even if certain philo-
sophically unsettled claims about them are indeterminate in truth value.
References
Cameron, R. (2008) ‘There Are No Things that are Musical Works’ British Journal
of Aesthetics 48: 295–314
Caplan, B. and Matheson, C. (2006) ‘Defending Musical Perdurantism’ British
Currie, G. (1989) An Ontology of Art (New York: St Martin’s Press)
Davies, D. (2004) Art as Performance (Oxford: Blackwell)
Davies, S. (1997) ‘John Cage’s 4’ 33”’ Australasian Journal of Philosophy 75: 448–62
——(2001) Musical Works and Performances: A Philosophical Exploration (New York:
Clarendon Press/Oxford University Press)
——(2003) Themes in the Philosophy of Music (Oxford: Oxford University Press)
——(2007) ‘Versions of Musical Works and Literary Translations’ in K. Stock (ed)
Philosophers on Music: Experience, Meaning and Work (Oxford: Oxford University
Press)
——(2008) ‘Musical Works and Orchestral Colour’ British Journal of Aesthetics
48: 363–75
Dodd, J. (2007) Works of Music: An Essay in Ontology (Oxford: Oxford University
Press)
Evans, G. (1984) ‘Can There Be Vague Objects?’ Analysis: 38
Field, H. (1973) ‘Theory Change and the Indeterminacy of Reference’ Journal of
Goehr, L. (1992) The Imaginary Museum of Musical Works: An Essay in the Philosophy
of Music (Oxford: Oxford University Press)
Gracyk, T. (1996) Rhythm and Noise—An Aesthetics of Rock (Durham, NC: Duke
University Press)
Kania, A. (2006) ‘Making Tracks: The Ontology of Rock Music’ The Journal of
Aesthetics and Art Criticism 64(4): 410–14
——(2008) ‘The Methodology of Musical Ontology: Descriptivism and its Impli-
cations’ British Journal of Aesthetics 48(4): 426–44
Kivy, P. (1983) ‘Platonism in Music: a Kind of Defense’ Grazer Philosophische
Studien 19: 109–29
306 JOSEPH G. MOORE
Kivy, P. (1987) ‘Platonism in Music: Another Kind of Defense’ American Philosophical

Quarterly 24: 245–52
Levinson, J. (1980) ‘What a Musical Work Is’ Journal of Philosophy 77: 5–28
——(1990) ‘What a Musical Work Is, Again’ in J. Levinson (ed) Music, Art, and
Metaphysics (Ithaca, NY: Cornell University Press)
Lewis, D. (1996) ‘Elusive Knowledge’ Australasian Journal of Philosophy 74(4): 549–67
Mag Uidhir, Christy and Magnus, P.D. (2011) ‘Art Concept Pluralism’ Metaphilo-
sophy 42: 83–97
Matheson, C. and Caplan, B. (2008) ‘Modality, Individuation, and the Ontology of
Art’ Canadian Journal of Philosophy 38(4): 491–517
Moore, J. (1999) ‘Misdisquoation and Substitutivity: When Not to Infer Belief
from Assent’ Mind 108(430): 335–66
——(2008) ‘A Modal Argument Against Vague Objects’ Philosophers’ Imprint 8(12):
1–17
Quine, W.V.O. (1969) Ontological Relativity and Other Essays (New York: Cornell
University Press)
Rohrbaugh, G. (2003) ‘Artworks as Historical Individuals’ European Journal of
Salmon, N. (1982) Reference and Essence (Princeton: Princeton University Press)
Thomasson, A.L. (2005) ‘The Ontology of Art and Knowledge in Aesthetics’ The
Journal of Aesthetics and Art Criticism 63(3): 221–9
van Fraassen, B. (1969) ‘Presuppositions, Supervaluations and Free Logic’ in
K. Lambert (ed) The Logical Way of Doing Things (New Haven: Yale University
Press)
Walton, K. (1970) ‘Categories of Art’ Philosophical Review 79: 334–67
Wolterstorff, N. (1980) Worlds and Works of Art (Oxford: Clarendon Press)
Index
absolutely general quantification 101 computer program 69–79, 71n6, 72–3, 77

abstract artifact 77–80 composer 64–5, 73–4, 77
adverbs of quantification 137 conceptual role 30, 32, 38
algorithm 70–3, 77 conservation 243, 248
analogical predication 136n contextualism 285–95
anaphoric predication 136–7, 141, 146–8, copying 174–6
150, 154–5 Cox, Renee 170n
anaphoric reference 136, 138–9, 140, 142, creation 10, 15, 33, 46, 62–5, 73–81, 172,
144, 149 233–7
Alexander, Samuel 233 critical practice 251–4, 255n, 260n
appropriation 250 curating 243–4, 248, 250–1
arithmetic 64, 66, 101n, 102–6
artform 62, 68, 70, 77, 79, 81 Danto, Arthur 50, 93–6, 176–7
art ontology David 225
revisionist 4 Davies, David 31, 33, 57n, 199–202, 205,
descriptivist 4 243, 253, 263n
deference view of 3 Davies, Stephen 64–5, 163, 169, 171n, 174,
independence view of 3 175, 268, 269, 285–6
reciprocity view of 18 definition 90, 99, 104–5
artist 62, 70, 72 deflationism 256, 261
artistic indication 49–61 dependence, ontological 64, 70, 77
artist’s sanction 245–7, 258 destruction 62–4, 73–81
Discobolus 226
Barnes, Elizabeth 190 discovery 64, 75, 80
Benjamin, Walter 176 Dodd, Julian 13, 14n, 33, 35–6, 56n, 64–5,
Borges, Jorge Luis 50, 170 108–9, 113–15, 118–20, 123, 161,
Boyd, Richard 36, 120–2 163–4, 168, 171–2, 174, 180, 198–9,
Brown, Lee 213 201–3, 265, 267–68, 267 n.6, 270,
270 n.6, 271, 272, 274, 274 n.7,
Cage, John 257–8 277–79, 280
Cameron, Ross 32–4, 162n, 167, 208–9, ducks 117–22
256n
Caplan, Ben 36, 172, 180, 188, 206, 252–3 eliminativism 30, 32, 208–11, 215–16
cardinal number 97–102 endurantism 252–4
Carlson, Gregory 127n, 137n, 139–42 essential properties 165–8
category, ontological 62, 71, 75, 77–8 Evans, Gareth 185, 188–90
class 66–7, 80 events 254–8
compliance class 67 Everett, Anthony 185–6, 189
empty class 66–7, 73 expressivism 294
equivalence class 66–7, 72–3, 76–7, 80
impure 76–7 family resemblance 89
collective predication 143–5, 147, 148–50, Ferreira, Michael 120
152–3 fictionalism 32, 34, 110, 116, 210–17, 257,
computer art 63, 67–73, 76 303–4
308 INDEX
Field, Hartry 211–12 Laocoön and His Sons 223

Fine, Kit 181 Levinson, Jerrold 33, 35, 44–5, 62, 64–5,
first-order predication 144–6, 150–1, 73–6, 79, 163, 169, 172, 204, 257–8,
153–4 267, 268, 278–9, 285
formalism 293–4 Lewis, David 163, 189
logical form 125–6, 146
games 255–7 Lopes, Dominic McIver 67–72
GEN operator 137–42, 145, 150
Ghiselin, Michael 117, 119 McGonigal, Andrew 191
Gödel, Kurt 103–5 Mag Uidhir, Christy 10n, 62n, 78, 116, 246–7
Gödelian indefinite extensibility 103–5 Magidor, Ofra 189
Goehr, Lydia 213–14, 290–1 Magor, Liz 244–6, 248, 254–5, 258
Goodman, Nelson 33, 62, 67, 75–6 Markosian, Ned 183
Gonzales-Torres, Felix 242, 255n, 259 materialism 207–11, 215–16
mathematical objects 66, 165–8
Hamilton, James 243n mathematics 63, 65–6, 71n
Hammerklavier Sonata 284–305 Matheson, Carl 172, 180, 206, 252–3
Harrison, Rachel 251–2 Mayr, Ernst 164
Hazlett, Allan 254–5 Mazurka in A minor 49, 52–5
higher-order predication 143–7, 151, 154 mereology 234
homeostatic property clusters 120–3 metaontology 181, 195
Hull, David 117, 119 methodology 29, 30, 32, 37, 38
metaphysical constraint 202–3, 206
identity 66–7, 70–3 modal 31, 34, 35, 36
improvization 53, 245–6, 258 Muller, Cathleen 188
impure set 63, 67, 72, 76 multiple artworks 264, 268–9
incompleteness theorems 103 musical interpretation 246, 251
indeterminacy 90, 97, 99, 104–6, musical materialism 129–34
295–6 musical endurantism 130, 131–4, 155
indeterminate identity 185–91, 194–5 musical perdurantism 130, 131
indefinite extensibility 88, 96–107 musical ontology 197–217, 252–4, 257–60
indicated structures 49–61, 74 musical value 212–15
inscription 66–7, 73, 75 musical works 62, 64–7, 73–7, 162–3
installation art 242–61
instance of a work 64–5, 66, 69–70, 72–4, natural number 88, 97, 102–6
264–6 necessary condition 87, 89–95, 97, 104
intention 77–9 necessity of identity 189–90
interpretation 245, 256 nominalism 34, 67, 75–6, 181–2, 185, 192–3,
intuitionistic logic 105 195, 205–10, 215–17
Irvin, Sherri 243n, 245, 258n, 260 norms 30, 44, 45, 46, 251, 254–6, 258–60
norm types 65, 265
judgment 29, 34, 40, 41, 42, 43, 44, 46 number 66, 79
Kania, Andrew 32–4, 38–41, 110, 115–16, obscure predication 128, 135–7, 154
198, 200, 202, 214 open-endedness 87–106
Kearns, Stephen 189 simple schema 91
Kivy, Peter 33, 64–6, 74, 163, 169, 170n, strong modal schema 92, 101
171–2 weak modal schema 92, 101, 104
Koslicki, Kathrin 143–7, 151, 154 weak modal non-relational
Kraut, Robert 37, 38, 39, 42, 43 schema 95–6, 101, 104
INDEX 309
ordinal number 87, 97–102 simple indication 49, 52–3

Ozymandias 49, 55 simple semantics 127–37, 140–1, 154, 155
ontological implications of 125–6, 128,
painting 70, 78, 242, 247–8, 258–9 155–6
parameters 242–3, 245, 247–8, 251–2, 254, social object 77–80
257n, 260–1 sonicism 13, 14n
particular 62, 66, 69, 79–80 sorites paradox 189
perdurantism 252–4 sound-structures 65, 74, 299–300
performance 62, 64–7, 76–7, 79 species 35–6, 44, 117, 164
performing arts 242–61 Starry Night 225
persistence 235–6 Stecker, Robert 200–1
photography 246–7, 250, 255–60 straightforward predication 128–35, 137,
Pierre Menard 50, 55 140–1, 154
Platonism 64–6, 73–5, 80–1, 198–208 Strawson, Peter 32, 40
Plessas, Angelo 68, 71 structuralism 286–95
pluralism 116 structure 32, 34, 40, 41, 42, 44, 45, 65–6,
pragmatic constraint 199–202, 205, 210 73–4, 80
pragmatism 30, 32, 34, 36, 42, 44, 46 sufficient condition 87, 89–91, 94, 97, 104
practice 29, 30, 31, 32, 33, 34, 35, 37, 39, 41, supervaluation 296–7
42, 43, 44, 45, 46 Szabó, Zoltan Gendler 203–4
Predelli, Stefano 127n, 139–41
proper class 99–100 The Thinker 237–9, 269–70
Thomasson, Amie 17n, 31, 38–9, 77–9,
qua objects 56 109–10, 182–5, 253, 255–7, 261
Quine, W.V.O. 31, 75–6, 295–6 Tillman, Chris 207–8, 252–4
token 63, 66, 75, 80
Rayo, Augustin 167 tradition-threads 300–1
realism 295, 301–5 truth 99, 102–6
repeatable 29, 34, 35, 36, 38, 45, 46, 62–3, truthmaking 181–5, 192–3, 195
66, 73, 75, 78–9 types 30, 33, 38, 39, 41, 42, 45, 63, 75–7,
Rodin, Auguste 237–9 79–80, 265, 267
Rohrbaugh, Guy 108–9, 111, 120, 161, Aristotelian 64–5
163–4, 168, 172, 252n, 270–3, 274n Doddian types 267–8, 274, 278–82
rule 44, 45 Platonic 62–6, 73–5, 80
Russell’s paradox 141–2, 144 Wollheimian types 274–82
Samuels, Richard 120 universal 62, 64–6, 79

Schnieder, Benjamin 186–8 unvoiced operators
Sculpture 242, 247–8, 258–9 GEN operator (see independent listing)
ambient space of 231–3 G (generic) operator 150, 152–5
cast 237–9 C (collective) operator 150, 152–3
creation vs. discovery of 233–7
holes 227–8 vagueness 297–8
solidity of 227–8 van Fraassen, Bas 211
temporality of 226 van Inwagen, Peter 183
three-dimensionality of 224–6 von Solodkoff, Tatjana 186–8
visual and tactile assessment of 228–31
set 63, 76, 87–8, 90, 97–102, 104–5 Weiner, Lawrence 245, 259
set theory 76, 87–8, 90, 97–102, 104–5 Weitz, Morris 87–90, 92–5, 97, 104
310 INDEX
Williams, J.R.G. 181 Wolterstorff, Nicholas 33, 54, 64–5, 67,

Winged Victory of Samothrace 225 163, 265, 267, 268, 269, 274,
Wittgenstein, Ludwig 89–90, 104 277, 281
Wollheim, Richard 33, 62, 75, 265, 268,
274–7 Yablo, Stephen 165

Art and Abstract Objects

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Art and Abstract Objects

Hochgeladen von

Copyright:

Verfügbare Formate

Art and Abstract Objects

This page intentionally left blank

Christy Mag Uidhir

List of Contributors vii

Introduction: Art, Metaphysics, and the Paradox of Standards 1

I. General Ontological Issues

II. Informative Comparisons

III. Arguments Against and Alternatives To

9. Platonism vs. Nominalism in Contemporary Musical Ontology 197

IV. Abstracta Across the Arts

ROSS P. CAMERON, Leeds University

aesthetics is an at best second-rate philosophical ﬁeld. In fact, I think such

I. Two Opposing Views

The Deference View: In all cases of relevant overlapping areas, aesthetics

practices unique to art-relevant domains requires carving out (perhaps

conventions. Contemporary metaphysics being unable to provide ontic

is the degree to which philosophical aesthetics holds its development and

II. Art-Abstracta and the Paradox of Standards

I. There are such things as art-abstracta.

The general impetus for and broad attraction to positing artworks of a

III. Resolutions from Deference and Independence

art-abstracta and to abstracta being causally-inert non-spatiotemporal

14 Most notably defended in Dodd (2007).

standards seems to secure a metaphysical tidy art ontology only to ﬁll it

of art-ontological enquiry being correctly adjusted so as to capture the

Of course, absent some strict causation requirement, this non-stand-

From the point of view of independence, resolution to the paradox of

IV. An Alternative View

end of the day those metaphysical commitments are incompatible with a

compatible with all issues central to contemporary metaphysics. Rather,

some controversial and sweeping meta-ontological position sufﬁciently

Where deference and independence marginalize, enervate, and insulate

artworld. Consequently, if one’s methodology were not consistent with and

constraints on ontological proposals without saying what meets them,

Orthodox views hold that species membership is a part-whole relation

projectable categories undergirding causally sustained generalization in one

This is, of course, precisely the point at which a thorough-going pragma-

contrast to which the former kind of description must be mere description,

be played with historically authentic instruments or that they ought not to

As I have tried to indicate, it is not the place of ontology to offer expla-

VI. A Final Suggestion

——(2008) ‘Defending “Defending Musical Perdurantism” ’ British Journal of

individual human that lies at the heart of such an artwork. A paradigmatic

should be distinguished from more ordinary kinds of indication, which we

possible future instantiation, he aims in addition to establish something

The key idea in the above formulation is that of ordaining a rule to be

8 Wolterstorff (1994, p. 120).

verbal world and a concrete individual human being, in other words, to

initiated types, such instantiation-exempliﬁcation occurs through musical

14 See the relevant essays in Levinson (1990a, 1996b, 2006).

will, intentionality, or goal-directedness, and amounting, more or less, to

——(2006) Contemplating Art (Oxford: Oxford University Press)

With these two examples, there is a case to be made that abstractness

common to two concrete pattern instances—the Pattern, as it were—seems

two unperformed works would be identical: hardly a welcome consequence

III. Computer Art

3 Wolterstorff is aware of this: compare Wolterstorff 1980, p. 100.

of microchips, depending on the architecture of the computer. At none

He observes that being produced by the same algorithm is a necessary7

(p. 216). The indicated structure (indicated by a person at a time in a socio-

different from discovering them, or drawing our attention to them, as

construe types as creatable and destructible, makes them utterly gratuitous

however, like algorithm-determined equivalence classes of displays or