Beruflich Dokumente
Kultur Dokumente
Contents
List of contributors
page ix
Introduction
Penelope Rush
part i
11
1 Logical realism
13
Penelope Rush
32
Jody Azzouni
49
Stewart Shapiro
72
Solomon Feferman
93
Penelope Maddy
Logical nihilism
109
Curtis Franks
128
Mark Steiner
part ii
vii
145
147
Contents
viii
160
Gyula Klima
178
Ermanno Bencivenga
11
189
Sandra Lapointe
part iii
12
specific issues
Revising logic
209
211
Graham Priest
13
224
14
233
Tuomas E. Tahko
References
Index
249
264
Contributors
ix
List of contributors
Introduction
Penelope Rush
This book is a collection of new essays around the broad central theme of
the nature of logic, or the question: what is logic? It is a book about logic
and philosophy equally. What makes it unusual as a book about logic is
that its central focus is on metaphysical rather than epistemological or
methodological concerns.
By comparison, the question of the metaphysical status of mathematics
and mathematical objects has a long history. The foci of discussions in the
philosophy of mathematics vary greatly but one typical theme is that of
situating the question in the context of wider metaphysical questions:
comparing the metaphysics of mathematical reality with the metaphysics
of physical reality, for example. This theme includes investigations into: on
exactly which particulars the two compare; how (if ) they relate to one
another; and whether and how we can know anything about either of
them. Other typical discussions in the eld focus on what mathematical
formalisms mean; what they are about; where and why they apply; and
whether or not there is an independent mathematical realm. A variety of
possible positions regarding all of these sorts of questions (and many more)
are available for consideration in the literature on the philosophy of
mathematics, along with examinations of the specic problems and attractions of each possibility.
But there is as yet little comparable literature on the metaphysics of
logic. Thus the aim of this book is to address questions about the
metaphysical status of logic and logical objects analogous to those that
have been asked about the metaphysical status of mathematical objects
(or reality). Logic, as a formal endeavour has recently extended far
beyond Freges initial vision, describing an apparently ever more complex realm of interconnected formal structures. In this sense, it may
seem that logic is becoming more and more like mathematics. On the
other hand, there are (also apparently ever more) sophisticated logics
1
Penelope Rush
describing empirical human structures: everything from natural language and reason, to knowledge and belief.
That there are metaphysical problems (and what they might be) for the
former structures analogous to those in the philosophy of mathematics is
relatively easily grasped. But there are also a multitude of metaphysical
questions we can ask regarding the status of logics of natural language and
thought. And, at the intersection of these (where one and the same logical
structure is apparently both formal and mathematical as well as applicable
to natural language and human reason), the number and complexity of
metaphysical problems expands far beyond the thus far relatively small set
of issues already broached in the philosophy of logic.
As just one example of the sorts of problems deserving a great deal more
attention, consider the relationship between mathematics and logic.
Questions we might ask here include: whether mathematics and logic
describe the same or similar in-kind realities and relatedly, whether there
is a line one can denitively draw between where mathematics stops and
logic starts. Then we could also ask exactly what sort of relationship this is:
is it one of application (of the latter to the former) or is it more complex
than this?
Another central problem for the metaphysics of logic is that of pinning
down exactly what it is that logic is supposed to range over. Logic has been
conceived of in a wide variety of ways: e.g. as an abstraction of natural
language; as the laws of thought; and as normative for human reason. But,
what is the thought whose structure logic describes; how natural is the
natural language from which logic is abstracted?; and to what extent does
the formal system actually capture the way humans ought to reason?
As touched on above, a key metaphysical issue is how to account for the
apparent double role applying to both formal mathematical and natural
reasoning structures that (at least the main) formal logical systems play.
This apparent duality lines up along the two central, indeed canonical
applications of logic: to mathematics and to human reason, (and/or human
thought, and/or human language). In many ways, the rst application
suggests that logic may be objective or at least as objective as mathematics, in the sense that, as Stewart Shapiro puts it (in this volume) we might
say something is objective if it is part of the fabric of reality. This in turn
might suggest an apparent human-independence of logic. The second
application, though, might suggest a certain subjectivity or intersubjectivity; and so in turn an apparent human-dependence of logic,
insofar as a logic of reason may appear dependent on actual human
thought or concepts in some essential way.
Introduction
Both the apparent objectivity and the apparent subjectivity of logic need
to be accounted for, but there are numerous stances one might take within
this dichotomy, including a conception of objectivity that is nonetheless
human-dependent. In Chapter 4, Solomon Feferman reviews one such
example in his non-realist philosophy of mathematics, wherein the
objects of mathematics exist only as mental conceptions [and] . . .
the objectivity of mathematics lies in its stability and coherence under
repeated communication. Others of the various positions one might take
up within this broad-brush conceptual eld are admirably explored in both
Stewart Shapiros and Graham Priests chapters, though from quite dierent stand points: Shapiro explores the nuances and possibilities in conceptions of objectivity, relativity, and pluralism for logic, whereas Priest looks
at these issues through the specic lens aorded by the question whether or
not logic can be revised.
There are, then, a variety of possible metaphysical perspectives we can
take on logic that, particularly now, deserve articulation and exploration.
These include nominalism; naturalism; structuralism; conceptual
structuralism; nihilism; realism; and anti-(or non-)realism, as well as
positions attempting to steer a path between the latter two. The following
essays cover all these positions and more, as defended by some of the
foremost thinkers in the eld.
The rst part of the book covers some of the main philosophical
positions one might adopt when considering the metaphysical nature of
logic. This section covers everything from an extreme realism wherein logic
may be supposed to be completely independent of humanity, to various
accounts and various degrees in which logic is supposed to be in some way
human-dependent (e.g. conceptualism and conventionalism).
In the rst chapter I explore the feasibility of the notion that logic is
about a structure or structures existing independently of humans and
human activity. The (typically realist) notion of independence itself
is scrutinised and the chapter gives some reasons to believe that there is
nothing in principle standing in the way of attributing such independence
to logic. So any benets of such a realism are as much within the reach of
the philosopher of logic as the philosopher of mathematics.
In the second chapter, Jody Azzouni explores whether logic can be
conceived of in accordance with nominalism: a philosophy which might
be taken to represent the extreme opposite of realism. Azzouni argues the
case for logical conventionalism, the view that logical truths are true by
convention. For Azzouni, logic is a tool which we both impose by convention on our own reasoning practices, and occasionally also to evaluate
Penelope Rush
them. But Azzouni shows that although there seems to be a close relationship between conventionality and subjectivity, logics being conventional
does not rule out its also applying to the world.
Stewart Shapiro, in the third chapter, argues the case for logical relativism or pluralism: the view that there is nothing illegitimate in structures
invoking logics other than classical logic. Shapiro defends a particular sort
of relativism whereby dierent mathematical structures have dierent
logics, giving rise to logical pluralism conceived of as [the] view that
dierent accounts of the subject are equally correct, or equally good, or
equally legitimate, or perhaps even (equally) true.
Shapiros chapter looks in some depth at the relationship between
mathematics and logic, identied above as a central problem for our
theme. But in particular, it investigates the extent to which logic can
be thought of as objective, given the foregoing philosophy. He oers a
thorough, precise, and immensely valuable analysis of the central concepts,
and claries exactly what is and is not at stake in this particular debate.
In the fourth chapter, Solomon Feferman examines a variety of logical
non-realism called conceptual structuralism. Feferman shares with Shapiro
a focus on the relationship between mathematics and logic, extending the
case for conceptual structuralism in the philosophy of mathematics to logic
via a deliberation on the nature and role of logic in mathematics. He draws
a careful picture of logic as an intermediary between philosophy and
mathematics, and gives a compelling argument for the notion that logic,
as (he argues) does mathematics, deals with truth in a given conception.
According to Fefermans account, truth in full is applicable only to
denite conceptions. On this picture, when we speak of truth in a
conception, that truth may be partial. Thus classical logic can be conceptualised as the logic of denite concepts and totalities, but may itself be
justied on the basis of a semi-intuitionist logic that is sensitive to
distinctions that one might adopt between what is denite and what is
not. Feferman shows how allowing that dierent judgements may be
made as to what are clear/denite concepts, aords the conceptual
structuralist a straightforward, sensible and clear understanding of the role
and nature of logic.
Penelope Maddy, in the fth chapter, oers a determinedly secondphilosophical account of the nature of logic, presenting another admirably
clear and sensible account, focusing in this case on the question why logic
is true and its inferences reliable. Second Philosophy is a close cousin of
naturalism as well as a form of logical realism and involves persistently
bringing our philosophical theorising back down to earth.
Introduction
Penelope Rush
Introduction
about logic still open today: e.g. Thom notes that for Aristotle, the types of
things that can belong to the categories are outside the mind or soul,
and so Kilwardbys analysis clearly relates to our modern question as to the
possible independence and objectivity of logic. The complexity of that
question is brought to the fore in Kilwardbys detailed consideration of
the various aspects under which stateables can be considered, and
according to which they may be assigned to dierent categories.
Thoms chapter goes on to oer a framework for understanding later
thinkers and traditions in logic, some of which (e.g. Bolzano in Lapointes
chapter) are also discussed in this part. His concluding section ably
demonstrates that understanding the history of our questions casts useful
light on the modern debate.
Gyula Klima also discusses strategies for dealing with the two way pull
on logic from its apparent abstraction from human reason and from its
apparent groundedness in the physical world. Klima focuses on the scholastics, comparing the semantic strategies of realists and nominalists
around Ockhams time. One of these was to characterise logic as the study
of second intentions concepts of concepts. Klima points out that when
logic is conceived of in this way, the core-ontology of real mindindependent entities could in principle have been exactly the same for
realists as for Ockhamist nominalists; therefore, what makes the
dierence between them is not so much their ontologies as their dierent
conceptions of concepts, grounding their dierent semantics.
Klima argues that extreme degrees of ontological and semantic diversity
and uniformity mark out either end of a range of possible positions
concerning the relationship between semantics and metaphysics, [from]
extreme realism to thoroughgoing nominalism and points out how the
conceptualisation of the sorts of things semantic values might be varies
according to where a given position sits within this framework. His chapter
illuminates the metaphysical requirements of dierent historical
approaches to semantics and the way in which the various possible metaphysical commitments we make come about via competing intuitions
regarding diversity: whether we locate diversity in the way things are or
in the way we speak of or conceptualise them.
In the next chapter, Ermanno Bencivenga picks up a thought Thom
touches on in his closing paragraph namely that our modern conception
of logic appears to have lost touch with the relevant ways in which actual
human reason can go wrong other than by not being valid. Oering a
Kantian view, Bencivenga suggests we adjust our conception of logic to
that of almost any structure we impose on language and experience, just so
Penelope Rush
Introduction
10
Penelope Rush
already in place and that (from this perspective) ought to be, between an
interesting set of metaphysical intuitions and those laws of logic we take to
be true.
In all, this book ranges over a vast terrain covering much of the ways in
which our beliefs about the role and nature of logic and of the structures it
describes both impact and depend on a wide array of metaphysical positions. The work touches on and freshly illuminates almost every corner of
the modern debate about logic; from pluralism and paraconsistency to
reason and realism.
part i
chapter 1
Logical realism
Penelope Rush
1. The problem
Logic might chart the rules of the world itself; the rules of rational human
thought; or both. The rst of these possible roles suggests strong similarities
between logic and mathematics: in accordance with this possibility, both logic
and mathematics might be understood as applicable to a world (either the
physical world or an abstract world) independent of our human thought processes.
Such a conception is often associated with mathematical and logical realism.
This realist conception of logic raises many questions, among which
I want to pinpoint only one: how logic can at once be independent of
human cognition in the way that mathematics might be; and relevant to
that cognition. The relevance of logic to cognition or, at the very least,
the human ability to think logically seems indubitable. So any understanding of the metaphysical nature of logic will need also to allow for a
clear relationship between logic and thought.1
The broad aim of this chapter is to show that we can take logical
structures to be akin to independent, real, mathematical structures; and
that doing so does not rule out their relevance and accessibility to human
cognition, even to the possibility of cognition itself.
Suppose that logical realism involves the belief that logical facts are
independent of anything human:2 that the facts would have been as they
1
Two things: note I do not claim we can or ought to show that logic underpins, describes, or arises
from cognition. In fact I think the relationship between thought and logic is almost exactly analogous
to that between thought and mathematics (see Rush (2012)), and I disagree with the idea that there is
any especially signicant connection between logic and thought beyond this. Two: while this chapter
deals with the notion of independence per se, it investigates this from the perspective of applying
that notion especially to logic. That is, my main aim here is to indicate one way in which the realist
conception of an independent logical realm might be considered a viable philosophical position but
one primary way I hope to do this is by showing how attributing independence to logic need be no
more problematic than attributing independence to anything else (e.g. by arguing that the realist
problem applies across any type of reality which is supposed to be independent).
See Lapointes characterisation as IND in this volume.
13
14
Penelope Rush
Folina was talking about mathematical realism, but the sort of logical realism I want to examine here
is directly analogous to mathematical realism in this respect.
Lapointe (this volume) explores a variety of reasons that may play a role in holding some version of
logical realism, so I wont go into these in depth here.
For more on the possible entities a logical realist might posit (e.g. meanings/propositions), see
Lapointe (this volume). Regardless of which entities are selected and where these are situated on the
Logical realism
15
abstractphysical scale of possible entities, just so long as the realist also posits IND (Lapointe, this
volume), theyll encounter some version of Benacerraf s or Sellars problem.
For more on the nuances of independence available to the realist, see Jenkins (2005) I take
essential independence to follow from modal independence, and I take modal independence as
characteristic of the sort of realism I want to explore.
16
Penelope Rush
As will become clear, I have a very particular notion of duality in mind here i.e. a (contradictory)
duality of object: one that is also two rather than a duality of an objects role, or aspect, or
components, etc: one that has two aspects/dimensions/components, etc.
Logical realism
17
Caveat: Id like to argue that the predicament can play this role just so far as the basic idea of an
independent reality existing at all can. It is the latter that I see the framework in Husserls ideas as
able to directly establish.
Or, again, to illustrate how conceiving logic as an independent objective structure akin to
mathematics need not be considered an especially problematic instance of the general idea of
independent reality itself, once that idea is eectively defended.
18
Penelope Rush
Thus both logic and mathematics, for Husserl, have a pure, abstract,
theoretical, denite, and axiomatic foundation. Further, Husserl
believed that:
one cannot describe the given phenomena like the natural number series or
the species of the tone series if one regards them as objectivities in any other
words than with which Plato described his ideas: as eternal, self-identical,
untemporal, unspatial, unchanging, immutable. (Hartimo 2010a: 115118,
italics mine)
Logical realism
19
Inextricability
Thanks to Curtis Franks for help with the expression of this point.
20
Penelope Rush
In the above quotes, both the presenting objects and the manner in
which they present give cognition its essential structure. It seems that
Husserl resists resolving the ambiguity in these phrases one way or
the other.
Husserls phenomenology of cognition is accomplished through a
prior conceptual step called the phenomenological reduction. This
reduction is related to Descartes method of doubt (e.g. in Husserl
1964: 23. A useful elaboration can be found in Teiszen 2010: 80). Teiszen
argues that for Husserl the crucial thing about the phenomenological
reduction was what remains even after we attempt, in Cartesian fashion,
to doubt everything. Teiszen makes the point that if we take a (certain,
phenomenologically mediated) transcendental perspective, we can uncover
in what remains (after Cartesian doubt) a lot more than an I who is
thinking. In particular, we can uncover direct apprehension of the ideal
objects of logic and mathematics (Teiszen 2010: 9) whose pure forms
extend far further than what Descartes ended up allowing as directly
knowable, and further than the knowable allowed for in Kants
philosophy.
Just as there is with what to make of the consciousness as consciousness
of idea, so too there is much controversy surrounding exactly what the
phenomenological reduction is and involves. To say that there is disagreement here among Husserl scholars is something of an understatement.
Indeed: there seem[s] to be as many phenomenologies as phenomenologists (Hintikka 2010: 91).
But the clarication of exactly what Husserl may have meant is not
relevant to my purpose here, which is to see if there are ideas we can draw
from Husserl that might help a realist philosopher of logic.
I pause to note, though, that Teiszens interpretation of the reduction as
a suspension or bracketing of the (natural) world and everything in it
(Teiszen 2010: 9) is standard; and the ideal objects recovered in Teiszens
consequent transcendental idealism (including their constituted mindindependence) are also standard for an established tradition of Husserl
scholarship (adhered to by Fllesdal, among others). But these ideal
objects are very far from the realist mind-independent realm that I want
to imagine has a place here (to hammer this point home, see Teiszen
2010: 18).
Again, it is the (possibly resolute) ambiguity in Husserls account
that allows for my alternate reading of phenomenology. Another case
in point: the description on essential lines of the nature of
consciousness . . . leads us back to the corresponding description
Logical realism
21
of the object consciously known (Husserl 1983: 359). The phrase: the
object consciously known is ambiguous. It can be read dierently
depending on each terms specic interpretation and on which terms
are emphasized: e.g. the consciously known can be read as the object as
we know it (i.e. a strictly constituted internal object); or as the
object that is known. It is the latter interpretation that opens the
possibility of an external element in the basic ingredients of the nature
of consciousness.
To reiterate: the interesting thing about Husserl for my purpose is
that in his ideas we can discern a (at least potential) role for an
independent objective other, while nonetheless focusing on experience
and consciousness: my thought is that if we can argue that intending
reality as it appears (i.e. in the case of the realist conception of logic: as
objective and independent) is itself constitutive of cognition and even
of the possibility of cognition itself; then we can see a way in which
objective independent reality is (complete with its attendant predicament) already there, structuring the essential nature of consciousness
and experience.
For me, the phenomenological reduction, or ruling out of all that can
be doubted, and the subsequent re-discovery of the world (ultimately)
demonstrates an important way that reality, in all of the ways it seems to
us to be (including being independent of us), in fact cannot be ruled
out. Thus, we can see in the basic elements of the phenomenological
analysis how objective, independent reality enters the picture as objective,
and independent not only as an object of consciousness, but as constituting consciousness itself. This is the case even if (or, as Husserl would
have it, especially if ) we try to focus only on pure experience or pure
consciousness.
Ill mention a couple of other perspectives that gesture in a similar
direction to my own before moving on.
From Levinas we get:
the fact that the in itself of the object can be represented and, in knowledge,
seized, that is, in the end become subjective, would strictly speaking be
problematic . . . This problem is resolved before hand with the idea of
the intentionality of consciousness, since the presence of the subject
to transcendent things is the very denition of consciousness. (1998: 114,
italics mine) [and]
the world is not only constituted but also constituting. The subject is no
longer pure subject; the object no longer pure object. The phenomenon
is at once revealed and what reveals, being and access to being. (1998: 118)
Penelope Rush
22
Once we get our heads around the idea that the presence of the subject to
transcendent things denes consciousness,11 it is not a huge leap to see how
this initial subjective/transcendent relationship (even if its just one of
mutual presence) can incorporate the entire problematic outlined above:
i.e. that the SellarsMeillassoux contradiction is built in just so far as it
describes that relationship. Recall that Husserl equates that problematic
with the problem of the possibility of cognition (p. 16 above): it should
now be apparent how his equation can be understood as a means by
which to understand (rather than resolve or dissolve) the natural,
scientic perspective, complete with its consequent dilemma. That is,
Husserls point:
The problem of the possibility of cognition is the traditional realist dilemma
12
Note that this need not go the other way: we can retain the phenomenological insight without the
inverse claim that the object itself depends on, or even is (either necessarily or always) present to,
consciousness.
Husserls name for something akin to Fregean sense, but also apparently akin to (though more negrained than) Fregean reference (for some interesting details on these subtleties, see Haddock 2010).
Logical realism
23
4. Overow
I want now to discuss the idea of the pregnant concept of evidence
(Husserl 1964: 46). Husserl says:
If we say: this phenomenon of judgement underlies this or that phenomenon of imagination. This perceptual phenomenon contains this or that
aspect, colour, content, etc., and even if, just for the sake of argument, we
make these assertions in the most exact conformity with the givenness of
the cogitation, then the logical forms which we employ, and which are
reected in the linguistic expressions themselves, already go beyond the
mere cogitations. A something more is involved which does not at all
consist of a mere agglomeration of new cogitationes. (1964: 401)
Elsewhere, he notes:
The epistemological pregnant sense of self-evidence . . . gives to an intention, e.g., the intention of judgement, the absolute fullness of content, the
fullness of the object itself. The object is not merely meant, but in the
strictest sense given. (Husserl 1970: 765)
The point I want to draw attention to is that Husserl takes both logical and
physical/perceptual objects as the sort of thing that in one sense or
another overow, or go beyond what is given to cogitation.
The word object must . . . be taken in a very broad sense. It denotes not
only physical things, but also, as we have seen, animals, and likewise
persons, events, actions, processes and changes, and sides, aspects and
appearances of such entities. There are also abstract objects . . . (Fllesdal,
in Fllesdal and Bell 1994: 135)
24
Penelope Rush
It is within this core identity we nd that which gives the noema its
pregnant sense of self evidence; that which makes what is given to
cognition overow cognition and any (e.g. formal) agglomeration of
new cognitiones. Other terms Husserl uses for this core identity include:
the object; the objective unity; the self-same; the determinable
14
15
16
Shim (2005) nicely characterizes hyletic data as the sensual stu of experience. He gives the
following helpful example of the process of precisication to contrast memory or fantasy with
genuine perception: In remembering the house I used to live in, I can precisify an image of a red
house in my head. The shape, the color and other physical details of that house must be lled in by
hyletic data. Now lets say I used to live in a blue house and not a red house. There is, however, no
veridical import to the precisications of my memory until confronted by the corrective
perception . . . there is no sense in talking about the veridical import in the precisications of [the
memory or] fantasy (pp. 219220). In the latter cases, we may mistake merely hyletic data for nonconceptual (or objectual) phenomena (p. 220). An analogous situation might be said (by a logical
realist) to occur for logical intuition when we encounter counter examples or engage directly with
the meaning of logical operators in these situations we can see a genuine role for veridical input
capable of correcting or precisifying our intuition. On the other hand, perhaps analogously to what
occurs in a fantasy or hallucination, we may mistake the mere manipulation of symbols for genuine
(veridical) comprehension.
Shim gives a sophisticated argument for the idea that what provides perceptual noemata with
overow is that they have both conceptual and non-conceptual content. My idea is similar, but,
as will be elaborated shortly, the duality I want to consider should not be rendered as (noncontradictory) aspects of one and the same object, but rather as a contradictory object; whereas
I think that Shim means the duality he proposes to be interpreted in the former sense.
Thanks to Graham Priest for pressing this point.
Logical realism
25
5. McDowell
It is useful to compare what has so far been drawn from Husserl to a
specic interpretation of McDowell.
Neta and Pritchard in their (2007) article make a point that helps situate
Husserls programme: they argue that one way to understand attempts
(specically McDowells, but their ideas extend to Husserls) to reach
beyond our inner world to an external realm is precisely by close examination of the assumptions we bring to the Cartesian evil genius thought
experiment. The argument they present demonstrates links between a
particular (perhaps natural) way of conceiving the distinction between
inner and outer, and the commonly held assumption that:
(R): The only facts that S can know by reection alone are facts that would
also obtain in Ss recently envatted duplicate. (p. 383)
Neta and Pritchard argue that McDowell rejects R on the basis that there is
something about our actual, embodied experience of the world that cannot
be replicated by stimulus, no matter how sophisticated, experienced by a
brain in a vat (compare this with Husserls dierentiation between genuine
pregnant perception and hyletic/sensuous data). The clue as to how
McDowell rejects (R) and to uncovering the similarities between his and
26
Penelope Rush
Logical realism
27
6. Eectively defending ~R
The important word in the preceding paragraph is somehow. Expanding
on the somehow, we can nd a sense in which neither McDowell nor
Husserl escapes or resolves the traditional, natural dilemma. Or rather, to
the extent that they can be said to, their solutions do not address this
original dilemma. Conversely, I want to suggest it is just to the extent that
they dont escape the dilemma that they may (via expansion on the
somehow) be taken as having oered a sort of solution wherein what
was unintelligible from the traditional/natural perspective, is made at least
a little intelligible. That is, their sort of insight might be taken as oering a
perspective from which the contradiction inherent in speaking of a reality
independent of humans altogether need not automatically undermine the
possibility of a relationship between the two.
To see this, we need to start by outlining the ways in which both
positions clearly [challenge] the traditional epistemological picture that
has (R) at its core.
Neta and Pritchard outline McDowells challenge to R this way:
McDowells acceptance of reectively accessible factive reasons . . . entails
that the facts that one can know by reection are not restricted to the
inner in this way, and can instead, as it were, reach right out to the
external world, to the outer. One has reective access to facts that would
not obtain of ones recently envatted duplicate, on McDowells picture. If
this is correct, it suggests that the popular epistemological distinction
between inner and outer which derives from (R) should be rejected,
or at least our understanding of it should be radically revised. (p. 386)
28
Penelope Rush
Logical realism
29
17
That is, I think arguments for the claim that Husserls and McDowells accounts do not
hypostatise ultimately fail (for examples of such arguments, see Hartimo 2010b, and Putnam
2003, particularly p. 178). Or, to the extent that they succeed, the accounts themselves are rendered
largely irrelevant to the philosophical problem I am addressing here.
30
Penelope Rush
Shim, Teiszen, and others see the duality (which Shim renders as conceptual/non-conceptual) as
residing strictly in the phenomenological attitude, and so Shim (2005) argues that the
phenomenological solution cannot neatly slot into a natural answer to scepticism. But I think
phenomenology is relevant to the natural answer to scepticism exactly insofar as it provides this
explicit way of dierentiating being in the (real) world from envattedness. This dierentiation
disrupts a neat holistic story, and so its lesson, carried through to science and the natural attitude, is
perhaps not a categorical mistake (Shim 2005: 225), but an alert as to the deciencies of a
philosophy that disallows any perspective other than its own. What we know from the
phenomenological attitude might resist reduction to naturalist/scientic knowledge, but it
nonetheless can oer an insight into the items with which the scientic/philosophical attitude is
concerned: e.g. reality, experience, and knowledge. It is exactly what makes the phenomenological
perspective both tempt and frustrate . . . the very philosophical desire it should have satised
(Shim 2005: 225), that can make it relevant to that desire, and can potentially stop a too quick,
neat, sealed holist answer from gaining complete purchase.
Logical realism
31
Thanks to Graham Priest, Curtis Franks, Tuomas Tahko, Sandra Lapointe, and Jody Azzouni for
helpful feedback on earlier drafts.
chapter 2
1. Introduction
Our logical practices, it seems, already exhibit truth by convention.
A visible part of contemporary research in logic is the exploration of
nonclassical logical systems. Such systems have stipulated mathematical
properties, and many are studied deeply enough to see how mathematics
analysis in particular and even (some) empirical science, is recongured
within their nonclassical connes.1 What also contributes to the appearance of truth by convention with respect to logic is that it seems possible
although unlikely that at some time in the future our current logic of
choice will be replaced by one of these alternatives. If this happens, why
shouldnt the result be the dethroning of one set of logical conventions for
another? One set of logical principles, it seems, is currently conventionally
true; another set could be adopted later.
Quine, nevertheless, is widely regarded as having refuted the possibility
of logic being true by convention. Some see this refutation as the basis for
his later widely publicized views about the empirical nature of logic.
Logical principles being empirical, in turn, invites a further claim that
logical principles are empirically true (or false) because they reect well (or
badly) aspects of the metaphysical structure of the world. Just as the truth
or falsity of the ordinary empirical statement There is a table in Miner
Hall 221B at Tufts University on July 3, 2012, reects well or badly how a
part of the world is, so too, the Principle of Bivalence is true or false because
it reects correctly (or badly) the worlds structure. Ill describe this
additional metaphysical claim one that Im not attributing to Quine
(by the way) as taking logical principles to have representational content.
Most philosophers think logical principles being conventional is
1
The families of intuitionistic and paraconsistent logics are the most extensively studied in this respect.
There is a massive literature in both these specialities.
32
33
2. Quines dilemma
Its really really sad that almost no one notices that Quines refutation of
the conventionality of logic is a dilemma. The famous Lewis Carroll
innite regress assails only one horn of this dilemma, the horn that
presupposes that the innitely many needed conventions are all explicit.
Quine (1936b: 105) writes, indicating the other horn:
It may still be held that the conventions [of logic] are observed from the
start, and that logic and mathematics thereby become conventional. It may
be held that we can adopt conventions through behavior, without rst
announcing them in words; and that we can return and formulate our
conventions verbally afterwards, if we choose, when a full language is at our
disposal. It may be held that the verbal formulation of conventions is no
more a prerequisite of the adoption of conventions than the writing of a
grammar is a prerequisite of speech; that explicit exposition of conventions
is merely one of many important uses of a completed language. So conceived, the conventions no longer involve us in vicious regress. Inference
from general conventions is no longer demanded initially, but remains to
the subsequent sophisticated stage where we frame general statements of the
conventions and show how various specic conventional truths, used all
along, t into the general conventions as thus formulated.
Quine agrees that this seems to describe our actual practices with many
conventions, but he complains that (Quine 1936b: 105106):
it is not clear wherein an adoption of the conventions, antecedently to their
formulation, consists; such behavior is dicult to distinguish from that in
which conventions are disregarded . . . In dropping the attributes of deliberateness and explicitness from the notion of linguistic conventions we risk
depriving the latter of any explanatory force and reducing it to an idle label.
2
Ted Sider, a contemporary proponent of the claim that logical idioms have representational content,
represents the positions as opposed in just this way; he (Sider 2011: 97) diagnoses the doctrine of
logical conventionalism as supporting the view that logical expressions do not describe features of
the world, but rather are mere conventional devices.
34
Jody Azzouni
We may wonder what one adds to the bare statement that the truths of
logic and mathematics are a priori, or to the still barer behavoristic statement that they are rmly accepted, when he characterizes them as true by
convention in such a sense.
3
4
See (Quine 1970b) for a reiteration of the rst challenge with respect to linguistic rules.
See (Lewis 1969: 78) but I draw this characterization from (Burge 1975: 3233).
35
(5) There is at least one alternative R 0 to R such that the belief that the
others conformed to R 0 would give almost everyone a good and
decisive practical or epistemic reason to conform to R 0 likewise; such
that there is a general preference for general conformity to R 0 rather
than slightly-less-than-general conformity to R 0 ; and such that there
is normally no way of conforming to R and R 0 both.
(6) (1)(5) are matters of common knowledge.
There are many problems with this approach indeed, its no exaggeration
to describe condition (6) as yielding the result that there are almost no
conventions in any human population anywhere. But can Quines challenges be met? Are tacit conventions cogent?
36
Jody Azzouni
Epstein (2006), for example, is worried. My thanks to him for conversations (and email exchanges)
about this topic that have inuenced the rest of this section.
I draw this example from (Epstein 2006: 4).
37
Is it conventional that we cook some of our food and dont eat everything raw? I think it is. Is the
alternative suboptimal? There is controversy about this, but I think it is: I think this is why the
alternative eventually died out among our progenitors (after thousands of years, that is). On the other
hand, some of the reasons for why the alternative died out (the greater likelihood of food poisoning,
the inadvertent thriving of parasites in ones meal, etc.) have been presumably eliminated by
technical developments in food processing. So the practice of eating all food raw neednt be as
suboptimal as it once was.
38
Jody Azzouni
honest, a cold hard look at our own practices reveals exactly the same
thing). Evolution takes a really long view of things even the extinction of
a population because it engages in a suboptimal practice may occur so
slowly that the conventional xation of that practice can occur for many
generations, at least.8 How suboptimal a practice can be (in relation to
alternatives) is completely empirical, of course, and turns very much on the
details of the practices involved (and the background context they occur in);
but optimality comparisons should play only a moderate role in an evaluation of what alternative candidates there are to a practice, and therefore in
an evaluation of whether that practice is conventional and in what ways.
(This will matter to the eventual discussion of the conventionality of logic:
that alternative logics are suboptimal in various ways wont bar them from
playing a role in making conventional the logic weve adopted.)
One last additional point about conventionality that Ive just touched
on in the last sentence. This is that it isnt so much entire practices that
are conventional, but aspects of them that are. Minor variations in a
practice are always possible, minor variations that we dont normally treat
as rendering the practice conventional because we dont normally treat
those variations as rendering the practice a dierent one. There are many
variations in how sticks can be rubbed together, for example. How we
describe a practice or label it (how we individuate it) will invite our
recognition of these variations as inducing conventionality or not. Its
conventional to rub two sticks together in such and such a way, but not
conventional (say) to rub two sticks together instead of doing something
else that doesnt involve sticks at all (in a context, say, where there are no
rocks). How we individuate practices correspondingly infects how and in
what ways we recognize a practice to be conventional; but this is hardly an
issue restricted to the notion of tacit convention, or a reason to think the
notion has problems.
A nice example, probably, is the arrangement of the lettered keys on computer keyboards. No doubt
the contemporary distribution of letters is suboptimal compared to alternatives; its clearly an inertial
result of the earlier arrangement of the keys on typewriters which was probably also suboptimal in
its time and relative to its context at that time. Im not suggesting, of course, that keyboard
conventions are contributing to a future extinction event although I have no doubt that a
number of conventions that we currently use are doing precisely that.
39
10
11
See the introduction to (Hurley and Chater 2005a&b) for an overview of work as of that date. See
the various articles in the volumes for details. The rst sentence of the introduction (Hurley and
Chater 2005a: 1) begins, dramatically enough, with this sentence: Imitation is often thought of as a
low-level, cognitively undemanding, even childish form of behavior, but recent work across a variety
of sciences argues that imitation is a rare ability that is fundamentally linked to characteristically
human forms of intelligence, in particular to language, culture, and the ability to understand other
minds. Its important to stress how recent these discoveries are only within the last couple of
decades.
One almost shocking development is that the study of these mechanisms is successfully taking place
at the neurophysiological level, and not at some more idealized (abstract) level as is the case with
most language studies to date, specically those of syntax.
See, e.g., (Chomsky and Lasnik 1995).
40
Jody Azzouni
13
See (Chomsky 2003), specically page 313. See (Millikan 2003), specically pages 3738. This
empirical question is the nub of their disagreement, as Millikan realizes (Millikan 2003: 37): If
[the childs language faculty] reaches a steady state, that will be only if it runs out of local
conventions to learn. I dont nd convincing Millikans arguments against the empirical
possibility of a (virtually nal) steady-state for the language faculty: They seem to turn only on
the sheer impression that theres always more language conventions for adults to acquire. But given
that the empirical question is about what actual subpersonal mechanisms are involved in language
acquisition and also in the use of the language by adults who have acquired a language, its hard to
see why sheer impressions of conventionality deserve any weight at all.
One can always introduce the appearance of massive ocial conventionality by individuating the
language practices nely enough e.g., minor sound-variations in the statistical norms of utterances
determining the individuation of utterance practices (recall the last paragraph of Section 3); but Im
assuming this trivial vindication of the conventionality of language isnt what either Burge or
Millikan have in mind when they presume it as evident that natural language is full of
conventionality.
41
See (Sher 2001) for discussion and for citations of earlier proponents of this approach to logic.
Although the axiomatic model anciently arose via Euclidean geometry, its striking that it wasnt
generally recognized when Euclidean geometry was translated entirely to a language-based
format how gappy those rules were. An early view was that a nonethymematic mathematical
proof was one without missing steps or gaps. But this view, based as it was on a picture of a
42
Jody Azzouni
16
17
conceptual relationship between the steps in a mathematical proof, remained purely metaphorical
(or, at best, promissory) until the notion of algorithm in the context of articial languages emerged
at the hands of Turing, Church, and others in the twentieth century.
See (Azzouni 2005: 1819). It should be noted that this dramatic aspect of informal-rigorous
mathematical proof is still with us despite the presence of formal systems that are apparently fully
adequate to contemporary mathematics. That is, informal-rigorous mathematical proof continues to
operate largely by conceptual implication supplemented, of course, with substantial
computational bits.
Nicely popularized by one of the major researchers in the area: See (Kahneman 2011).
43
basis for our capacities for mathematics, and for reasoning generally, there is
no echo in our neuropsychological capacities to reason, and to prove, of the
semantic/syntactic apparatus the contemporary view of logic (and even its
competitors) provides.18 That apparatus is an all-purpose topic-neutral piece
of algorithmic machinery; how we actually reason, by contrast, involves
quite topic-specic, narrowly applied, highly componentalized, mental tools.
This means that the role of formal logic can only be a normative one; it has
emerged as a reasoning tool that we ocially impose on our ordinary
reasoning practices and that we (at times) can use to evaluate that reasoning.19
The foregoing, if right, makes the conventionality of logic quite plausible even if its an optimal logic, compared to competitors.20 The foregoing, if right, also makes plausible the emergence of classical logic as
explicitly conventional in the twentieth century; and it makes plausible its
role as tacitly conventional (at least in mathematical reasoning) for earlier
centuries before sets of rules for logic became explicit. I turn now to
discussing some of the reasons philosophers have for denying logic such a
conventional status. The rst kind of objection Ill consider turns on how
the notion of truth is used in the characterization of validity; next Ill
evaluate certain arguments that have been oered for why logical principles
have a (metaphysical) representational role.
19
20
See (Carey 2009), especially chapter 4 also see (Dehaene 1997) for good introductions to this
remarkable and important empirical literature.
Ive argued that this role of formal logic has emerged in the course of the twentieth century; it rst
occurred in mathematics but has spread throughout the sciences in large part because of the
mathematization of those sciences. See (Azzouni 2013), chapter 9, as well as (Azzouni 2005) and
(Azzouni 2008a) for discussion. I should stress that there are several psychological and historical
contingencies that seem involved in why the tacit employment of logical consequence in
mathematical practice turned out to be in the neighborhood of a rst-order and classical one: one
of those, I suggest (Azzouni 2005), is the psychological impression (on a case-by-case basis) that the
introduction and elimination rules for the logical idioms (and, or, not, etc.) are contentpreserving, an impression that isnt sustained for even quite short inference patterns, such as modus
ponens or syllogism.
Some philosophers argue that classical rst-order logic isnt optimal because of its representational
drawbacks: proponents of higher-order logics (e.g., Shapiro), on the one hand, think that it cant
represent mathematical concepts such as nite, proponents of one or another paraconsistent
approach (e.g., Priest) think it cant represent certain global concepts, e.g., all sentences. Although
Ive weighed in on these debates, they dont matter for the issue of whether logic is conventional
precisely because its been established in Section 3 of this chapter that suboptimality in relation to
competitors doesnt bar a practice from nevertheless being an alternative candidate.
44
Jody Azzouni
deductive reasoning preserving truth: If the premises are true then the
conclusion is true (must be true) as well. Many philosophers have taken
truth-preservation to be a characterization of classical logical principles. If
the notion of truth, in turn, is a correspondence notion, then it would
seem to follow that classical logic is semantically rooted in metaphysics, in
whats true about the world. And, it might be thought that what follows
from this is that logic cannot be conventional. This argument-strategy fails
for a large number of reasons; for current purposes, Ill focus on only three
of its failures. The rst is that a characterization of deduction as truthpreserving fails to single out any particular set of logical rules it fails to
even require that a set of logical principles be consistent! The second is
that, in any case, even if a characterization of logic in terms of truthpreservation singled out only classical logic (and not its alternatives), that
wouldnt rule out the conventionality of classical logic: suboptimality of
alternatives is no bar to their rendering a practice conventional. The last
reason is that truth, in any case, is too frail an idiom to root logic
semantically in the world. This is because it functions perfectly adequately
in discourses that bear no relationship to what exists.
The rst claim is easy to prove. Relevant is that the truth idiom is
governed by Tarski biconditionals: given a sentence S and a name of that
sentence S, S is true i S. Also relevant is that this condition cant be
supplemented by adding conditions to either wing of the Tarski biconditionals that arent equivalent to the wings themselves.21 But these points
are sucient to make the truth idiom logically promiscuous: its compatible
with any logical principles whatsoever. Let R be any set of logical principles.
And supplement R with the following inference schema T: S TS ,
and TS S. If the original set of rules is syntactically consistent (as, e.g.,
Priors tonk isnt), then so is the supplemented version. That R is truthpreserving follows trivially, regardless of whether R is consistent or
not: If U V according to R, then, using T, we can show: U V i
TU TV holds in [R, T].
Notice that a characterization of a choice of logic being legislatedtrue is licensed by the foregoing: Start by choosing ones logic, and then
supplement that choice with the T-schema. The resulting logic has been
legislated-true. It might be thought that more substantial uses of the
truth idiom, in semantics and in model theory, cant be executed in the
context of a nonclassical logic. But this isnt true either. In particular,
21
45
a model theory characterized metalogically using intuitionistic connectives in the metalanguage is homophonic to classical model
theory.22
The second point has already been established: Imagine (contrary to
what has just been shown) that a population adopts a suboptimal set of
logical principles ones strictly weaker than ours. (Intuitionist principles,
for example.) Then one possible result would be a failure to know all sorts
of things, both empirically and in pure mathematics, that we proponents
of classical logic know. Lets say that this is suboptimal;23 but this is hardly
fatal. And so the conventionality of logic isnt threatened by the presumed
suboptimality of other candidates.
Lastly, a number of philosophers have thought that the Tarski biconditionals all by themselves characterize truth as a correspondence notion.
There are many reasons to think they are wrong about this. Among them is
the fact that if a consistent practice of using nonreferring terms, such as
Hercules or Mickey Mouse is established, such a practice remains
consistent if its augmented with the T schema. Regardless of whether
the truth idiom functions as a correspondence notion for certain discourses, it wont function that way in this discourse. That shows that talk
of truth has to be supplemented somehow to give it metaphysical traction.
All by itself, it doesnt do that job.
The point generalizes, of course. In trying to determine whether logic is
conventional, some philosophers focus on specic statements like Either it
is raining or it is not raining, and worry about whether this statement is
about the world or not; more dramatically, some philosophers worry about
whether the supposed conventionality of logic yields the result that we
legislate the truth of a statement like this.24
But this misses the point. The claim that logic is conventional is
orthogonal to the question of whether logical truths have content
(worldly or otherwise), or (equivalently?) whether they are or arent about
the world. No doubt some philosophers have thought these claims
linked especially philosophers (like the paradigmatic positivists inuenced by Wittgensteinian Tractarian views) who are driven by epistemic
22
23
24
46
Jody Azzouni
26
Although the discussion is murky (or perhaps just metaphorical), this seems to be part of the project
undertaken by (Sider 2011), when he speaks of joint-carving logical notions, e.g., on page 97.
See (Kleene 1971), for lots of explicitly indicated examples.
47
(i)
(ii)
(iii)
In defense of iii): a disjunction states simply that one or the other of its
disjuncts holds; to legislate-true a disjunction one would need to legislatetrue one of its disjuncts. . . . It is open, of course, for the defender of truth
by convention to supply a notion of legislating-true on which the arguments premises are false. The challenge, though, is that the premises seem
correct given an intuitive understanding of legislate-true.
One of the oldest (but still quite popular) ways of begging the question
against proponents of alternative logics (as well as a popular way of begging
the question against logical conventionalism) is to implicitly adopt a lofty
metalanguage stance, and then use the very words that are under contention against the opponent. That doing this is so intuitive evidently
contributes to the continued popularity of the fallacy.
Some readers may be tempted to deny that this is a fallacy. They may
want to speak as Walker (1999: 20) does:
Anyone who refuses to rely on modus ponens, or on the law of noncontradiction, cannot be argued with. If they insist on their refusal there
48
Jody Azzouni
is therefore nothing to be done about it, but for the same reason there is no
need to take them seriously.27
28
See (Lewis 2005), where a similar refusal on similar grounds to debate the law of non-contradiction
is expressed; see (van Inwagen 1981) for the same maneuver directed towards substitutional
quantication. As I said: its a popular maneuver with many illustrious practitioners.
Metalogical debates, in particular, are ones where proponents can easily debate one another on
common ground, as many clearly do in the philosophical literature. See (Azzouni and Armour-Garb
2005) for details.
chapter 3
50
Stewart Shapiro
For Beall and Restall, the variable x ranges over types of cases. Classical logic
results from the Generalized Tarski Thesis if cases are Tarskian models;
intuitionistic logic results if cases are constructions, or stages in constructions (i.e., nodes in Kripke structures); and various relevant and paraconsistent logics result if cases are situations. So Beall and Restall take logical
consequence to be relative to a kind of case, and the General Relativistic
Schema is apt. For them, the law of excluded middle is valid relative to
Tarskian models, invalid relative to construction stages (Kripke models).
Beall and Restall call their view pluralism, eschewing the term
relativism:
we are not relativists about logical consequence, or about logic as such. We
do not take logical consequence to be relative to languages, communities of
inquiry, contexts, or anything else. (p. 88, emphasis in original)
It seems that Beall and Restall take relativism about a given subject matter
to be a restriction of what we here call folk relativism to those cases in
which the independent variable ranges over languages, communities of
inquiry, or contexts (or something like one of those). Of course, those are
the sorts of things that debates concerning, say, morality, knowledge, and
modality typically turn on. Here, we do not put any restrictions on the sort
of variable that the independent variable can range over. However, there
is no need to dispute terminology. To keep things as clear as possible, I will
usually refer to folk-relativism in the present, quasi-technical sense.1
1
John A. Burgess (2010) also attributes a kind of (folk) relativism to Beall and Restall: For pluralism
to be true, one logic must be determinately preferable to another for one clear purpose while
determinately inferior to it for another. If so, why then isnt the notion of consequence simply
51
1. Relativity to structure
Since the end of the nineteenth century, there has been a trend in
mathematics that any consistent axiomatization characterizes a structure, one at least potentially worthy of mathematical study. A key
element in the development of that trend was the publication of David
Hilberts Grundlagen der Geometrie (1899). In that book, Hilbert provided (relative) consistency proofs for his axiomatization, as well as a
number of independence proofs, showing that various combinations of
axioms are consistent. In a brief, but much-studied correspondence,
Gottlob Frege claimed that there is no need to worry about the
purpose relative (p. 521). Burgess adds, [p]erhaps pluralism is relativism but relativism of such a
harmless kind that to use that word to promote it would dramatise the position too much. The
present label folk-relativism is similarly meant to cut down on dramatic eect.
52
Stewart Shapiro
consistency of the axioms of geometry, since the axioms are all true
(presumably of space).2 Hilbert replied:
As long as I have been thinking, writing and lecturing on these things,
I have been saying the exact reverse: if the arbitrarily given axioms do not
contradict each other with all their consequences, then they are true and the
things dened by them exist. This is for me the criterion of truth and
existence.
The correspondence is published in Frege (1976) and translated in Frege (1980). The passage here is
in a letter from Hilbert to Frege, dated December 29, 1899.
53
Otherwise, the value d would not be unique, for any function. Recall,
however, that in any eld, every element distinct from zero has a multiplicative inverse. It is easy to see that a nilsquare cannot have a multiplicative inverse, and so no nilsquare is distinct from zero. In other words, there
are no nilsquares other than 0:
8 2 0 ! :: 0 , which is just
8 2 0 ! : 6 0 :
So, to repeat, zero is not the only nilsquare and no nilsquare is distinct
from zero. Of course, all of this would lead to a contradiction if we also
had (8x)(x = 0_x 6 0), and so smooth innitesimal analysis is inconsistent with classical logic. Indeed, :(8x)(x = 0_x 6 0) is a theorem of
the theory (but, since the logic is intuitionist, it does not follow that
(9x):(x = 0_x 6 0)).
Smooth innitesimal analysis is an elegant theory of innitesimals,
showing that at least some of the prejudice against them can be traced to
the use of classical logic Robinsons non-standard analysis notwithstanding. Bell shows how smooth innitesimal analysis captures a number of
intuitions about continuity, many of which are violated in the classical
theory of the reals (and also in non-standard analysis). Some of these
intuitions have been articulated, and maintained throughout the history
of philosophy and science, but have been dropped in the main contemporary account of continuity, due to Cantor and Dedekind. To take one
It follows from the principle of micro-aneness that every function is dierentiable everywhere on its
domain, and that the derivative is itself dierentiable, etc. The slogan is that all functions are smooth.
It is perhaps misleading to call the nilsquares a region or an interval, as they have no length.
54
Stewart Shapiro
55
2. What is objectivity?
Intuitively, a stretch of discourse is objective if the propositions (or
sentences) in it are true or false independent of human judgment,
4
See, for example, da Costa (1974), Mortensen (1995), (2010), Priest (2006), Brady (2006), Berto
(2007), and the papers in Batens et al. (2000). Weber (2009) is an overview of the enterprise.
56
Stewart Shapiro
preferences, and the like. Many of the folk-relative predicates are characteristic of paradigm cases of non-objective discourses. Whether something
is tasty, it seems, depends on the judge or standard in play at the time. So
taste is not objective (or so it seems). Whether something is rude depends
on the ambient location, culture, or the like. So etiquette is folk-relative
and, it seems, not objective. Etiquette may not be subjective, in the sense
that it is not a matter of what an individual thinks, feels, or judges, but,
presumably, it is not objective either. It is not independent of human
judgment, preferences, and the like.
One would be inclined to think that simultaneity and length are
objective, even though both are folk-relative, given relativity. As is the
case with much in philosophy (and everywhere else), it depends on what
one means by objective. We are told that whether two events are
simultaneous, and whether two rods are of the same length, depends on
the perspective of the observer. Does that undermine at least some of the
objectivity? But, vagueness and such aside, time and length do not seem to
depend on anyones judgment or feelings, or preferences. A given observer
can be wrong about whether events are simultaneous, even for events
relative to her own reference frame.
One might say that a folk-relative predicate P is objective if, for each
value n of the independent variable, the predicate P-relative-to-n does not
depend on anyones judgment or feelings. For example, if a given subject
can be wrong about P-relative-to-n, then the relevant predicate is objective.
However, even an established member of a given community can be wrong
about what is rude in that community. But one would not think that
etiquette is objective, even when restricted to a given community.
Clearly, to get any further on our issue, we do have to better articulate
what objectivity is, at least for present purposes. Again, objectivity is tied to
independence from human judgment, preferences, and the like. There is a
trend to think of objectivity in straightforward metaphysical terms. It must
be admitted that this has something going for it. The idea is that something, say a concept, is objective if it is part of the fabric of reality. The
metaphor is that the concept cuts nature at its joints, it is fundamental.
Theodore Sider (2011) provides a detailed articulation of a view like this,
but the details do not matter much here.
Presumably, taste and etiquette are not fundamental; tastiness and
rudeness do not cut nature at its joints (whatever that means). Does logic,
or, in particular, logical validity cut nature at its joints? It is hard to say,
without getting beyond the metaphor. What are the joints of reality?
Does it have such joints? How does logic track them?
57
One might argue that there can be at most one logic that is objective, in
this metaphysical sense. Sider does argue that at least parts of logic are
fundamental. As it happens, the logic he discusses is classical, but, so far as
I know, there is no argument supporting that choice of logic. It might be
compatible with his overall program that, say, parts of intuitionistic logic
or a relevant logic are fundamental instead. But perhaps two distinct logics
cannot both be fundamental. Contraposing, if the present folk-relativism
about logic is correct, then logic is not objective, in the foregoing
metaphysical sense.
For what it is worth, I would not like to tie objectivity to such deep
metaphysical matters as Sider-style fundamentality. First, things that are
not so fundamental can still be objective. Intuitively, the fact that the
Miami Heat won the NBA title in 2012 is objective (like it or not), but (I
presume) it is hardly fundamental. One can call a proposition objective if it
somehow supervenes on fundamental matters, but that requires one to
accept a contentious metaphysical framework, and to articulate the relevant notion of supervenience.
More important, perhaps, several competing philosophical traditions
have it that there simply is no way to sharply separate the human and
the world contributions to our theorizing. Protagoras supposedly said
that man is the measure of all things. On some versions of idealism, not to
mention some postmodern views, the world itself has a human character.
The world itself is shaped by our judgments, observations, etc. Perhaps
such views are too extreme to take seriously. A less extreme position is
Kants doctrine that the ding an sich is inaccessible to human inquiry. We
approach the world through our own categories, concepts, and intuitions.
We cannot get beyond those, to the world as it is, independently of said
categories, concepts, and intuitions.
On the contemporary scene, a widely held view, championed by
W. V. O. Quine, Hilary Putnam, Donald Davidson, and John Burgess,
has it that, to use a crude phrase, there simply is no Gods eye view to be
had, no perspective from which we can compare our theories of the world
to the world itself, to gure out which are the human parts of our
successful theories and which are the world parts (see, for example,
Burgess and Rosen 1997). On such views, the world, of course, is not of
our making, but any way we have of describing the world is in human
terms. As Friedrich Waismann once put it:
What rebels in us . . . is the feeling that the fact is there objectively no
matter in which way we render it. I perceive something that exists and put
58
Stewart Shapiro
it into words. From this, it seems to follow that something exists independent of, and prior to language; language merely serves the end of
communication. What we are liable to overlook here is the way we see a
fact i.e., what we emphasize and what we disregard is our work.
(Waismann 1945: 146)
This KantQuine orientation may suggest that there simply is no objectivity to be had, or at least no objectivity that we can detect. Perhaps
objectivity is a awed property, going the way of phlogiston and caloric, or
witchcraft. If this is right, then there simply is no answering the question of
this paper folk-relativism or no folk-relativism. Logic is not objective,
since nothing is. Despite having sympathy with the Kant Quine orientation, I would resist this rather pessimistic conclusion. There may not be
such a thing as complete objectivity whatever that would be but it still
seems that there is an interesting and important notion of objectivity to
be claried and deployed. There seems to be an important dierence a
dierence in kind between statements like the atmosphere contains
nitrogen and statements like the Yankees are disgusting. The distinction
may be vague and even context dependent, but it is still a distinction, and,
I think, an important one. Our question concerns whether the present
folk-relative logic falls on one side or the other of this divide (or perhaps on
or near its borderline).
Crispin Wrights Truth and objectivity (1992) contains an account of
objectivity that is more comprehensive than any other that I know of,
providing a wealth of detailed insight into the underlying concepts. Wright
does not approach the matter through metaphysical inquiry into the fabric
of reality, wondering whether the world contains things like moral properties, funniness, or numbers. He focuses instead on the nature of various
discourses, and the role that these play in our overall intellectual and
social lives.
As Wright sees things, objectivity is not a univocal notion. There are
dierent notions or axes of objectivity, and a given chunk of discourse can
exhibit some of these and not others. The axes are labeled epistemic
constraint, cognitive command, the Euthyphro contrast, and the
width of cosmological role. In a previous paper, (Shapiro 2000), I argue
that logic easily passes all of the tests. The conclusion is (or was) that, on
each of the axes, either logic is objective (if anything is) or matters of logic,
such as validity and consistency, lie outside the very framework of objectivity and non-objectivity, since most of the tests presuppose logic. That is,
to gure out whether a given stretch of discourse is objective, on this or
that axis, one must do some logical reasoning or gure out what is
59
3. Epistemic constraint
Epistemic constraint is an articulation of Michael Dummetts (1991a)
notion of anti-realism. According to one of Wrights formulations, a
discourse is epistemically constrained if, for each sentence P in the
discourse,
P $ P may be known: p: 75
It is perhaps ironic (or at least interesting) that Resnik argues against pluralism and relativism about
logic. He claims that there ought to be but one logic; the logic he favors is classical.
Actually, if the background logic is intuitionistic, there is a dierence between the absence of
unknowable truths and the truth of the biconditional: P $ P may be known. That dierence
does seem to bear on Wrights argument that if epistemic constraint fails in the sense that there are,
or could be, unknowable propositions in that area then the discourse is objective, but we will not
pursue this further here.
60
Stewart Shapiro
truth is to grant that, if the world co-operates, the truth or falsity of any
such statement may be settled beyond our ken. So . . . we are forced to
recognise a distinction between the kind of state of aairs which makes such
a statement acceptable, in light of whatever standards inform our practice of
the discourse to which it belongs, and what makes it actually true. The
truth of such a statement is bestowed on it independently of any standard
we do or can apply . . . Realism in Dummetts sense is thus one way of
laying the essential groundwork for the idea that our thought aspires to
reect a reality whose character is entirely independent of us and our
cognitive operations. (p. 4)
61
may be unknowable whether the runner left base before the ball was
caught, from the perspective of the home plate umpire. For example, it
may be too dark or no human can see distinctions that ne, or whatever.
So matters of temporal order, from a given reference frame, are not
epistemically constrained. And, intuitively, matters of temporal order are
objective, vagueness aside.
The point here is that with folk-relative discourses, we can only ask
about epistemic constraint for statements that have the relevant parameters
fully specied, at least implicitly. So the central question is whether there
can be unknowable truths concerning whether the argument (form) is
valid in a given logic L?
In eect, this matter was dealt with in my earlier paper (Shapiro 2000),
and also in Shapiro (2007), which concerns mathematics. Classical logic
was in focus then, but to some extent, the argument generalizes. Whether
there are unknowable truths in this area depends on what one means by
unknowable. If we do not idealize on the knowers, then of course there
can be unknowable truths. Suppose that our argument is an instance of
&-elimination in which the premise and conclusion each have, say, 10100
characters. Then is valid in, say, classical logic, but no one can
know that, since no one can live long enough to check that is an instance
of &-elimination.
So, to give epistemic constraint a chance of being fullled, we have to
idealize on the knowers. One sort of idealization is familiar. We assume
that our knowing subjects have unlimited (but still nite) time, attention
span, and materials at their disposal, and that they do not make any simple
computation errors. These idealizations are common throughout mathematics, and we take them to be conceptually unproblematic (and thus we
set aside issues concerning rule-following, as in, say Kripke (1982)). Then,
if L is classical rst-order logic, or intuitionistic logic, or most of the
relevant logics, and is an arbitrary argument form (with nitely many
premises and conclusions), then a statement that is valid in logic L is true
if and only if that fact is knowable (by one of our ideal agents). That is
because each of those logics has an eective and complete deductive
system.
Things are not so clear if the logic in question is classical second-order
logic, since its consequence relation is not eectively enumerable. Nor are
things so clear for statements that a given argument is not valid in one of
the aforementioned logics. Invalidity is not recursively enumerable, and so
checking invalidity is not a matter of running an algorithm. So if we are to
insist that all matters of logic are epistemically constrained, once the logic
62
Stewart Shapiro
4. Cognitive command
Assume that a given area of discourse serves to describe mind-independent
features of a mind-independent world. In other words, assume that the
discourse in question is objective, in an intuitive, or pre-theoretic sense.
Suppose now that two people disagree about something in that area. It
follows that at least one of them has misrepresented reality, and so something went wrong in his or her appraisal of the matter. Suppose, for
example, that two people are arguing whether there are seven, as opposed
to eight, spruce trees in a given yard. Assuming that there is no vagueness
concerning what counts as a spruce tree and no indeterminacy concerning
the boundaries of the yard, or whether each tree is in the yard or not, it
follows that at least one of the disputants has made a mistake: either she
did not look carefully enough, her eyesight was faulty, she did not know
what a spruce tree is, she misidentied a tree, she counted wrong, or
something else along those lines. The very fact that there is a disagreement
suggests that one of the disputants has what may be called a cognitive
shortcoming (even if it is not always easy to gure out which one of them it
is that has the cognitive shortcoming).
In contrast, two people can disagree over the cuteness of a given baby or
the humor in a given story without either of them having a cognitive
shortcoming. One of them may have a warped or otherwise faulty sense of
taste or humor, or perhaps no sense of taste or humor, but there need be
nothing wrong with his cognitive faculties. He can perceive, reason, and
count as well as anybody.
The present axis of objectivity turns on this distinction, on whether
there can be blameless disagreement:
A discourse exhibits Cognitive Command if and only if it is a priori that
dierences of opinion arising within it can be satisfactorily explained only
in terms of divergent input, that is, the disputants working on the basis of
63
So what of logic, again assuming the correctness of the foregoing folkrelativism? Let be a given argument form, and consider two folks who
disagree or seem to disagree whether is valid. One says it is and the
other says it is not. Our question breaks into two, depending on whether
we x the logic. For our rst type of case, let be an instance of excluded
middle or double-negation elimination, and consider the dispute
between advocates of classical logic and advocates of intuitionistic logic.
The inference is valid in classical logic, invalid in intuitionistic logic. For
the other sort of case, we x the logic and ponder disputes concerning that
logic. We imagine two folks who disagree or seem to disagree whether
is valid in L, where L is, say, a particular relevant logic.
We start with the second sort of case, disagreements that concern a xed
logic. I would think that there is room for blameless disagreement concerning how a given argument, formulated in natural language, should be
rendered in a formal language. However, such issues would take us too far
aeld, broaching matters of the determinacy of meaning, the slippage
between logical terms and their natural language counterparts, and the
intentions of the arguer. It is not so clear whether a disagreement in how
to render a natural language argument is excusable as a result of
vagueness . . . or in the standards of acceptability, or variation in personal
evidence thresholds or the like.
So let us set such matters aside, and just assume that our target
argument is fully formalized. One of our characters says that is valid
64
Stewart Shapiro
in the given logic L and the other says that is invalid in that logic L.
Do we know (a priori) that at least one of them has a cognitive
shortcoming?
Suppose that the logic L is either dened in terms of a deductive system
or that there is a completeness theorem for it. So L can be classical rstorder logic, intuitionistic logic, or one of the various relevant and paraconsistent logics that are given axiomatically. So our disputants dier on
whether there is a deduction whose undischarged premises are among the
premises of and whose last line is the conclusion of . So, up to
Churchs thesis, our disputants dier over a 1-sentence in arithmetic,
one in the form (9n), where is a recursive predicate. So our question
concerning cognitive command for this logic L reduces to whether cognitive command holds for these simple arithmetic sentences. I would think
that cognitive command does hold here. One of our disputants has made
(what amounts to) a simple arithmetic error, and that surely counts as a
cognitive shortcoming. But I will rest content with the reduction. Cognitive command holds in this case if and only if it holds for 91-sentences (or,
equivalently, 1-sentences).
Now suppose that our xed logic L is not complete. Say it is secondorder logic, with standard, model-theoretic semantics. In that case, the
question at hand reduces to set theory. Suppose, for example, that our
target argument has no premises and that its conclusion is, in eect, the
continuum hypothesis (see Shapiro 1991: 105). So is valid if and only if
the continuum hypothesis is true. So, in eect, our disputants dier over
the truth of the continuum hypothesis. Is that dispute cognitively blameworthy? Surely, that would take us too far aeld (but see Shapiro 2000,
2007, 2011, 2012), and we will leave this case with the reduction.
Let us briey consider the analogues of our question concerning
cognitive command with our other examples of folk-relative predicates.
Suppose that two judges dier on whether a certain event a occurred
before another event b from the same frame of reference (putting aside the
fact that this discourse is not epistemically constrained). Assume, for
example, that the two judges are in the same reference frame. Then,
unless the disagreement is excusable as a result of vagueness . . . or in the
standards of acceptability, or variation in personal evidence thresholds,
at least one of them exhibits a cognitive shortcoming. She did not look
carefully enough, or did not time the events properly, or forgot something. So cognitive command holds, and, of course, matters of temporal
order from a xed frame of reference are intuitively objective. The same
goes for matters of length.
65
Now consider two folks who disagree over whether a given food is tasty
for one and the same subject. Suppose, for example, that Tom and Dick
dier over whether licorice is tasty to Harry. To keep things simple,
assume that neither Tom nor Dick is Harry. Toms and Dicks judgments
would presumably be based on what Harry has told them and their
observations of his reactions when eating licorice. We should assume that
Tom and Dick have exactly the same body of such evidence (since
otherwise one of them has the cognitive shortcoming of lacking relevant
evidence). And we should set aside matters of vagueness . . . standards of
acceptability, [and] variation in personal evidence thresholds. Tom and
Dick may have come to opposite conclusions because they weighed certain
pronouncements or reactions dierently. In this case, perhaps, neither of
them has a cognitive shortcoming each is cognitively blameless. If so,
cognitive command fails.7 I take it that talk about taste in general
concerning what is tasty (simpliciter) is a paradigm of a non-objective
discourse, but I am not sure whether discourse about Harrys taste is
objective, intuitively speaking. Maybe we have a borderline case.
Returning to matters logical, Ive saved the hardest sort of situation for
last. That is on prima facie disagreements when the logic is not held xed.
To focus on a specic example, let be an instance of the law of
excluded middle (with no premises). Let h be a classicist who says that
is valid and let b be an intuitionist who insists that is not valid. Is
this a disagreement that is (cognitively) blameless? If so, then cognitive
command fails here, and this aspect of logic falls on the non-objective
side of this particular axis (assuming that cognitive command tracks a sort
of objectivity).
According to the foregoing folk-relativism, h and b are both right. Each
has spoken a truth, and so presumably there is nothing to fault either of
them. So each is (cognitively) blameless, at least concerning this particular
matter. The only question remaining is whether they disagree. Here we
encounter a matter that is treated extensively in the philosophical literature, and I must report that the issues are particularly vexed. There does
not seem to be much in the way of consensus as to what makes for a
disagreement. John MacFarlane (2014), for example, articulates several
7
A referee for Shapiro (2012) suggested that the failure of cognitive command does not distinguish
cases which are not objective from those in which evidence is scant. The situation sketched above,
with Tom and Dick, is not that dierent (in the relevant respect) from cases in science where
available evidence must be evaluated holistically say in cosmology. Two scientists might both be in
reective equilibrium, having assigned slightly dierent weights to various pieces of evidence.
Cognitive command might fail there, too, despite science being a paradigm case of objectivity.
66
Stewart Shapiro
Of course, this is not to say that these terms are like the standard indexicals in every manner.
67
that the content of the word left is dierent in the two contexts. In the
rst, it means something like to the left from Jills perspective and in
the second it means to the left from Jacks perspective. And they can
both be correct intuitively one of them is correct just in case the
other is.
Non-indexical contextualism, about a given term, is the view that its
content does not vary from one context of use to another, but the extension
can so vary according to a parameter determined by the context of
utterance.9 Suppose, for example, that a graduate student sincerely says
that a local roller coaster is fun, and her Professor replies No, that roller
coaster is not fun, it is lame. According to a non-indexical contextualism
about fun (and lame), each of them utters a proposition that is the
contradictory of that uttered by the other so they genuinely disagree. Yet,
assuming both are accurately reporting their own tastes, each has uttered a
truth, in his or her own context. For the graduate student, at the time,
the roller coaster is fun, since it is fun-for-the-graduate-student. For the
professor, the roller coaster is not fun, since it is not fun-for-the-professor.
Indeed, it is lame-for-the-professor.
Finally, assessment-sensitive relativism, sometimes called relativism
proper, about a term agrees with the non-indexical contextualist that
the content of the term does not vary from one context of use to another,
and so, in the above scenario, the relativist holds that the graduate student
and the professor each express a proposition contradictory to one expressed
by the other. However, for the assessment-sensitive relativist, the term gets
its extension from a context of assessment. Suppose, for example, that a third
person, a Dean, overhears the exchange between the graduate student and
professor and, assume that the roller coaster is not fun-for-the-Dean.
Then, from the context of the Deans assessment, the student uttered a
false proposition and the professor uttered a true one. And, from the
graduate students context of assessment, the Professor uttered a false
proposition, and from the Professors context of assessment, the student
uttered a false proposition.
According to MacFarlane, the dierence between non-indexical
contextualism and assessment-sensitive relativism is made manifest by
the phenomenon of retraction. That dierence does not matter here, and
we can lump non-indexical contextualism and assessment-sensitive
9
Nearly all terms have dierent extensions in dierent possible worlds. That is not the sort of
contextual variation envisioned here. For terms subject to non-indexical contextualism, the
relevant contextual parameter is for a judge, a time, a place, etc.
68
Stewart Shapiro
69
Again, the key idea is that each logic is tied to a specic language.
Presumably, the meaning of the logical terms diers in the dierent
languages.
So the DummettQuineCarnap perspective has it that we have a kind
of indexical contextualism here. The logical terms themselves have dierent contents for our characters h and b. Using a subscript-C to indicate a
classical connective and a subscript-I for the corresponding intuitionistic
connective, we have that h holds that _C:C is valid, while b holds that
_I:I is invalid. This is the same sort of situation as with Jack and Jill
and the salt. There is no disagreement between h and b unless it be over
whether the other has a coherent meaning at all. If they are suciently
open-minded, h and b might agree that _C:C is valid and that _I:I
is invalid. So we do not have a failure of cognitive command.
The DummettQuineCarnap perspective is not shared by all. Beall
and Restall (2006), for example, insist that their pluralism concerns
the notion of validity for a single language, with a single batch of
logical terms. So there is not, for example, a separate _C and _I.
There is just _. Restall (2002: 432) puts the dierence with Dummett
QuineCarnap well:
If accepting dierent logics commits one to accepting dierent languages
for those logics, then my pluralism is primarily one of languages (which
come with their logics in tow) instead of logics. To put it graphically, as a
pluralist, I wish to say that
A, :A C B, but A, :AR B
A and :A together, classically entail B, but A and :A together do not
relevantly entail B. On the other hand, Carnap wishes to say that
A, :C A B, but A, :R AB
A together with its classical negation entails B, but A together with its
relevant negation need not entail B.
70
Stewart Shapiro
I presume that Beall and Restall did not intend to make a claim about the
semantics of an established term of philosophical English. However, the
presence of the subscript x in the statement of the thesis might indicate that
the word valid has a sort of elided constituent, a slot where a logic can be
lled in. This suggests a sort of indexical contextualism about the word valid.
The same idea is suggested by the use of subscripts in the above passage from
Restall [2002], when he is using his own voice. He says that, for him:
A, :A C B, but A, :A R B:
71
chapter 4
73
74
Solomon Feferman
2. Conceptual structuralism
This is an ontologically non-realist philosophy of mathematics that I have
long advanced; my main concern here is to elaborate the nature and role of
1
75
logic within it. I have summarized this philosophy in Feferman (2009) via
the following ten theses.2
1. The basic objects of mathematical thought exist only as mental
conceptions, though the source of these conceptions lies in everyday
experience in manifold ways, in the processes of counting, ordering,
matching, combining, separating, and locating in space and time.
2. Theoretical mathematics has its source in the recognition that these
processes are independent of the materials or objects to which they
are applied and that they are potentially endlessly repeatable.
3. The basic conceptions of mathematics are of certain kinds of relatively simple ideal-world pictures that are not of objects in isolation
but of structures, i.e. coherently conceived groups of objects interconnected by a few simple relations and operations. They are communicated and understood prior to any axiomatics, indeed prior to
any systematic logical development.
4. Some signicant features of these structures are elicited directly from
the world-pictures that describe them, while other features may be
less certain. Mathematics needs little to get started and, once started,
a little bit goes a long way.
5. Basic conceptions dier in their degree of clarity or deniteness. One
may speak of what is true in a given conception, but that notion of
truth may be partial. Truth in full is applicable only to completely
denite conceptions.
6. What is clear in a given conception is time dependent, both for the
individual and historically.
7. Pure (theoretical) mathematics is a body of thought developed
systematically by successive renement and reective expansion of
basic structural conceptions.
8. The general ideas of order, succession, collection, relation, rule, and
operation are pre-mathematical; some implicit understanding of
them is necessary to the understanding of mathematics.
9. The general idea of property is pre-logical; some implicit understanding of that and of the logical particles is also a prerequisite to the
understanding of mathematics. The reasoning of mathematics is in
principle logical, but in practice relies to a considerable extent on
various forms of intuition in order to arrive at understanding and
conviction.
2
This section is largely taken from Feferman (2009), with a slight rewording of theses 5 and 10.
76
Solomon Feferman
77
78
Solomon Feferman
Extensionality 8X 8 Y [8n(n 2 X $ n 2Y ) ! X = Y ]
Comprehension For any denite property P(n) of members of N,
9 X 8nn 2 X $ Pn:
79
sets are characterized by axioms in full second-order logic; that is, any two
structures satisfying the same such axioms are isomorphic.3 On that
account, the proofs of categoricity in one way or another then
appeal prima facie to the presumed totality of arbitrary subsets of any
given set.4
Even if the deniteness of S(N) is open to question as above, we can
certainly conceive of a world in which S(N) is a denite totality and
quantication over it is well-determined; in that ideal world, one may
take for the property P in the above Comprehension Principle any formula
of full second-order logic over the language of arithmetic. Then a number
of theorems can be drawn as consequences in the corresponding system
PA2, including purely arithmetical theorems. Since the truth denition for
arithmetic can be expressed within PA2 and transnite induction can be
proved in it for very large recursive well-orderings, PA2 goes in strength far
beyond PA even when that is enlarged by the successive adjunction of
consistency statements transnitely iterated over such well-orderings.
What condence are we to have in the resulting purely arithmetical
theorems? There is hardly any reason to doubt the consistency of PA2
itself, even though by Gdels second incompleteness theorem, we cannot
prove it by means that can be reduced to PA2. Indeed, the ideal world
picture of (N, S(N), 2) that we have been countenancing would surely lead
us to say more, since in it the natural numbers are taken in their standard
conception. On this account, any arithmetical statement that we can prove
in PA2 ought simply to be accepted as true. But given that the assumption
of S(N) as a denite totality is a purely hypothetical and philosophically
problematic one, the best we can rightly say is that in that picture,
everything proved of the natural numbers is true.
Those who subscribe to this set-theoretical view of the categoricity results may dier on whether
the existence of the structures in question follows from their uniqueness up to isomorphism.
Shapiro (1997), for example, is careful to note repeatedly that it does not, while Isaacson (2011)
apparently asserts that it does (cf., e.g., Isaacson 2011, p. 3). In any case, it is of course not a logical
consequence.
In general, proofs of categoricity within formal systems of second-order logic can be analyzed to see
just what parts of the usual impredicative comprehension axiom scheme are needed for them. In the
case of the natural number structure, however, it may be shown that there is no essential dependence
at all, in contrast to standard proofs. Namely, Simpson and Yokoyama (2012) demonstrate the
categoricity of the natural numbers (as axiomatized with the induction axiom in second-order form)
within the very weak subsystem WKL0 of PA2 that is known to be conservative over PRA (Primitive
Recursive Arithmetic). By comparison, it is sketched in Feferman (2013) how to establish categoricity
of the natural numbers in its open-ended schematic formulation in a simpler way that is also
conservative over PRA. For an informal discussion of the categoricity of initial segments of the
cumulative hierarchy of sets in the spirit of open-ended axiom systems, see D. Martin (2001, sec. 3).
80
Solomon Feferman
Incidentally, all of this and more comes into question when we move
one type level up to the structure (N, S(N), S(S(N)), 21, 22) in which
Cantors continuum hypothesis may be formulated. A more extensive
discussion of the conception of that structure and the question of its
deniteness in connection with the continuum problem is given in
Feferman (2011). We shall also see below how taking N and S(N) to be
denite but S(S(N)) to be open-ended can be treated in suitable formal
systems.
81
that has the same set of valid formulas as for the case that is the rst
uncountable cardinal. In view of the leap over the case = , one may
suspect that the requirement that the set of valid formulas be given by
some eective set of axioms and rules of inference is not sucient to
express completeness in the usual intended sense. We need to say something more about how such axioms and rules of inference ought specically to be complete for a given quantier. The key is given by Gentzens
(1935) system of natural deduction NK (or sequent calculus LK) where
each connective and quantier in the classical rst-order predicate calculus
is specied by Introduction and Elimination rules for that operation only.
Moreover, for each pair of such rules, any two connectives or quantiers
satisfying them are equivalent, i.e. they implicitly determine the operator
in question. So a strengthened condition on a proposed addition by a
generalized quantier Q to our rst-order language is that it be given by
axioms and rules of inference for which there is at most one operator
satisfying them. That was the proposal of Zucker (1978) in which he gave a
theorem to the eect that any such quantier is denable in the rst-order
predicate calculus. In particular, that would apply to the Lindstrm
quantiers. However, there were some defects in Zuckers statement of
his theorem and its proof; I have given a corrected version of both in
Feferman (to appear). To summarize: we have fully satisfactory semantic
and inferential criteria for a logic to deal with structures whose domains are
rst-order and that are completely denite in the sense described above,
and these limit us to the standard rst-order classical logic.
Let us turn now to conceptions of structures with second-order or
higher-order domains, such as (N, S(N), 2, . . .) where the ellipsis indicates
that this augments an arithmetical structure on N such as (N, 0, Sc, <, , ).
Again, if S(N) is considered as a denite totality, the classical notion of
truth is applicable and the semantics of second-order logic must be
accepted. But as is well known there is no complete inferential system
that accompanies that, since again the arithmetical structure is categorically
axiomatized in this semantics and in consequence the set of its truths is not
eectively enumerable. In any case, as I have argued above, S(N) ought not
to be considered as a denite totality; to claim otherwise, is to accept the
problematic realist ontology of set theory. As Quine famously put it,
second-order logic is set theory in sheeps clothing. Boolos (1975, 1984)
tried to get around this via a reduction of second-order logic to a nominalistic system of plural quantication. This was incisively critiqued by
Resnik in his article Second-order logic still wild: Boolos is involved in a
circle: he uses second-order quantication to explain English plural
82
Solomon Feferman
quantication and uses this, in turn, to explain second-order quantication (Resnik 1988, p. 83).
Though the Lindstrm quantiers are restricted to apply to rst-order
structures and thus bind only individual variables they may well be dened
using higher-order notions in an essential way, in particular those needed
for the cardinality quantiers. Another example where the syntax is rstorder on the face of it but the semantics is decidedly second-order is IF
(Independence Friendly) logic, due to Hintikka (1996). This uses formulas in whose prenex form the existentially quantied individual variables are declared to depend on a subset of the universally quantied
individual variables that precede it in the prex list. Explanation of the
semantics of this requires the use of quantied function variables; over any
given rst-order structure (D, . . .) those variables are interpreted to range
over functions of various arities with arguments and values in D. Indeed,
Vnnen (2001, p. 519) has proved that the general question of validity of
IF sentences is recursively isomorphic to that for validity in full secondorder logic. Thus, as with the Lindstrm quantiers, the formal syntax can
be deceptive. See Feferman (2006) for an extended critique of IF logic.
83
Various methods of realizability, initially introduced by Kleene in 1945, can be used to give precise
independence results for such schemes, but are still not complete for intuitionistic logic. Cf. Troelstra
and van Dalen (1988, Ch. 4.4).
84
Solomon Feferman
partially ordered set, (ii) D is a function that assigns to each k in K a nonempty set D(k) such that if k k 0 then D(k) D(k 0 ), and (iii) v is a
function into f0, 1g at each k in K, each n-ary relation symbol R in the
language and n-ary sequence of elements of D(k), such that if k k 0 and
d1,. . .,dn 2 D(k) and v(k, R(d1,. . .,dn)) = 1 then v(k 0 , R(d1,. . .,dn)) = 1. One
motivating idea for this is that the elements of K represent stages of
knowledge, and that k k 0 holds if everything known in stage k is known
in stage k 0 . Also, v(k, R(d1,. . .,dn)) = 1 means that R(d1,. . .,dn) has been
recognized to be true at stage k; once recognized, it stays true. The domain
D(k) is the part of a potential domain that has been surveyed by stage k;
the domains may increase indenitely as k increases or may well bifurcate
in a branching investigation so that one cannot speak of a nal domain
in that case.
The valuation function v is extended to a function v(k, A(d1,. . .,dn)) into
f0, 1g for each formula A(x1,. . .,xn) with n free variables and assignment
(d1,. . .,dn) to its variables in D(k); this is done in such a way that if k k 0
and d1,. . .,dn 2 D(k) and v(k, A(d1,. . .,dn)) = 1 then v(k 0 , A(d1,. . .,dn)) = 1.
The clauses for conjunction, disjunction, and existential quantication are
just like those for ordinary satisfaction at k in D(k). The other clauses are
(ignoring parameters): v(k, A ! B) = 1 i for all k 0 k, v(k 0 , A) = 1 implies
v(k 0 , B) = 1; v(k, ) = 0; and v(k, 8x A(x)) = 1 i for all k 0 k and d in
D(k), v(k 0 , A(d)) = 1. As above, we identify :A with A ! ; thus v(k, :A) = 1
i for all k 0 k, v(k 0 , A) = 0. We say that k forces A if v(k, A) = 1; i.e. A is
recognized to be true at stage k no matter what may turn out to be known
at later stages. A formula A(x1,. . .,xn) is said to be valid in a model (K, ,
D, v) if for every k in K and assignment (d1,. . .,dn) to its free variables in
D(k), v(k, A(d1,. . .,dn)) = 1. Then the completeness theorem for this
semantics is that a formula A is valid in all Kripke models i it is provable
in the rst-order intuitionistic predicate calculus. We shall see in the next
section how Kripke models can be generalized to take into account
dierences as to deniteness of basic relations and domains.
Satisfying as this completeness theorem may be, there remains the
question whether one might not add connectives or quantiers to those
of intuitionistic logic while retaining some form of its semantics. Though
intuitionistic logic is part of classical logic, the semantical and inferential
criterion above for classical logic doesnt apply because of the dierences in
the semantical notions. But just as for the classical case, on the inferential
side each of the connectives and quantiers of the intuitionistic rst-order
predicate calculus is uniquely identied via Introduction and Elimination
rules in Gentzens natural deduction system NJ. Even more, Gentzen rst
85
formulated the idea that the meaning of each of the above operations is
given by its characteristic inferences. Actually, Gentzen claimed more: he
wrote that the [Introduction rules] represent, as it were, the denitions
of the symbols concerned (Gentzen 1969, p. 80). Prawitz supported this
by means of his Inversion Principle (Prawitz 1965, p. 33): namely, it follows
from the normalization theorem for NJ that each Elimination rule for a
given operation can be recovered from the appropriate one of its Introduction rules when that is the last step in a normal derivation. Without
subscribing at all to this proposed reduction of semantics to inferential
roles, we may ask whether any further operators may be added via suitable
Introduction rules. The answer to that in the negative was provided by the
work of Zucker and Tragesser (1978) in terms of the adequacy of what they
call inferential logic, i.e. of the logic of operators that can simply be marked
out by Introduction rules. As they show, every such operator is dened in
terms of the connectives and quantiers of the intuitionistic rst-order
predicate calculus. To be more precise, this is shown for Introduction rules
in the usual sense in the case of possible propositional operators, while in
the general case of possible operators on propositions and predicates now
in accord with the BHK interpretation proof parameters and constructions on them are incorporated in the Introduction rules, but those
are eventually suppressed.6
Incidentally, as Zucker and Tragesser show (p. 506), not every propositional operator given by
simple Introduction rules has an associated Elimination rule; a counterexample is provided by
(A ! B) _ C.
86
Solomon Feferman
87
Solomon Feferman
88
result in this case is that the following are of the same proof-theoretical
strength: SOO AC (), COO QF-AC (), and ID1, the theory of
arbitrary arithmetical inductive denitions. It is known that the latter
has the same proof-theoretical strength in intuitionistic logic as in
classical logic.
6.3
89
9
10
There is a considerable literature on semi-intuitionistic theories of sets including the power set
axiom going back to the early 1970s. See Feferman (2010, sec. 7.2) for references to the relevant
work of Poszgay, Tharp, Friedman, and Wolf.
Mathias (2001) proved that KP Pow (V = L) proves the consistency of KP Pow, so the usual
argument for the relative consistency of (V = L) doesnt work.
Michael Rathjen (2014) has recently veried this conjecture.
Curiously, this quote is from Lewis book, Parts of Classes, which oers a revisionary theory of classes
that diers from the usual mathematical conception of such.
90
Solomon Feferman
For an interesting social institutional account of mathematics see Cole (2013); this diers from
conceptual structuralism in some essential respects while agreeing with it in others.
91
should be put into question when it comes to the universe of sets. One
criticism of conceptual structuralism that has been made is that its not
clear/denite what mathematical concepts are clear/denite, and making
that a feature of the philosophy brings essentially subjective elements into
play.12 Actually, conceptual structuralism by itself, as presented in the
theses 110, takes no specic position in that respect and recognizes that
dierent judgments (such as mine) may be made. Once such are considered, however, logic has much to tell us in its role as an intermediary
between philosophy and mathematics. As shown in the preceding section,
one can obtain denitive results about formal models of dierent standpoints as to what is denite and what is not. Moreover, the results can be
summarized as telling us that to a signicant extent, the unlimited (de
facto) application of classical logic in mainstream mathematics i.e., the
logic of denite concepts and totalities may be justied on the basis of
a more rened mixed logic that is sensitive to distinctions that one might
adopt between what is denite and what is not.13 In other words, once
more they show that, at least to that extent, you can have your cake and
eat it too.
There are other dimensions of mathematical practice that reward metamathematical study motivated by the philosophy of conceptual
structuralism. One, in particular, that I have emphasized over the years
is the open-ended nature of certain principles such as that of induction for
the integers and comprehension for sets. This accords with the fact that in
the development of mathematics what concepts are recognized to be
denite evolve with time. Thus one cannot x in advance all applications
of these open-ended schematic principles by restriction to those instances
denable in one or another formal language, as is currently done in the
study of formal systems. This leads instead to the consideration of logical
models of practice from a novel point of view that yet is susceptible to
metamathematical study. One such is via the notion of the unfolding of
open-ended schematic axiom systems, that is used to tell us everything that
ought to be accepted if one has accepted given notions and principles.
Thus far, denitive results about the unfolding notion have been obtained
by Feferman and Strahm (2000, 2010) for schematic systems of non-nitist
and nitist arithmetic, resp., and by Buchholtz (2013) for arithmetical
12
13
In particular, this criticism has been voiced by Peter Koellner in his comments on Feferman (2011);
cf. http://logic.harvard.edu/EFI_Feferman_comments.pdf.
These kinds of logical results can also be used to throw substantive light on philosophical discussions
as to the problem of quantication over everything (or over all ordinals, or all sets) such as are found
in Rayo and Uzquiano (2006).
92
Solomon Feferman
chapter 5
The Second Philosopher is introduced in (Maddy 2007), and her views on logic detailed in Part III of
that book. The discussion here reworks and condenses the presentation there (see also (Maddy to
appear)).
For more on the Second Philosophers approach to mathematics, see (Maddy 2011).
93
94
Penelope Maddy
eectiveness and improve them as she goes. When I propose to treat the
question of the reliability of the coin inference as an ordinary question,
I have in mind to examine it from the Second Philosophers point of view.
She holds no prior convictions about the nature of the question; she sees it
simply as another of her straightforward questions about the world and her
investigations of it.
The rst thing shes likely to notice is that neither the reliability of the
coin inference nor the truth of the corresponding ifthen statement3
depends on any details of the physical composition of the item in her
hand or the particular properties that characterize dimes as opposed to
other coins. She quickly discerns that whats relevant is entirely independent of all but the most general structural features of the situation: an object
with one or the other of two properties that lacks one must have the other.
In her characteristic way, she goes on to systematize this observation for
any object a and any properties, P and Q, if Qa-or-Pa and not-Qa, then
Pa and from there to develop a broader theory of forms that yield such
highly general forms of truth and reliable inference. In this way, shes led
to consider any situation that consists of objects that enjoy or fail to enjoy
various properties, that stand and dont stand in various relations; she
explores conjunctions and disjunctions of these, and their failures as well;
she appreciates that one situation involving these objects and their interrelations can depend on another; and eventually, following Frege, she
happens on the notion that a property or relation can hold for at least
one object, or even universally suppose she dubs this sort of thing a
formal structure.4
Given her understanding of the real-world situations shes out to
describe in these very general, formal terms, she sees no reason to suppose
that every object has precise boundaries is this particular loose hair part
of the cat or not? or that every property (or relation) must determinately
hold or fail to hold of each object (or objects) is this growing tadpole now
a frog or not? She appreciates that borderline cases are common and fully
determinate properties (or relations) rare. Thinking along these lines, shes
led to something like a Kleene or Lukakasiewicz three-valued system: for a
given object (or objects), a property (or relation) might hold, fail, or be
indeterminate; not-(. . .) obtains if (. . .) fails and is otherwise indeterminate; (. . .)-and-(__) obtains if both (. . .) and (__) obtain, fails if one of
them fails, and is otherwise indeterminate; and so on through the obvious
3
4
95
clauses for (. . .)-or-(__), (there is an x, . . .x. . .) and (for all x, . . .x. . .).
A formal structure of this sort validates many of the familiar inference
patterns for example, the introduction and elimination rules for not,
and, or, for all, and there exists; the DeMorgan equivalences; and the
distributive laws but the gaps produce failures of the laws of excluded
middle and non-contradiction (if p is indeterminate, so are p-or-not-p and
not-( p-and-not-p).5 The subtleties of the Second Philosophers dependency relation undercut many of the familiar equivalences: not-(the rose is
red)-or-2 2 = 4, but 2 2 doesnt equal 4 because the rose is red.
Fortunately, modus ponens survives: when both (q depends on p) and
p obtain, q cant fail or be indeterminate. Suppose the Second Philosopher
now codies these features of her formal structures into a collection of
inference patterns; coining a new term, she calls this rudimentary logic
(though without any preconceptions about the term logic). She takes
herself to have shown that this rudimentary logic is satised in any
situation with formal structure.
This is a considerable advance, but it remains abstract: whats been
shown is that rudimentary logic is reliable, assuming the presence of formal
structure. Common sense clearly suggests that our actual world does
contain objects with properties, standing in relations, with dependencies,
but the Second Philosopher has learned from experience that common
sense is fallible and she routinely subjects its deliverances to careful
scrutiny. What she nds in this case is, for example, that the region of
space occupied by what we take to be an ordinary physical object like the
coin does dier markedly from its surroundings: it contains a more dense
and tightly organized collection of molecules; the atoms in those molecules
are of dierent elements; the contents of that collection are bound together
by various forces that tend to keep it moving as a group; other forces make
the region relatively impenetrable; and so on. Similarly, she conrms that
objects have properties, stand in relations, and that situations involving
them exhibit dependencies.
Now it must be admitted that there are those who would disagree, who
would question the existence of ordinary objects, beginning with
Eddington and his famous two tables:
One of them is familiar to me from my earliest years. . . . It has extention; it
is comparatively permanent; it is coloured; above all it is substantial. By
5
Here, briey, the distinction between logical truths and valid inferences matters, because the gaps
undermine all of the former. Inferences often survive because gaps are ruled out when the premises
are taken to obtain.
96
Penelope Maddy
substantial I do not mean merely that it does not collapse when I lean upon
it; I mean that it is constituted of substance. (Eddington 1928: ix)
Table No. 2 is my scientic table. . . . It . . . is mostly emptiness. Sparsely
scattered in that emptiness are numerous electric charges rushing about
with great speed; but their combined bulk amounts to less than a billionth
of the bulk of the table itself. Notwithstanding its strange construction it
turns out to be an entirely ecient table. It supports my writing paper as
satisfactorily as table No. 1; for when I lay the paper on it the little electric
particles with their headlong speed keep on hitting the underside, so that
the paper is maintained in shuttlecock fashion at a nearly steady level. If
I lean upon this table I shall not go through; or, to be strictly accurate, the
chance of my scientic elbow going through my scientic table is so
excessively small that it can be neglected in practical life. (Eddington
1928: x)
The Second Philosopher naturally wonders why this should be so, why the
so-called scientic table isnt just a more accurate and complete description of the ordinary table.6 In fact, it turns out that substance in
Eddingtons description of table No. 1 is a loaded term:
It [is] the intrinsic nature of substance to occupy space to the exclusion of
other substance. (Eddington 1928: xii)
There is a vast dierence between my scientic table with its substance (if
any) thinly scattered in specks in a region mostly empty and the table of
everyday conception which we regard as the type of solid reality . . . It
makes all the dierence in the world whether the paper before me is poised
as it were on a swarm of ies . . . or whether it is supported because there is
substance below it. (Eddington 1928: xixii)
Some writers reject the ordinary table on the grounds that its boundaries would be inexact. As weve
seen, the Second Philosopher is happy to accept this sort of worldly vagueness.
97
realize that its supporting the paper or resisting my elbow arise very
dierently than I might have at rst imagined that one couldnt come
to realize this, that is, without also coming to realize that there is no such
thing as table No. 1. But why should this be so? Why should our initial
conceptualization be binding in this way? For that matter, is it even clear
that our initial conceptualization includes any account at all of how and
why the table supports paper or resists elbows? The Second Philosopher
sees no reason to retract her belief in ordinary macro-objects.7
So lets grant the Second Philosopher her claim that formal structure
as she understands it does turn up in our actual world. This means not
only that rudimentary logic applies in such cases, but that it does so
regardless of the physical details of the objects composition, the precise
nature of the properties and relations, any particular facts of spatiotemporal location, and so on. This observation might serve as the rst step
on a path toward the familiar idea, noted earlier, that questions like
these are peculiarly philosophical: the thought would be that if the
correctness of rudimentary logic doesnt depend on any of the physical
details of the situation, if it holds for any objects, any properties and
relations, etc., then it must be quite dierent in character from our
ordinary information about the world; indeed, if none of the physical
details matter, if these truths hold no matter what the particular contingencies happen to be, then perhaps theyre true necessarily, in any
possible world at all and if thats right, then nothing particular to
our ordinary, contingent world can be whats making them true.
By a series of steps like these, one might make ones way to the idea
that logical truths reect the facts, not about our world, but about a
platonic world of propositions, or a crystalline structure that our
world enjoys necessarily, or an abstract realm of meanings or concepts,
or some such distinctively philosophical subject matter. Many such
7
Eddingtons two tables may call to mind Sellars challenge to reconcile the scientic image with the
manifest image. In fact, the manifest image includes much more than Eddingtons table No. 1 it is
the framework in terms of which, to use an existentialist turn of phrase, man rst encountered
himself (Sellars 1962: 6) but Sellars does come close to our concerns when he denies that manifest
objects are identical with systems of imperceptible particles (Sellars 1962: 26). He illustrates with the
case of the pink ice cube: the manifest ice cube presents itself to us as something which is pink
through and through, as a pink continuum, all the regions of which, however small, are pink (Sellars
1962: 26), and of course the scientic ice cube isnt at all like this. Here Sellars seems to think, with
Eddington, that science isnt in a position to tell us surprising things about what it is for the ice cube
to be (look) pink; he seems to agree with Eddington that some apparent features of the manifest ice
cube cant be sacriced without losing the manifest ice cube itself. Indeed the essential features they
cling to are similar: a kind of substantial continuity or homogeneity. The Second Philosopher
remains unmoved.
98
Penelope Maddy
options spring up in the wake of this line of thought, but ordinary facts,
ordinary information about our ordinary world has been left behind,
and ordinary inquiry along with it weve entered the realm of philosophy proper.
But suppose our Second Philosopher doesnt set foot on this path.
Suppose she simply notices that nothing about the chemical makeup of
the coin is relevant, that nothing about where the coin is located is
relevant, that only the formal structure matters to the reliability of the
rudimentary logic shes isolated. From here she simply continues her
inquiries, turning to other pursuits in geology, astronomy, linguistics,
and so on. At some point in all this, she encounters cathode rays and
black body radiation, begins to theorize about discrete packets of energy,
uses the quantum hypothesis to explain the photo-electric eect, and
eventually goes on to the full development of quantum mechanics. And
now shes in for some surprises: the objects of the micro-world seem to
move from one place to another without following continuous trajectories;
a situation with two similar particles A and B apparently isnt dierent
from a situation with A and B switched; an object has some position and
some momentum, but it cant have a particular position and a particular
momentum at the same time; there are dependencies between situations
that violate all ordinary thinking about dependencies.8 Do the objects,
properties, relations, and dependencies of the quantum-mechanical
micro-world enjoy the formal structure that underlies rudimentary logic?
The Second Philosopher might well wonder, and sure enough, her doubts
are soon realized. In a case analogous to, but simpler than position and
momentum, she nds an electron a with vertical spin up or vertical spin
down, and horizontal spin right or horizontal spin left (Ua or Da) and
(Ra or La) but for which the four obvious conjunctions (Ua and Ra) or
(Ua and La) or (Da and Ra) or (Da and La) all fail. This distributive law
of rudimentary logic doesnt obtain!
Were now forced to recognize that those very general features the
Second Philosopher isolated in her formal structures actually have some
bite. Though it wasnt made explicit, an object in a formal structure was
assumed to be an individual, fundamentally distinct from all others; having
a property like location, for example was assumed to involve having a
particular (though perhaps imprecise) property a particular location, not
just some location or other. These features were so obvious as to go
unremarked until the anomalies of quantum mechanics came along to
8
For more on these quantum anomalies, with references, see (Maddy 2007, III.4).
99
demonstrate so vividly that they can in fact fail.9 Those of us who ventured
down that path the Second Philosopher didnt take were tempted to think
that her formal structure is to be found in every possible world, but it turns
out it isnt present even in every quarter of our own contingent world!10
Rudimentary logic isnt necessary after all; its correctness is contingent on
the very general, but still not universal, features isolated in the Second
Philosophers formal structure.
Weve focused so far on the metaphysics what makes these inferences
reliable, these truths true? but theres also the epistemology how do we
come to know these things? If we followed the philosophers path and
succeeded in dismissing the vicissitudes of contingent world as irrelevant
well before the subsequent shocks dealt the Second Philosopher by quantum mechanics, then we might continue our reasoning along these lines: if
logic is necessary, true in all possible worlds, if the details of our contingent
world are beside the point, then how could coming to know its truths
require us to attend to our experience of this world?11 Again a range of
options ourish here, from straightforward theories of a priori knowledge
9
10
11
In yet another twist in the tradition of Eddington and Sellars, Ladyman and Ross (2007) begin from
this observation that the micro-world doesnt seem to consist of individual objects then go on to
classify the ordinary table, along with the botanists giant redwoods and the physical chemists
molecules, as human constructs imposed for epistemological book-keeping (p. 240) on an entirely
objectless world. I suspect that this disagreement with the Second Philosopher traces at least in part
to diering pictures of how naturalistic metaphysics is to be done. The Second Philosophers
metaphysics naturalized simply pursues ordinary science and ends up agreeing with the folk, the
botanist and the chemist that there are tables, trees and atoms, that trees are roughly constituted by
biological items like cells, cells by chemical items like molecules, molecules by atoms, and so on. She
doesnt yet know, and may never know, how to extend this program into the objectless microworld, but she has good reason to continue trying, and even if she fails, she doesnt see that this
alone should undermine our belief in the objects of our ordinary world. In contrast, the naturalized
metaphysics of Ladyman and Ross is the work of naturalistic philosophical under-labourers
(p. 242), designed to show how two or more specic scientic hypotheses, at least one of which
is drawn from fundamental physics, jointly explain more than the sum of what is explained by the
two hypotheses taken separately (Ladyman and Ross 2007: 37) and its this project that delivers
the surprising result that ordinary objects are constructed by us. From their perspective, the Second
Philosopher metaphysics naturalized is just more science: the botanist and the physical chemist
make no contribution to ontology; metaphysics only begins when their hypotheses are unied with
fundamental physics. From the Second Philosophers perspective, theres no reason to suppose that
ordinary objects are human projections or to insist that assessments of what there is must involve
unication with fundamental physics. Indeed, from her perspective, given our current state of
understanding (see below), quantum mechanics is perhaps the last place we should look for
ontological guidance!
This incidentally removes another sort of skeptical challenge to the Second Philosophers belief in
ordinary macro-objects, namely, the charge that an inquiry starting with objects with properties,
etc., will inevitably uncover objects with properties.
An inference from necessary to a priori is less automatic in our post-Kripkean age, when many
philosophers recognize a posteriori necessities, but logical truth seems a poor candidate for this sort
of thing. In any case, what Im tracing here are tempting paths, not conclusive arguments.
100
Penelope Maddy
13
14
This is reminiscent of many of Quines descriptions of his epistemology naturalized, but Quine
also tends to fall back on more traditional philosophical formulations, asking how we manage to
infer our theory of the external world from sensory data (see Maddy 2007, I.6; for more).
I can only give the smallest sampling of this work here. For more, with references, see (Maddy 2007,
III.5).
E.g., does the child lift a cloth to nd a desirable object shes seen hidden there?
101
102
Penelope Maddy
can be in two places at once or that separate units can occupy the same
space, and they expect them to travel in continuous trajectories. In the
words of Elizabeth Spelke, a pioneer in this eld, the infants objects are
complete, connected, solid bodies that persist over occlusion and maintain
their identity through time (Spelke 2000: 1233):
Putting together the ndings from studies of perception of object boundaries and studies of perception of object identity, young infants appear to
organize visual arrays into bodies that move cohesively (preserving their
internal connectedness and their external boundaries), that move together
with other objects if and only if the objects come into contact, and that
move on paths that are connected over space and time. Cohesion, contact,
and continuity are highly reliable properties of inanimate, material objects:
objects are more likely to move on paths that are connected than they are to
move at constant speeds, for example; and they are more likely to maintain
their connectedness over motion than they are to maintain a rigid shape.
Infants perception appears to accord with the most reliable constraints on
objects. (Spelke et al. 1995: 319320)
E.g., Hateld (2003) argues that the ndings of Spelke and her collaborators only establish that
young infants perceive bounded trackable volumes not individual material objects. Of course,
Spelke (e.g., in Spelke et al. (1995), cited by Hateld) does allow that the infants object concept
continues to develop in early childhood, so there is room here for clarication of levels or degrees of
object perception. A question like this would prompt the Second Philosopher to get down to sorting
things out, but Im not so idealized an inquirer and leave these further investigations to others.
103
With this in mind, its less surprising that the beginnings of the childs
identication of objects by their properties comes a couple of months later
than their identication by the more straightforward spatiotemporal
means, and perhaps even that this new development apparently coincides
with the acquisition of their rst words property nouns like ball and
duck!
So as not to belabor this fascinating developmental work, let me just
note that similar studies have shown that young infants detect conjunctions and disjunctions of object properties, the failure of properties or
relations, simple billiard-ball style causal dependencies, and so on. Its also
notable that many of these abilities found in young infants are also present,
for example, in primates and birds. This suggests an evolutionary origin,
and clearly the advantages conferred by the ability to track objects spatiotemporally, to perceive their properties and relations, to notice dependencies, would have been as useful on the savanna as they are in modern life.
All this leaves the Second Philosopher with two well-supported hypotheses: the ability to detect (at least some of ) the formal structure present in
the world comes to humans at a very early age, perhaps largely due to our
evolutionary inheritance; whether by genetic endowment, normal maturation, or early experience, the primitive cognitive mechanisms underlying
this ability are as they are primarily because humans (and their ancestors)
interact almost exclusively with aspects of the world that display this
formal structure. From here its a short step to the suggestion that the
presence of these primitive cognitive mechanisms, all tuned to formal
104
Penelope Maddy
18
105
106
Penelope Maddy
20
Though these idealizations involve falsication in her description of the physical world, they are
satised in the world of classical mathematics: excluded middle holds and the dependencies are
logical. For more on the ontology of mathematics, see (Maddy 2011).
There is serious disagreement between various writers over the source of the indeterminacy: is it
purely linguistic or does the world itself include borderline cases and fuzzy objects? Here the Second
Philosopher sides with the latter, but this shouldnt aect the brief discussion here, despite the
formulation in the quotation in the next clause above.
107
classical logic with care,21 as one should any idealization, rather than switch
to a less viable logic.22
Advocates of various conditional logics protest the Second Philosophers
other bold idealization: replacement of real dependencies with the simple
material conditional. There are many proposals for a more substantial
conditional, far too many to consider here (even if my slender expertise
allowed it), but perhaps the conditional of relevance logic can be used as one
representative example. The motivation here speaks directly to the falsication in question: the antecedent of a conditional should be relevant to the
consequent.23 To return to our earlier example, the redness of this rose isnt
relevant to the fact that 2 2 = 4, despite the truth of the corresponding
material conditional (if the rose is red, then 2 2 = 4). Of course, as before,
the Second Philosopher fully appreciates that the material conditional is a
falsication, that the rose inference is an anomaly, but the pertinent
questions are whether or not the falsication is benecial and benign, and
whether or not the relevance logician has something better to oer. Again
I think that for now, we do best to employ our classical logic with care.
So we see that some deviant logics depart from the Second Philosophers
classical logic by rejecting her idealizations,24 and that our assessment then
depends on the extent to which the falsications introduced are benecial and
benign, and on the systematic merits of the proposed alternative. But not all
deviant logics t this prole; some concern not just the idealizations of classical
logic, but the fundamentals of rudimentary logic itself. Examples include
intuitionistic logic which rejects double negation elimination quantum
21
22
23
24
Sorensen (2012) credits this approach to H. G. Wells: Every species is vague, every term goes cloudy
at its edges, and so in my way of thinking, relentless logic is only another name for a stupidity for a
sort of intellectual pigheadedness. If you push a philosophical or metaphysical enquiry through a
series of valid syllogisms never committing any generally recognized fallacy you nevertheless
leave behind you at each step a certain rubbing and marginal loss of objective truth and you get
deections that are dicult to trace, at each phase in the process (Wells 1908: 11).
Williamson (1994) also advocates retaining classical logic, but his reason is quite dierent: because
there is no real vagueness, because apparent borderline cases really just illustrate our ignorance of
where the true borderline lies. This strikes many, including me, as obviously false.
Relevance logicians are particularly unhappy with what they call explosion, the classical oddity that
anything follows from a contradiction. For related reasons, full relevance logic rejects even some
rudimentary logical inferences not involving the conditional, like disjunctive syllogism, but I leave
this aside here. (For a bit more, see Maddy 2007: 292, footnote 24.)
Some other deviant logics respond to idealizations of language rather than the worldly features of
rudimentary logic: e.g., free logicians counsel us to reject the falsifying assumption that all naming
expressions refer. Here, too, our assessment depends on the eectiveness of the idealization and the
viability of the alternative. In practical terms, leaving aside the various technical studies in the theory
of free logics, Im not sure using a free logic is readily distinguishable from being careful about the
use of existential quantier introduction in the context of classical logic. In any case, our concern
here is with worldly idealizations, not linguistic ones.
108
Penelope Maddy
logic which rejects the distributive laws and dialetheism which holds
that there are true contradictions. Given the connection of rudimentary logic
with the Second Philosophers formal structure, the challenge for each of
these is to understand what the world is like without this formal structure,
what the world is like that this alternative would be its logic.25 Of the three,
intuitionistic logic comes equipped with the most developed metaphysical
picture, but its suited to describing the world of constructive mathematics,
not the physical world.26 Quantum logic at rst set out to characterize the
non-formal-structure of the micro-world, but in practice it has not succeeded
in doing so;27 the problem of interpreting quantum mechanics remains
open. And dialetheism faces perhaps the highest odds: as far as I know, its
defenders have focused for the most part on the narrower goal of locating a
compelling example of a true contradiction in the world, perhaps so far
without conspicuous success.28 The Second Philosopher tentatively concludes that rudimentary logic currently has no viable rivals as the logic of
the world, and that classical logic likewise stands above its rivals as an
appropriate idealization of rudimentary logic for everyday use.
In sum, then, the Second Philosophers answer, an ordinary answer to
the question of why that coin must be foreign, is that the coin and its
properties display formal structure and the inference in question is reliable
in all such situations. This answer doesnt deliver on the usual philosophical expectations: the reliability of the inference is contingent, our knowledge of it is only minimally a priori at best. The account itself results from
plain empirical inquiry, which may lead some to insist that it isnt
philosophy at all. Perhaps not. Then again, if the original question
why is this inference reliable? counts as philosophical and its not clear
how else to classify it then the answer, too, would seem to have some
claim to that honoric. But the Second Philosopher doesnt care much
about labels. After all, even Second Philosophy and Second Philosopher
arent her terms but mine, used to describe her and her behavior. In any
case, philosophy or not, I hope the Second Philosophers investigations do
tell us something about the nature of that inference about the coin.29
25
26
27
29
Our interest here is in the logic of the world, not the logic that best models something else, as, e.g.,
paraconsistent logic (a variety of relevance logic) might serve to model belief systems (see Maddy
2007: 293296).
See the discussion of Creator Worlds in (Maddy 2007: 231233, 296) and (Maddy to appear, II).
28
See (Maddy 2007: 276279, 296).
See (Maddy 2007: 296297).
My thanks to Patricia Marino for helpful comments on an earlier draft and to Penelope Rush for
editorial improvements.
chapter 6
Logical nihilism
Curtis Franks
1. Introduction
The idea that there may be more than one correct logic has recently
attracted considerable interest. This cannot be explained by the mere fact
that several distinct logical systems have their scientic uses, for no one
denies that the logic of classical mathematics diers from the logics of
rational decision, of resource conscious database theory, and of eective
problem solving. Those known as logical monists maintain that the
panoply of logical systems applicable in their various domains says nothing
against their basic tenet that a single relation of logical consequence is
either violated by or manifest in each such system. Logical pluralists do
not counter this by pointing again at the numerous logical systems, for
they agree that for all their interest many of these indeed fail to trace any
relation of logical consequence. They claim, instead, that no one logical
consequence relation is privileged over all others, that several such relations
abound.
Interesting as this debate may be, I intend to draw into question the
point on which monists and pluralists appear to agree and on which their
entire discussion pivots: the idea that one thing a logical investigation
might do is adhere to a relation of consequence that is out there in the
world, legislating norms of rational inference, or persisting some other
wise independently of our logical investigations themselves. My opinion is
that xing our sights on such a relation saddles logic with a burden that it
cannot comfortably bear, and that logic, in the vigor and profundity that
it displays nowadays, does and ought to command our interest precisely
because of its disregard for norms of correctness.
I shall not argue for the thesis that there are no correct logics. Although
I do nd attempts from our history to paint a convincing picture of a
relation of logical consequence that attains among propositions (or sentences, or whatever) dubious, I should not know how to cast general doubt
109
110
Curtis Franks
on the very idea of such a relation. By drawing this point into question
I mean only to invite reection about what work the notion of a correct
logic is supposed to be doing, why the debate about the number of logical
consequence relations is supposed to matter to a logician, and whether the
actual details of logic as it has developed might be dicult to appreciate if
our attention is overburdened by questions about the correctness of logical
principles. Rather than issue any argumentative blows, I propose merely to
lead the reader around a bit until his or her taste for a correct logic sours.
I emphasize that this really is a matter of perspicuity. One should not think that the phenomena
described below are artifacts of peculiar features of propositional logic. They are nearly all
consequences of decisions about lem that are invariant across a wide spectrum of logics. Consider:
the structural subsumption of lem applies also to the predicate calculus; the admissible propositional
rules of Heyting Arithmetic (and of Heyting Arithmetic with Markovs principle) are exactly those of
IPC (Visser 1999); the disjunction property holds for Heyting Arithmetic (Kleene 1945) and for
intuitionistic ZermeloFraenkel set theory (Myhill 1973).
Logical nihilism
111
this calculus by disallowing dne but now introduce a new rule that allows
one to write down any instance of lem in any context, f will be derivable
in all contexts in which ::f is derivable.
In (193435), Gentzen observed to his own surprise2 that the sequent
calculus presentation of CPC admits an even more elegant modication
into a presentation of IPC than the one just described. One simply
disallows multiple-clause succedents and leaves the calculus otherwise
unchanged. Thus lem and with it the entire distinction between
intuitionistic and classical logic is subsumed into the background structure
of the logical calculus. All the inference rules governing the logical particles
(^, _, , :) and all the explicit rules of structural reasoning (identity, cut,
weakening, exchange, and contraction) are invariant under this transformation. Thus it appears that a duly chosen logical calculus allows a precise
analysis of what had been thought of as a radical disagreement about the
nature of logic. When classicists and intuitionists are seen to admit
precisely the same inference rules, their disagreement appears in some
ways quite minor, if more global than rst suspected.
Exactly how minor, on closer inspection, is the dierence between these
supercially similar calculi CPC and IPC? Not very. Even before
Gentzens profound analysis, Gdel (1932) observed that IPC satised
the disjunction property: formulas such as f _ are provable only if
either f or is as well. At rst sight this might appear to be no more than
a restatement of the intuitionists rejection of lem. After all, that rejection
was motivated by the idea that instances of lem are un-warranted when
neither of their disjuncts can be independently established. But, one might
think, if any disjunction is warranted in the absence of independent
verication of one of its disjuncts, those like f _ :f are, so rejecting
lem should lead to something like the disjunction property. This reasoning
strikes me as worthy of further elaboration and attention, but it should be
unconvincing as it stands. For one thing, the formal rejection of lem only
bars one from helping oneself to its instances whenever one wishes. The
gap between this modest restriction and the inability ever to infer any
disjunction of the form f _ :f at all, unless from a record of that
inference one could eectively construct a proof either of f or of :f, is
a broad one. More, there are innitely many formulas f unprovable in
2
Gentzen described the fact that lem prescribes uses of logical particles other than those given by their
introduction and elimination rules as troublesome. The way that in the sequent calculus the logical
rules are quarantined from the distinction between classical and intuitionistic logic he called
seemingly magical. He wrote, I myself was completely surprised by this property . . . when rst
formulating that calculus (Gentzen 1938: 259).
112
Curtis Franks
Logical nihilism
113
114
Curtis Franks
Logical nihilism
115
It seems to me that this attitude derives from wanting to preserve the naive
procedural understanding of the logical connectives. The situation ought
rather, I counter, lead one to appreciate the subtlety of procedurality
exhibited in intuitionism. For the logical lesson to be learned is that
in the absence of lem the context of inference takes on a new role. Thus
f means that given any background of assumptions from which f is
provable, a proof of f can be transformed into a proof of under those
same assumptions, and this understanding does not reduce, as it does with
logics insensitive to context, to the idea that any proof of f can be
transformed into a proof of . This irreducibility strikes me as a very
desirable property for many purposes. I should like to know more about
the conditions that lead to it.
From this point of view, it is natural to ask whether there are logics that,
unlike classical logic, admit a constructive interpretation but, like classical
logic, are not sensitive in this way to context. Perhaps the constructive
nature of IPC derives from its context-sensitivity. Surprisingly, Jankovs
logic, IPC:f _ : :f, appears to undermine any hope of establishing a
connection between these phenomena. Consider the Medvedev lattice of
degrees of solvability. The setting is Baire space (the set of functions
from to ) and the problem of producing an element of a given subset of
this space. By convention, such subsets are called mass problems, and their
elements are called solutions. One says that one mass problem reduces to
another if there is an eective procedure for transforming solutions of the
second into solutions of the rst. If one denes the lattice
of degrees of
reducibility of mass problems, it happens that under a very natural
corresponds to the set of theorems
valuation, the set of identities of
of Jankovs logic, so that the theory of mass problems provides a
constructive interpretation of this logic.3 However, Jankovs logic is structurally complete (Prucnal 1976). Thus one sees that the so-called weak law
of excluded middle preserves the context insensitivity of CPC despite,
standing in the place of full lem, allowing for a procedural semantics.
In my graduate student years, several of my friends and I were thinking
about bounded arithmetic because of its connections with complexity
theory and because the special diculty of representing within these
theories their own consistency statements shed much light on the ne
details of arithmetization. We had a running gag, which is that formal
theories like PA are awfully weak, because with them one cant draw very
many distinctions. The implicit punchline, of course, is that the bulk of
3
116
Curtis Franks
the distinctions one can draw in theories of bounded arithmetic are among
statements that are in fact equivalent. Lacking the resources to spot these
equivalences is no strength!
Something perfectly analogous happens in the case of substructural
logics. There are theorems of CPC that are unprovable in IPC, but not
vice versa, so the latter logic is strictly weaker. Moreover, CPC proves all
sorts of implications and equivalences that IPC misses. But if we stop
believing for a moment, as the discipline of logic demands we do, and ask
about the ne structure of inter-dependencies among the formulas of
propositional logic, IPC delivers vastly more information. Consider just
propositional functions of a single variable p. In CPC there are exactly
four equivalence classes of such formulas: those inter-derivable with p, :p,
p _ :p, and p ^ :p. In IPC the equivalence classes of these same formulas
exhibit a complicated pattern of implications, forming the innite Rieger
Nishimura lattice.
One thought one may have is that IPC should be considered an expansion of CPC: every classical tautology can be discovered with IPC via the
negative translation of Gdel (1933) and Gentzen, so with IPC one gets all
the classical tautologies and a whole lot more. (Gdel at times suggested
something like this attitude.) But, of course, neither the negative translation nor the very idea of a classical tautology arises within intuitionistic
logic. The thought that I encourage instead is this: The logician is loath to
choose between classical and intuitionistic logic because the phenomena of
greatest interest are the relationships between these logical systems. Who
would have guessed that the rejection of a single logical principle would
generate so much complexity an r.e. set of admissible rules with no nite
basis, an innite lattice of inter-derivability classes?
The intuitionist and the classicist have very ne systems. Perhaps with
them one gains some purchase on the norms of right reasoning or the
modal structure of reality. The logician claims no such insight but
observes that one can hold xed the rules of the logical particles and, by
merely tweaking the calculus between single conclusion and multi conclusion, watch structural completeness come in and out of view. The same
switch, he or she knows, dresses the logical connectives up in a constructive, context-sensitive interpretation in one position and divests them of
this interpretation in the other. These connections between sequent
calculus, constructive proof transformation, structural completeness, and
lem are xtures from our logical knowledge store, but they cannot
seriously be thought of as a network of consequences in some allegedly
correct logic.
Logical nihilism
117
118
Curtis Franks
4
5
Logical nihilism
119
...n
0 sup 1 2
n<
The theory of primitive recursive arithmetic extended with a principle of transnite induction to 0.
120
Curtis Franks
one way or the other. Their disinterest in the analogous results about
fragments of PA is just evidence that they harbor no skepticism about these
theories. We recognize Gentzens analysis of rst-order quantiers as one
of the deepest results in the history of logic as soon as, and no sooner than,
we stop believing.
I now wish to respond more directly to the objection that opened this
section. Of course it is true that in the study of logical systems one must
engage in reasoning of some sort or another. This reasoning can possibly
be described by one or a few select logical systems. But why should anyone
assume that this amount of reasoning is anything more than ways of
thinking that have become habitual for us because of their proven utility?
Further, why should anyone assume that there is any commonality among
the principles of inference we deploy at this level over and above the fact
that we do so deploy them?
To expand on the rst of these points, it may be helpful to draw an
analogy between rudimentary logic and set theory. Often it is thought that
decisions about which principles should govern the mathematical theory of
sets should be made by appealing to our intuitions about the set concept
and even about the cumulative hierarchy of sets. Doubtless such appeals
have gured centrally in the development of set theory. But the history of
the subject suggests that a complete inversion of this dynamic has also been
at play. Kanamori (2012: 1) explains:
[L]ike other elds of mathematics, [set theorys] vitality and progress have
depended on a steadily growing core of mathematical proofs and methods,
problems and results . . . from the beginning set theory actually developed
through a progression of mathematical moves, whatever and sometimes in
spite of what has been claimed on its behalf.
Logical nihilism
121
122
Curtis Franks
knew in advance of all evidence that some such property attains, then it
might be reasonable to look for it in whatever classes of inference rules we
happen to nd collected together. But we have no such foreknowledge
and, in fact, the evidence suggests that the arrangement of our toolkit is a
highly contingent matter. Are we not better o shedding this vestigial
belief that among all the intricate and interesting consequence relations out
there, some have a special normative status? Can we not get by with the
understanding that principles of inference with a rather wide range of
applicability dier from those suited only to specic inference tasks only in
having a wider range of applicability? Had this been the understanding
that our culture inherited, would anything we have learned from studying
logic lead us to question it?
At the end of Platos Phaedrus, Socrates explains that prior to investigating the essence of a thing, it is important to devise an extensionally
adequate denition of that thing so that we will be in agreement about
what we are investigating. This attitude seems right to me, and it seems to
me that the familiar debates about, for example, where logic leaves o and
mathematics begins violate this principle unabashedly. Suppose it were
clear to everyone that some but not all patterns of reasoning are inescapable and furthermore that it were easy to tell which these are. Then we
would have good reason to label this logic in distinction to patterns of
reasoning that we all recognize as the province of some special science or
particular application. It would be reasonable to wonder what accounts for
the privileged role that these principles play in our lives. As things stand,
however, many of us seem instead to assume that there simply must be
patterns of reasoning that dier in kind from others. Typically, our minds
are already made up about the psychological or metaphysical circumstances
that underwrite this dierence. This is what Wittgenstein stressed with his
observation that the crystalline purity of logic was not the result of an
investigation, that instead it was a requirement (Wittgenstein 1953:
107). Driven by this assumption, we thrash about looking for some
extensional denition that we can hang our ready-made distinction on.
These denitions are simply unconvincing on their own. They can satisfy
only people who cannot tolerate the thought that there is no line to
be drawn.
When Gentzen began his study of logic, he parted ways with his
predecessors7 by not rst dening logical validity and then seeking out
logical principles that accord with that denition. He simply observed that
7
There are historical precedents for Gentzens attitude: Aristotle, Condillac, Mill.
Logical nihilism
123
10
This is explicit in (Gentzen 193435) and even more vividly depicted in section 4 of (Gentzen 1936).
Actually Gentzen vacillated over the inclusion of the principle of mathematical induction,
ultimately deciding against it.
For details of this conception of completeness, see section 3 of (Franks 2010).
124
Curtis Franks
valid or valid in some other way than general reasoning, and even if they
did, we should be inclined to ignore these reports if they did not reect in
mathematical practice. For these reasons Gentzen could never bring himself to describe the distinction between inference rules that appear in the
predicate calculus and those that belong specically to arithmetic as a
distinction between the logical and the non-logical. He only thought that
he had designated a logical system, one that by design encodes some of the
inferences he was bound to make when reasoning about it but whose
logical interest derives solely from what that reasoning brings to light.
Contrast this with one of the more famous attempts to demarcate the
logical: Quines defense of rst-order quantication theory. Second-order
quantication, branching quantiers, higher set theory, and such can each
be dissociated from logic for failing to have sound and complete proof
systems, for violating the compactness and basic cardinality theorems, and
other niceties. There is even a (1969) theorem, due to Lindstrm, to the
eect that any logic stronger than rst-order quantication theory will fail
to exhibit either compactness or the downward LwenheimSkolem theorem. Second-order logic, Quine concluded, is just mathematics in
sheeps clothing because by using second-order quantiers one is already
committed to non-trivial cardinality claims (Quine 1986: 66).
How true will these remarks ring to someone who doesnt know in
advance that they are expected to distinguish logical and mathematical
reasoning? Quines consolation is telling: We can still condone the more
extravagant reaches of set theory, he writes, as a study merely of logical
relations among hypotheses (Quine 1991: 243). I should have thought that
this accolade, especially in light of the intricate sorts of logical relations that
set-theoretical principles bear to one another and that set theory bears to
other systems of hypotheses, would be used rather to enshrine a discipline
squarely within the province of logic. For if we never suspected that among
the plenitude of logical relations are a privileged few that capture the true
inter-dependencies of propositions, what else would we mean by logic
than just the sort of study Quine described?
As to the properties that characterize rst-order quantication theory, it
should now go without saying that from our perspective Lindstrms
theorem, far from declaring certain formal investigations extra-logical,
exemplies logic. So too do results of Henkin (1949) and others to the
eect that second-order quantication theory and rst-order axiomatic set
theory each are complete with respect to validity over non-standard
models. For a nal example, I can think of none better than the recent
result of Fan Yang that Vnnens system of dependence logic (with
Logical nihilism
125
branching quantiers), extended with intuitionistic implication (the context sensitive and constructive operator that results from the denial of lem),
is equivalent to full second-order quantication theory (Yang 2013).
Wittgenstein (108) expected his readers to recoil from the suggestion
that we shed our preconceived ideas by turning our whole examination
around. Rather than impose our intuitions about logic on our investigations by asking which principles are truly logical, let us rst ask if a close
look at the various inference principles we are familiar with suggests that
some stand apart from others. If some do, then let us determine what it is
that sets them apart. But the question he puts in our mouths But in that
case doesnt logic altogether disappear? suggests that we know deep
down that our empirical investigation is bound to come up empty. Various
criteria will allow us to demarcate dierent systems of inference rules to
study, but when none of these indicate more than a formal or happenstance association we will nd ourselves hard pressed to explain why any
one of them demarcates the logical.
I am more optimistic than Wittgenstein. The conclusion that I expect
my reader to draw from the absence of any clear demarcation of the logical
is not that there is no such thing as logic. Let us agree instead that no one
part of what logicians study, contingent and evolving as this subject matter
is, should be idolized at the expense of everything else. Logic outstrips our
preconceptions both in its range and in its depth.
4. Conclusion
Traditional debates about the scope and nature of logic do not do justice to
the details of its maturation. In asking whether certain inferential practices
are properly logical or more aptly viewed as part of the special sciences, for
example, we ignore how modern logic has been shaped by developments in
extra-logical culture. Similarly, questions about whether logic principally
traces the structure of discursive thought or the structure of an impersonal
world presuppose a logical subject matter unaected by shifts in human
interest and knowledge.
I mean, by saying this, not just to suggest that the principles of
rudimentary logic are contingent, not dierent in kind from principles
that we use only some of the time or very rarely and only for specic tasks.
I do urge this attitude. But the caution against mistaking our default,
multi-purpose habits of reasoning for something monumental is only
preparation for a second, more valuable reaction. One should warm up
to the trend of identifying logic with the specialized scientic study of the
126
Curtis Franks
relationships among various systems and their properties. This is, after all,
how logicians use the word. Our preference to ignore questions about a
logics correctness stems not only from an interest in exploring the properties of possible logical systems in full generality but also from an appreciation, fostered by the study of logic, that no one such system can have all
the properties that might be useful and interesting.
In closing, let me re-emphasize that the idea of a true logic, one that
traces the actual inter-dependencies among propositions, is unscathed by
all I have said. Part of the diculty in questioning that idea is that it is a
moving target: argue against it, you feel it again in that very argument;
close the door, it will try the window. But this very circumstance only
underlines the fact that the idea is a presupposition, nothing that
emerges from any discovery made in the study of logic. For the same
reason that we can marshal no evidence against it, we see that if we can
manage to forget it our future discoveries will not reveal to us that we
have erred.
This realization, coupled with the observation that a xation on the true
logical relationships out there hinders the advancement of logic, certainly
recommends nihilism on practical grounds. The question that remains is
whether we are capable of sustaining a point of view with no direct
argumentative support.
The proper antidote to our reexive tendencies will surely extend an
analysis of modern logic and include a rehearsal of the subjects history.
I cannot oer that here.11 I can only mention that logic as a discipline has
evolved often in deance of preconceived notions of what the true logical
relations are. Logic has been repeatedly reconceived, not as a fallout from
our better acquaintance with its allegedly eternal nature, but in response to
the changing social space in which we reason. There is reason neither to
expect nor to hope that logic will not be continually reconceived. Such
reconceptions have been and likely will again be fundamental, so that what
makes the moniker logic apt across these diverse conceptions is not an
invariable essence.
In these pages I have indicated instead logics modern contours, highlighting the fact that the deepest observations logic has to oer come with
no ties to preconceptions about its essence. The richness of logic comes
into view only when we stop looking for such an essence and focus instead
on the accumulation of applications and conceptual changes that have
11
For the details of the evolution of one central concept that of logical completeness in the past
two centuries, see (Franks 2013).
Logical nihilism
127
chapter 7
128
129
He certainly meant to say that this is true because in our world it is most
convenient to have a universal quantier, not because logic is itself empirical.
I believe that this part of what he says there, though not all of it, can be
attributed to him well before 1939.4 Even in 1939, Wittgenstein told the class:
To say A reality corresponds to 2 2 = 4 is like saying A reality
corresponds to two. It is like saying a reality corresponds to a rule, which
would come to saying: It is a useful rule, most useful we couldnt do
without it for a thousand reasons, not just one.5 (LFM: 249)
Such a view of mathematics places it on a par with the rules of logic. Both are
grammatical rules, the dierence being which vocabulary the rules govern.
This is not to say, with Frege and Russell, that mathematics is logic. The
rules of logic are used to prove mathematical theorems, to be sure, but this
does not make mathematics into logic: logic is used in every discourse.
During the period 19367, Wittgenstein began to study in earnest the
concept of rule-following which was to loom so large in his Philosophical
investigations. The connection between rules and regularities (Regelmigkeit)
becomes manifest to those who study his notebooks.
Rules are norms which evaluate what happens or what is done by people;
regularities are what happen most of the time, or what people do most of
the time when they are trained the same way. Rules label the deviations
from these regularities mistakes, abnormalities, perturbations6 (in
4
This further idea, expressed, as I say, in 1939, I would not want to attribute to Wittgenstein in the
early 1930s (which is what I am discussing here) and I will discuss it below, in the context of (what
I will call) his revolution of 1937. It is reminiscent of Nelson Goodmans relativism concerning
natural kinds (grue).
The idea that the relationship between the empirical world and mathematical propositions is that the
former makes the latter useful is not replaced in 1939, but augmented by a much deeper connection
between mathematical propositions and empirical reality, which we will discuss later.
I dont mean to say that the mathematical technique of perturbation theory is normative. I bring
the subject of perturbations in because Wittgenstein himself does:
Suppose we observed that all stars move in circles. Then All stars move in circles is an
experiential proposition, a proposition of physics Suppose we later nd out they are not
quite circles. We might say then, All stars move in circles with deviations or All stars move
in circles with small deviations. (LFM IV: 43)
If it is a calculation we adopt it as a calculation that is, we make a rule out of it. We make
the description of it the description of a norm we say, This is what we are going to
compare things with. It gives us a method of describing experiments, by saying that they
deviate from this by so much. (LFM X: 99)
130
Mark Steiner
physics) and the like. There are a number of possible explanations of the
utility of stigmatizing deviations in this way, at least in the areas of language,
logic, and mathematics.7 Society has an interest in rendering certain practices as uniform as possible, and adding negative and positive incentives
may do the trick.8 This account is plausible in the areas of language and
mathematics, which is our topic here.
The so-called rule-following paradox, as Wittgenstein himself labels it
in PI, is (and is intended to be) a paradox only for academic philosophers.
I9 use this term to refer to those who take the goal of philosophy to
explain10 human practices like rule-following i.e., almost all philosophers
besides Wittgenstein. The explanations emerge from diverse philosophies,
from mentalism to physicalism, but all agree that there must be some fact
about a person, beyond the regularities of his behavior, in virtue of which
we can say he is following a specic rule at a specic time and not another.
The explanation that follows here follows that of Saul Kripke, though,
as will become clear, it diers from his in some crucial details.11 Kripke
attributes a skeptical argument concerning rule-following to Wittgenstein, and has drawn much criticism on this account. I agree with Kripke
that Wittgenstein did construct a skeptical argument, but I hold that the
argument is supposed to be valid only for academic philosophy, as distinct
from Wittgensteins own philosophy.
7
Wittgenstein never attempted to found an account of ethics on this basis there are regularities in
the way people treat one another, and moral norms arise from stigmatizing deviational behavior
and it is an interesting question why.
When someone whom I am afraid of orders me to continue the series, I act quickly, with perfect
certainty, and the lack of reasons does not trouble me. (PI: 212) I believe that this passage reects
actual occurrences in Wittgensteins life. After the publication of the Tractatus, Wittgenstein left
academics and went into school teaching. Ray Monk (Monk 1991, pages 195196, 232233) reports
that Wittgenstein used to inict corporal punishment on his pupils if he thought they were not
applying themselves to the arithmetic lessons he was giving them. Not enough has been said about
the connection between the rule-following arguments in Philosophical investigations and
Wittgensteins short-lived experience as a schoolteacher, which came to an end, when one of his
pupils lost consciousness as a result of being struck by Wittgenstein.
Wittgenstein himself referred to academic philosophy in a letter, but not in his published or (so
far as I know) unpublished works. Felix Mhlhlzer draws my attention to the following passage
from Zettel, 299:
We say: If you really follow the rule in multiplying, it MUST come out the same. Now,
when this is merely the slightly hysterical style of university talk, we have no need to be
particularly interested. . .
10
11
Wittgenstein condemns this kind of philosophy in one of the most famous passages in Philosophical
investigations: We must do away with explanation and description must take its place.
(Philosophical investigations, 109; in the 4th edition this is translated: All explanation must
disappear. . .)
Cf. (Kripke 1982).
131
Since it turns out that there is no such fact which can serve as an
explanatory criterion for rule-following, academic philosophers are faced
with a paradox since it now follows that there is no such thing as rulefollowing at all.12 This could be considered a skeptical argument, though an
ad hominem one.13
Let us now examine the arguments Kripke brings to support the
contention (on behalf of Wittgenstein) that there is no fact in virtue of
which somebody is following a rule. Kripke adduces two arguments
for this, of which only one is actually in Wittgenstein. This is the
normative argument: to ascribe rule-following to someone is to assert
that someone is acting according to a norm, i.e. following the rule
correctly as we say. The gap between is and ought then implies
that no fact or state of the person at time t could be identied with
following the rule at t. The situation is dierent in other cases of
explanation and reduction in science.14 Disposition terms (e.g. solubility) can in principle be reduced to state descriptions of a substance,
which actually replace the disposition term. This is the so-called place
holder theory of dispositional terms. Given the normative nature of
rule-following, i.e. its social nature, it applies to what interests society:
behavior. Thus it cannot be reduced to, or identied with, but only
correlated with, an underlying state description either of the mind or of
the brain; that is Wittgensteins argument.
Kripke has another argument, the so-called innity argument,
according to which any state of the brain,15 for example, is necessarily
nite; while rule-following commits the trainee to innitely many
12
13
14
15
The case is formally similar to Freges foundation of arithmetic upon logic. When Russells
paradox showed that Freges logic is inconsistent, Frege overreacted by saying arithmetic
totters.
Kripke compares Wittgensteins skeptical argument to that of David Hume, and the comparison
is just, but not in the way that Kripke imagines: both Wittgenstein and Hume use skeptical
arguments to dispose of various kinds of academic philosophy, without themselves being skeptics.
As I have argued in the text above, Humes skeptical argument disposes of necessary connections
between events, which are used in rationalist explanations of causal reasoning. It is a skeptical
argument only for them, because they hold that, without the necessary connections there is no
causal reasoning at all. See here (Steiner 2009: 26 ).
I am here oering my own opinions, not those of Wittgenstein. In fact, I am not at all sure that
Wittgenstein distinguished clearly between dispositions in science and abilities in humans, since in
Philosophical investigations, 193194, he claims that academic philosophers make the same kind of
mistakes in discussing the abilities of humans with dispositions of machines. See also Philosophical
investigations, 182, where he compares to t (said of bodies in holes) and to be able, to
understand, said of humans.
As my colleague Oron Shagrir has cogently argued, Kripke seems to be thinking of a brain state as
the physical realization of a nite digital computer.
132
Mark Steiner
applications of the rule as in the rule add two. Not only is this
argument not in Wittgenstein, it couldnt be, as I will argue below.16
The rule-following paradox is, then, a paradox only for academic
philosophy. For Wittgenstein himself there is no paradox to begin with.
For paradox to loom, our ordinary discourse about some topic must be
seen to lead to catastrophe. For example, Zenos paradoxes began with
ordinary conceptions of motion and showed that they lead to inconsistency, or to the conclusion that no motion is possible. Wittgensteins
account of rule-following involves the claim that all rules are supervenient
upon regularities. In the case of the rule add 2 which surfaces in
Philosophical investigations how do we know that our trainee is following
this rule if he manages to go 2, 4, 6, 8, . . ., 1000 or another rule which
says that after 1000 one starts adding 4? We dont know, but our
experience, both as students AND as teachers, is that almost all who
produce this series go on the same17 way: to 1,002 and not to 1,004.
(The claim is not that we reect upon these regularities that would be
a misinterpretation but that the regularities make our practice in this
regard possible and coherent.) This regularity allows us to attribute the rule
add two (the rule which is hardened from just this regularity) with
great condence to our trainee, and to call his response erroneous or
perhaps provocative if he says next 1,004. (We may lter out frivolous
responses by warning the trainee that he will be severely sanctioned if he
doesnt give the right answer.) Wittgenstein expressed this idea quite
clearly in 1939:
Because in innumerable cases it is enough to give a picture or a section of
the use, we are justied in using this as a criterion of understanding, not
making further tests, etc. (LFM I: 21)
17
In fairness, however, I must add that my own presentation of Wittgensteins normative argument
that there is no fact in virtue of which somebody is following a rule, also improves the argument
somewhat. The distinction I draw between disposition terms which are in principle reducible to
state descriptions and rule-following ascriptions which are not is not in Wittgenstein. On the
contrary, Wittgenstein tends to see rule-following as precisely a disposition, but denies that
disposition terms are reducible to underlying state descriptions. Since I think Wittgenstein is
mistaken here, not only about the issue itself, but also about how to make his own argument
(a malady which many philosophers are prone to), I have made the necessary adjustments. Kripkes
innity argument, on the other hand, is one which in fact contradicts basic Wittgensteinian
insights and has no place even in an improved Wittgensteinian corpus.
I will deal below, page 133, with the objection that there is no objective meaning to the same and
hence that the claim is circular.
133
goes on to give the same responses that we would give, i.e. agreement
with society. On the contrary, society is itself predicated on empirical
regularities of its members. The criterion, then, is simply that our trainee
has successfully followed the rule 2 up to now. It is true that without
behavioral regularity upon training, this criterion would have no point.
But to apply the everyday criterion, one does not have to reect on this
regularity, or even know about it.
Wittgenstein argues in Philosophical investigations that the notion of a
rule and that of regularity are picked up simultaneously as a result of our
common training, so that there is no circularity in saying that a rule is
founded on regularity even though detecting a regularity requires an ability
of ours to follow rules. The same training teaches the concept of the
same. For this reason our previous statement that People trained the
same act the same is not susceptible to skeptical doubt.18
In the passages of Philosophical investigations we are discussing, though
not necessarily in RFM, Wittgenstein is employing a very simple concept
of applying a rule. We may understand Wittgenstein as saying that
applying a rule is simply following it (correctly). Since rules are grounded
in regularities, it is the ability to continue the series by doing what almost
everybody does when placed in the same situation, which grounds the
ability to apply the rule. In principle, there is no dierence between rules
of logic and grammatical rules. The hardness of the logical must is a kind
of projective superstition, much as the superstition that Hume thought he
had exposed in the idea there is necessary connection between causes and
their eects. It is similar to the superstition of thinking that when one is
reading he is having a characteristic experience of being inuenced or
being guided by the text. (PI, 170.)
Since rules are norms, there is no equivalence between saying that
somebody is following a rule and saying that his behavior falls under the
underlying regularity. Saying that somebody is following a rule is simply
evaluating his behavior, not describing it even though the evaluation
results from observing his previous behavior and responding to it in light
of our own training in following rules, and the regularities that are
instilled by that training. In other words, we can describe the criteria
that a teacher is using to evaluate the students behavior as successfully
following a rule, even though the teacher may not be aware of these
criteria. In fact one of the purposes of Wittgensteins analysis in
18
134
Mark Steiner
20
PI, 151. I understand the Private Language Argument of Wittgenstein as saying that what is
called referring to our mental states is more like expressing them than naming them.
Can one say Where there is no doubt there is no knowledge either?
135
Let us go back to the law of contradiction. We saw last time that there is a
great temptation to regard the truth of the law of contradiction as something which follows from the meaning of negation and of logical product
and so on. Here the same point arises again. (LFM: 211).
We can say, then, that logical laws arise in a two step process. First, the
child is trained in the use of words like not. The training induces a
regularity in this use, a regularity which society reinforces as correct
usage. Within this regularity, however, there arises a subregularity, when
the rules for using not are to be applied to special cases like double
negation. Most trainees nd themselves using double negation as they
would armation. This regularity is then put in the archives as a law
of logic.
Something similar happens in arithmetic, according to Wittgenstein. In
applying the rules for division to 1/7, most procient students nd themselves repeating the sequence 0.142857142857. . .22 In fact, most procient
students in dividing m by n always get a nite decimal or a repeating
decimal. This subregularity is then converted into a rule in itself, a law or
21
22
This is a point that Quine also made during the very same period. See (Quine 1936a), reprinted in
(Quine 1976).
See LFM, p. 123.
136
Mark Steiner
23
This is not Wittgensteins only interpretation of geometry: see (Mhlhlzer 2001) for another one.
137
One might think that the innite expansion of 1/7 is determined25 from
the beginning by the rules for division that are learned in school (or were
once learned in school). But the rules cannot outstrip the regularities that
are their basis, and the regularities, being regularities of human beings
cannot go on forever, and in fact, at some nite point, the regularities will
peter out: the deviation will increase to the extent that no rule could be
founded on human practice.
Mathematics to the rescue of mathematics: the theorem gives a schematic picture of doing the division. Using a pigeonhole principle it is
clear that the algorithm will run out of remainders, and thus that the
24
Wittgenstein asserts:
It is the use outside mathematics, and so the meaning of the signs, that makes the sign-game
into mathematics. (RFM, V: 2)
Here we have the extreme anti-formalist statement that the applications of mathematics give
meaning to its language.
In case the message has been missed, Wittgenstein relays it again at once:
What does it mean to obtain a new concept of the surface of a sphere? How is it then a concept
of the surface of a sphere? Only insofar as it can be applied to real spheres. (RFM, V: 4)
25
Mark Steiner
138
rst remainder, 3, will recur, and thus that the whole cycle will start again.
This induces the mathematician and the rest of us to label as wrong any
calculation which does not lead to a repeating decimal; it overrules the
nave use of the school rules. This kind of proof is characteristic of
mathematics:
It is just the same with 1:7 = 0.142857142. . . You say, This must give
so-and-so.
Suppose it doesnt.
Suppose what doesnt?
Here I am adopting a new criterion for seeing whether I divide this
properly and that is what is marked by the word must. But it is a
criterion which I need not have adopted. For just as bricks measured with
all exactness might give a curve (space is curved), so 1 : 7 = 0. . . . looked
through with all exactness might give something else. But it hardly ever does
[my italics i.e., we have noticed an empirical regularity]. And now Ive
made up a new criterion for the correctness of the division. And I have
made it up because it has always worked. If dierent people got dierent
things, Id have adopted something dierent. (LFM XIII: 129)
By nitism Wittgenstein always means what is now called strict nitism, according to which it
is incorrect or false to assert There are innitely many natural numbers.
139
We now can see where Kripke went wrong in attributing the innity
argument to Wittgenstein. The argument was supposed to defeat the idea
that following a rule is identiable with some (perhaps dispositional)
state of the brain. When we say that somebody is following the rule 2
or even plus, we are saying that he is committed to innitely many
(correct) responses to the question, What is . . . 2? But the brain, being
nite, cannot produce innitely many answers to questions of this kind.
Kripke discusses a number of possible responses to this argument and nds
fault with them all. He does not realize, however, that the major premise of
his argument is in direct conict with a basic feature of Wittgensteins
account of arithmetic: the idea that adopting an algorithm like plus
determines in some physical, mental, or metaphysical way ones response
to innitely many exercises is nothing but covert Platonism, in many ways
worse than the Platonism of objects.
These reections reect on the application of logic to arithmetic. By the
application of logic to arithmetic I mean simply the substitution of
arithmetic propositions in the variables (or schematic letters, if you prefer)
of logical rules or truths. Consider the law of the excluded middle, a law
of the Propositional Calculus, p_ p. An application of this would
be: Either the Goldbach conjecture is true or its negation is true. The
Goldbach conjecture states that every even number greater than 2 is the
sum of two primes (e.g. 8 = 5 3). The conjecture has been shown to hold
for very large numbers, and there are corollaries of the conjecture which
have been proved. But no proof of the full conjecture has been given,
though most mathematicians are persuaded that it is true. (There are
pseudo-probabilistic arguments for this, based on the fact that as the
numbers get larger, the probability that a given number can be partitioned into two primes rises monotonically, since the number of the
partitions themselves rises.)
The intuitionists hold that it is a form of metaphysics to assert the law
of excluded middle for such a case. To assert it here is to presuppose that
the natural numbers form a closed totality, or what Aristotle called an
actual innite, so that we can say that either there is, or is not, a
counterexample to the Goldbach conjecture in this closed totality. If we
think of the natural numbers through the metaphor of becoming,
rather than being, then the present absence of a proof or of a refutation of the Goldbach conjecture means only that the truth of the
conjecture is not determined, and the law of the excluded middle cannot
be asserted. As an invalid rule of inference, it is thus banished from
classical mathematics.
140
Mark Steiner
I actually deny this, and have given examples of mathematical propositions that were assertible even
when there was no proof of them in (Steiner 1975). But I will take for granted that Wittgenstein
agrees with Dummett on this point, an agreement that has a solid basis in the corpus.
141
Mhlhlzer protests this translation, which has become entrenched in the philosophical
Wittgenstein discourse, and insists that the right phrase is multi-colored.
142
Mark Steiner
143
The real harm will not come in unless there is an application, in which case
a bridge may fall down or something of that sort. (LFM: 211)
For as long as an actual inconsistency does not turn up, Wittgenstein held,
we need not worry that the bridges will fall down. Like any other
mathematical proposition, inconsistency is either a rule, or nothing. As
long as it is not a rule, i.e. a proven theorem, physical applications go on as
before.
But lets look at this a little closer. Wittgenstein discusses whether a
bridge could fall down because somebody divided by zero. This is certainly
possible; consider the equation x2 = x. 93 percent of precalculus students at
City College of the City University of New York, in a recent test, divided
by x and got the (only) answer x = 1.29 Not knowing about the solution
x = 0 could, in some scenarios, indeed cause a bridge to fall down. Much
more sophisticated cases could be constructed in which somebody does
not know he is dividing by zero.
Is this a case, however, of an inconsistency of a formal system, or is it
just a simple mistake in informal mathematics? One could imagine a case of
teaching students an axiomatic number theory in which cancellation of
zero is possible, in other words an inconsistent system. The students might
not even notice that ac = bc ! a = b yields 1 = 2 if we allow c to be zero,
because they have little cause to divide by zero. But it is hard to think of an
actual case in which a hidden contradiction in a formal axiomatic system
caused bridges to fall down.
A good example of this quandary is the theory of quantum electrodynamics (QED), pioneered by, among others, Schwinger and Feynman.30 The
calculations aorded by this theory are remarkably accurate, but nobody
knows how to base the calculations in a consistent axiomatic mathematical
system. In fact, there are mathematical physicists who think it cannot be
done. One reason is as follows. In calculating the probability of events in
29
30
144
Mark Steiner
part ii
chapter 8
1. Aristotle
There is not much in Aristotles own writings that bears directly on our
question. Four passages are noteworthy.
147
148
Paul Thom
Here Aristotle leaves two positions open: either statements do not change
truth-value at all, or else any change in their truth-value is due to a change
in something external to them, namely the things which the statements
are about.
Second, in the De Interpretatione we nd Aristotle apparently proposing
a general semantic theory according to which the meaning of spoken and
written utterances is to be found in the existence of mental items that
somehow correspond to them:
Now spoken sounds are symbols of aections in the soul, and written marks
symbols of spoken sounds. . . . Just as some thoughts in the soul are neither
true nor false while some are necessarily one or the other, so also with
spoken sounds. For falsity and truth have to do with combination and
separation. (Aristotle 1963: 16a2)
Here, the meaning of spoken and written language is derived from aections in the soul, and truth and falsity are seen as residing primarily in the
combination or separation of mental items.
Third, there is a remark in the Posterior Analytics which, again, seems to
point to the soul as the locus of truth and demonstration.
By contrast, it is always possible to nd fault with external arguments (i.e.
spoken or written ones): For demonstration is not addressed to external
argument but to argument in the soul since deduction is not either. For
one can always object to external argument, but not always to internal
argument. (Aristotle 1994: 76b23)
149
These are scattered remarks. Aristotle doesnt show how they could be
combined in a coherent theory of terms, propositions and arguments. We
do not nd such a theory in Aristotle; we nd only some materials that
seem to have the potential for theoretical development.
An interpreter of Aristotle, faced with this situation, might try to
develop a theory in one of two ways. One option would be to enlist
elements drawn from Aristotles metaphysics or his account of scientic
knowledge. Another would be to import non-Aristotelian ideas. We will
see that both approaches were used by later Aristotelians in their eorts to
esh out Aristotles sketchy remarks.
One obvious place to look for theoretical help in this enterprise is the
Philosophers division of all beings into the ten categories (substances,
quantities, relatives, qualities etc). From the standpoint of the theory of the
categories, our question becomes: Do the objects of logic belong to any of
the Aristotelian categories, and if they do, to which category or categories
150
Paul Thom
2. Robert Kilwardby
The thirteenth-century English philosopher and churchman Robert
Kilwardby commented extensively on Aristotles logic, as well as composing a treatise On the origin of the sciences and a set of questions on the four
books of Peter Lombards Sentences. Over the course of his career he
showed a continuing interest in the nature of the objects of logic, and
indeed the nature of logic itself.
In his early question-commentary on the Prior Analytics Kilwardby takes
the view that logic is one of the language-related sciences along with
grammar and rhetoric. In the works rst sentence he adopts Boethiuss
characterisation of logic as an art of discoursing (Kilwardby 1516: 2ra).1 He
goes on to consider the meaning of the words proposition [propositio,
Aristotles protasis] and syllogism [syllogismus] as they occur in Boethiuss
translation of Aristotles text, distinguishing propositions from statements
[enuntiationes]. A statement is put forward on its own account, a proposition on account of the conclusion it is intended to support. A statement
expresses what is in the speakers soul, and accordingly is dened as that
which is either true or false since truth and falsity reside in the soul
(Kilwardby 1516: 4rb).2 In his other writings Kilwardby will generally
preserve this distinction, reserving the term proposition for the premise
of an argument.
He asks whether a syllogism should be dened as a kind of process,
rather than a kind of discourse (following Aristotles denition). He agrees
that there is a sense in which a syllogism is a mental process, but says that
this is a metaphorical sense (Kilwardby 1516: 4vb).3 And it must indeed be
regarded as a transferred usage for someone whose starting-points are
Aristotles usage of syllogism to mean a kind of discourse and Boethiuss
characterisation of logic as a science of language.
In his later work On the rise of the sciences logic is no longer characterised
purely as a linguistic science, and the syllogism is no longer a purely
linguistic phenomenon. Logic is there presented under two guises. It is a
science of reason as well as being a language-related science:
1
2
3
151
Paul Thom
152
Here is his question:
King 2010: Abelard describes this as signifying what the sentence says, calling what is said by the
sentence its dictum (plural dicta). To the modern philosophical ear, Abelards dicta might sound like
propositions, abstract entities that are the timeless bearers of truth and falsity. But Abelard will have
nothing to do with any such entities. He declares repeatedly and emphatically that despite being
more than and dierent from the sentences that express them, dicta have no ontological standing
whatsoever. In the short space of a single paragraph he says that they are no real things at all and
twice calls them absolutely nothing. They underwrite sentences, but they arent real things. For
although a sentence says something, there is not some thing that it says. The semantic job of
sentences is to say something, which is not to be confused with naming or denoting some thing. It is
instead a matter of proposing how things are, provided this is not given a realist reading.
153
Again, the twelfth-century Ars Burana denies that enuntiabilia belong to any
of the Aristotelian categories.They exist, but belong to a category of their own
(Ars Burana, 208).6
But Kilwardby doesnt have these versions in mind when he refers to the
view that the stateables are not to be found in any Aristotelian category.
Rather, he is thinking of the version of the view advanced by the English
theologian Alexander of Hales (Hales 19511957: 1 d.39 n.1). Alexander held
that the ontological type to which a statement belongs depends on whether
the statement expresses an essential or an accidental predication. In the
former case the statement is nothing other than its subject, and thus
belongs to the same Aristotelian category as its subject. Thus the statement
Fido is a dog is a substance, and is the very same substance as Fido. In the
case of accidental predications, the statement can be reduced to the
Aristotelian categories in one of two ways: either it reduces to the category
in which its accidental predicate is located, or partly to that category and
partly to the category of the subject. Thus Fido is white turns out either
to be a quality (and then it is the quality of whiteness) or partly a quality
and partly a substance (and then it is partly Fido and partly whiteness)
(Kilwardby 1986: 1 q.90: 70).
In opposition to this view, Kilwardby holds that compositions, stateables and the other objects of logic can be assigned to the Aristotelian
categories in their own right without having to be reduced to the categories
to which their subjects and predicates belong. His view involves a complex
reduction to the Aristotelian categories.
Every thing, he declares, is either divine or human. The products of
nature he includes among the divine, along with things that issue from
God by himself. Human things, in his parlance, do not include what issues
from humans solely in virtue of their existence as natural beings, but only
what comes about through human activity in the form of industry or skill.
He classes the objects of logic, not among divine things, but among
human things in this narrow sense (Kilwardby 1986: 1 q.90: 102).
Among such things he distinguishes those that are internal to a human
and those that are external. The former include actions of combining,
dividing or reasoning, as well as the corresponding acts which he calls
6
Anon 1967: If you ask what kind of thing it is, whether it is a substance or an accident, it must be
said that the sayable [enuntiabile], like the predicable, is neither substance nor accident nor any kind
of other category. For it has its own mode of existence [suum enim habet modum per se existendi]. And
it is said to be extracategorial, not, of course, in that it is not of any category, but in that it is not of
any of the ten categories identied by Aristotle. Such is the case with this category, which can be
called the category of the sayable [praedicamentum enuntiabile].
154
Paul Thom
combinations, divisions, reasonings etc. The human things that are external include utterances, the making of works and the works made (e.g. the
making of a house, and the house that is made). This distinction between
what is internal to the human and what is external appears to rest on a
distinction between doing and making. While making can be considered
as a kind of doing, it can also be distinguished from other kinds of doing
insofar as it involves the production of something, or at least a process
aimed at the production of something. Thus when we mentally combine
or separate concepts, or when we reason in our heads, we do not thereby
produce anything external to ourselves: we have done something but we
havent made anything. But when we utter something, or build a house,
we do produce something external, we make something. If this is what
Kilwardby means, then the acts which he distinguishes from actions, and
which he also considers to be internal, cannot be products of those actions.
Being purely internal, they have no product. It is clear that the relation
between acts and actions should be similar to the relation between works
and the making of works. But works stand to the making of works in more
than one relation. The relation of product to process is one such relation,
but it is of no use to us here because doings which are not makings have no
products. There is, however, another relation connecting works to their
making: the relation of completion. All actions, in principle, have completions; and it is these completions, I believe, that Kilwardby refers to as acts
or things-done. Thus the human things that are the objects of logic include
completed acts of stating and reasoning, as well as the actions that have
those acts as their completions. According to Kilwardby, all of these are
things of reason. They are secondarily in a category, because they are
founded on things of nature in one of two ways. In the case of makings
and actions of reason, they are founded on things of nature in the sense
that the latter constitute their subject matter. In the case of things-done or
made by reason and art, they are founded on things of nature in the sense
that they are certain relations or accidental conditions of things of nature.
Kilwardby takes both of these senses to indicate that the things of reason
and art have things of nature as their subjects; and he means here the
metaphysical subject that underlies these things of reason and of art. Thus,
while it is the things of nature that are primarily and of themselves in the
categories, the things of reason and art can be assigned to the categories in
a secondary sense, via the things of nature that are their metaphysical
subjects (Kilwardby 1986: 1 q.90: 111).
Stateables and arguments, whether completed or incomplete, may exist
in writing, in speech or merely in thought; and Kilwardby applies the
155
156
Paul Thom
157
3. Later thinkers
Jakob Schmutz argues that scholastic ideas were transmitted to the early
modern period along two paths. The rst of these paths, which he calls the
idealistic main road, took the subject-matter of logic to be the activity of
the mind. The second path, the realistic by-pass, took logic to deal with
independent objects and structures (Schmutz 2012: 249). We have seen a
version of the rst path in the writings of Kilwardby. Kilwardby was a
moderate realist. But other versions of this path can be found in nominalists like William Ockham, for whom the objects of logic are individual
written, spoken, or mental tokens.7 Walter Burley, who opposed Ockhams views in most matters, appears to be working within the second
path: for him, propositions are either complexes depending on mental acts
of composition and separation, or intentional complexes existing in the
mind, or complexes existing outside the mind, which are signied by those
mental complexes. These extra-mental propositions [propositiones in re] are
the causes of truth of mental propositions (Cesalli 2007: 234).
The second path is taken up in the nineteenth century by Bernard
Bolzano then by Frege. Bolzano believed in propositions in themselves
(Stze an sich), and held that it is the job of logicians to describe these
entities and their properties (Lapointe, this volume). He outlines his
notion of a proposition as follows:
One will gather what I mean by proposition as soon as I remark that I do
not call a proposition in itself or an objective proposition that which the
grammarians call a proposition, namely, the linguistic expression, but rather
simply the meaning of this expression, which must be exactly one of the
two, true or false; and that accordingly I attribute existence to the grasping
of a proposition, to thought propositions as well as to the judgments made
in the mind of a thinking being (existence, namely, in the mind of the one
who thinks this proposition and who makes the judgment); but the mere
proposition in itself (or the objective proposition) I count among the kinds
of things that do not have any existence whatsoever, and never can attain
existence. (Bolzano 2004: 40)
The objects of logic, on Bolzanos view, are not human things and are not
grounded in the things of nature. As Rusnock and George say, It should
be possible, [Bolzano] thought, to characterize propositions, ideas, inferences, and the axiomatic organization of sciences without reference to a
thinking subject (Rusnock and George 2004: 177).
7
158
Paul Thom
For Bolzano, propositions are not human things, they do not exist in the
mind or in language or in any way at all, and they are objective not relative.
His view is designed to pare down our conception of the objects of logic to
a bare minimum so that propositions are understood simply as that which
is true or false, and arguments are understood as congurations of
propositions.
4. Concluding remarks
In his essay in the present volume Graham Priest asks whether logic can be
revised, whether this can be done rationally, and if so how. And he
distinguishes logic as something that is taught, logic as something that is
used, and logic as the correct norms of reasoning (Priest, this volume).
I would like to add a few comments on Priests questions.
The history of logic contains plenty of examples of logicians proposing
to revise what hitherto had been accepted as the correct norms of
reasoning. Some of the great logicians Abelard and Ockham along with
the well-known greats of the nineteenth century saw themselves as not
just revising but reforming logic. Sometimes these reforms are motivated
by a sense that accepted logics are erroneous or in other ways inadequate to
accepted ideals of what logic should be. And sometimes what motivates a
reforming logician is a new vision of what logic should be. I think that the
major reformers of the nineteenth century had this sort of motivation.
Looking at the traditional logic of their day, which was a watered-down
version of medieval logic, usually along the lines of Schmutzs idealist
road, they worked with a vision of logic as an objective science. We benet
today from the fruits of that vision. But it can be salutary occasionally at
least to look back to the dierent aims of the idealist logicians of the high
Middle Ages.
The reason why Kilwardby and other idealist medieval logicians conceived of the objects of logic as human things is to be found in the aims
which they thought logic should have. In treating logic as an art, they were
committed to thinking that it should teach us how to construct good
denitions, divisions and arguments. So the objects of logic had to include
human activities of dening, dividing and arguing.
Everyone agrees that an argument is faulty if it allows the conclusion to
be false while the premises are true; and accordingly any good logical
theory has to include among its norms that one should not argue from
truths to a falsehood. Faults and norms go together.
159
chapter 9
161
So, in the end, the semantic notions of truth and logical validity in
predicate logic, being dependent on what the correlates of our universal
terms are, demand at least a certain semantic clarication of the issue of
universals. Contemporary conventional wisdom that we can glean from
ordinary logic textbooks would tell us that those correlates are sets, the
extensions or denotations of common terms. (See, e.g., Hurley 2008:
8284) And if we press the issue of what sets are, then we are told that they
are possibly completely arbitrary collections of just any sorts of things, yet
somehow they are abstract entities. Clearly, ordinary logic text books can
just stop there. After all, they are not supposed to go into the metaphysical
problems of abstract entities: qua logic texts, they are just supposed to
provide some validity-checking machinery, and need not worry about the
possible ontological qualms of metaphysicians these machineries involve,
just like elementary math texts, as such, need not worry about the
ontological status of mathematical entities when they concern themselves only with providing reliable methods of calculation or construction.
This sort of attitude of the logician toward the metaphysical issues raised
by his subject is almost as old as the subject itself, as is testied by
Porphyrys famously raising the fundamental questions concerning universals just in order to set them aside as pertaining to deeper enquiries, but
not to logic. (Spade 1994: 1) And of course it is one of the famous ironies of
the history of ideas that it was precisely on account of these questions that
medieval logicians got so much involved in these deeper enquiries that
John of Salisbury in his Metalogicon (John of Salisbury 2009: 111116) had
to complain about how his contemporaries endless debates over these
issues confuse, rather than instruct, their students of introductory logic.
But despite the pedagogical validity of Johns objection to this practice,
one cannot really blame those logicians who get involved in these issues;
after all, as we shall see, the answers to Porphyrys questions determine to a
large extent the construction of logical semantics in general, and thus the
understanding of the relationship between the subject matters of logic and
metaphysics in particular.
162
Gyula Klima
terms at all? Platos realist answer, namely, that the dierence between
universal and singular terms hinges on the ontological dierence between
the kinds of entities these terms primarily name, rests on a relatively
simplistic understanding of the semantic relations of these terms: i.e. the
notion that their meaning consists in naming these dierent kinds of
entities in the same way. In fact, generalizing on this idea we might say
that on a realist conception semantic dierences are accounted for in terms
of the ontological dierences of the semantic values of syntactical items of
dierent categories, and not in terms of the dierences in the semantic
functions of these items themselves: on this approach, in realism we can
have semantic uniformity at the expense of ontological diversity.
By contrast, those medieval thinkers who were convinced by Aristotles
and Boethiuss arguments against platonic universals (by John of Salisburys time practically everybody (Klima 2013a: n. 27)) would account for
the semantic diversity of singular and common terms not on the basis of
the ontological dierences of the kinds of entities these terms denote, but
rather in terms of how they denote the same kind of entities, namely,
individuals, the only kind of real entities there are. Thus, on this understanding of the Aristotelian view, we can have ontological uniformity on the
basis of semantic diversity. As we shall see, the two formulae just italicized
can be regarded as the two extremes of a whole range of possible positions
concerning the relationship between semantics and metaphysics, ranging
from extreme realism to thoroughgoing nominalism. Indeed, let me call
the theoretical extreme of extreme realism the position that holds that all
semantic dierences are ontological dierences: dierent items in semantically dierent syntactical categories dier in what kinds of entities their
semantic values are and not in what kinds of semantic functions relate
them to their semantic values. By contrast, on the other theoretical
extreme we have the position of extreme nominalism, which would hold
that all dierent items in semantically dierent syntactical categories dier
only in the kinds of semantic functions that relate them to their semantic
values, but all those semantic values are ontologically of the same kind, the
same, single kind of entities (or just the one single entity) there is. But in
order to see how actual historical positions can be arranged on this
theoretical scale, we should get into some further details concerning each
extreme.
On the platonic view, as we could see, the semantic relation between
common and singular terms and their semantic values would be of the
same kind: namely, denoting a single entity. What would make the
dierence would be just the further ontological relation of the entity
163
164
Gyula Klima
3. Scholastic conceptualisms
To see this issue in a little more detail, we should see exactly how the pieces
of the theory presented so far t together in this tradition of medieval logic,
which I like to call via antiqua semantics, in contrast to a radically
dierent medieval logical tradition that emerged from the works of William Ockham, John Buridan, and their fellow nominalists, which I refer to
as via moderna logic (Klima 2011a, 2013a). As we shall see, both of these
approaches to logical semantics are basically variations on what may still be
called conceptualism; however, they are based on radically dierent
165
conceptions of what concepts are and how they are related to their
objects, and accordingly give rise to very dierent constructions of logical
semantics.
The easiest way to make this contrast is through the analysis of an
example. Take one of the staples of scholastic lore: Every man is an
animal. This is an armative, universal categorical proposition (in the
medieval sense of proposition, meaning sentence-token), both terms of
which are common or universal terms, joined by a copula and determined
by a universal sign of quantity (a universal quantier, as we would say). On
the common via antiqua analysis, the subject and predicate terms of this
proposition, its categorematic terms, have their semantic property of
signifying human and animal natures, respectively, on account of being
subordinated to the respective concepts our minds abstracted from their
individuating conditions in the humans and animals we have been exposed
to. Thus, although whatever it is on account of which I am a man (i.e., a
human being, regardless of gender) is a numerically distinct item from
whatever it is on account of which you are a man, the concept we
abstracted from humans we have been exposed to in forming our concept
of man abstracts from any individual dierences (individuating conditions). This is precisely the reason why this concept will represent not
only the humans we have been exposed to, but any past, present, future
and merely possible humans, that is to say, whatever it is that did, does,
will or can satisfy the condition of being human, whatever this condition
is, and whatever means we have (or dont have) for verifying the satisfaction of this condition (which would be a question of epistemology and not
of semantics). Accordingly, the corresponding term (man in English or
homo in Latin) can stand for any of these individuals in a proposition.
Indeed, this is what it does in this proposition: it stands or (to use the
Anglicized form of the scholastic technical term commonly used in the
secondary literature) supposits for all human beings that presently exist.
(For an overview of scholastic theories of properties of terms, including
supposition, see Read 2011) The reason why this term supposits only for
presently existing humans is the present tense of the copula, which restricts
the supposition (reference) of the term to present individuals that actually
satisfy the condition of its signication, namely, those individuals that
actually have human nature signied in general by this term. By contrast,
with dierent tenses or modalities, or when construed with verbs and their
derivatives that signify acts of the cognitive soul (i.e., sensitive or intellective, as opposed to the purely vegetative, soul) that are capable of targeting
objects beyond the presently existing ones (such as memory, imagination,
166
Gyula Klima
167
168
Gyula Klima
169
170
Gyula Klima
The second strategy, as we could already see, was based on the idea that
the minds dierent ways of conceiving of mind-independent entities of
external reality produces certain mind-dependent, intentional objects, the
objective concepts, the information contents of our mental acts, by means of
which we variously conceive ultimately those mind-independent objects
that satisfy the criteria of applicability set by these objective concepts, or
intentions. This is precisely why in this tradition the subject matter of logic
was generally characterized as the study of second intentions, that is, of
concepts of concepts (such as the concepts of subject, predicate, proposition, negation, or the ultimately targeted notion of valid consequence).
So, the core-ontology of real mind-independent entities could in principle
have been exactly the same for these realists as for Ockhamist
nominalists.
In fact, both late-medieval realists and nominalists were conceptualists, but based on a rather dierent conception of concepts and their role in
logic, semantics, and epistemology. In this connection, it is informative to
compare Ockhams earlier, ctum-theory of universals with that of the via
antiqua conception discussed so far. For the important dierence between
the two is that even if Ockhams cta are ontologically on the same footing
as the objective concepts of the realists (they are beings of reason), and they
would be best characterized in the same way, namely, as the objective
information content of individual mental acts, they do not have the same
role in Ockhams theory.
In fact, as prompted by the arguments of his confrere, Walter Chatton,
Ockham came to realize that cta did not play any signicant role in his
logic at all, and so, grabbing his famous razor, he painlessly cut them out
from his ontology. The reason why Ockham could do so is that for him
the universality of universal representations (whether cta or universally
representing mental acts) consists merely in their indierent representation
of a number of individuals (in the case of a natural kind, all past, present,
future, and merely possible individuals of the same kind). However, this
indierent representation is due not to some abstracted condition of having
a certain nature that individuals of a given kind satisfy, but, as a matter of
brute fact, to the indierence of the causal impact of one individual or
another of the same natural kind on the human mind.
Accordingly, for Ockham, there is no question whether there is a real
distinction between the nature of an individual represented by a universal
concept and the individual itself (as this emerged as a metaphysical
question in the via antiqua), because what these concepts indierently
represent are just the individuals themselves. Therefore, for him, the
171
supposita of the terms subordinated to these concepts are not the individuals that actually have these natures relative to the time connoted by the
copula (as was conceived in the via antiqua), but simply the individuals
represented by the concept that are actual at that time. As a result, terms in
the predicate position do not signify inherent natures either, so Ockham
and his followers endorse the identity-theory of predication, as opposed to
the inherence-theory. According to the identity-theory, an armative predication is true, just in case the terms of the proposition supposit for the
same thing or things. But this is obviously not a general denition
of truth. In order to achieve a truth-denition on this approach, one
should provide similar satisfaction clauses for all logically dierent
proposition-types, such as negatives, universals, particulars, not to mention
the propositional complexes, such as conjunctions, disjunctions, etc.
(Klima 2009: c. 10).
As can be seen, on this nominalist approach, just as terms do not gure
into the calculation of truth-values with their intensions, but their extensions, so too, the truth of propositions themselves is not determined in
terms of their intension or signication, but solely by the extensions (sets of
supposita) of their categorematic terms. Accordingly, nominalist semantics
as such has no use for enuntiabilia or complexe signicabilia, as is brilliantly
illustrated by the logic of John Buridan.
On Buridans theory, propositional signication is simply the set of all
signicata (and connotata) of a propositions categorematic terms, which of
course yields a very coarse-grained conception of propositional signication. In fact, on this conception, contradictory propositions must signify
the same, although dierently, on account of the concept of negation
included in the one, but not in the other of the contradictory pair of
propositions (Klima 2009: c. 9). However, Buridan does not have to care
much. On his account, truth is not a function of signication, so, two
propositions of the same signication can have opposite truth-values.
Thus, when he needs a more ne-grained semantics of propositional
signication (as in intentional contexts) he can always refer to the diversity
of the corresponding propositions on the mental level, where, of course, in
line with his nominalist ontology, the mental propositions in question are
just inherent qualities, individual acts of individual human minds, just as
are the concepts entering into their semantic make-up (Klima 2009: c. 8).
So, nominalist semantics can aord to be based on an entirely homogeneous, parsimonious ontology (containing only two or three distinct
categories of entities, namely, substances, quantities sometimes identied
with substances or qualities, as by Ockham and qualities). However, this
172
Gyula Klima
parsimonious homogeneity is achieved at the expense of massive semantic diversity, assigning some of these entities various, distinctive semantic
functions, especially on the mental level. However these semantic functions are always dened in terms of the extensions of these mental items,
the formal concepts inherent in individual human minds, to the exclusion
of items in the ontological limbo of the objective concepts of the
older model.
Still, even the nominalist version of scholastic conceptualism could
maintain that logic is the study of second intentions without lapsing into
subjectivism, conventionalism, or skepticism, let alone psychologism features
that in a modern context are so often associated with conceptualism. Well,
how come? Actually, answering this question will allow us to draw some
general conclusions concerning both major versions of scholastic conceptualism sketched out here, and some general lessons we can learn from
these scholastic theories concerning the subject matter of logic and
metaphysics.
173
gure out just why. But that is an issue of epistemology, not semantics. As
far as meaning is concerned, on this conception you can have the same
concept as I do only if our individual mental acts latch onto the same kind
of objects in the same way, carrying exactly the same information. To be
sure, one of us may have a better understanding of the nature of the thing
or things thus conceived, on account of being more aware of the relationships among this concept and others, picking out the same nature dierently, on account of other, more specic or more generic information, as
when one of us knows the genuine quidditative denition of the kind of
thing in question. But regardless of whether either of us has this denition
in mind or knows what it would be, we can be said to have a concept of
this kind of thing as such, only if we managed to form the objective
concept of its essence, which must be the same for both of us, or we just
do not have this concept at all (Cf. Aquinas 2000: Sententia Metaphysicae,
lib. 9. l. 11. n. 13.).
From this it should also be clear that these objective concepts are nonconventionally objective. For what determines the information content of
our abstracted, simple concepts is what kinds of things they are abstracted
from, that is to say, the nature or essence of those things themselves. To be
sure, we can construct complex concepts out of these simple ones as we
wish and agree to express them by words we wish (ad placitum, as the
scholastics said), but whether the concept we both abstracted from samples
of H2O will apply to all and only samples of that kind of thing (even if we
cannot infallibly identify all such samples in all possible scenarios) is clearly
not a matter of our wishes.
Finally, even if our psychological mechanisms require that when we
form these simple concepts and their combinations our minds work with
their own individual, subjective mental acts, their formal concepts; the
logical relations among these mental acts are not a matter of the causal or
other psychological relations among them, but a matter of the relations of
their objective semantic contents, the relations among their objective
concepts. So, no wonder scholastic thinkers working in this tradition
would identify the subject matter of logic as those second intentions or
objective concepts of our objective concepts that express precisely these
objective conceptual relations. (Schmidt 1966; Natalis 2008)
Therefore, it should also be clear that the laws of logic in this framework
are supposed to be fundamentally dierent from the laws of psychology.
For while the former are the laws of the logical relations among objective
concepts, the latter are the laws of the causal relations among formal
concepts. Thus, whereas logic can be normative, prescribing the laws of
174
Gyula Klima
175
176
Gyula Klima
177
chapter 10
178
179
an objective status? And, if so, how do they acquire it? No answers to such
questions can be attempted without a substantive view of what objectivity,
and an ontology, are. Since this is not the place to defend my Kantian,
transcendental-idealist position on the matter, I will simply state it before
moving on.2 (Though I must note that, here, the matter dealt with is not
innocent: an ontology is a logic of being, hence what ontological status a
logic has is not independent of what logic it is.)
A transcendental philosophy as described and practiced by Kant is itself
a logic. It is not intended to decide such factual questions as whether there
is a God or humans are free, but to address semantical issues like what the
meaning of God or freedom is. The reason why the formidable epithet
transcendental is attached to it is precisely the misunderstanding
I alluded to above: if you think that logic only deals with (some) conjunctions, adverbs, and pronouns, then you are forced to qualify this narrow
concern as general logic and to conjure up some other name for the full
line of business.
Within the semantical space where the (transcendental) logical enterprise is located, one can take dierent words as primitives and establish a
network of semantical relations and dependencies based on those primitives and involving other words, each time resulting in (the beginning of ) a
dierent transcendental philosophy/logic; as more such structure is
exposed, the meanings of the words involved will become correspondingly
better established and clearer. If we want, we can even talk about concepts: clusters of largely interchangeable words resonating with a common
theme, not necessarily spoken but suggestively intimated by the resonance.
A transcendental realism (TR) is a transcendental philosophy/logic
that takes a cluster of largely interchangeable words including object,
substance, thing, and existence as primitives, and then turns to
the (hopeless) task of dening words like experience or knowledge
on that basis. A transcendental idealism (TI) my chosen course is a
transcendental philosophy/logic that takes its cue from a dierent
cluster including experience, representation, and consciousness,
and then moves to dening object and existence. Not surprisingly,
a TI has a lot more to say about objectivity what makes an object an
object than a TR: of primitives we will forever be dumb and, though
occasionally that incapacity is depicted as mystical depth, the bottom
line is that no interesting account of what primitives mean is forthcoming. In a TI, however, objectivity belongs to a derived cluster; hence its
2
180
Ermanno Bencivenga
181
in the corner was not an elephant; it was an armchair; but the meanings of
elephant and armchair are not disrupted by this mishap. And yet, it is
not always that easy; for, what about the semantical necessity, until circa
1905, that a wave is not a particle? And about the logical clash that ensued
when we were forced to deny that necessity: a clash whose logical character
would be missed by more parochial characterizations of logic? If, on the
other hand, you want to insist on a parochial characterization and attribute
that clash to the empirical realm, I urge you to consider what happened a
few years earlier, when the very logic of sets blew up in peoples face. As
very unfortunate happenings of this kind can never nally be ruled out,
logics align themselves with worlds, in the following way:
A logic cannot be a theory of meaningless discourse (of alogos). But any
word we use can only be meaningful if our whole discourse is meaningful:
if all words we use belong in an ideal complete dictionary that sets
consistent, connected relations among them once again, it is an all-ornothing aair. Only a logic associated with this kind of dictionary would
be objective in the sense of possibly describing a world of objects (would be
a real, not an apparent, logic), independently of the data that gave empirical content to its entries. As no such dictionary can ever be at hand, we are
never in possession of a logic but only of something we presume to be a
fragment of one, and which is always at risk of dissolving into the stu
dreams are made of. A logic (like a world) is worse than a territory
constantly under threat of being conquered by enemies: it is constantly
under threat of vanishing into thin air.
That being the case, a major consequence follows, of a sign opposite to
the Quinean puritanism mentioned earlier. Just as, in the absence of a
complete system of representations or a complete world, we are to maximize the consistency, connectedness, and inclusiveness of what systems of
representations or of intentional objects we do have, in the absence of a
logic we are to maximize our closeness to one, walking away from the
depopulated citadels of the propositional and the predicate calculi toward a
ner and ner appreciation of the logical distinctions between crowd
and mob, or magenta and scarlet. In a true Kantian vein,
completeness will be not actual but set as a task, so logic will graduate
from a tenseless doctrine into a concrete practice ready to uncover semantical treasures under any rock, and carry semantical threads around any
corners. The routine of jotting down a few axioms, formally interpreting
them by translating them into the stock language of set theory, and
vindicating them by proving a completeness theorem will be shunned
in favor of the completeness that really matters: the one that is never
182
Ermanno Bencivenga
For a systematic account of the contrast between Aristotelian and Hegelian logic, see my 2000.
See the following passages from Hegels 1995: in his metaphysics, physics, psychology, etc., Aristotle
has not formed conclusions, but thought the concept in and for itself (p. 217; translation modied);
it must not be thought that it is in accordance with . . . syllogisms that Aristotle has thought. If
Aristotle did so, he would not be the speculative philosopher that we have recognized him to be
(p. 223); Like the whole of Aristotles philosophy, his logic really requires recasting, so that all his
determinations should be brought into a necessary systematic whole (p. 223). While he thus
acknowledged the comprehensive character of Aristotles logic, however, Hegel did not see it as an
alternative to his own, as I do, but rather as a step toward the latter.
183
both and think of something else entirely. That Kants criteria of objective
identity be spatiotemporal shows him committed to this Aristotelian route:
one and the same thing cannot be at two dierent locations at the same
time. But, when it comes to the semantics of regulative ideas, including the
ones that determine the criteria of objectivity, his inclination seems to be
proto-Hegelian, witness his derivation of positive from negative freedom, or of reciprocal action from simultaneity.5
With Hegel, on the other hand, contradiction is not a threat: it is an
opportunity. When the semantics of a word faces a bifurcation between
contradictory options, its fate is to take both, and its job is to evolve in
such a way that both options be present in a dialectical overcoming of their
contrast. Light is both particles and waves: the two are complementary
descriptions of one and the same complex reality, indeed belong to the
very substance of that reality, which is nourished (adds to its concreteness, Hegel would say) by their antagonism. Therefore the world that no
one will ever experience but of which everyone takes herself to be experiencing a portion is a monistic one: as not even contradictions can divide,
no two things are radically divided; all divisions are but chapters of one
story. And the very unfortunate events that might bring this logic to a crisis
will not be the surfacing of contradictions, as is the case with its analytic
counterpart. It will rather be the confronting of occurrences (the Holocaust, say) that simply cannot be integrated within one and the same
comprehending, rationalizing, spiritual narrative.
I said that these are the two most obvious examples of the search for a
logic. There are countless, less obvious, others; except that they are not to
be found where one would be most likely to look for them. As I pointed
out already, individual calculi cannot be regarded as logics, unless they are
part of an ambitious program that extends over a substantial area of
experience, indeed potentially all of it. But, whereas most of what falls
under the academic discipline of logic does not qualify as logic for me, a lot
of traditional philosophy does. Transcendental philosophy is not a new
way of doing philosophy initiated by Kant: it is a new way of looking at
what philosophy has always done, without much awareness and hence
with considerable self-deception. Of course pre-Kantian, and many postKantian, philosophers typically took themselves to be establishing factual
claims like the existence of God or human freedom, but the way they did
5
For the former, see Groundwork of the Metaphysics of Morals, in Kant 1996, p. 94: The preceding
denition of freedom is negative and therefore unfruitful for insight into its essence; but there ows
from it a positive concept of freedom. . .. For the latter, see my 1987, p. 149.
184
Ermanno Bencivenga
185
would imagine that, when reason speaks, everyone will stop and listen.
Not so: behavior occurs in empirical reality, Kant thinks, subject to
empirical laws, so only empirical factors like temperament, education, or
emotions can have motivating force there. Reason has none. One can only
hope (if one takes reasons side) that those empirical factors will promote
what reason would want to see done; that moral feeling, say, will ally itself
with rational judgment and make the agent move the right way in this
case, take his distance from the stigmatized behavior. If the agent decides
otherwise, there is nothing reason can do. It can call the agent irrational,
even deny him the status of an agent; but the (non?)agent need not be
impressed by any of this. In fact, he can appropriate words like reason
and rational and provide them with his own semantics; and there will be
no forcing him to recognize that as an error. Reason (whatever that is) is
playing its own game and, however consistent and connected the game
might be, one can always, simply, opt out of it.
Same thing here. Every transcendental philosophy/logic sets out its own
game, to be played by its own set of rules. Now suppose that, by the rules
current in a particular game, I prove that, than which nothing greater can
be thought, necessarily to exist. If I am a believer, I rejoice in thus seeing
my faith conrmed and I generously broadcast my proof to all others, so
that they can see the light also. And I am puzzled when many of those
others, instead of coming to a harmonious, reasonable agreement with me,
use their disbelief as the premise of a modus tollens and start looking for
what is wrong with my proof. Eventually they might focus on something
I took to be included in the semantics of greater: that existing, say, is
greater than not existing. And they might deny it: adopt an alternative
account of greater. What can I do then? Clearly, they are playing a
dierent game; and, no matter how loud I protest, there is no convincing
them that my game is the one they should be playing. This last judgment is
internal to my game, and of course from that internal perspective it looks
irrefutable. From the outside, it just looks like something else one
could say.
In and by itself (more about this qualication shortly), logic has no
persuasive force. Despite the metaphors of constraint that are invariably
brought up in its wake, it can constrain no one. If anything, the practice of
logic (as opposed to the often deceptive theory of it) has a liberating eect.
You felt constrained to making a certain inferential step (say, from something being necessary to its being necessarily so); but, when you bring
logical acuity and attention to bear upon it, you realize that it was a matter
of habit, that you can dislodge yourself from that straightjacket and make
186
Ermanno Bencivenga
the step not a forced but an optional one that you are free to go either
way, you have a choice in the matter that you had missed at rst. Think of
the story long told about Girolamo Saccheri:7 of how he wanted to rm
up, once and for all, the necessity of Euclids fth postulate (to withdraw
the option of having it, or not having it, as an independent assumption)
and ended up unwittingly freeing thought from Euclidean fetters.
What, then, is the use of a logic? How is the search for its objectivity
ever going to pay o? Its value judgments, I said, are internal to the game
the logic is playing; it is only from within that game that certain principles
appear secure, certain inferential steps apodeictical, certain objections
untenable. So it is only internally that a logic, in and by itself, has a use.
The development of my transcendental philosophy/logic will be like the
development of an organism: a realization of its own potential and a
functional interaction of all its components. Repeatedly, I will come upon
theoretical options, and the game I am playing (I decided to play,
I committed myself to playing) will sometimes determine my choice of
one of them, in which case I will naturally accept that choice, and
sometimes not, in which case I will reect on what else I want to add to
the rules of the game in order to have it cover more ground, to make it
more delicately responsive to the rugged terrain on which I must travel.
And the places I get to by traveling on that terrain will retroact on my
initial commitments: I will regard those commitments as conrmed to the
extent that I approve of my destination; I will correct them to the extent
that I nd it unwelcome. There will even be surprises along the way:
locations I never thought I would reach but my rules irresistibly take me
to, either to be more powerfully reassured that I am on the right track, or
more anxiously aware that I must be doing (assuming) something wrong.
A logic is a self-organizing structure, self-enclosed and self-referential,
that provides the bare scaolding of a world and, if given enough data,
even a large part of its actual construction. (So, as anticipated earlier, a
logic includes its own ontology.) Luigi Pirandello called it a corrosive,
infernal little machine8 because from any imagined variation in the
existing circumstances it could engineer, one step after the other, the
most horrid outcomes; and he considered it something to be afraid of.
For me, the fear at issue here is the one that always accompanies freedom.
7
A Jesuit priest and professor of mathematics at Pavia, who published in 1733 Euclides ab omni naevo
vindicatus, a presumed reductio proof of Euclids fth postulate from his other assumptions, long
regarded as the rst (unintended) development of a non-Euclidean geometry.
Pirandello 1990, pp. 11081109.
187
Also, crucially, because of her interlocutors deference to her authority. See the following footnote.
188
Ermanno Bencivenga
Similar points could be made about ethos. The character and competence of a speaker, in and by
themselves, will have no power to persuade an audience unless the latter feels respect for them. So, as
appropriate to a discussion of persuasion that is, of how an audience can be manipulated , it is
always pathos (the emotions the speaker is able to instigate in the audience) that works if anything
does; and the real distinction is among the emotions that are in play. Both logos and ethos will only
be successful if the speaker can raise the emotions akin to them, and if these emotions, under the
circumstances, prevail over conicting ones.
chapter 11
1. Framework
The term logical realism, as it is commonly understood, picks out a
family of views that are committed to at least two theses.1 The rst, let
us call it (LF), is that there are logical facts. Here (LF) is construed in
the widest possible sense to include any theory that assumes that there is a
fact of the matter when it comes to the truth-value of claims about logic. (LF)
can thus be cashed out in more or less robust terms. Take for instance the
putatively true claim that modus ponens is a valid principle of inference.
The realist may be committed to there being something whatever this
turns out to imply that makes the claim that modus ponens is valid true.
Or she may understand the idea that the validity of modus ponens is a fact
to mean merely that the corresponding claim is true. Both interpretations
of (LF), and every other one in between, raise a number of questions that
go beyond the scope of the present chapter. (For instance: what is truth?
What is it for a fact to make true a truth?) What is relevant here is the
following: whatever she understands logical facts to be, what makes the
adherent to (LF) a realist about logic is a further assumption (IND), that
logical facts are independent of our cognitive and linguistic make-up and
practices; they are independent of our minds and languages. In this sense,
for the logical realist the truth or falsity of logical claims is objective.
History oers a number of theoretical alternatives to logical realism.
Whats common to nihilism, pragmatism and pluralism,2 for instance, is
the fact that they deny (LF). By contrast, the proponents of naturalism
(of which there are many variants, including logical psychologism) and
My thanks to Matt Carlson, Nicholas F. Stang, Penny Rush, David Sanson, Ben Caplan and Peter
Hanks, Julie Brumberg and Teresa Kouric for their input on previous versions of this chapter.
1
This characterization of logical realism draws on Resniks 2000: 181.
2
I revert to the denition of pluralism given by Stewart Shapiro in the chapter included in this
collection.
189
190
Sandra Lapointe
191
The theory of eidetic variation and Wesensschau Husserl eventually committed to is an instance
of this kind of epistemology, and this explains in good part why, in many circles, his theories
eventually fell into disrepute. Gdel adopted a similar view, and one directly inspired by Husserl.
(Cf. Kennedy 2012.) The idea that the realist might be bound to adopt an epistemology seems to be a
common objection to the doctrine as a whole. In her chapter in this volume Penny Rush argues for
the potential of phenomenology as regards this problem.
192
Sandra Lapointe
But it will be further asked what the ground is for this connection, since there is a reality in it which
does not mislead. The reply is that it is grounded in the linking together of ideas. In response to this
193
and that God displays (some of ) these truths to us.5 The former commits
Leibniz to truths that are independent of human minds (and language).
And taken together, these two assumptions explain how human knowledge
is possible on Leibnizs view. Leibnizs primary concern in introducing
propositio is not for what there is, but for how we acquire knowledge.
Leibnizs grounds to commit to the existence of proposition-like entities
are thus (in part) that the supposition of such entities and the further
assumption that a benevolent God exists! is required to provide a
coherent theory of knowledge. Similarly, Poppers grounds for thinking
that there is objective knowledge in (Popper 1968) for instance
whatever their merit, is that this allegedly explains certain features of the
sciences, such as the relatively autonomous character of scientic theories
and problems. Whether they are epistemological or otherwise, as long as
the logical realists grounds for believing in the existence of logical facts are
not themselves logical, I will call the kind of realism she adopts external or
extra-semantic.
Ones grounds to subscribe to (LF) and (IND) and to the idea that there
are proposition-like entities, in particular, need not be external. What
often underlies ones commitment to logical facts may correspond to
(implicit) theoretical desiderata or aims. Desiderata and aims are types of
grounds in the relevant sense: they have explanatory value when it comes
to accounting for the ontological commitments that come with a logical
theory. Let us call the kind of logical realism that would underlie such a
theory internal. Historically, many instances of realism in logic have been
internal. The exact nature of the grounds that underlie the internal realists
commitment to logical facts vary. It may be that the logician desires to see
certain intuitions satised or certain epistemic purposes fullled by the
logical theory. Why precisely these intuitions and purposes ought to be
satised by the theory is bound to be a matter of contention, but theres a
case to be made to the eect that they pervade logic and its philosophy.6
What I call intuitions here correspond to certain claims that seem
more certain, more epistemically salient or otherwise accessible to the
5
6
it will be asked where these ideas would be if there were no mind, and what would then become of
the real foundation of this certainty of eternal truths. This question brings us at last to the ultimate
foundation of truth, namely to that Supreme and Universal Mind who cannot fail to exist and whose
understanding is indeed the domain of eternal truths. . .That is where I nd the pattern for the ideas
and truths which are engraved in our souls. IV.xi.447 (my emphasis added).
IV.v.397. I wish to thank Chloe Armstrong for an informative discussion concerning this point.
The exact nature of the distinction between intuitions and purposes would benet from a closer
investigation, but this would go beyond the scope of the present chapter.
194
Sandra Lapointe
agent (though there might not be a fact of the matter as to whether they
really are). The logical realist may be convinced, for instance, that truth,
whatever it is, is immutable, in the sense that it cannot be changed or
destroyed and she wont regard the theory as adequate unless the immutability of truth is a consequence of it. Since she is also likely to hold the
belief that individual sentence- and thought-tokens do not persist indenitely (for they dont) and thus cannot be the fundamental bearers of truth
(and falsity), she might deem it necessary to introduce ontologically robust
abstract entities, precisely in order to satisfy this intuition. If that is the
case, then As (desire to satisfy this) intuition has explanatory value when it
comes to accounting for her commitment to proposition-like entities:
there is a denite sense in which A believes that there are propositions
because she believes that truth is immutable.
Quantifying over meanings may also serve certain more or less clandestine purposes within the theory. The logical realist may, for instance, be
guided by the fact that systematically including instances of (1), above, in a
semantic theory (surreptitiously) introduces a paraphrastic procedure
that can be used to clarify natural language sentences or make them
more exact.7 If ones motive, be it explicitly or not, in introducing the
semantic operator means and in quantifying over propositions are the
(stealthy) epistemic gains that come from translations of this type, ones
grounds to commit to propositions are subservient to the semantic theory
and the type of realism they embrace is internal.
Admittedly, in certain cases, it could be unclear whether ones grounds
are internal or external. Take the case in which As belief that there are
logical facts is the consequence of certain assumptions concerning
the relation between language and the world. A may believe that there
are objective logical facts because A believes (TM):
(TM) The truth of a claim implies its correspondence to something that
makes it true (or the existence of a truth-making relation), whatever this
turns out to be.
Assuming that some claims about logic are true, (TM) implies the existence of entities that full this truth-making role in logic. Commitment to
(TM) explains the commitment to logical facts. Indeed, (TM) epitomizes
7
See (Lepore and Ludwig 2006). They write: The assignment of entities to expressions, which was to
be the key to a theory of meaning, turns out to have been merely a way of matching object-language
expressions with metalanguage expressions thought of as used (in referring to their own meaning), so
that we are given an object-language expression and a matched metalanguage expression we
understand, in a context which ensures that they are synonymous (Lepore and Ludwig 2006: 31).
195
9
On this, see (Mulligan, Simons and Smith 1984: 289).
Cf. (Bolzano 1837, 21: 84).
Cf. (Bolzano 1837, 125: 7). That truth is immutable that is true is not a relativized predicate is
thus an intuition Bolzanos theory seeks to satisfy and one of the grounds that motivates his
commitment to logical realism.
196
Sandra Lapointe
197
As Bolzano sees it, the main reason for positing propositions is their
usefulness for certain theoretical purposes, in particular for the purpose
of reaching satisfactory denition of logical notions (understood broadly):
The usefulness of the distinction [between propositions in themselves and
thought propositions] manifests itself in tens of places and in the most
surprising way in that it allows the author to determine objectively a
number of concepts that had not been explained before or that were
explained incorrectly. For instance, the concept of experience, a priori,
possibility, necessity, contingency, probability, etc. (1839: 128)
198
Sandra Lapointe
But now the question arises whether someone who rejects the concept of
propositions in themselves and accepts only that [for instance] of thought
propositions could nonetheless admit of a connection amongst the latter
more or less like the one Bolzano describes as objective. And this, we think,
should be answered in the armative. (Bolzano 1841: 68)
On the face of it, the kind of semantic descriptivism Bolzano adopts seems
incompatible with the claim that someone who rejects the notion of
proposition could still admit his denitions of logical notions. Indeed such
statement contradicts what Phonsk, Bolzanos close collaborator, seems
to have assumed in the New Anti-Kant, namely that:
All will be lost if they cannot grant us this concept [of a proposition], if they
keep representing truths in terms of certain thoughts, appearances in the
mind of a thinking being . . . (Phonsk 1850: 5)
199
15
The same holds for what he describes as the objective connections between propositions,
including formal properties. More on this in the next section.
That proposition is a simple concept is something Bolzano suggests at (1837, 128: 8) when he
writes: From the mere fact that representations are the components of propositions we cannot infer
that the concept of a representation must be simpler than that of a proposition. On the contrary,
there is a lot to say for the idea that this mark which I use in 48 merely as an explanation of the
concept of a representation is the actual denition of the latter. At (1837, 48: 216) Bolzano had
written: Anything that can be part of a proposition in itself, without being itself a proposition,
I wish to call a [representation] in itself. This will be the quickest and easiest way of conveying my
meaning to those who have understood what I mean by a proposition in itself .
200
Sandra Lapointe
example, every one can grasp which concept is designated by the word point
on the basis of the following propositions: the point is the simple object in
space, it is the limit of the line without being part of the line, it is extended
neither lengthwise nor according to width, nor according to depth, etc. As
is well known, this is the means through which we all learn the rst
signications of our mother tongue. (1810, II, 8: 5455)
17
18
Cf. Bolzano (1837, 12: 51) where he explains (my emphasis): The clearest denitions say hardly
more than that we consider the form of propositions and ideas when we keep an eye only on what
they have in common with many others, that is, when we speak of entire species or genera of the
latter. . . . one calls a species or genus of proposition formal if in order to determine it one only needs to
specify certain parts that appear in these ideas or propositions while the rest of the parts which one calls the
stu or matter remain arbitrary.
Cf. (Bolzano 1837, 9: 42f ).
Bolzano often speaks of propositions containing variable representations and he does not always
revert to schemata to indicate variability. If [Caius] is taken to be variable in [Caius who has
mortality, has humanity], the latter can in principle be designated by the schematic expression
X who has humanity, has mortality; and [Caius is Caius] by A is A.
201
comes dening formal notions. Take for instance Bolzanos claims that there
are logically analytic or tautological propositions. Bolzano writes:
The following are some very general examples of analytic propositions
which are also true: A is A, An A which is a B is an A, An A which
is a B is a B, Every object is either B or non-B, etc. Propositions of the
rst kind, i.e., propositions cast in the form A is A or A has (the
attribute) a are commonly called identical or tautological propositions.
(Bolzano 1837, 148: 84)
and so on.
It is certainly not incongruous for Bolzano to claim that A is A the
schematic expression is logically analytic. Indeed, it would seem that one
need understand what it means for a proposition to belong to such a class
or to fall under such a form in order to understand how it itself can be
said to be analytic. The proposition: [Caius is Caius], for instance, falls
under the form A is A: A is A is a determinate connection of words or
signs through which the class to which [Caius is Caius] belongs can be
represented.20 To say that the individual proposition [Caius is Caius] is
19
20
In one of his numerous historical digressions, Bolzano notes that the Latin word forma . . . was in
fact used as equivalent to the word species, i.e. the word class (Bolzano 1837, 81: 391).
See (Bolzano 1837, 81: 393).
202
Sandra Lapointe
4. Bolzanos logic
Internal grounds to adhere to logical facts or in Bolzanos case to fully
edged semantic entities are typically certain desiderata or aims the
theory is meant to full. In Bolzanos case, one of the main purposes in
introducing propositions in themselves is to achieve precise and satisfactory denitions. By way of consequence, on Bolzanos own account
the success of the endeavour depends on whether his commitment to
propositions allows him to deliver a good theory of logic, or at least
one that is preferable to its rivals. To a large extent, Bolzano succeeds. It
is not only that the Theory of Science is furnished with rich and
remarkably well-articulated distinctions and theoretical innovations
but also that he set out to redene the very nature of a logical investigation in a way that is largely consistent with well-established contemporary endeavours.
As Bolzano sees it, at its core, the purpose of logic is to tell us what it
means for something to follow from something else, i.e. what it means for
an inference to be valid or for a claim to be the consequence of some other
claim(s). As an explanation of what it means for a truth to follow from
others, Bolzanos views on deducibility (Ableitbarkeit) are comparatively
close to the ones that have become standard following Tarski in the
twentieth century. Bolzano denes deducibility in the following terms:
203
To logicians and philosophers of logic today, the idea that the aim of logic
is to dene validity via the elaboration of a theory of logical consequence is
unremarkable. Pointing to the similarities (and dissimilarities) between
Bolzanos denition of deducibility and Tarskis denition of logical
consequence has become commonplace in the literature.22 This goes to
show that at least some of the desiderata and aims that underlie Bolzanos
logic rest on the kind of intuitions that have proven to be enduring. This
should be emphasized for at least two reasons. First, when he published the
Theory of Science in 1837, Bolzanos views on deducibility were perfectly
anachronistic. For one thing, by the end of the eighteenth century it had
become usual for philosophers to think of logic as invested in the study of
reason through an investigation of thought and to conceive of such an
investigation to involve the study of mind-dependent operations and
products. Though the methodologies underlying these investigations
varied widely contrast Lockes empirical approach in the Essay on Human
Understanding with Kants transcendental philosophy they largely contributed to either discredit formal logic as a discipline23 or, at best, to
convey the opinion that it could not be improved on.24 In this light,
Bolzanos eorts toward a new logic based on an objective doctrine of
21
22
23
24
For a more detailed discussion of Bolzanos theory of deducibility, see (Lapointe 2011: 7290).
See, for instance, (van Benthem 1985; George 1986; Siebel 2002; Lapointe 2011).
See (George 2003: 99s).
See Kants famous claim that logic is closed and complete (1781: Bviii).
204
Sandra Lapointe
inferences in themselves constitutes an important break from his immediate modern predecessors.
Second, while Bolzano reaches back to Aristotle, his approach to the
denition of validity also marks an important departure from Aristotle and
most of his (early) traditional scholastic commentators.25 Aristotle introduces the notion of a good deduction (i.e. syllogism) in the Prior
Analytics. He writes:
A deduction (syllogismos) is speech (logos) in which, certain things having
been supposed, something dierent from those supposed results of necessity
because of their being so. (Prior Analytics I.2, 24b1820)
Let us call this the intuitive Aristotelian notion of validity. Contemporary attempts at a denition of logical consequence one may think of
Tarski-type model-theoretic denitions in particular are generally understood to account for the intuitive Aristotelian notion of validity. The same
holds for Bolzanos. What makes Bolzanos account historically distinctive
is the assumption that a good denition of the intuitive Aristotelian notion
of validity needs the support of a semantic theory. In this, his denitional
strategy ought to be contrasted with that of much of the Aristotelian
tradition itself. Aristotle and his early medieval successors are mostly
known for their understanding of validity as epitomized in traditional
syllogistic theories. But traditional syllogistic denitions of validity are
not concerned with providing a semantic account of validity.26 The
standard and paradigmatic methodology behind traditional syllogistic
theories of valid inference, and the one that is best known, is two-pronged.
It rst consists in making a list of all possible forms of arguments (syllogisms) and then in identifying those forms whose instances eectively full
the intuitive Aristotelian denition of validity. In order to determine
whether a particular inference is valid, one is thus required to determine
whether it instantiates one of the forms identied as valid.
There are at least three problems, from Bolzanos perspective, with this
approach. First, traditional syllogistic denitions of validity suppose that
there is a nite (and implausibly small) number of possible forms of
inference. Bolzano is right. If we follow the teachings of the Schoolmen,
25
26
Here I am not concerned with comparing the Bolzanian and Aristotelian conception of the object of
logic (see Thom, this volume, for such a discussion) but their views on validity.
Here, I exclude from what I call syllogistic tradition the theories of consequentia that emerged in
the fourteenth century those we nd in Occam and Buridan, for instance. The latter were
attempts to generalize syllogistic and aimed at providing a new insight into the intuitive Aristotelian
notion based on semantic considerations. On this topic, see (Novaes 2012). Something similar holds
for Abelard. I am grateful to Julie Brumberg-Chaumont for this precision.
205
As Bolzano sees it, one need not suppose that the number of (valid) forms
of inferences is nite or that it is linked to a determinate syntax, for
instance that it can only be dened for inferences that have only two
categorical premises.28
Moreover, the three above problems concerning traditional syllogistic
treatments of validity are linked to a fourth more general one. There are
various ways of xing the extension of a concept, not all of which amount
to denition. The mere fact of knowing which inferential forms satisfy the
27
28
This is even more obvious when one reads the beginning of the second book of the Prior
Analytics, which was devoted to the relationship between premises and conclusion as regards
their truth-value.
See (Phonsk 1850: 115f ). Some passages of the Prior Analytics suggest that Aristotle was aware of
the problem. See for instance (Prior Analyics I, 32). But Aristotle himself did not provide a
systematic account of what it is for an inference that is not a syllogism to result of necessity.
Sandra Lapointe
206
5. Conclusion
In light of what precedes, Bolzanos internal realism is vindicated: Bolzanos
positing of propositions in themselves allows him to articulate a theory of
deducibility that could do what the syllogistic theories of his predecessors
could not: provide us with a general semantic theory of validity. Nonetheless,
as those acquainted with recent scholarship know, there are problems with
Bolzanian deducibility. (See, e.g. Siebel 2002.) For one, despite Bolzanos
claim to the contrary, his denition of deducibility fails to capture what is
usually taken to be the modal insight that underlies the intuitive Aristotelian
notion of validity, namely the idea that the conclusion of a good argument
results of necessity. Consequently, it overgenerates. Bolzanian deducibility
systematically includes inferences that are merely materially valid. I say
systematically because if it is the case that all As are Bs, then X is B is
invariably deducible from X is A. For instance, on Bolzanos account:
X is no taller than three metres
is deducible from:
X is a man
207
with respect to X. This failure may strike as the result of a misunderstanding (coupled with contentious exegetical choices). Bolzano interprets the
relevant passage of the Prior Analytics in the following terms:
Since there can be no doubt that Aristotle assumed that the relation of
deducibility can also hold between false propositions, the results of necessity
can hardly be interpreted in any other way than this: that the conclusion
becomes true whenever the premises are true. Now it is obvious that we
cannot say of one and the same collection of propositions that one of them
becomes true whenever the others are true, unless we envisage some of their
parts as variable. For propositions none of whose parts change are not
sometimes true and sometimes false; they are always one or the other.
Hence when it was said of certain propositions that one of them becomes
true as soon as the others do, the actual reference was not to these propositions themselves, but to a relation which holds between the innitely many
propositions which can be generated from them, if certain of their representations are replaced by arbitrarily chosen other representations. The
desired formulation was this: as soon as the exchange of certain representations makes the premises true, the conclusion must also become true. (1837,
155: 129)
208
Sandra Lapointe
part iii
Specic Issues
chapter 12
Revising logic
Graham Priest
1. Whats at issue
Much ink has been spilled over the last few decades in disputes between
advocates of classical logic that is, the logic invented by Frege and
Russell, and polished by Hilbert and others and advocates of nonclassical logics such as intuitionist and paraconsistent logics. One move
that is commonly made in such debates is that logic cannot be revised.
When the move is made, it is typically by defenders of classical logic.
Possession, for them, is ten tenths of the law.
The point of this chapter is not to enter into substantive debates about
which logic is correct though relevant methodological issues will transpire in due course. The point is to examine the question of whether logic
can be revised.1 (And let me make it clear at the start that I am talking
about deductive logic. I think that matters concerning non-deductive logic
are much the same, but that is an issue for another occasion.) Three
questions, then, will concern us:
Can logic be revised?
If so, can this be done rationally?
If so, how is this done?
Unfortunately, debates about the answers to these questions are often
vitiated by a failure to observe that the word logic is ambiguous. Only
confusion results from running the senses of the word together. Once the
appropriate disambiguations are made, some of the answers to our questions are obvious; some are not. It pays, for a start, to be clear about which
are which.
1
Thanks go to Hartry Field for many enjoyable and illuminating discussions on the matter. We taught
a course on the topic together in New York in the Fall of 2012. Many of my views were claried in the
process.
211
Graham Priest
212
Logica docens
Logica utens
Logica ens
What each of these is will require further discussion and clarication. But
as a rst cut, we may characterise them as follows.
Logica docens (the logic that is taught) is what logicians claim about
logic. It is what one nds in logic texts used for teaching. Logica utens (the
logic which is used) is how people actually reason. The rst two phrases are
familiar from medieval logic. The third, logica ens (logic itself ) is not. (I
have had to make the phrase up.) This is what is actually valid: what really
follows from what.
Of course, there are important connections between these senses of
logic, as we will see in due course. But the three are distinct, both
intensionally and extensionally, as again we will see.
I will proceed by discussing each of these senses of logic, and asking
each of our three target questions about them. We have, then a nine-part
investigation.
2. Logica docens
2.1
Can it be revised?
Let us start with logica docens. The discussion of this will form the longest
part of the essay, since it informs the discussion with respect to the other
two parts. The question of whether the logica docens can be revised is,
however, the easiest to deal with. It can be revised because it has been
revised.
The history of logic in the West has three great periods.2 The rst was in
Ancient Greece, when logic was founded by Aristotle, the Megarians, and
the Stoics. The second was in the new universities of Medieval Europe,
such as Oxford and Paris, where Ockham, Scotus, and Buridan ourished.
The third starts in the late nineteenth century, with the rise of mathematical logic, and shows no signs yet of ending. Between these three periods
were periods of, at best, mainly maintaining what was known, and at worst
forgetting it. Much of Greek logic was forgotten in Europe, but fortunately
2
The history of logic in the East has its own story to tell, but that will not be our concern here.
Revising logic
213
preserved by the great Arabic scholars such as Al Farabi and Ibn Rushd.
Most of medieval logic was simply wiped out by the rise of the Enlightenment, and the consequent obliteration of Scholasticism. It is only in
the twentieth century that we have started to rediscover what was lost
in this period.
At any rate, one needs only a passing acquaintance with logic texts in the
history of Western logic to see that the logica docens was quite dierent in
the various periods. The dierences between the contents of Aristotles
Analytics, Paul of Venices Logica Magna, the Port Royale Logic, or the
Art of Thinking, Kants Jsche Logik, and Hilbert and Ackermanns
Principles of Mathematial Logic would strike even the most casual observer.
It is sometimes suggested that, periods of oblivion aside, the development of logic was cumulative. That is: something once accepted, was never
rejected. Like the corresponding view in science, this is just plain false. Let
me give a couple of examples.
One of the syllogisms that was, according to Aristotle, valid, was given
the name Darapti by the Medievals, and is as follows:
All As are Bs
All As are C s
Some Bs are C s
As anyone who has taken a rst course on modern rst-order logic will
know, this inference is now taken to be invalid.3
For another example: Classical logic is not paraconsistent; that is, the
following inference (Explosion) is valid for all A and B: A, :A B. It is
frequently assumed that this has always been taken to be valid. It has not.
Aristotle was quite clear that, in syllogisms, contradictions may or may not
entail a conclusion. Thus, consider the syllogism:
No As are Bs
Some Bs are As
All As are As
Graham Priest
214
Logica docens, then, has been revised, and not in a cumulative fashion. The
next question is whether revision can be rational.
Arguably, not all the changes in the history of logic were rational (or
perhaps better: occurred for reasons that were internal to the subject).
Thus, logic fell into oblivion in the early Middle Ages in Western Christendom because the institutions for the transmission of philosophical texts
collapsed. And later Medieval logic was written o on the coat-tails of the
rejection of Scholasticism during the Enlightenment.5
However, many changes that did arise were the result of novel ideas,
reason, argument, debate. These are the things of which rational change
are made. This should be pretty obvious with respect to the only change
that most logicians are now familiar with: the rise of mathematical logic.
In the mid nineteenth century, text book logic (traditional logic) was
a highly degenerate form of medieval logic: essentially, Aristotelian
syllogistic with a few medieval accretions, such as immediate inferences like modus ponens. But this was a period in which high standards
of rigour in mathematics were developing. Mathematicians such as
Weierstrass and Dedekind were setting the theory of numbers on a
rm footing. And when it came to examining the reasoning required in
the process, notably by Frege, it became clear that traditional logic did
not seem to be up to the job. Hence Frege invented a logic that did
much better: classical logic. The extra power of this logic made it much
preferable rationally; and within 50 years it had replaced traditional
logic as the received logica docens.
I will come back to this in the next section. For the present, let us move
on to our third question.
4
5
For references and further discussion on all these matters, see (Priest 2007: sec. 2).
Actually, my knowledge of the history of these periods is pretty sketchy; but I think that these claims
are essentially correct.
Revising logic
2.3
215
Logic as theory
216
Graham Priest
the case that there is an innitude of numbers, does it follow that there
is an innitude of numbers? Classical logic says yes; intuitionist logic
says no.6
In other words, a pure logic with its canonical application is a theory of
the validity of ordinary arguments: what follows (deductively) from what.
How to frame such a theory is not at all obvious. Many approaches have
been proposed and explored. One approach is to take validity to be
constituted modally, by necessary truth-preservation (suitably understood). Another is to dene validity in terms of probabilistic constraints
on rational belief. Perhaps the most common approach at present is to take
a valid inference to be one which obtains in virtue of the meanings of (at
least some of ) the words employed in it. This strategy has itself two ways
in which it can be implemented. One takes these meanings to be spelled
out in terms of truth conditions, giving us a model-theoretic account of
validity; the other takes these meanings to be spelled out in inferential
terms, giving us a proof-theoretic account of validity.
It is clear that a theory of validity is no small undertaking. It requires an
account of many other notions, such as negation and quantication.
Moreover, depending on the theory in question, it will require an articulation of other important notions, such as truth, meaning, probability. No
wonder it is hard to come up with plausible such theories!
At any rate, it is crucial to distinguish between logic as a theory (logic
docens, with its canonical application), and what it is a theory of (logica
ens). In the same way we must clearly distinguish between dynamics as a
theory (e.g., Newtonian dynamics) and dynamics as what this is a theory of
(e.g., the dynamics of the Earth). This is enough to dispose of the Quinean
charge (still all too frequently heard): change of logic means change of
subject.7 If one changes ones theory of dynamics, one can still be
reasoning about the same thing: the way the Earth moves.
2.4 What is the mechanism of rational revision?
With this substantial prolegomenon over, we can now address the question
of the mechanism of rational change of logica docens. As we have seen, a
pure logic with its canonical appication is essentially a theory of validity
and its multitude of cognate notions. How do we determine which theory
is better? By the standard criteria of rational theory choice.
6
Revising logic
217
Matters are spelled out in detail on (Priest 2006a: ch. 8), and especially, (Priest to appear).
218
Graham Priest
A word of warning: it would be wrong to infer that classical logic did not
have its problems. It had its own ad hoc hypotheses (to deal with the
material conditional, for example). It had areas where it seemed to perform
badly (for example, in dealing with vague language). And why should one
expect a logic that arose from the analysis of mathematical reasoning to be
applicable to all areas of reasoning? It was just these things which left the
door open for the development of non-classical logics. That, however, is
also a topic for another occasion.9 We have seen, at least in outline, what
the mechanism of rational change for a logica docens is.
3. Logica utens
3.1
What is this?
So much for the discussion of logica docens. Let us now turn to the next
disambiguation. Before we address our three questions, however, there is
an important preliminary issue to be addressed. What exactly is logica
utens?
I said that it is the way that people actually reason. This may make it
sound like a matter of descriptive cognitive psychology; but it is not this,
for the simple reason that we know that people often reason invalidly. Set
aside slips due to tiredness, inebriation, or whatever. We know that people
actually reason wrongly in systematic ways.10
To take just one very well established example: the Wason Card Test.
There is a pack of cards. Each card has a letter on one side and a positive
integer on the other. Four cards are laid out on the table so that a subject
can see the following:
A
The subject is then given the following conditional concerning the displayed situation: If there is an A on one side of the card, there is an even
number on the other. They are then asked which cards should be turned
over (and only those) to check this hypothesis. The correct answer is:
A and 3. But a majority of people (even those who have done a rst course
in logic!) tend to give one of the wrong answers: A, or A and 4.
Exactly what is going on here has occasioned an enormous literature,
which we do not need to go into. The experiment, and ones like it, show
that people can reason wrongly systematically. Of course, people are able
9
10
Revising logic
219
Graham Priest
220
4. Logica ens
4.1
Can it be revised?
It is quite compatible with this point that sometimes truth may be internal to a practice for
example, within classical and intuitionist pure mathematics. See (Priest 2013).
Revising logic
221
But the object of a social scientic theory may not change when the
theory does, for all that. (Many basic laws of psychology are, presumably,
hard-wired in us by evolution.) Whether the truth of validity-claims can
change will depend on what, exactly, constitutes validity. Let me illustrate.
Suppose that one held a divine command theory of validity: something is
valid just if God says so. Then, God being constant and immutable, what
is valid could not change. On the other hand, suppose that one were to
subscribe to the dentist endorsement view of validity: what is valid is
what 90 per cent of dentists endorse. Clearly, that can change.
These theories are, of course, rather silly. But they make the point: the
truth of validity-claims may or may not change, depending on what
validity actually is. An adequate answer to our question would therefore
require us to settle the issue of what validity is, that is, to determine the
best theory of validity. That is far too big an issue to take on here.12
I shall restrict myself in what follows to some remarks concerning the
model-theoretic and proof-theoretic accounts of validity. According to
the rst, an inference is valid i every model of the premises is a model
of the conclusion. But a model is a structured set, that is, an abstract
object, the premises form a set, another abstract object, and the premises
and conclusions themselves are normally taken to be sentence types, also
abstract objects. According to the second, an inference is valid if there is a
proof structure (sequence or tree), at every point of which there is a
sentence related to the others in certain ways. But a proof structure is an
abstract object, as, again, are the sentences.
In other words, validity, on these accounts, is a realtionship between
abstract objects. As usual, we may take these all to be sets. If this is so,
then, at least if one is a standard platonist about these things, the truth of
claims about validity cannot change.13 Claims about mathematical objects
are not signicantly tensed: if ever true true, always true.
4.2 Can meanings change?
That is not an end of the matter, though. The propositions about validity
may not change their truth values. But we express these in language. It
might be held that the words involved may change their meanings and,
moreover, do this in such a way that the truth values of the sentences
12
13
I have said what I think about the matter in (Priest 2006a: ch. 11).
Certain kinds of constructivists may, of course, hold that the truth about numbers and other
mathematical entities may change for example, as the result of our acquiring new proofs.
222
Graham Priest
involved may change. If this is the case, then the sentences expressing
validity claims can change their truth values.
Can meanings change in such a way as to aect truth value? Of course
they can. When Nietzsche wrote The Gay Science, it was a reference to the
art of being a troubadour. Nowadays, one could hear it only as concerning
a study of a certain sexual preference. In modern parlance, Nietzsche did
not write a book about (the) gay science.
Now, could there be such change of meaning in the case we are
concerned with? Arguably, yes. In both a proof-theoretic and a modeltheoretic account of validity, part of the machinery is taken as giving an
account of meanings notably, of the logical connectives (introduction or
elimination rules, truth conditions). If we change our theory, then our
understanding of these meanings will change. This does not mean that the
meanings of the vernacular words corresponding to their formal counterparts changes. You can change your view about the meaning of a word,
without the word changing its meaning. However, if one revises ones
theory, and then brings ones practice into line with it, in the way which
we noted may happen, then the usage of the relevant words is liable to
change. So, then, will their meanings assuming that meaning supervenes
on use (and some version of this view must surely be right). So the
sentences used to express the validity claims, and maybe even which
propositions the language is able to express, can change.14
It might be thought that this makes such a change a somewhat trivial
matter. Suppose we have some logical constant, c, which has dierent truth
or proof conditions according to two dierent theories. Can we not just
use two words, c1 and c2, which correspond to these two dierent senses?
Perhaps we can sometimes; but certainly not always: for meanings can
interact. Let me illustrate. Suppose that our logic is intuitionist. Then
Peirces law, ((A ! B) ! A) ! A, is not logically valid. But suppose that
we now decide to add a new negation sign to the language, which behaves
as does classical negation. Then Peirces law becomes provable. The
extension is not conservative. Another case: given many relevant logics,
the rules for classical negation can be added conservatively, as can the
natural introduction and elimination rules for a truth predicate. But the
14
A pertinent question at this point is whether the meaning of follows deductively from or however
this is expressed can itself change. Perhaps it can; and if it does, this adds a whole new dimension
of complexity to our investigation. However, I see no evidence that the meaning of the phrase (as
opposed to our theories of what follows from what) has changed over the course of Western
philosophy. So I ignore this extra complexity here.
Revising logic
223
5. Conclusion
Let me end by summarising the main conclusions we have reached, and
making a nal observation.
A logica docens may be revised rationally, and this happens by the
standard mechanism of rational theory choice. A logica utens may be
changed by bringing it into line with a logica docens; and if the docens is
chosen rationally, so is the utens. The answer to the question of whether or
not the logica ens may change depends on ones best answer to the question
of what validity is. However, under the model- or proof-theoretic accounts
of validity, the answer appears to be: no. This does not mean, however,
that the sentences used to express these facts may not change. And a
rational change of logica utens may occasion such a change.
Now the observation. The rational logica utens depends on the rational
logica docens. The true logica docens depends on the facts of validity. And
assuming a model- or proof-theoretic account of meaning, the language
available to express these may depend on the logica utens. It is clear that we
have a circle. If one were a foundationalist of some kind, one might see this
circle as vicious: there is no privileged point where one can ground the
entire enterprise, and from which one can build up everything else.
However, I take it that all knowledge, about logic, as much as anything
else, is situated.16 We are not, and could never be, tabulae rasae. We can
start only from where we are. Rational revision of all kinds then has to
proceed by an incremental and possibly (Hegel notwithstanding) neverending process.
15
16
chapter 13
1. Introduction
There are a variety of reasons why we would want a paraconsistent
account of logic, that is, an account of logic where an inconsistent
theory does not have every sentence as a consequence. One relatively
standard motivation is epistemic in nature.1 There is a high probability
that we will come to hold inconsistent beliefs or inconsistent theories
and we would like some account of how to reason from an inconsistent
theory without everything crashing. Another motivation, rooted in the
philosophy of logic or language, is that we want a proper account of
entailment or relevant implication, where there is a natural sense in
which inconsistent claims do not (relevantly) entail arbitrary propositions where not every claim follows from arbitrary inconsistency.2
A third motivation, the one which will occupy our attention here, is
metaphysical or semantic. One might, for various reasons, endorse that
there are true contradictions, or as they are sometimes called, truthvalue gluts true sentences of the form ^ : , claims which are both
true and false. We shall say that a glut theorist is one who endorses
glutty theories theories that are negation-inconsistent with the full
knowledge that they are glutty.
There are dierent kinds of metaphysical commitments that can lead
one to be a glut theorist. One route towards glut theory arises from
views about particular predicates of a language or the properties that
those predicates express. Along these lines, a familiar route towards glut
theory holds that certain predicates like is true, is a member of , or
exemplies are essentially inconsistent: they cannot be (properly)
1
For work in this tradition, see Rescher and Manor (1970); Schotch et al. (2009); Schotch and
Jennings (1980).
For work in this tradition, see Anderson and Belnap (1975); Anderson et al. (1992); Dunn and Restall
(2002); Mares (2004); Slaney (2004).
224
225
LP is the gap-free extension of FDE, the logic of tautological entailments; it is the dual of the familiar
glut-free extension of FDE called strong Kleene or K3. See Dunn (1966, 1976), Anderson and
Belnap (1975), and Anderson et al. (1992).
226
for any predicate P, d(P) = P , P , where P [ P = D.
The only dierence from the standard LP treatment appears here, in the
form of a restriction that captures the distinction between the antinomic
(i.e., essentially glutty) and essentially classical predicates:
4
For purposes of accommodating glutty theories, the propositional logic LP was rst advanced in
Asenjo (1966) under the name calculus of antinomies; it was later advanced, for the same purpose,
under the name logic of paradox by Priest (1979), who also gave the rst-order logic under the same
name (viz., LP). What we are calling LA is the rst-order (conditional-free) logic advanced by
Asenjo and Tamburino (1975), which was intended by them to be a rst-order extension of Asenjos
basic propositional logic. Due to what we call the LA Predicate Restriction (see page 227) LA isnt a
simple rst-order extension of Asenjos propositional LP as will be apparent below (see 4).
For simplicity, we focus entirely on unary predicates. Both LA and LP cover predicates of any arity,
but focusing only on the unary case suces for our purposes.
227
if P is in , then the intersection P \ P must be empty;
if P is in , then the intersection P \ P must be non-empty.
As above, the Ais are the essentially classical predicates, while the Bis are
those which are antinomic.6
||v is the semantic value of a sentence with respect to a variable
assignment v, which is dened in the standard recursive fashion. (We leave
the relevant interpretation implicit, as it will always be obvious.) For atomics:
8
0 if I t 2
= P and I t 2 P
>
<
= P
jPtjv 1 if I t 2 P and I t 2
>
1
:
otherwise:
2
4. Contrast: LA and LP
We begin with formal contrast. While both logics are paraconsistent (just
let jjv 21 and jjv = 0, for at least some formulae and ), there are
some obvious but noteworthy formal dierences between the logics LA
6
The presentation in Asenjo and Tamburino (1975) is rather dierent; but we present their account in
a way that aords clear comparison with LP.
228
Since any predicate Bi in must have at least some object in the intersection of its extension and anti-extension, it follows that something is in its
extension.
4.1 Metaphysics, formal and material consequence
How are we to understand the logical dierences between LA and LP? For
a rst pass, they might be naturally understood as arising from dierent
notions of consequence: namely, material and formal consequence. The
distinction may not be perfectly precise, but it is familiar enough.7
Material consequence relies on the matter or content of claims, while
formal consequence abstracts away from such content. Example: there is
no possibility in which Max is a cat is true but Max is an animal is not
true; the former entails the latter if we hold the meaning the matter, the
content of the actual claims xed. But the given entailment fails if we
abstract away from matter (content), and concentrate just on the standard
rst-order form: Cm does not entail Am.
The notion of formal consequence delivers conclusions based on logical
form alone. Material consequence essentially requires use of the content of
the claims or the meaning of things like predicates that appear in them.
7
See Read (1994, Ch. 2) wherein Read provides a defense of material consequence as logical
consequence, and also for further references.
229
230
predicates are like that: potentially glutty, one and all, but none antinomic none essentially glutty.
There is another metaphysical route to LP. We might not start with any
commitments about the nature of any predicates, their meaning, the
properties they express, and whether or not they are essentially inconsistent. One might start with the commitment that ones theory is both true
and inconsistent, while remaining agnostic about where to locate the
origins of the inconsistency. There is no reason to think that this position
excludes a material approach to consequence. Its just that such a view
lacks any particular metaphysical commitments that would motivate a
restriction on predicates like the LA predicate restriction.
Of course, from the material point of view, LA and LP far from exhaust
the possibilities. So far weve mentioned fairly strong, all-or-nothing
approaches. On a material approach to consequence, the proponent of
LA is committed to all predicates being essentially classical or glutty, while
the proponent of LP is committed to all predicates being potentially
classical or glutty. Mixed approaches are available. These are achieved by
adding obvious combinations to the LA predicate restriction for
example, some antinomic, some essentially classical, some neither, etc.
We leave these to the reader for exploration.
We turn (briey) to an issue peculiar to the logics under discussion:
detachment or modus ponens.
5. Detachment
A salient problem for LP is that there is no detachable (no modus-ponenssatisfying) conditional denable in the logic (Beall et al. (2013)); and thus,
historically, LP has been viewed as unacceptably weak for just that reason.
A lesson one might try to draw from the above observations is that LP can
be improved by shifting focus to the material notion of consequence. But
this is not quite right. Though one fragment of LA diers from LP in that
it satises detachment, LA is like LP in that detachment doesnt hold
generally: arguments from and to have counterexamples.
On this score, Asenjo and Tamburino (1975), along with Priest (1979,
2006b), have a solution in mind. The remedy is to add logical resources to
the base framework to overcome such non-detachment.8 But the remedy
8
Until very recently, Beall (2013), all LP-based glut theorists focused their eorts on the given task:
adding logical resources to the base LP framework to overcome its non-detachment. Whether this is
the appropriate response to the non-detachment of LP is something we leave open here.
231
n1 o
,1
if jjv 0 and jjv 2
2 o
n
1
1
if jjv and jjv 2
,1
2
2
otherwise
The trouble, however, comes from Currys paradox. Focusing on the settheoretic version (though the truth-theoretic version is the same), Meyer
et al. (1979) showed that, assuming standard structural rules (which are in
place in LP and LA! and many other logics under discussion), if a
conditional detaches and also satises absorption in the form
! ! !
then the given conditional is not suitable for underwriting nave foundational principles. In particular, in the set-theory case, consider the set
c fx : x 2 x ! g
But, now, since the AsenjoTamburino arrow satises the given absorption rule, we quickly get
c2c!
9
232
6. Closing remarks
Philosophy, over the last decade, has seen increasing interest in paraconsistent approaches to familiar paradox. One of the most popular
approaches is also one of the best known: namely, the LP-based approach
championed by Priest. Our aim in this chapter has been twofold: namely,
to highlight an important predecessor of LP, namely, the LA-based
approach championed rst by Asenjo and Tamburino, and to highlight
the salient dierences in the logics. Weve argued that the dierences in
logic reect a dierence in both background philosophy of logic and
background metaphysics. LA is motivated by a material approach to logical
consequence combined with a metaphysical position involving antinomic
predicates, while LP is compatible with both a formal and material
approach to consequence and can be combined with a large host of
metaphysical commitments (including few such commitments at all).12
10
11
12
We note that Asenjo himself noticed this, though he left the above details implicit. We have not
belabored the details here, but it is important to have the problem explicitly sketched.
We note, however, that Beall has recently rejected the program of nding detachable conditionals
for LP, and instead defends the viability of a fully non-detachable approach (Beall (2013)), but we
leave this for other discussion.
We note that Priests ultimate rejection of LP in favor of his non-monotonic LPm (elsewhere called
MiLP) reects a move back in the direction of the original AsenjoTamburino approach, where
one has restricted detachment and the like, though the latter logic (viz., LA) is monotonic. We
leave further comparison for future debate. For some background discussion, see Priest (2006,
Ch. 16) and Beall (2012) for discussion.
chapter 14
For a recent discussion on this topic, see Sher (2011), who examines the idea that logic is grounded
either in the mind or in the world, and defends that it is grounded in both hence logic has a dual
nature. See also the opening chapter of this volume.
See Chateaubriand Filho (2001, 2005) for a version of the metaphysical interpretation of logical truth
partly similar to mine.
Models are to be interpreted in a wide sense: they may for instance be interpretations, possible
worlds, or valuations. We will return to this ambiguity concerning model below.
I should mention that I will omit discussion of Carnap and Quine on logical truth, as their debate is
not directly relevant for my purposes. However, see Shapiro (2000) for an interesting discussion of
Quine on logical truth.
233
234
Tuomas E. Tahko
235
between sentences and the world.5 Yet, Tarskis (1944: 342343) initial
considerations on the meaning of the term true explicitly take into
account an Aristotelian conception of truth, where correspondence with
the world is central. Davidson (1973: 70) as well seems to have some
sympathy for the idea that an absolute theory of truth is, in some sense,
a correspondence theory of truth, although he insists that the entities that
would act as truthmakers here are nothing like facts or states of aairs, but
sequences (which make true open sentences).
I will not aim to settle the status of the correspondence theory here, but
it will be necessary to discuss it in some more detail. I suggest adopting an
understanding of the correspondence relation which is neutral in terms of
our theory of truth. It is this type of weak correspondence intuition that
I believe central to the metaphysical interpretation of logical truth. But it
should be stressed that the correspondence intuition itself is not necessarily
expressive of realism (Daly 2005: 9697). For instance, Chris Dalys
suggested denition of the intuition is simply that a proposition is true if
and only if things are as the proposition says they are. Daly explains the
neutrality of (his version of ) the correspondence intuition as follows:6
Consider the coherence theorist. He may consistently say If <p> is true, it
has a truthmaker. <p> corresponds to a state of aairs, namely the state of
aairs which consists of a relation of coherence holding between <p> and
the other members of a maximal set of propositions. Consider the pragmatist. He may consistently say, If <p> is true, it has a truthmaker. <p>
corresponds to a state of aairs, namely the state of aairs of <p>s having
the property of being useful to believe. It is controversial whether there
exist states of aairs. Let that pass. My point here is that the coherence
theory and the pragmatic theory are each compatible with the admission of
states of aairs. Furthermore, each of these theories is compatible with the
admission of states of aairs standing in a correspondence relation to truths.
(Daly 2005: 97)
Furthermore, the idea that the T-schema or the correspondence theory are somehow expressive of
realism has been forcefully disputed. See for instance Morris (2005) for a case against the connection
between realism and correspondence; in fact Morris argues that correspondence theorists should be
idealists. See also Gmez-Torrente (2009) for a discussion about Tarskis ideas on logical
consequence as well as on Etchemendys critique of Tarskis model-theoretic account.
The angled brackets describe a proposition, following Horwich (1998).
236
Tuomas E. Tahko
Crivelli also denes a stricter sense of correspondence, which can be found in Aristotle. But
sometimes Aristotles view on truth is also considered as a precursor to deationism about truth,
so we shouldnt put too much weight on the historical case. For a more historically inclined
discussion, see Paul Thoms chapter in this volume.
Admittedly, once we explicate isomorphism, reference to propositions, states of aairs or something
of the sort could easily re-emerge. This shouldnt worry us too much, because it is likely that we want
a structured mapping from something to reality. The reason to opt for isomorphism here is merely to
keep the door open for ones preferred (structured) ontology.
237
Note that the question concerning which model is right is not, strictly speaking, a question for the
logician. For instance, as Burgess (1990: 82) notes, it is the metaphysicians task to determine the
correct modal logic, as this depends on our understanding of (metaphysical) modality. In contrast,
the question about the right sense of logical validity remains in the realm of logic.
238
Tuomas E. Tahko
It has been suggested to me (by Penny Rush) that relative truth may be problematic because of its
underlying metaphysical commitment to relativism, rather than not being up to the job of giving a
metaphysical interpretation of logical truth at all. This may indeed be the case. I have attempted to
preserve ontological neutrality while at the same time making it clear that I am presently only
interested in putting forward a realist interpretation of logical truth. But I will set this issue aside for
now, whether or not it is possible to combine relative truth and realism.
239
This is, of course, somewhat controversial, but as we will see, there are
reasons to think that only metaphysical modality is tting for the task. In
any case, more needs to be said about how the space of metaphysical
possibilities is restricted. We will return to this in the next section.
We are now in the position to dene a provisional sense of logical truth
which I propose to call metaphysical:
(ML) A sentence is logically true if and only if it is true in every genuinely
possible conguration of the world.
The above formulation diers from Aristotles only with regard to the
qualication regarding the same domain here the domain is the set of
genuinely possible congurations of the world. How do we know whether
LNC is true in this sense? I have previously argued (Tahko 2009) that we
do have a good case for the truth of LNC in the metaphysical sense the
primary opponent here is Graham Priest (e.g., 2005, 2006b).11 I will not
11
See also Berto (2008) for an attempt to formulate a (metaphysical) version of LNC which even the
dialetheist must accept. Bertos idea, to which I am sympathetic, is that LNC may be understood as
240
Tuomas E. Tahko
repeat my arguments here, but it may be noted that this is not strictly a
question for logic. For instance, Priests most celebrated arguments in favor
of true contradictions (in the metaphysical sense) concern the nature of
change and specically motion, the paradoxical nature of which is supposedly demonstrated by Zenos well-known paradoxes. Although these
paradoxes can quite easily be tackled by mathematical means, the relevant
question is whether change indeed is paraconsistent.12 The answer to this
question requires both metaphysical and empirical inquiry. I will return to
this point briey below, but rst I wish to say something about the
methodology of logical-cum-metaphysical inquiry.
In terms of ML, demonstrating the falsity of LNC would rst require a
genuinely possible conguration of the world where LNC fails. That is, it
is not enough that we have a model where LNC is not true, such as
paraconsistent logic, but we would also need to have some good reasons to
think that the world could have been arranged in such a way that the
implications of the metaphysical interpretation of LNC do not follow. This
point deserves to be emphasized, for it would be much easier to show that
a paraconsistent model can be useful in modelling certain phenomena, or
interpreted in such a way that it is compatible with all the empirical data.
But what is required here is that LNC, fully interpreted in the metaphysical
sense, can be shown to fail.
Note that we may also ask whether LNC is necessary, i.e., are there any
possible worlds in which LNC does not hold even if we did have a good
case for its truth in the actual world? In fact, this is the question we should
begin with, since if LNC is necessary, then it could not fail in the actual
world either. However, it is not clear how we could settle this question
conclusively, given that we are dealing with the metaphysical interpretation of LNC. Moreover, I do think that there could (in an epistemic
sense) be possible worlds in which LNC fails, and hence I take the debate
about LNC seriously. Yet, I am uncertain about whether such a paraconsistent possible world is in fact a genuinely possible conguration, as I will
go on to explain.13 In any case, if a possible world in which LNC is not true
12
13
a principle regarding structured exclusion relations (between properties, states of aairs, etc.), and
the world is determinate insofar as it conforms to this principle.
For discussion regarding Zenos paradoxes, see for instance Sainsbury (2009: Ch. 1).
It is worth pointing out here that in my proposed construal, the distinction between absolute truth
and truth in a model is not quite so striking for dialetheists. The idea, which I owe to Francesco
Berto, is that the world cannot be a model, because it contains everything, and theres no domain of
everything, on pain of Cantors paradox. The result is that something can be a logical truth in the
sense of being true in all models, without being true in the absolute sense, for the world is not a
model. My proposed treatment of this issue proceeds by understanding absolute truth in terms of
241
14
metaphysical modality, but the dialetheist could, in principle, endorse paraconsistent set theory and
posit that absolute truth is just truth in the world-model the model whose domain is the world.
I should add that cashing out these constraints is, I think, a much more complicated aair than the
traditional KripkePutnam approach to metaphysical a posteriori necessities suggests.
242
Tuomas E. Tahko
243
be found.15 This is what Priest has attempted to show with the case of
change and Zenos paradoxes, but I remain unconvinced. As I have argued
(Tahko 2009), Priests examples can all be accounted for in terms of
semantic rather than metaphysical dialetheism a distinction developed
by Edwin Mares (2004). The idea is that there may be indeterminacy in
semantics, but this does not imply that there is indeterminacy in the world.
Only the latter type of indeterminacy would corroborate the existence of a
genuinely possible paraconsistent conguration of the world. Since I have
not seen a convincing case to the eect that such a conguration is
genuinely possible, I take it that LNC is a good candidate for a metaphysically necessary principle. If I am right, this means that a paraconsistent
possible world could not have turned out to accurately represent the actual
world. The fact that there are paraconsistent models has no direct bearing
on this question. I do not claim to have settled the status of LNC once and
for all, but I think that a strong empirical case for the truth of LNC can be
made, on the basis of the necessary constraints for the forming of a stable
macrophysical world, i.e., the emergence of stable macrophysical objects.
I have developed the preceding line of thought before with regard to the
Pauli Exclusion Principle (PEP) (Tahko 2012), and electric charge (Tahko
2009). For instance, as PEP states, it is impossible for two electrons (or
other fermions) in a closed system to occupy the same quantum state at the
same time. This is an important constraint, as it is responsible for keeping
atoms from collapsing. It is sometimes said that PEP is responsible for the
space-occupying behavior of matter electrons must occupy successively
higher orbitals to prevent a shared quantum state, hence not all electrons
can collapse to the lowest orbital. Here we have a principle which captures
a crucial constraint for any genuinely possible conguration of the world
that contains macroscopic objects. Whether or not there are genuinely
possible congurations that do not conform to PEP is an open question,
but it seems unlikely that such a conguration could include stable
macroscopic objects.
Consider the form of PEP: it states that two objects of a certain kind
cannot have the same property (quantum state) in the same respect (in a
closed system) at the same time. Compare this with Aristotles formulation
of LNC: the same attribute cannot at the same time belong and not
belong to the same subject in the same respect (Metaphysics 1005b1920).
LNC is of course a much more general criterion than PEP it concerns
15
I have in mind concrete objects in the rst place; see Estrada-Gonzlez (2013) for a case to the eect
that there are abstracta which violate LNC in this sense.
244
Tuomas E. Tahko
one thing rather than things of a certain kind but its underlying role is
evident: if any fermion were able to both be and not be in a certain
quantum state at the same time, then PEP would be violated and macroscopic objects would collapse. If LNC is needed to undergird PEP, then we
have a strong case in favor of the metaphysical interpretation of LNC in
worlds that contain macrophysical objects, given the necessity of PEP for
the forming of macrophysical objects. This is of course not sucient to
establish the metaphysical necessity of either principle, but it is an interesting result in its own regard.
If this is indeed what pluralism about logical truth amounts to, then it
appears that anyone who accepts multiple classes of cases is a pluralist
about logical truth. But what does being true in a case mean? On the face
of it, one might think that it means exactly the same as being true in a
model, that is, we are talking about a type of relative truth familiar from
Davidson. This would imply that anyone who accepts multiple classes of
models will also be a pluralist about logical truth. Pluralism about logical
truth would then mean only that there are multiple models, and we can
talk about logical truth separately in each one of these models. But this
would be a rather uninteresting sense of logical pluralism, at least from the
point of view of the metaphysical interpretation of logical truth. However,
as Hartry Field has recently pointed out, this cannot be what Beall and
245
Restall have in mind. Moreover, Field suggests two reasons why modeltheoretic accounts are irrelevant to logical pluralism:
One of these reasons is that by varying the denition of model, this
approach denes a large family of notions, classically valid, intuitionistically valid, and so on; one neednt accept the logic to accept the notion of
validity. A classical logician and an intuitionist can agree on the modeltheoretic denitions of classical validity and of intuitionist validity; what
they disagree on is the question of which one coincides with genuine
validity. For this question to be intelligible, they must have a handle on
the idea of genuine validity independent of the model-theoretic denition.
Of course, a pluralist will contest the idea of a single notion of genuine
validity, and perhaps contend that the classical logician and the intuitionist
shouldnt be arguing. But logical pluralism is certainly not an entirely trivial
thesis, whereas it would be trivial to point out that by varying the denition
of model one can get classical validity, intuitionist validity, and a whole
variety of other such notions. (Field 2009: 348)
Field goes on to suggest that Beall and Restall must have meant that there
is an implicit requirement for interpreting truth in a case, namely, that
truth in all cases implies truth. Field then argues that this will not produce
an interesting sort of logical pluralism as the pluralist notion of logical
consequence suggested by Beall and Restall does not capture the normal
meaning of logical consequence. But it should be noted that Beall and
Restall (2006: 36 .) do say something about the matter. Specically, they
suggest that on one reading of case (the TM account), Tarskian models
are to be understood as cases. Another reading (the NTP or necessary truthpreservation account) takes possible worlds to be cases. Beall and Restall
(2006: 40) add that the existence of a possible world that invalidates an
argument entails the existence of an actual (abstract) model that invalidates
the argument.
So, it is not clear that Fields critique is accurate, as Beall and Restall do
suggest that there is a case that corresponds with the actual world on the
TM account it is a Tarskian model and on the NTP account it is a possible
world. The latter is of immediate interest to us, given that the metaphysical
246
Tuomas E. Tahko
interpretation of logical truth also makes use of the possible worlds jargon.
Yet, Beall and Restall do not provide an interpretation of possible worlds,
so it is not quite clear what the connection, if any, between the NTP
account and the metaphysical interpretation of logical truth is.
Connecting all this with the analysis provided in the previous section,
one might suggest that classes of cases are sets of metaphysically possible
worlds, distinguished in terms of logical truths that are true in each set of
possible worlds. Only one possible world is actual, but the logical truths
that are true in the actual world will also be true in all worlds which are in
the same set of possible worlds, i.e., these worlds may dier in other
regards, but they are close to the actual world in the sense that all the
logical truths are shared.
Accordingly, pluralists about logical truth, in the metaphysical sense,
hold that there are distinct sets of possible worlds in which dierent logical
truths hold. The metaphysical interpretation of logical truth can accommodate this sense of logical pluralism, provided that possible worlds are
interpreted appropriately this also enables us to preserve CI.16 However,
accommodating pluralism in the metaphysical interpretation of logical
truth does require a revision in our original denition (ML), which dened
a sentence as logically true if and only if it is true in every genuinely
possible conguration of the world. Since in this view of logical pluralism
there can be proper subsets of genuinely possible congurations with
dierent laws of logic, we must revise ML as follows:
(ML-P) A sentence is logically true if and only if it is true in every possible
world of a given subset of possible worlds representing genuinely possible
congurations of the world.
ML-P can of course also accommodate the situation where the laws of
logic are the same across all subsets of genuinely possible congurations,
i.e., logical monism in that case the relevant subset of possible worlds
would not be a proper subset of the genuinely possible congurations.
An alternative formulation of ML-P is possible, dismissing subsets
altogether. We could understand logical pluralism by giving dierent
interpretations to genuinely possible congurations.17 This formulation
16
17
Why is interpreting logical truth on the basis of metaphysical possibility the only way to preserve
CI? Because weve seen that only by restricting our attention to metaphysically possible worlds can
we preserve a sense of correspondence between logical truth and genuinely possible congurations of
the world. Only metaphysically possible worlds are suciently constrained to take into account all
the governing principles such as metaphysical a posteriori necessities.
Thanks to Jesse Mulder for suggesting this type of formulation.
247
248
Tuomas E. Tahko
to dier in spin (i.e., to have dierent spin quantum numbers). Accordingly, when we observe electron A having spin-up, we immediately know
that any electron, B, in the same orbital as A must have spin-down.
Moreover, there can be only two electrons in the same orbital and they
must always have opposite spin.
If cases such as the one for a real world modus ponens can be found,
then we may indeed have a rich enough set of logical laws to constitute a
logic, enabling the suggested interpretation of logical pluralism. The
resulting subset could be called WLNCMP.
This hardly exhausts the debate about logical pluralism, but it appears
that there are ways, perhaps several ways, to accommodate pluralism about
logical truth within the metaphysical interpretation.
5. Conclusion
In conclusion, I have demonstrated that there is a coherent metaphysical
interpretation of logical truth, and that this interpretation has some
interesting uses, such as applications regarding logical pluralism. It has
not been my aim to establish that this interpretation of logical truth is the
correct one, but only that it is of special interest to metaphysicians. I have
assumed rather than argued for a type of realism about logic for the
purposes of this investigation, but I contend that for realists about logic,
one interesting interpretation of logical truth is the one sketched here.18
18
Thanks to audiences at the University of Tampere Research Seminar and the First Helsinki-Tartu
Workshop in Theoretical Philosophy, where earlier versions of the paper were presented. In
particular, Id like to thank Luis Estrada-Gonzlez for extensive comments. In addition,
I appreciate helpful comments from Franz Berto and Jesse Mulder. Thanks also to Penny Rush
for editorial comments. The research for this chapter was made possible by a grant from the
Academy of Finland.
References
Anderson, A. R., and Belnap, N. D. 1975. Entailment: The logic of relevance and
necessity, volume i. Princeton University Press
Anderson, A. R., Belnap, N. D., and Dunn, J. M. 1992. Entailment: The logic of
relevance and necessity, volume ii. Princeton University Press
Anon. 1967. Ars Burana, in L.M. de Rijk, Logica modernorum vol. ii: The origin
and early development of the theory of supposition. Assen: Van Gorcum
Aquinas, T. 2000. Opera omnia, edited by E. Alarcn. Pamplona: University of
Navarre, www.corpusthomisticum.org/iopera.html
Aristotle. 1963. Categories and de interpretatione, translated with notes by J. l.
Ackrill. Oxford: Clarendon Press
1984. Metaphysics, translated by W. D. Ross, revised by J. Barnes. Princeton
University Press
1993. Metaphysics books gamma, delta and epsilon, translated with notes by
C. Kirwan, second edition. Oxford: Clarendon Press
1994. Posterior analytics, translated with a commentary by J. Barnes, second
edition. Oxford: Clarendon Press
2009. Prior analytics, translated and with a commentary by G. Striker. Oxford
University Press
Asenjo, F. G. 1966. A calculus of antinomies, Notre Dame Journal of Formal
Logic 7(1), 103105
Asenjo, F. G., and Tamburino, J. 1975. Logic of antinomies, Notre Dame
Journal of Formal Logic 16(1), 1744
Azzouni, J. 2005. Is there still a sense in which mathematics can have
foundations, in G. Sica, (ed.) Essays on the foundations of mathematics and
logic. Monza, Italy: Polimetrica International Scientic Publisher
2006. Tracking reason: Proof, consequence, and truth. Oxford University Press
2008a. Why do informal proofs conform to formal norms? Foundations of
Science 14, 926
2008b. Alternative logics and the role of truth in the interpretation of
languages, in D. Patterson, (ed.) New essays on Tarski and philosophy.
Oxford University Press, 390429
2010. Talking about nothing: numbers, hallucinations and ctions. Oxford
University Press
249
250
References
2013. Semantic perception: How the illusion of a common language arises and
persists. Oxford University Press
Azzouni, J. and Armour-Garb, B. 2005. Standing on common ground, Journal
of Philosophy 102(10), 532544
Barker, C. 2013. Negotiating taste, Inquiry 56, 240257
Batens, D., Mortensen, C., and Van Bendegem, J.P. (eds.) 2000. Frontiers of
paraconsistent logic. Dordrecht: Kluwer Publishing Company
Beall, Jc 2009. Spandrels of truth. Oxford University Press
2012. Why Priests reassurance is not reassuring, Analysis 72(3), 739741
2013. Free of detachment: logic, rationality, and gluts, Nos 47(3), 114
Beall, Jc and Restall, G. 2006. Logical pluralism. Oxford University Press
Beall, Jc, Forster, T., and Seligman, J. 2013. A note on freedom from detachment
in the Logic of Paradox, Notre Dame Journal of Formal Logic 54(1)
Beebee, H. and Dodd, J. (eds.) 2005. Truthmakers: The contemporary debate.
Oxford University Press.
Bell, J. 1998. A primer of innitesimal analysis. Cambridge University Press
Benacerraf, P. 1973. Mathematical truth, Journal of Philosophy 70, 66179
Bencivenga, E. 1987. Kants Copernican revolution. New York: Oxford University
Press
2000. Hegels dialectical logic. New York: Oxford University Press
2007. Ethics vindicated: Kants transcendental legitimation of moral discourse.
New York: Oxford University Press
Benthem, J. van, 1985. The variety of consequence, according to Bolzano,
Studia Logica 44(4), 389403
Berto, F. 2007. How to sell a contradiction: Studies in Logic 6. London: College
Publications
2008. and material exclusion, Australasian Journal of Philosophy
86(2), 16590
Blackburn, S. 1984. Spreading the word. Oxford: Clarendon Press
Bolzano, B. 1810. Beytrge zu einer begrndetereu Darstelluns der Mathematik.
Prague: Widtmann
1837. Wissenschaftslehre. Sulzbach: Seidel
1839. Dr Bolzano und seine Gegner. Sulzbach: Seidel
1841. Wissenschaftslehre (Logik) und Religionswissenschaft in einer beurtheilenden
Uebersicht. Sulzbach: Seidel
19852000. Wissenschaftslehre, critical edition by Jan Berg. Stuttgart: FrommanHolzboog
2004. On the mathematical method and correspondence with Exner, translated by
Rusnock, P., and George, R. Amsterdam-New York: Rodopi
Boolos, G. 1975. On second-order logic, Journal of Philosophy 72, 509527.
(Reprinted in Boolos 1998, pp. 3753)
1984. To be is to be the value of some variable (or some values of some
variables), Journal of Philosophy 81, 430450. (Reprinted in Boolos 1998,
pp. 5472)
1998. Logic, logic, and logic. Cambridge, MA: Harvard University Press
References
251
Brady, R. 2006. Universal logic. Stanford, CA: Centre for the Study of Language
and Information
Brady, R. and Rush, P. 2009. Four Basic Logical Issues, The Review of Symbolic
Logic 2(3), 488508
Buchholtz, U. 2013. The unfolding of inductive denitions, PhD Dissertation,
Dept. of Mathematics, Stanford University
Burge, T. 1975. On knowledge and convention, in Foundations of mind (2007),
3237. Oxford University Press
Burgess, J. A. 1990. The sorites paradox and higher-order vagueness, Synthese 85,
417474
2010. Review of Beall and Restall 2006, Philosophy and Phenomenological
Research 81, 519522
Burgess, J. P., and Rosen, G. 1997. A subject with no object: Strategies for
nominalistic interpretation of mathematics. Oxford University Press
1999. Which modal logic is the right one? Notre Dame Journal of Formal Logic
40, 8193
Carey, S. 2009. The origin of concepts. Oxford University Press
Carnap, R. 1934. Logische Syntax der Sprache, Vienna, Springer, translated as The
logical syntax of language, New York, Harcourt, 1937; another edition edited
by Amethe Smeaton, Patterson, New Jersey, Littleton, 1959
Cesalli, L. 2007. Le ralisme propositionnel: smantique et ontologie des propositions
chez Jean Duns Scot, Gauthier Burley, Richard Brinkley et Jean Wyclif. Paris:
Vrin
Chateaubriand Filho, O. 2001. Logical forms: Part i truth and description.
Campinas: UNICAMP
2005. Logical forms: Part ii logic, language, and knowledge. Campinas:
UNICAMP
Chomsky, N. 2003. Reply to Millikan, in L. M. Antony and N. Hornstein
(eds.) Chomsky and his critics. Oxford: Blackwell Publishing, 308315
Chomsky, N. and Lasnik, H. 1995. The theory of principles and parameters, in
Noam Chomsky The minimalist program. Cambridge, MA: The MIT Press,
13127
Cole, J. 2013. Towards an institutional account of the objectivity, necessity, and
atemporality of mathematics, Philosophia Mathematica 21, 936
Condillac, E. 1780. La logique (Logic), translated by W. R. Albury. New York:
Abaris
Crivelli, P. 2004. Aristotle on truth. Cambridge University Press.
da Costa, N. 1974. On the theory of inconsistent formal systems, Notre Dame
Journal of Formal Logic 15, 497510
Daly, C. 2005. So wheres the explanation? In Beebee and Dodd 2005,
85103.
Davidson, D. 1973. In defense of convention T, in H. Leblanc (ed.) Truth,
syntax and modality. Amsterdam: North-Holland. Reprinted in his Inquiries
into truth and interpretation. Oxford University Press.
Dehaene, S. 1997. The number sense. Oxford University Press
252
References
References
253
254
References
References
255
256
References
References
257
Ladyman, J. and Ross, D. (with D. Spurrett and J. Collier). 2007. Every thing
must go, Oxford University Press
Lapointe, S. 2011. Bolzanos theoretical philosophy. Houndsmill: Palgrave
2014. Bolzano, Quine and logical truth in G. Harman, and E. Lepore (eds.)
Blackwell companion to Quine. London: Blackwell
Lepore, E. and Ludwig, K. 2006. Ontology in the theory of meaning,
International Journal of Philosophical Studies 14, 325335
Levinas, E. 1998. Discovering existence with Husserl. Illinois: Northwestern
University Press
Lewis, D. 1969. Convention: A philosophical study. Cambridge University Press
1991. Parts of classes. Oxford University Press
2005. Quotation, in G. Priest, Jc Beall, and B. Armour-Garb (eds.) The law of
contradiction: New essays. New York: Oxford University Press, 1
Lindstrm, P. 1966. First order predicate logic with generalized quantiers,
Theoria 32, 186195
1969. On extensions of elementary logic, Theoria 35, 111
MacFarlane, J. 2005. Making sense of relative truth, Proceedings of the
Aristotelian Society 105, 321339
2009. Nonindexical contextualism, Synthese 166, 231250
2014. Assessment sensitivity: Relative truth and its applications. Oxford University
Press
Maddy, P. 2007. Second philosophy. Oxford University Press
2011. Defending the axioms. Oxford University Press
(to appear) The logical must. Oxford University Press
Marenbon, J. 2012. (ed.) The Oxford handbook of medieval philosophy. Oxford
University Press
Mares, E. D. 2004. Semantic dialetheism, in G. Priest, Jc Beall, and
B. Armour-Garb (eds.) The law of non-contradiction. Oxford: Clarendon
Press, 264275
2007. Relevant logic: A philosophical interpretation. Cambridge University Press
Martin, C. J. 2001. Obligations and liars, in M. Yrjnsuur (ed.), Medieval
formal logic: Obligations, insolubles and consequences. Springer, 6394
2010. The development of logic in the twelfth century, in Pasnau 2010,
129145
2012. Logical consequence, in Marenbon 2012, 289311
Martin, D. A. 2001. Multiple universes of sets and indeterminate truth values,
Topoi 20, 516
Mathias, A. R. D. 2001. The strength of Mac Lane set theory, Annals of Pure
and Applied Logic 110, 107234
McDowell, J. 1994. Mind and world. Cambridge, MA: Harvard University Press
Meillassoux, Q. 2011. Contingency & the absolutization of the one, A lecture
delivered at the Sorbonne for a symposium. Source: http://speculativeheresy.
wordpress.com/2011/03/24.
Meyer, Robert K., Routley, R. and Dunn, J. Michael. 1979. Currys paradox,
Analysis 39, 124128
258
References
References
259
260
References
References
261
262
References
References
263
Index
264
Index
Field, H, 211, 244245
rst-order logic, 4243, 61, 64, 74, 80, 82, 213,
226
Fllesdal, D, 20, 92
formalism, 74
Frege, G, 1, 41, 94, 128, 131, 157, 195, 214
and Russell, 129, 211
correspondence with Hilbert, 5152
Fregean, 22
realism, 196
Gentzen, G, 111112, 116, 122124
meaning, 85
proof, 118120
system of natural deduction, 81, 84
geometry, 8, 52, 93, 123, 128, 136, 208
application, 136, 215
axioms, 52
Euclidean, 41, 77, 128, 215
non-Euclidean, 186, 215
Gdel, K, 87, 92, 111, 116, 118119, 142, 191
completeness theorem, 42, 80
incompleteness theorem, 79
Goldbach conjecture, 76, 139141
Hateld, G, 102
Hilbert
Hilbertian, 8
Hilbert, D, 5152, 110, 118119, 211
Hilbertian, 52, 55
Hilberts program, 142
space, 72, 104
Husserl, E, 17, 2528, 30
cognition, 22
concept of evidence, 2324
conception of logic, 1819
Husserlian, 26
logical realism, 190
phenomenological reduction, 1921
transcendence, 17
idealization, 71, 106107
classical logic, 106108, See logic, classical
of rudimentary logic, 5, See logic, rudimentary
technique of, 105106
throughout mathematics, 61
implication, 41, 73
intuitionistic, 125
incompleteness, 113114
theorem, 79, See Gdel
independence, 3, 13
conceptions of, 26
essential and modal, 15
human-, 2, 15
IF Independence Friendly, 82
265
mind-, 20, 56
of facts, 14
of logic, 7
of logical truth, 29
proofs, 51
realist, 3, 1518
results, 83
intuitionism, 115, 140
intuitionist validity, 245
intuitionistic, 69, 74, 112
analysis, 52, 54
consistency, 59
intuitionistic logic, 116
intuitionistically, 45, 70, 113
logic, 46, 50, 52, 54, 57, 6061, 6364, 74, 77,
8283, 108, 111112
predicate calculus, 8486
propositional calculus, 110
semi-, 86, 89, See logic, intuitionist
semi-intuitionism, 85
Jankovs logic, 115
Kant, 20, 41, 57, 94, 180, 183, 187, 195, 203, 208,
213
Anti-, 198
ethics, 184185
Kantian, 7, 179, 181, 183
Kant-Quine, 58, 71
KF-structure, 94
Ladyman, J, 99
Ladyman, J and Ross, D, 99, 104
language acquisition, 40, 103
law of excluded middle (LEM), 29, 50, 65, 74,
8788, 90, 139142, 144
weak, 115
law of non-contradiction (LNC), 9, 29, 48, 239,
242
Lewis Carroll regress, 33
logic
applied, 215
canonical application, 2, 215216, 220
classical, 45, 4245, 50, 5253, 77, 81, 84,
8688, 107, 111113, 115, 211, 214, 216, 218,
228
application to mathematics, 91
idealization, 106108
rise of, 214
valid in, 6061, 63
conditional, 107
content-containment model of, 42
deviant, 69, 104, 106107
intuitionist, 215216, 219
mathematical, 7273, 212, 214, 217
266
Index
logic (cont.)
medieval, 147, 158159, 161, 164, 212214
Megarian, 212
non-classical logic, 5, 49, 211, 218
paraconsistent, 50, 5455, 64, 68, 108, 211, 215,
225, 240, See paraconsistency
Port Royale, 213
pure, 18, 178, 215216
relevance, 54, 107
rudimentary, 5, 95, 97100, 104108, 120121, 125
rule-governed model of, 4142
semi-intuitionist, 4
substitution model of, 42
traditional, 214, 217
logica docens, 212216, 218, 220, 223
logica ens, 212, 216, 220, 223
logica utens, 212, 218219, 223
logical
connectives, 23, 115116, 222
consequence, 8, 51, 5960, 79, 109110, 112,
123, 235
Beall and Restalls, 50, 244245
Bolzano, 203204, 207
in mathematical practice, 43
material approach to, 232
Reads defense of material, 228
traditional denition of, 192
conventionalism, 3, 33, 47, 190
inference, 5, 41, 93, 100, 134
pluralism, 4, 9, 217, 237, 244248
realism, 4, 8, 1315, 189192, 195197, 208, 233
schemata, 41
logical validity, 50, 56, 121123, 161, 237, 247
logicism, 74
MacFarlane, J, 51, 6567
Maddy, P, 5, 121
mathematical
objects, 1, 90, 221
proof, 42, 120, 123
realism, 14
reality, 1, 15
McDowell, J, 2530
meaning
of a mathematical proposition, 137
of all, 78
of logical operations, 24, 85
of logical particles, 112
of logical predicates, 225, 228
of logical terms, 69, 135
of proposition, 150
of spoken and written utterances, 148149
of stateable, 155
Medvedev lattice, 115
metalogical, 45
metalogical debates, 48
mirror neuron, 39
model theory, 4445, 226227
model-theoretic, 42, 64, 80, 204, 207, 235236,
238, 245
account of validity, 221223
modus ponens, 43, 48, 95, 113, 136, 189190, 214,
228, 230, 247248
monism, 51, 54, 62, 217, 246
naturalism, 34, 74, 189
necessity, 42, 159, 186, 204207, 241,
See possibility
causal, 174
epistemic, 208
follows of, 205
logical, 35, 134, 174, 241
metaphysical, 242, 244
natural, 175
semantical, 181
non-realist, 34, 74
norms of reasoning, 158
objectivity, 3, 14, 5660, 6566, 71, 174, 179180,
184, 186
axes of, 62
criteria of, 183
of logic, 7
of mathematics, 76
open-texture, 71
paraconsistency, 910, See logic, paraconsistent
Peano Arithmetic, 73, 87
Piaget, 100
Plato, 18, 122, 162, 164, 166, 168, 176, 184
platonic, 18, 25, 74, 97, 162
platonism, 5, 135, 139140
platonist, 136, 147, 221
pluralism, 9, 49, 51, 69, 189, 247, See logical:
pluralism
possibility, 70, 197, 228, 247, See necessity
genuine, 241
logical, 174
metaphysical, 241, 247
of cognition, 13, 1617, 2122
of logic, 32
Priest, G, 3, 9, 158, 225226, 229230, 232
arguments against LNC, 240, 243
principle of bivalence, 32
principles and parameters model, 39
quantum mechanics, 9899, 104, 108, 144
Quine, W.V.O.
on second-order logic, 124
substitutional procedure, 195
Index
Quine, W.V.O., 32, 81, 100, 124, 135, 166
challenge, 3235, 3840, 48
Dummett-Quine-Carnap, 6869
holism, 144
Kant-Quine, 58, 71
Putnam and Davidson, 57
Quinean, 178, 181, 216
rationality, 168, 184
relativism, 49, 51, 129, 190, 220, 238
folk-, 4951, 5760, 63, 6566, 68, 7071, 190
logical, 49, 51
proper, 6768, 70
rule-following, 61, 129132, 137, 142
second-order logic, 61, 64, 73, 79, 81, 128
classical, 74
full, 79, 82
Quine on, 81, 124
semantics, 81
Sellars, W, 22, 97, 99, 196
objection, 1415
set theory, 6, 64, 81, 88, 92, 124, 136, 181, 231
background, 42
development of, 120121
Kripke-Platek, 88
paraconsistent, 241
satisability in, 52
Zermelo-Fraenkel (ZF), 73
Shapiro, S, 24, 9, 43, 61, 7879, 118, 189, 233
smooth innitesimal analysis, 5254
Spelke, E, 102103
structuralism, 3, 74
conceptual, 34, 78, 80, 9091
in-re, 74
modal, 74
Tarski, A, 195, 202, 234236, 238
biconditionals, 4445
denition of logical consequence, 203
Generalised Tarski Thesis, 50, 70
T-schema, 233
267
-type, 204
Tarskian conception, 238
Tarskian model, 50, 207, 245
theory choice, 9, 216, 223
truth
absolute, 233234, 237239, 241
by convention, 32, 4748
in a model, 233
logical, 3, 9, 29, 41, 46, 95, 99, 233, 240, 242
all, 105
rst-order, 42
ground of, 93
interpretation, 233234, 248
metaphysical, 89, 233, 235239, 244247
pluralism about, 244245
Quine on, 233
realist, 238
reecting facts, 97
preservation, 44, 160, 222, 245
preserving, 44, 160
relative, 234, 236238, 244
truth tables, 46
T-schema, 44, 46, 233235
vagueness, 56, 6165
logics of, 106
real, 107
worldly, 96
Waismann, F, 57, 71
Wason Card Test, 218219
Wittgenstein, L, 5, 125, 144, 195, 219
and physics, 144
means by postulate, 144
on mathematics, 128129
on rule-following, 129132, 143
rejection of Hilberts program, 142
Steiner on, 6
Wittgensteinian, 45, 132, 141
Wright, C, 49, 5860, 6263, 70
Zermelo-Fraenkel set theory, 73