1


The Semantics of Existence

Friederike Moltmann
IHPST (Paris1/ENS/CNRS)
fmoltmann@univ-paris1.fr

1. Introduction

The notion of existence has puzzled philosophers for a very long time, and a great range of
views about that notion can be found throughout the history of philosophy. While some
philosophers think that existence and ontological commitment can and perhaps should be
pursued independently of the linguistic form of the relevant sentences, the linguistic form of
statements of existence has also often been taken to be revealing as to the ontological issues
themselves. In any case, it is a worthwhile and important project to see what the actual
semantics of sentences expressing existence in fact is, and it is what the present paper aims to
pursue. This paper explores the linguistic semantics of statements of existence with respect to
its philosophically relevant aspects, but it also yields a number of results that are of interest to
linguistic semantics as such, for example concerning the distinction between stage-level and
individual-level predicates and the semantics of stative and eventive verbs.
The verb exist is of course a central expression for making statements about existence.
Many philosophers have expressed particular views concerning that expression (or
occurrences of it), though at the same time exist has hardly been a subject of study in
linguistic semantics, it seems because of its apparent ‘technical’ and thus marginal status. This
paper explores the semantics of sentences with verbs of existence, not just with the verb exist,
but also other predicates of existence, such as occur or obtain and the somewhat related
expression real. It appears that from the point of view of natural language semantics, exist and
other existence verbs do not in fact behave that exceptionally, but pattern in a number of
respects together with other classes of predicates, in ways that are revealing for the
philosophical issues themselves. I will focus on verbs of existence when they occur
predicatively, as in (1) and (2):

(1) a. The man we talked about exists.

b. The golden mountain does not exist.
 2




(2) Electrons exist.

These examples illustrate in what ways exist seems so peculiar as a predicate. In (1a), exist
appears to apply trivially, stating that an existing man exists. In (1b), exist is said to be false of
the subject referent – an object that is said not to exist. In (2), exist seems to not act as a
predicate at all, but to express existential quantification
While there is a major philosophical tradition according to which existence statements are
not semantically subject-predicate statements, more recently a number of philosophers have
defended the view that exist is in fact a first-order extensional predicate, at least with singular
terms as subjects. In this paper I will explore this view in its full generality. I will argue that
existence predicates such as exist and others like occur and obtain have a particular lexical
meaning, which matches the particular nature of the entities they apply to. At the same time,
exist has certain features that make it applicable independently of the type of entities in
particular contexts. I will argue that exist acts as a first-order extensional predicate also in (2),
where the bare plural has in fact the status of a kind-referring term rather than being
quantificational.
Sentences with the verb exist as in (1) and (2) have a very different semantics from certain
other sentences that can be used to express existence, in particular there-sentences and
existentially quantified sentences. There-sentences and quantificational sentences may involve
a significantly greater domain of entities than what exist could be true of. This may suggest
that exist is on a par with the adjectival predicate real, but in fact the two expressions are
fundamentally different linguistically and carry different ontological implications.
I will first point out a range of differences between there-sentences and sentences with
existence predicates and propose particular lexical analyses of different existence predicates. I
then will give an account of existence statements with bare plurals and mass nouns as
involving kind reference. Finally, I will compare the predicate exist to the expression real.

2. Existence statements and there-sentences

In philosophy, there are two opposing views on existence. On one view, existence is a
univocal concept and closely tied to existential quantification and counting: if there is one
thing and there is another thing, even of a very different kind, then there are two things. On
the other view, things of different kinds may ‘exist’ in fundamentally different ways. While
there have been philosophical considerations put forward for the one view or the other, it
 3




appears that natural language in fact supports both views.1 There are two distinct types of
sentences corresponding to the two views: there-sentences as well as sentences with simple
existential quantifiers (some) reflect the first notion of existence; existence statements, which
may contain a variety of existence predicates (such as exist, occur, take place, obtain), reflect
the second notion. While there are different existence predicates for existence statements,
exist at the same time has special semantic features that make it suitable as a univocal
concept, applying to entities of any sorts.
There-sentences and existence statements differ in several linguistic respects: with respect
to their syntactic structure, with respect to the ‘ontological commitment’ they carry, and with
respect to any constraints on the kinds of entities they may be ‘about’.
There-sentences consist in there, followed by a verb (such as the copula be), a weak NP
(an NP carrying existential import), and possibly a 'coda', a predicative expression of some
sort; in (3a), the coda is empty giving the statement an existential interpretation; in (3b), the
coda is a location modifier, giving the statement a locational interpretation:2


(3) a. There are black swans.
b. There is [a man] [in the garden].

Existence statements have a very different syntactic structure: they are subject-predicate
sentences with a verb of existence as predicate, such as exist, occur, or obtain, and any kind of
NP (not just a weak NP) as subject, such as a singular or plural terms or a bare (that is,
determinerless) plural or mass noun:

(4) a. The president of France exists.

b. The students exist.
c. Vulcan does not exist.
d. Natural numbers exist.































































1
See van Inwagen (1998) for a philosophical discussion of the two views and a defense of the former, and
McDaniel (to appear a, b) for a recent defense of the latter. The latter view was also that of Aristotle and Ryle.
2
There-sentences may also contain an implicit location restriction. Thus (1b) can be understood as in (1a) in a
particular context:

(1) a. There are exactly three scientists in this laboratory that can solve the problem.
b. There are exactly three scientists that can solve the problem. (can be understood as ‘in this laboratory’)


 4




The common view about (4c) is that it is a quantificational statement just like (3a). I will later
argue that (4c) is in fact entirely parallel to (4a) (on its most natural reading).
There-sentences and existence statements are not just syntactically different; they also
display fundamental semantic differences.3 The first difference is one of ‘ontological
commitment’: There-sentences can quantify over past, merely possible, and merely
intentional objects, objects that exist (or other existence predicates) could not be true of:

(5) a. There are historical buildings that no longer exist.

b. There are possible buildings that do not actually exist.
c. There are imaginary buildings that do not exist.
(6) a. There are buildings built in the past that no longer exist.
b. There are buildings I might have built that do not exist.
c. There are buildings John thought of that do not exist.

In these sentences, it is the particular construction and semantic context that enables
quantification over such 'nonexistent' objects. In (5), existential quantification over past,
possible and intentional objects is made possible by the intensional adjectives historical,
possible, and imaginary. In (6), it is made possible by the use of relative clauses containing a
modal verb (requiring a relative-clause-internal interpretation of the head of the relative
clause buildings). Without such modifiers the sentences could not possibly be true:


(7) ?? There are buildings that do not exist.

The contrast between (5a, 6a) and (7) also shows that ordinary sortal predicates like building
are existence-entailing, just like exist itself. Exist cannot be true of past and merely possible
objects because such objects are not in the domain of the presently actual ones. Exist cannot
be true of intentional objects, either. Intentional, ‘nonexistent’ objects are highly controversial
and require an in-depth discussion that goes beyond the scope of this paper (McGinn 2000,
author, ms). In this paper, I will assume that natural language does permit intentional objects
as denotations of noun phrases, as is particularly plausible for sentences like (5c) and (6c),
though not too much in the subsequent discussion hinges on this assumption. I will follow
McGinn (2000) in taking intentional objects to be the denotations of NPs that fail to have an






























































3
Priest (2005) takes there-sentences and exist-sentences to pattern the same semantically and to be distinct from
quantification (which for him allows for a grater domain of quantification), erroneously, I think.
 5




actual referent; intentional objects then are things that essentially do not exist, that is, for
which exist is defined, but is not true of (unlike of fictional characters (author, ms)).4
Ordinary existential quantification shares with there-sentences the ability to quantify over
‘nonexistent’entities: some can be used to quantify over past, possible, and intentional objects,
while stating that they do not exist:

(8) a. Some historical buildings no longer exist.

b. Some possible buildings do not actually exist.
c. Some imaginary buildings do not exist.

It is clear from these examples that the ‘ontological commitment’ that may be expressed by
there-sentences or existentially quantified sentences has to do with the use of an existence-
entailing predicate, not the there-construction or quantification as such.
The second difference concerns different types of entities. There-sentences differ from
existence statements by allowing quantification over any kind of entity:

(9) There were many objects / events / facts / situations / …

The same holds for ordinary quantificational sentences, which allow for quantification over
any type of entity with the same kinds of quantifiers (some, a, …). By contrast, different
existence predicates (exist, occur, take place, obtain, remain) are generally restricted to
particular kinds of entities. Exist generally can apply only to entities that are not events, in
particular when it involves a temporal specification (such as past tense). There are instead
specific existence predicates for events, such as occur, happen, and take place:5


(10) a. The murder * existed / ok occurred.
b. John’s speech * existed / ok took place this morning.


A-temporal exist can apply to events, but it is a use of exist that is to be distinguished from
tensed exist, and I will turn to it in the next section. I will come to the specific lexical
meanings of (tensed) exist and occur later, in Section 4.































































4
For a critical discussion of McGinn’s (2000) view see van Inwagen (2008).
5
I will disregard the difference between occur, happen, and take place in what follows.
 6




Entities such as facts, possibilities, and situations might be said to ‘exist’, but they more
naturally ‘obtain’ or ‘remain’. Obtain and remain are existence predicates that are restricted to
abstract entities of the sort of facts and situations. Why do obtain and remain not apply to
ordinary objects? There is one common property of the kinds of entities that are facts or
situations, and that is that they are not in the world, but ‘at’ a world, as Strawson (1950) put
it.6 A plausible view of what makes an entity be at a world rather than in the world, is that
such an entity is an abstraction from objects in the world and properties of them as well as
relations among them: it is an abstract entity and thus lacks a location, but it may be entirely
dependent on concrete objects as well as properties of them and relations among them. This
means that the predicates obtain and remain locate entities at a world, whereas exist locates
entities, at least primarily, in the world (unless used in a derivative way).7
Another existence predicate applies specifically to laws: a law exists in a state s just in case
it is valid in s; in the case of laws, existence means validity. Yet another existence predicate is
live or be alive. Live applied to biological entities presupposes a body, but the biological
entity exists just as long as it has a ‘life’. Again, the predicate exist can somewhat marginally
be applied to biological organisms (the horse still exists, the man still exists).
Even though exist in existence statements is obviously a predicate syntactically, there is a
significant philosophical tradition of proposals according to which the existence verb in such
sentences does not have the function of predicating a property of the subject referent. This
tradition includes Kant’s view that existence is not a property as well as Frege’s view
according to which exist is a second-order predicate, predicated of a concept denoted by the
subject. More recently, however, some philosophers tried to do justice to the fact that exist in






























































6
This is a view not shared by all philosophers; Austin (1979) famously disagreed with Strawson (1950) on the
ontological status of facts.


7


This also holds for the possibilities that are the referents of terms of the sort the possibility that S. Possibilities
may obtain or remain:

(1) a. The possibility that John is no longer chairman obtains.

b. The possibility that John will become chairman remains.

What are possibilities as the referents of terms like the possibility that S ? Possibilities obviously are only partial
possible worlds, as characterized by the content of the that-clause, and in fact possibilities as referents of terms
of the sort the possibility that S can only be epistemic possibilities, not, say, physical, deontic, or metaphysical
possibilities. Thus, the possibility that John is no longer chairman exists, because his not being chairman any
longer is not excluded by what the interlocutors are supposed to know. But there can be no such thing as the
possibility that John can lift the table, understood as physical possibility; or the possibility that we may enter the
room, understood as deontic possibility; and the possibility that I do not exist does not obtain because of what is
metaphysically possible, but rather it is excluded because of what I know. Possibilities are in fact not on a par
with possible objects which have being independently of what is actually the case and what is actually known.


 7




existence statements is a predicate linguistically and have argued that at least with singular
terms as in (4a,b) exist acts as a predicate also semantically (Salmon (1987, 1998), Miller
(1986), McGinn (2000)). The fact that exist is not true of all the objects one can quantify over
or make reference to was a crucial motivation for Salmon to take it to be a predicate; the fact
that different existence predicates apply to different kinds of entities gives a good piece of
further support for the view. It means that existence predicates have a particular lexical
content, imposing particular conditions on the objects they can apply to.

3. A secondary meaning of exist

Different kinds of existence predicates appear to attribute particular ‘ways of being’ to

particular types of entities, whereas existential quantification just ‘counts’ entities in whatever
way they are. Yet not all existence predicates are alike in that respect. Even though exist
occurs as a predicate with a lexical meaning restricting it to entities with no temporal parts, it
has special features making it in some sense applicable to the larger domain of entities as
such.
This peculiarity of exist manifests itself in three kinds of contexts. First, exist appears to be
able to apply to intentional entities of any sort in negative existence statements and questions,
as in (11) or, in fact, to entities that might have been merely intentional entities, as in (12):

(11) a. The Third World War does not exist.

b. Does the Third World War exist?
(12) The medieval war mentioned in the book exists.

Crucially, on this use, exist is tensed, and applies even to objects in the past without temporal
specification.
Other existence predicates cannot apply in this way: occur, happen, and take place can
never apply to enduring objects or facts and obtain can never apply to events or enduring
objects, live can never apply to non-biological entities, and is valid cannot apply to other
things than laws. This difference is obviously a difference in the presuppositional part of the
lexical meanings of the predicates in question: exist is false of entities if they are merely
intentional; all other existence predicates are false of entities if they are intentional entities of
the presupposed type.
 8




Exist is special also in its ability to occur with bare plurals and mass nouns that may
describe entities of any sort:

(13) a. Great wars still exist.

b. Political protests do not exist in Bhutan.

Finally, exist differs from other existence predicates in its ability to occur in there-
sentences:

(14) a. There exists a man that can solve the problem.

b. There exists a man in Germany that can solve the problem.

Among the other existence predicates only locational or existential remain are possible in
there-sentences:

(15) a. There remained three people in the garden.

b. There remains no other possibility.

All other existence predicates do not fare that well in there-sentences:

(16) a. ? There occurred three murders this morning.

b. ?? There took place a protest in this country.
c. ?? There obtains no other possibility.
d. ?? There live still three men of her family.

There are different views of how there-sentences are interpreted and these observations
obviously bear on the issue. On one view, discussed in Higginbotham (1987), what follows
there is just interpreted like an ordinary existence statement, in virtue of a movement of the
weak NP to sentence-initial position, to replace there. Thus, (15a) is interpreted as in (17a)
and (15b) as in (17b):

(17) a. Three people remained in the garden.

b. No other possibility remains.
 9




There are problems for this view, though. One of them is the inability of most predicates of
existence to occur after there, whether they are existence predicates of objects, events, or
situations.8 Ordinary existence predicates, with their ordinary meaning, simply do not occur in
there-sentences, and exist is entirely exceptional in that respect. According to Higginbotham,
be in there-sentences is in fact the be of existence, a problematic suggestion because be is
hardly used for expressing existence anymore. A more plausible view of be in there-sentences
than as existential is as locational be (as in John is in Paris), so that the coda is considered a
specification of a location in a somewhat extended sense (including a domain of entities
sharing a property).
If the verb in there-sentences is a predicate of location, also the possibility of remain in
there-sentences, as opposed to other existence predicates, is explained: only remain, not the
other existence predicates, has a locational use (John remained in Paris). Given this
generalization, exist in there-sentences should also be considered locational, acting somewhat
pleonastically as equivalent to locational be. I take this to be a secondary meaning of exist, a
meaning available only in a context in which exist cannot occur as a true existence predicate
with its usual lexical meaning (which I turn to in the next section). On the locational use, exist
expresses a relation between entities (of any sort) and locations; the location is specified by
the overt location modifier or in the absence of a location modifier, it will be the actual world,
as in (14a).
If exist has a secondary meaning as a locational predicate, this can explain its application to
objects not satisfying the sortal requirements of exist as an existence predicate. In there-
sentences, exist occurs with its locational meaning and thus does not impose a sortal
requirement at all. Also, if exist in negative existentials as in (14a) has in fact a locational
meaning, it will not require that objects it applies to be enduring objects. Rather, here the
presupposition is simply that the subject stands for a non-intentional or an intentional object;
in the former case, exist is true of the object, in the latter it is not:

(18) [exist]w ([NP]) = 1 if [NP] is a non-intentional object,

[exist]w([NP]) = 0 if [NP] is an intentional object,
[exist]w([NP]) = undefined otherwise, , i.e. if [NP] is empty.































































8
Another argument against the view is that in many languages verbs occur after existential pleonastic pronouns
that have no use as existence predicates. Thus, French has avoir ‘have’ in existential sentences, and German
geben ‘give’.


 10




The secondary location meaning will also be relevant when specifying the particular meaning
exist has with bare plurals and mass nouns.

4. The primary meaning of exist and the meaning of occur

We can now turn to the particular lexical content of exist and occur. The difference between
exist and occur indicates that exist does not just locate entities in a world at a time; rather it
does it in a way that has to do with a fundamental difference between objects and events.9
Objects and events differ fundamentally in their ability to have temporal parts, at least
according to one important view10: according to that view, events are perduring entities with
temporal parts, whereas objects are enduring objects that have only spatial parts. An object is
an entity that is wholly present at each instance of its lifespan. By contrast, at a given moment
that is a proper part of an event’s duration, only a temporal part of the event is present, not the
whole of the event. Thus, in first approximation, the conditions on the application of exist
may be stated as follows:


(19) a. For an entity x that cannot have temporal parts,
exist is true of x at a time t in a world w iff for any subinterval t' of t, (the whole of) x
is present at t’ in w.
b. For an entity x that can have temporal parts,
occur is true of x at a time t in a world w iff for any proper part t' of t, only some
proper part of x is present at t' in w.


Stating the content of exist and occur in terms of such application conditions on objects does
not yet account, though, for some crucial difference between the two verbs, namely the fact































































9



Exist can also apply to a temporal stage of an individual:

(1) The Berlin of the 1920ies does not exist anymore.

This means that a temporal stage of an individual still has the ontological status of an object, not that of an event;
that is, its parts are not temporal.


10
There are also philosophical views, for example that of Sider (2001), according to which both objects and
events can have temporal parts and thus according to which there is no fundamental ontological difference
between objects and events. Natural language does not seem to reflect that view, though.
 11




that exist is a stative verb and occur an eventive verb. This difference is particularly apparent
from the possibility of applying the progressive to occur, but not to exist:

(20) a. * The building is currently existing.

b. The protest is finally occurring / taking place.

The difference is also reflected in the corresponding nominalizations. The existence of the
building clearly describes a state, whereas the occurrence of the protest describes an event.
The latter can have typical event properties, such as ‘being sudden’, which the former cannot.
Given the Davidsonian view on which events act as implicit arguments of verbs, the event
or state that nominalizations stand for is quite simply the implicit event or state argument of
the verb from which the nominalization was derived. Given that view, exist is a stative verb
which describes states that would also be the referents of NPs with the corresponding
nominalization, such as the existence of the president of France, whereas occur is an
accomplishment or achievement verb that describes events that would also be the referents of
NPs with the corresponding nominalization, such as the occurrence of the murder.
‘The occurrence of the murder’ is not the same event as ‘the murder’: The latter may have
been done with an axe, may have been grisly and brutal; the former cannot be any of that;
though it could be sudden, unexpected, or early. An occurrence of an event e in fact is an
event that does not have any inherent qualitative properties and in that respect generally
differs from e; it is an event that is entirely constituted by transitions from the temporal
location of one part of e to the temporal location of another part of e. Thus, occur when
applied to an event e describes another event that consists in the transitions among the
‘presences’ of the parts of e at relevant subintervals that belong to the duration of e. Exist
when applied to an object x, by contrast, describes a state that is the presence of the whole of
x during the time in question. With Davidsonian event arguments, the lexical meaning of exist
and occur can thus be characterized as below:

(21) a. For a world w, an entity x without temporal parts, and an interval t, <e, x> ∈ [exist]w,t
iff e consists in the presence of (the whole of) x in w at t’ for any subinterval t’ of t.
b. For a world w, an entity e with temporal parts, and an interval t,
<e, e’> ∈ [occur]w,t iff e consists of transitions from the presence of e’’ in w at t to
the presence of e’’’ in w at t’’ for any minimal parts e’ and e’’ of e for which there are
subsequent subintervals t’ and t’’ at which e’ and e’’ take place.
 12




Thus, the lexical content of exist involves mapping an object onto a (non-qualitative) state of
the object at a time. The lexical content of occur involves mapping an event onto a non-
qualitative event that reflects the temporal part structure of the former.
The difference in the kinds of events that exist and occur describe can explain a further
difference between the two existence predicates, namely a difference in location modification.
Existence statements with occur do allow for location restrictions:

(22) The murder occurred in Munich.

By contrast, location restrictions (modifying the verb or entire sentence, not the subject) are
generally not possible in existence statements with exist. With singular terms and strong
quantifiers, such modifiers are completely excluded:

(23) a.* The man we talked about exists in another city.

b. * Mary does not exist in Germany.
(24) a. * Every cat we talked about exists in this city.
b. * Most people mentioned in this book exist in Germany.
c. * The only man who can solve the problem exists in Germany.

Exist-sentences with singular terms do not allow an implicit location restriction either:

(25) a.* The man we talked about exists. (to be understood: ‘in another city’).
b. * Mary does not exist. (to be understood: ‘in Germany’).
(26) a. ?? At least five million people exist. (to be understood: ‘in this country’).
b.?? Several universities exist. (to be understood: ‘in this city’).

There are two exceptions to the constraint against location modifiers in existence statements,
First, with weak quantifiers as subjects, location modifiers are at least marginally acceptable:


(27) a. ?? At least five million people exist in this country.
b.?? Several universities exist in this city.
c. Several / Exactly three hundred / At least ten cats exist in this village


 13




Second, exist-sentences with bare plurals or mass nouns as subjects are unproblematic with
location modifiers:

(28) a. Giraffes exist only in Africa.

b. Wild ponies do not exist in Germany.
c. Pure air does not exist in China anymore.

Both cases can be accounted for on the basis of the secondary locational meaning of exist.
The relative tolerance of location modifiers in (27) can be related to the acceptability of
location modifiers in the corresponding there-sentences:

(29) a. There exist at least five million people in this country.

b. There exist three scientists in this city.

The sentences in (27) in fact arguably are derived from the same underlying syntactic
structure as a there-sentence, by moving the weak NP in postverbal position into the position
that would, without movement, be spelled out as there. In that construction, recall, exist can
occur only with its locational meaning, not as a true existence predicate.11
That meaning is also at stake in exist-sentences with bare plurals or mass nouns as
subjects, as I will argue in the next section.
Not all adverbial modifiers are unacceptable in exist-sentences. Temporal modifiers are
unproblematic in existence statements with exist, just as in there-sentences:

(30) a. One building in this village still exists.

b. There is still one building in the destroyed village.

The restriction against location modifiers with exist can be explained given recent linguistic
work on the semantics of stative verbs. Given a Davidsonian event semantics, adverbial
modifiers will be predicates of an implicit event argument. Since events are generally spatio-
temporally located, exist either does not have an event argument or else its event argument is































































11
Williams (1984) argued that the subject position of there-sentences and the postverbal weak NP are linked by
coindexing; the proposal would be that such coindexing would permit movement of the postverbal NP into
subject position, before ‘there-insertion’.


 14




of a sort that would not be spatially located. Whereas some semanticists such as Katz (2003)
propose that stative verbs in general lack an event argument (but perhaps have a time
argument instead), I will follow Maienborn (2001, 2007) according to whom also stative
verbs have an implicit event argument position, namely for states. Moreover, I will follow
Maienborn (2007) in distinguishing two kinds of stative verbs. Stative verbs such as sit, wait,
stand, or sleep take concrete states as arguments (‘Davidsonian states’, as Maienborn calls
them). Such verbs allow for location modifiers, as well as manner modifiers, and they can act
as infinitival complements of perception verbs:

(31) a. John sat in the corner.

b. John sat awkwardly.
c. Mary saw John sit in the corner.

By contrast, stative verbs such as own, belong, resemble, and know do not allow for location
modifiers or manner adverbials, and they cannot act as infinitival complements of perception
verbs:

(32) a. * John resembles Mary in Germany.

b. * John resembles Mary with effort.
c. * Bill saw John resemble Mary.

Maienborn calls the state arguments of such verbs ‘Kimian states’, following Kim’s (1980)
conception of events. According to that conception, events are implicitly defined in terms of
their existence and identity conditions, on the basis of individuals, properties and times,

(33) a. For a property P, an object o and a time t, the event f(P, o, t) exists if Pt(o).
b. Two events f(P, o, t) and f(P’, o’, t’) are identical iff P = P’, o = o’, t = t’.

Kimian events are abstract; they are in fact obtained by abstraction in a Fregean sense. (33a,b)
are abstraction principles that provide identity conditions for events, but no such properties as
location or duration. Kimian events are thus not ‘in’, but ‘at’ the world.
It is generally agreed that the conditions in (33) actually define facts rather than events.
But it is a definition that, with one modification, is also suited for states. States are like facts
in that they naturally 'obtain', rather than exist and thus are just as abstract as facts, not being
 15




‘in’, but ‘at’ the world. States have a duration, though. This can be captured by taking states to
be obtained just from a property and an individual, with an additional condition specifying the
duration of a state, as in (34c):

(34) a. For a property P and an individual o, the state s(P, o) obtains just in case for some
time t, Pt(o).
b. Two states s(P, o), s(P’, o’) are identical in case P = P’ and o = o’.
c. A state s(P, o) obtains at a time t just in case Pt(o).

Given their abstract status, as entities implicitly defined by the conditions in (34), states will
fail to have a location, will not allow for particular manifestations, and are not possible
objects of perception. More precisely this holds for ‘Kimian states’, which I will now call
abstract states to distinguish them from concrete states.
Exist clearly classifies with abstract state verbs. Besides not permitting location modifiers,
exist does not allow for manner adverbials, unless exist is coerced into a ‘concrete way of
being’ reading (on which exist would also take location modifiers):12

(35) a.?? The man we talked about exists quietly / discretely / secretly.
b. ?? The animal exists peacefully in the forest.

Exist moreover cannot be used in an infinitival complement of perception verbs:

(36) * John saw the house exist for a long time.

The incompatibility of location modifiers with exist thus follows from exist taking abstract
states as event arguments. But what kinds of abstract states would these be, from what sort of
property would they be obtained, given the definition in (34)? The property on which the
abstract state argument of a stative verb should be based cannot be the property expressed by
the stative verb itself; otherwise there would be a circularity: the argument of the verb would
then in fact be dependent on the content of the verb itself. In the case of exist, the property
should be a time-relative property -- or a relation between properties and times, which holds































































12
See Maienborn (2007) for the possibility of coercion of state verbs into an eventive reading.
 16




of a time t and an object x in case at any subinterval of t (the whole of) x is present, as in
(37a). An event argument of exist relative to a time and an individual will then be as in (37b):

(37) a. P = λtλx[∀t’ < t AT(x, t’)]

b. For a time t, an event e and an individual x, <e, x>∈ [exist]t iff e = f(P(t), x)

We can now turn to occurrences. Occurrences may also seem abstract: they are not
qualitative, and as we have seen, do not permit manner adverbials or instrumentals. However,
occurrences, like basically all events, still can be located in space:

(38) John’s murder occurred / took place in Germany.

Moreover, they can be the object of perception:

(39) John saw the murder occur this morning.

Even if occurrences cannot have a range of qualitative features, they are still concrete, in the
sense of being 'in' the world, with a spatio-temporal location, perceivability, and causal
efficaciousness. Thus occurrences cannot be conceived of as abstract states, as being obtained
from properties and objects by abstraction. Occurrences need to be conceived of as
qualitatively 'thin' but still concrete events. I suggest that events in general should be
construed in terms of the notion of a concrete state, as transitions among concrete states.
Concrete states in turn may be conceived of as being composed of individuals and particular
features of those individuals. Qualitatively thick events will generally be transitions among
concrete states composed of individuals and qualitative features. By contrast, occurrences can
be viewed as transitions from compositions of event parts with the temporal feature ‘being at
a particular time’, to other compositions of this sort. If we take c to be the relevant
composition function, then the occurrence relation can be characterized as follows:

(40) For events e and e’ and a time t, <e, r’> ∈ [occur]t iff e = transit(c(e1, λx[AT(x, t1)),
c(e2, λx[AT(x, t2)), …) and e1, …, en are relevant temporal parts, with t1, …, tn as
their duration.
 17




Given this account, occurrences are as concrete as other events; they will just involve less
qualitatively specified concrete states.

5. Other modes of existence

The different kinds of verbs of existence with their restrictions to particular kinds of entities
support the old philosophical view according to which there are different modes of existence
for different kinds of entities. There is one other mode of existence that is interesting to
consider in relation to natural language, and that is that of dependent existence, recently
discussed by McDaniel (to appear a, b). Both properties and tropes according to Aristotle
involve a notion of ‘existence in’, that is, a two-place notion of existence that holds between
substances and dependent entities (on Aristotle’s view). Another form of dependent existence
involves entities like holes, entities constituted by the absence of material. Natural languages
appear in fact to display a particular construction expressing dependent existence, such as the
have-construction in English. The object position of have is, like there-sentences, subject to
the indefiniteness effect, and the have-construction can express both the particular relation
between a substance and a property or trope and that between a substance and a hole:

(41) a. John has the property of being happy.

b. John has intelligence / fear / joy.
(42) The box has a hole.

This suggests that the have-construction expresses dependent existence, whereas there-
sentences express location-bound existence (in the broadest sense). (Though of course the
have-construction is polysemous and in addition expresses other dependence relations than
ontological ones, such as possession and kinship (John has a car, John has a child)).

6. Existence statements with bare plurals and mass nouns

6.1. Bare plurals and mass nouns as kind-referring terms

Let us now turn to existence statements with bare plurals or mass nouns as subjects:

(43) a. Giraffes exist.

 18




b. White gold exists.

The general view in the philosophical literature is that such sentences have a fundamentally
different logical form from those with singular terms in that they are quantificational
sentences. That is, exist in such sentences would contribute to the expression of existential
quantification rather than acting as a predicate of individuals.
There is a range of evidence that shows that this view is mistaken. It appears that with the
verb exist bare plurals and mass nouns are not (or at least not generally) quantificational NPs,
expressing existential quantification. Rather they are kind terms in the sense of Carlson
(1977). Thus, giraffes in (43a) stands for the kind whose instances are particular giraffes and
white gold in (43b) for the kind whose instances are particular quantities of gold. With exist,
bare plurals and mass nouns are kind-referring terms just as with kind predicates such as rare,
widespread, or extinct, that is, instance-distribution predicates:

(44) a. Giraffes are rare.

b. Dinosaurs are extinct.

Carlson (1977) had argued that bare plurals and mass nouns are in fact always kind-
referring, but that they trigger different readings with different classes of predicates. There are
two classes of predicates that are particularly important in that respect: so-called stage-level or
episodic predicates on the one hand and individual-level or characterizing predicates on the
other hand (Carlson 1977, Kratzer 1995). Roughly speaking, episodic predicates express
properties perceived as temporary, and characterizing predicates properties perceived as
permanent (and thus in particular essential properties). Carlson (1977), who introduced the
distinction in the context of the semantics of bare plurals and mass nouns, called the first class
‘stage-level’ because they are predicates that when predicated of an individual can be taken to
be predicated of just a temporal stage of the individual, whereas individual-level predicates
were so-called because they can be understood only as predicates predicated of an individual
as such. The terms ‘episodic predicate’ and ‘characterizing predicate’ avoid making a
distinction between individuals and their temporal stages (Krifka et al. (1995)).
The two classes of predicates characteristically exhibit different readings with bare plurals
and mass nouns. With episodic predicates, bare plurals and mass nouns trigger existential
quantification over the instances of the kind, as in (45), whereas with characterizing
predicates, they trigger generic or universal quantification, as in (46):
 19




(45) a. Firemen are available.

b. Water is nearby.
(46) a. Apples are healthy.
b. Water is transparent.

The logical form of (45a) would thus be as in (47a) and the one of (45b) as in (47b), where I
is the instantiation relation and Gn is a suitable generic quantifier (cf. Krifka et al. 1995):

(47) a. ∃x (x I [firemen] & available(x))

b. Gn x (x I [apples]  healthy(x))

It is also generally held that bare plurals and mass nouns allow for a generic (habitual) reading
with episodic predicates; thus (45a) also has the reading: for any firemen x, x is available. By
contrast, characterizing predicates do not allow for an existential reading.
Carlson’s view according to which bare plurals and mass nouns are always kind-referring
is not universally accepted. More common in fact is the view that in addition to being kind-
referring, bare plurals and mass nouns may also have an interpretation on which they express
existential quantification (with episodic predicates), a view I will turn to further below.
However, whatever view one may adopt, the various arguments for kind reference do apply to
bare plurals and mass nouns with the predicate exist; thus establishing bare plurals and mass
nouns are kind-referring in that particular context.
First, definite anaphora behave with bare plurals differently from NPs that clearly express
existential quantification, such as three dinosaurs:

(48) a. Dinosaurs do not exist. But they once did exist.

b. Three dinosaurs do not exist. * But they (three dinosaurs or other) once did exist.

They in (48a) stands for the entire kind, the denotation of dinosaurs, but they in (48b) can only
stand for particular instances of the kind, which is unsuitable for the predicate once did exist.
Furthermore, bare plurals and mass nouns in exist-sentences do not take wide scope over
negation or other quantifiers:

(49) a. Dinosaurs do not exist anymore. (* for some dinosaurs x, x does not exist anymore)
 20




b. Two dinosaurs do not exist anymore. (ok: for two dinosaurs x, x does not exist
any more)

If dinosaurs in (49a) stands for a kind, then not can only deny the holding of the predicate of
the entire kind, not of just some instances. By contrast, two dinosaurs in (49b) can take scope
over not.
Another piece of evidence for kind reference with exist is that bare plurals and mass nouns
in the subject position of a sentence with exist can be modified by a relative clause whose
predicate is an instance-distribution predicate:

(50) Dinosaurs, which used to be widespread in Europe, do not exist anymore.

Further evidence comes from temporal modifiers. Still and no longer in (51a, b) are
understood so as to qualify the entire lifespan of the kind and not that of particular instances:

(51) a. Dolphins still exist.

b. Dinosaurs no longer exist.

The same point is made by aspectual predicates such as continue and cease:

(52) a. Dinosaurs continued to exist.

b. Dinosaurs ceased to exist.

Only the entire kind can be said to continue or cease to exist.

Yet another argument for kind reference with exist comes from location modifiers, which,
recall, are possible in exist-sentences with bare plurals or mass nouns as subjects. This is
again seen below:

(53) a. Giraffes exist only in Africa.

b. Syphilis does not exist in Europe anymore.

I will come to an explanation for the possibility of location modifiers on the basis of kind
reference at the end of this section.
 21




A final piece of evidence that bare plurals are kind-referring with exist is that singular
indefinite NPs and other existentially quantified NPs are significantly less natural as subjects
of exist-sentences:

(54) a. ? A giraffe exists.

b. ? Some giraffes exist.

Intuitively, (54a) is about the existence of a single giraffe, not about giraffes as such; similarly
(54b) states that for some x that are giraffes, x exist. It does not just say that there are giraffes
or that giraffes exist. That is, in (54a) and (54b), a and some and exist do not merge into a
single ‘existential quantifier’, but rather exist is predicated of entities that a giraffe or some
giraffes quantifies over. The logical form of (54a) would thus be as in (55a), whereas the
logical form of (54a) would be as in (55b), for the kind of giraffes k:

(55) a. ∃x(giraffe(x) & exist(x))

b. exist(k)

We can thus conclude that bare plurals and mass nouns in exist-sentences are kind terms.
The view that bare plurals or mass nouns in existence statements are kind-referring has almost
never been pursued in the philosophical literature, which generally takes those NPs to be
quantificational (somehow merging their semantic contribution with that of exist).13
There is some evidence that bare plurals and mass nouns may also sometimes act as
existential quantifiers even in exist-sentences. First of all, there is somewhat indirect evidence,
coming from languages such as French and Italian which do not have bare plurals or mass
nouns. In French and Italian, definite plurals and mass nouns are used as kind terms (les
giraffes ‘the giraffes’, l’or ‘Gold’), and NPs with de are used for existential quantification.
(Jean a acheté des livres ‘John has bought books’, Jean a bu du vin ‘John drank wine’). In
exist-sentences, often both options are possible:































































13


One exception is Geach (1968), who suggests that exist can apply to bare plurals as well as singular terms for
the same reason that a predicate like disappear can apply to the two kinds of terms:

(1) a. John disappeared.

b. Dinosaurs disappeared.

Disappear in (1b) is a typical kind predicate, an instance-distribution predicate, see the end of this section.
 22




(56) a. Les nombres naturels existent.

b. Des nombres naturels existent.
‘Natural numbers exist.’

However, when the nominal does not describe a ‘natural class’ or kind (that is, a maximal
collection of resembling particulars), the second option is better:

(57) Des nombres primes entre 10 et 15 / ?? Les nombres primes entre 10 et 15 existent.
‘Prime numbers between 10 and 15 exist.’

In that case, English would also use the bare plural, as indicated in the translation of (57).
There is also evidence from English that bare plurals in such contexts are existentially
quantified, rather than kind-referring, and that is that simple bare plurals such as natural
numbers and bare plurals such as prime numbers between 10 and 15 behave differently with
respect to anaphora support, as illustrated below:

(58) a. * Prime numbers between 10 and 15 exist. I did not think that they would exist.
b. Natural numbers exist; I did not think that they would exist.

Kind referring they as in the second sentences cannot take existentially quantified bare plurals
as antecedent. Thus, some bare plurals or mass nouns with exist are in fact used as
existentially quantified NPs, rather than as kind-referring, namely precisely those for which
the descriptive content of the NP would make kind reference implausible.14

6.2. Kind reference of bare plurals and mass nouns with exist

In general, exist holds of a kind (as denoted by a bare plural or mass noun) just in case there
are instances of the kind that ‘exist’. Existential quantification is clearly involved in cases like
the following:

(59) a. Electrons exist.































































14


See also Chierchia (1998) on conditions on bare plurals and mass nouns to be interpreted by existential
quantification rather than kind reference.
 23




b. Unicorns exist.
c. Prime numbers exist.
d. White gold exists.
e. True justice exists.

That is, a kind exists in virtue of there being actual instances.

The existential reading obviously classifies exist as an episodic predicate. Before
discussing why exist should classify as episodic, let us consider whether exist can also display
a generic reading, as would be expected for stage-level predicates. Such a reading generally
goes along with focusing the predicate, and indeed the reading seems to be possible with
suitable examples. Thus (59a) can also be read: ‘electrons really exist, and do not just have a
theoretical status in science’; the generic quantifiers here is naturally restricted contextually,
as roughly in ‘any electron predicted by theory really does exist;’
There is another class of examples that naturally display a universal reading with exist, and
these are mathematical examples. Mathematical examples figure prominently in a recent
paper by Fine (to appear), who points out that the statement in (60a) is intuitively stronger
than that in (60b):

(60) a. Integers exist.

b. Natural numbers exist.

(60a) appears to claim the existence of every integer, not just some integer (which may in fact
be a natural number). By contrast, (60b) claims the existence of only every natural number.
However, on the view on which existence statements express existential quantification, the
converse holds: (60b) would make a stronger statement than (60a).
Fine’s use of such examples was to make a general point about ontological commitment,
namely that statements of ontological commitment in general involve universal quantification
and thus require exist to act as a predicate roughly equivalent to real. If there are statements of
ontological commitment that are universally quantified, as (60a) and (60b) appear to be, then,
as Fine argues, statements of ontological commitment in fact involve a domain of entities
which may or may not be real and state which ones in them exist, that is, are real. For that
reason, exist has to be a predicate, in fact a predicate roughly synonymous with real.
 24




Statements of ontological commitment to kinds thus presuppose a domain of ‘light entities’

and involve predication of a property of existence of such entities.15 The logical form of (60a),
Fine argues, is then as in (61a), or equivalently as in (61b):

(61) a. For every x (integer(x)  x exist)

b. For every x (integer(x)  real(x))

(Later I will argue that the semantics of real is in fact fundamentally different from that of
exist, and that it in particular triggers a different reading with bare plurals.)
It does not seem that the universal reading of the examples in (61) is exactly the same
phenomenon as the generic reading which exist as an episodic predicate is predicted to
display. The universal reading displayed by (61a, b) corresponds to an entirely neutral
intonation and moreover, it appears to be the only reading available. In such examples, the
universal reading in fact seems to be a secondary effect of what the sentences actually mean, a
meaning that would involve existential quantification, not universal quantification. The
universal reading in (60a) and (60b) appears to be a pragmatic effect of the particular kind of
entity the bare plural stands for: in the case of clearly defined mathematical sequences,
accepting one simply means accepting all. Note that the same effect seems to be displayed by
the corresponding there-sentences:

(62) a. There really are integers.

b. There really are natural numbers.

Also (62a) seems to make a stronger statement than (62b).16































































15
Exist also occurs with definite plurals:

(1) The integers exist.

Such sentences have only a universal reading, on which (1) is equivalent to (2):

(2) Every integer exists.

But the logical form of (1) does not consist in universal quantification. Rather it involves plural reference and an
obligatorily distributive reading of exist:

(3) a. For a plurality x, exists(x) iff for every member y of x, exist(y).

b. A predicate P is obligatorily distributive iff for any plurality x, P(x) iff for any member y of x, P(y).

That is, (1) consists in a plural description and a predicate that when applied to the plurality denoted by the
description automatically applies to each instance.
 25




The universal-quantification effect shows up also with certain kinds of mathematical sets
or classes:

(63) a. Geometrical figures exist.

b. Triangles exist.
c. Equilateral triangles exist.

The universal effect is somewhat less obvious in (63b) and still less obvious in (63c). The
reason is that (63c), and perhaps (63b), does not answer a sufficiently plausible question of
mathematical ontology concerning the ontological nature of a geometrical form, but rather
asks for an example of a geometrical form with a particular specification. The universal-
quantification effect in (63a, b) should thus be traced to a pragmatic implication rather than
the logical form of the sentences themselves.17
Fine certainly is wrong in maintaining that exist always triggers a universal reading with
bare plurals. With non-mathematical kinds, an existential reading is clearly the natural
reading, whereas with mathematical ‘kinds’ a universal reading appears to be a pragmatic
effect rather than a matter of interpretation. What then about the generic reading that is
available with exist? Such a reading does seem to be of the kind Fine envisioned: here a
domain of ‘light’ entities in the domain of discourse is presupposed and exist states that they
really exist.































































































































































































16


The universal quantification effect is particularly strong if the sentence states that a kind exists in virtue of
there being an instance. If the sentence explicitly quantifies over instances only and claims that at least one
exists, the universal quantification effect is less strong. Thus, the sentences below do not differ in the way (40a,
b) do.

(1) a. Some integer exists.

b. Some natural number exists.


17
The universal-quantification effect shows up also with believe in, a predicate expressing objectual ontological
commitment (Szabo 2003):

(1) a. John believes in unicorns.

b. John believes in integers.
c. John believes in natural numbers.

According to (1a), John is right in his belief if there are some unicorns, whereas for John to be right in his
belief according to (1b) and (1c) all integers / natural numbers need to exist.
Note, though, that believe in does not always require the existence of all or some instances of the kind for the
agent to be right in his belief. Given (2), John can be right in his belief even if there is in fact no actual instance
of true justice:

(2) John believes in true justice.



 26




By displaying an existential reading as well as, with some effort, a generic reading, exist
classifies as an episodic predicate. The question then is, what sense can one make of the
classification of exist as an episodic predicate.

6.3. The individual-stage level distinction and exist

The characterizing-episodic distinction is known to be a notoriously problematic distinction.

There are many predicates that seem to express temporary properties but that do not trigger an
existential reading of bare plurals. Nervous, happy, or sick are examples, which contrast thus
with predicates like available, audible, or visible, which do trigger an existential reading.
Moreover, many predicates that one might classify as characterizing predicates may trigger
existential readings at least in object position, such as belong, own, contain, and include.
Whether such verbs should be considered characterizing or episodic, clearly the object
position of such predicates goes along with an existential reading, whereas the subject
position goes along with a generic reading:

(64) a. Very rich men own expensive cars.

b. Shells contain pearls.

In fact, the existential reading that many episodic predicates seem to trigger does not seem
tied to a kind-referring use of a bare plural, but rather an existentially quantified use. This is
particularly clear when using the anaphora test for kind reference. The anaphora test indicates
that with ‘standard’ episodic predicates such as buy or disappear no kind reference is
involved:

(65) a. ?? John bought apples. Mary bought them too.

b. ?? Gold coins have disappeared. They have never disappeared before.

This also holds for the ‘paradigmatic’ episodic adjectives available and visible:

(66) a. ? Firemen are available. They were not available yesterday.

b. ? Stars are visible in the sky. They were not visible yesterday night.
 27




In fact, it has been observed that the existential reading of such examples goes along with
stress on the subject, which is indicative of an interpretation as existential quantification
rather than kind reference (whereas with exist stress was on the predicate) (cf. Krifka et al.
1995).
The same observation about anaphora can be made for the object position of own and
contain:

(67) a. ?? Very rich men own expensive cars, though 200 years ago they could not own them.
b. Shells contain pearls. ?? But actually they do not always contain them.

Yet there are also cases of episodic predicates that involve kind reference and trigger an
existential reading, such as taste and live from:

(68) a. Mary has tasted black beans. Joe has tasted them too.
b. Mary lived from black beans. Joe lived from them too.

Thus, existential readings with kind-referring bare plurals appear significantly more limited
than is often assumed, and it is just one of a range of lexically specific readings that predicates
may display with kind-referring bare plurals. Yet there is, in such a more restricted range of
cases, a correlation between episodic predicates and an interpretation of the bare plural
involving existential quantification.
But how can exist classify as an episodic predicate when its application obviously is never
limited to a temporal stage of an individual, but applies to the individual’s entire life-span?
There is one obvious criterion classifying exist as an episodic predicate in the case of concrete
individuals, and that is that exist expresses an accidental property.18 Concrete individuals must
have the essential properties they have, but they need not have existed. Identifying episodic
predicates with those expressing accidental properties and characterizing ones with those
expressing essential properties is not unproblematic, though. Existence could not be an































































18


Another criterion may take into account the entire period during which a predicate could be true of an
individual, including the time after the individual’s existence, such as the time of the individual’s influence. Exist
then comes out a predicate that may apply to only a subinterval of such a period. But also predicates expressing
essential properties, except for those relating to the individual’s influence (for example, is a philosopher), would
come out as episodic.


 28




accidental property of abstract objects (such as mathematical objects), which exist necessarily
(if they exist).19 20
However, it appears that in general a predicate is classified as either episodic or
characterizing not on the basis of whether the property it expresses is an essential property of
the object in question or not. Rather a predicate appears to be classified as episodic or
characterizing on the basis of certain kinds of, let’s say, primary objects to which it applies
and then the classification is carried over to other objects to which the predicate may also
apply, possibly with an extended meaning. For example, colors are properties that are
essential with some entities (paint), but not others (tables), yet bare plurals and mass nouns
always display a universal, not an existential reading with color adjectives. Thus (69a) and
(69b) could not be true in virtue of an existential reading; they are false because only a
universal reading is available:

(69) a. Paint is red.

b. Tables are red.

Red always classifies as a characterizing predicate.

In general, then, predicates classify either as episodic or characterizing as such, not when
applied to one sort of object as opposed to another. Predicates appear to decide whether they
are characterizing or episodic on the basis of one type of object and then carry over their
classification also when applied to other types of objects. For red the type of object on the
basis of which the classification as episodic or characterizing is made are essentially red
objects, for exist it is concrete objects that exist only accidentally, not abstract objects.

6.4. The formal semantics of existence statements with bare plurals or mass nouns

Turning now to the logical form of exist-sentences with bare plurals or mass nouns, it appears
that they do not express simple existential quantification over instances with the predicate
exist applying to some instances; that is, (70a) is not quite equivalent to (70b):






























































19


There are also, it has been argued, abstract objects that exist only accidentally, namely fictional characters
conceived of as abstract artifacts (cf. Thomasson 1999). Laws and institutions may be other cases.


20


One might argue that that exist classifies as a characterizing predicate with abstract objects and that this is in
fact the source of the universally quantified reading that exist exhibits with mathematical objects. However, I
found that speakers generally have the intuition of a primarily existential reading, with the universal
understanding being a secondary effect. Moreover, predicates in general do not seem to change their
classification as episodic or characterizing depending on the type of object they apply to (see below).
 29




(70) a. Electrons exist.

b. Some electrons exist.

(70a) claims the existence of a kind and not just the existence of some individual instance –
though a kind ‘exists’ in virtue of there being instances.
That the kind-related application of an episodic predicate is not entirely reducible to the
application of the very same predicate to some instance can be observed with other predicates
as well, such as discover or recognize:

(71) a. Joe discovered white gold -- in virtue of coming across some instances.
b. John recognized tuberculosis (by examining an instance of it).

Given this observation and the fact that the existential reading of kind predicates with
episodic predicates is much more restricted than generally supposed, it is not surprising that
the kind-related interpretation of exist is not derived by simple existential quantification over
instances. Rather, the kind-related interpretation may be derived in some other way, as should
be the case for episodic predicates in general.
To identify the kind-related interpretation of exist, let us recall that spatial modifiers are
possible in exist-sentences with kind terms:

(72) a. Giraffes exist only in Africa.

b. Syphilis does not exist in Europe anymore.

In such cases it is also obvious that exist cannot just apply to the kind in virtue of applying,
with its primary meaning, to an instance of the kind: exist does not apply to particulars with a
spatial modifier. With spatial modifiers, the derived meaning of exist is not that of an episodic
predicate, but rather that of an instance-distribution predicate. Instance-distribution predicates,
for example widespread, rare, or extinct, express properties concerning the spatial
distribution, and as such they allow for spatial restrictions:

(73) a. Ants are widespread in Europe.

b. Dinosaurs are extinct.
 30




It appears that predicates in general may obtain a meaning extended to kinds which is that of
an instance-distribution predicate, rather by generic quantification over instances (for
characterizing predicates) or existential quantification over instances (for episodic predicates).
Another such case is the verb disappear in English. As noted by Geach (1968), disappear is a
predicate that also applies to individuals and kinds:

(74) a. John has disappeared.

b. Dinosaurs have disappeared.

Disappear should be an episodic predicate. But with a bare plural as in (74b), it does not have
a derived meaning consisting of existential quantification over instances and application of
disappear (or a closely related condition) to them, but rather it expresses a condition on the
distribution of the instances as a whole, namely that there are no instances at all.21 In the case
of disappear, the kind interpretation is obviously derived by applying the ordinary lexical
meaning of disappear to the entire collection of the actual instances
The kind-related interpretation of exist can obviously not be obtained by applying exist
with its primary meaning to the kind. Rather the kind-related interpretation appears to be
obtained from the locational secondary meaning of exist. Using the secondary meaning, the
kind-related meaning of exist can be given as below, where I is the instantiation relation:

(75) a. For a kind k and a location l, existk(k, l) iff for some i, i I k, exist2(i, l).
b. For a kind k, existk(k) iff for some i, i I k, exist2(i).

If bare plurals and mass nouns are kind-referring in exist-sentences, then it is clear why
temporal modifiers and aspectual predicates apply in the way they do when exist applies to a
kind term. Temporal modifiers and aspectual predicates shift the time of evaluation for exist






























































21


Having a derived kind meaning as an instance-distribution predicate is a lexically idiosyncratic property of
particular episodic predicates in particular languages. For example, in German, verschwinden ‘disappear’ fails to
act as an instance-distribution predicate with bare plurals:

(1) Dinosaurier sind verschwunden.

‘Dinosaurs have disappeared.’

(1) has only the reading on which for some dinosaurs x, x has disappeared. It cannot mean that dinosaurs are
extinct. English disappear also displays the derived meaning of an episodic predicate in sentences such as:

(2) John gets upset when books disappear.



 31




as a kind predicate, allowing at each time different instances of the kind to satisfy the relevant
condition.

7. The adjective real

Exist, we have seen, is a predicate of individuals that holds of an individual if the individual is
a presently persisting one, and it is false of past and merely possible objects as well as
intentional objects. Existence thus is not a feature of certain entities, and the lack of existence
does not consist in the absence of such a feature, but rather it has to do with the relation of
entities to time and with their quasi-representational status. This is also evident when
comparing exist to the adjective real. The two expressions differ fundamentally linguistically
and in the kinds of ontological notions they involve.
One major linguistic difference between the predicate exist and the predicate real consist
in their different readings with bare plurals and mass nouns. Whereas exist generally triggers a
reading involving existential quantification, real goes along with universal quantification.
This is illustrated by the contrast between (76a) and (76b) as well as the contrast between
(77a) and (77b):

(76) a. Prime numbers exist.

b. Prime numbers are real.
(77) a. Electrons exist.
b. Electrons are real.

Obviously, real classifies as a characterizing predicate, not an episodic predicate like exist.
But how can real be a characterizing predicate and differ from exist? If real is contrasted
with the adjective possible, it is not obvious at all that it should classify as a characterizing
predicate. It would depend on the philosophical view. If what distinguishes a real from a
merely possible object is that real objects belong to the actual world whereas merely possible
objects belong to other possible worlds, real should be as episodic a predicate as a predicate
of location, or exist for that matter. But if real applies to entities whose nature is in some way
different from that of merely possible objects (the latter being conceptual entities of some
sort, let’s say), then real would be a characterizing predicate. At the same time, real should
not really apply to past object, but the nature of past objects is not different from the nature of
 32




presently existing objects. Thus, using philosophical considerations alone does not seem to
get us a clear classification of real.
A simpler explanation for the status of real as a characterizing predicate can be obtained
when taking a closer look at its linguistic behavior. Real is an adjective that is in fact not very
felicitous on its own. It naturally occurs only as a modifier of a sortal noun:

(78) a. A is a real object.

b. B is a real person.
c. C is a real witch.
d. D is a real solution.

In that function, real competes with ‘intensional’ adjectives such as intentional, fictional,
mythical, and possible, which also naturally occur as modifiers of sortal nouns:

(79) a. A is a fictional person.

b. B is a mythical dragon.
c. C is an intentional object.
d. D is a possible house.

Some of those adjectives appear to have the semantic status of operators: if an object x is a
‘real object’, this means that really (in reality), x is an object; when an object x is a ‘fictional
person’, this means that in some piece of fiction, x is a person; when an object x is a ‘mythical
dragon’, this means that according to some myth, x is a dragon. In the last two cases, x must
be an object that has the status of a fictional character: the operator here serves to permit
ascription of properties as attributed in the story or myth.
Intentional and possible are better analysed not as operators but as intensional functors
mapping a sortal noun onto an intentional or possible object. If x is an intentional object, then
x is constituted by intentionality alone; x is not such that it is ‘intended’ to be an object. If x is
a possible house, then x is not something such that it is possibly a house; rather, x belongs to
some possible world in which it is a house.
The intensional-functor analysis in fact can be applied to (79a) and (79b) as well, thus
providing a uniform semantics of intensional adjectives. Mythical would be a functor mapping
a sortal N to a mythical character x constituted by some myth, such that according to that
myth x has N, and similarly for fictional.
 33




The explanation of the status of real as a characterizing predicate is then obvious: real in
general is followed by an explicit or silent sortal noun N and the resulting complex predicate
a real N is a characterizing predicate because N is: sortal nouns are always characterizing
predicates.22
The difference between exist and real is also reflected in the corresponding
nominalizations: in (80a), the existentially quantified complement of reality takes obligatory
wide scope, whereas in (80b) the existentially quantified complement of existence can take
narrow scope:23

(80) a. John denies the reality of a witch.

b. John denies the existence of a witch.

(80a) has only the reading on which for some witch x, John denies the reality of x, and (80b)
has only the reading on which John denies that a witch exists. In (80a), the denial targets a
feature of a given witch, that of its reality; in (80b), the denial focuses on there being a witch,
not on a given feature that a witch may or may not have. ‘Reality’ is something only an entity
can have that already has some form of being, whereas ‘existence’ is something that goes
along with the entity itself.24 The reality of a witch acts as a term referring to a feature of an
individual whose existence is presupposed. In that respect it is just like any other term
referring to a ‘feature’ or ‘trope’, for example the qualification of a candidate, which
contrasts in the same way with the existence of a candidate below:































































22


What if N is a phase sortal, such as teacher or passenger? In real teacher and real passenger, real obviously
targets the ‘phase content’ of the noun, not the sortal content (a real teacher is a person that really is a teacher,
not someone that really is a person and teaches). Thus real would rather make an ‘episodic’ contribution to the
complex nominal. Nonetheless real teacher and real passenger classify as characterizing predicates in that they
trigger generic quantification with a bare plural:

(1) Employees of this organization are teachers.

The characterizing status of real teacher or real passenger thus should be traced to the sortal content of the head
noun, not the contribution of real.


23


There is no obligatory wide scope of the complement with the bare plural, though, as was pointed out to me
by Richard Kayne:

(1) John denies the reality of witches.

24


Apparently, for Avicenna existence was an accidental ‘feature’ of entities, that is, a trope. Entities for
Avicenna are possibilia defined by their ‘essence’.




 34




(81) a. John denies the qualification of a candidate.

b. John denies the existence of a candidate.

Thus, existence and reality are fundamentally different notions, or so natural language tells us.

8. Conclusion

The predicate exist has puzzled philosophers for a long time. The purpose of this paper was
to indicate how much linguistic semantics can contribute to illuminate the semantic behavior
of exist as well as other existence predicates. The semantic analysis of exist (and other
existence predicates) that was proposed explains the various semantic peculiarities of exist as
quite systematic, as falling within independently established generalizations about predicates
of particular sorts.

Appendix:
The linguistic plausibility of views on which exist is not a first-order extensional
predicate

1. Exist as a second-order predicate

For Frege exist is a second-order predicate: it takes a predicate or concept-denoting expression

and states that its extension is nonempty. Thus, in (1), exist would state that the extension of
king of France is nonempty:25

(1) The king of France exists.

This view has been criticized extensively by philosophers (Salmon 1987, Miller 1986). I will
restrict myself here simply to the question of the linguistic plausibility of the view. For the
view to be plausible linguistically, there should be independent motivations that a predicate
such as exist can go along with a concept-denoting expression as subject and moreover that
definite descriptions can act as concept-denoting. While a function of definite descriptions as






























































25


Similarly, Russell takes exist to be a predicate of ‘propositional functions’.
 35




predicates and thus as concept-denoting is plausible (Graff 2001), proper names do not seem
to generally be able to be reinterpreted that way.
Turning then to the predicate exist, the question to ask is: are there other second-order
predicates in natural language that go together with concept-denoting subjects or
complements, and does exist behave like those predicates? A second-order predicate in natural
language would be a predicate that requires predicates as complement or subject, that is,
adjectives or predicative NPs. There is an important class of predicates in natural language
that act that way, namely copula verbs:

(2) John became a man.

The question thus is, does the subject of a sentence with exist have the same predicative status
as the complement of a copula verb? The answer is clearly negative. Exist does not occur, on
the relevant interpretation, with complements that could qualify as predicates, such as
indefinite singular NPs. Thus (3a) does not just claim the nonempty extension of the concept
denoted by man, in the way (3b) and (3c) do:

(3) a. A man exists.

b. Men exist.
c. There is a man.

To the philosophical objections that have been raised against the analysis of exist as a second-
order predicate with singular terms, we can thus add that exist as a second-order predicate is
implausible also linguistically.

2. Exist as an intensional predicate

The second view takes exist to be an intensional predicate. That is, exist as in (1) would apply
to the intension of the subject, let’s say an individual concept, a (partial) function from
possible worlds to individuals:

(4) The king of France exists.

 36




Exist on this account is true of an individual concept I at a time t in a world w iff I is defined
at t in w. The question then is, does exist really classify as an intensional predicate?
When it comes to the notion of an intensional verb, different types of intensional or
apparent intensional verbs need to be distinguished. I will make use of distinctions introduced
in Moltmann (1997, 2008).
One type of intensional verb, verbs like resemble or compare to, simply takes predicative
complements. Such verbs are thus second-order predicates of another sort than copula verbs,
and we have seen that exist does not pattern with those.
The second type of intensional verb includes need, look for, recognize, and own. Such
verbs are characterized by a certain nonspecific reading of the complement and a general
restriction to weak quantifiers as complement. The latter constraint manifests itself in that
even if uniqueness were to be fulfilled (in the relevant worlds), an indefinite is required for
the intensional reading (Moltmann 1997):

(5) a. John needs a wife.

b. ?? John needs his wife.
(6) a. The institute needs a director.
b. ?? The institute needs the director.

Obviously, exist is not subject to the restriction to weak quantifiers.

Another peculiarity of intensional verbs of the second type (shared by the first type in fact)
is the possibility of the complement being substituted by a ‘special quantifier’, a quantifier
such as something, nothing, or the pronoun that (Moltmann 1997, 2008). Special quantifiers
are neutral in gender, and they can replace the complement of an intensional verb even if that
complement is not neutral:

(7) a. John needs something / * someone – a wife.

b. Something / * someone is urgently needed – a plumber.

It is easy to see that exist does not pattern with intensional verbs in that respect. The subject of
an exist–sentence cannot generally be replaced by a special quantifier, namely not when the
subject fails to be neutral:

(8) * Something / Someone does not exist anymore -- the king of France.
 37




Thus, exist does not classify as an intensional verb like need.

There is another type of potential intensional verbs that exist is more likely to pattern with,
namely verbs that have been taken to apply to individual concepts, as denoted by certain
definite NPs (Montague 1973). Thus, the predicates change and is rising have been taken to
apply to individual concepts in the sentences below:26

(9) a. The director changed.

b. The temperature is rising. (Montague 1973)
c. The president is elected every four years.

However, exist cannot apply to the kinds of entities that change, is rising, and is elected every
four years can apply to:

(10) a. ?? The director always exists. (meaning ‘there is always some director or other’)
b. ?? The president now exists; he did not exist for a few years.
c. ?? The director exists, though he did not exist before the election a few days ago.

There are some cases where exist appears to apply to the intension of an NP, though in fact
even in those cases exist applies in fact to something else. First, exist apparently applies to the
intension of the complement of an intensional verb like nee, ask for, and see in cases such as
the following:

(11) a. What John needs exists.































































26


It is not very clear, though, that these are truly cases of intensional verbs, that is, verbs that apply to the
intension rather than the extension of an argument. The subject of sentences with such verbs as predicate cannot
generally be replaced by a special quantifier. Thus (1b) as a continuation of (1a) is infelicitous:

(1) a. The president is elected every four years.

b. Something is elected every four years.

Moreover, replacement of the subject by the special pronoun that is not acceptable, but only replacement by
the non-neutral pronoun he is:

(2) The president is elected every four years. He / * That is not elected every three years.

This is an indication that the subject in (9a) – (9c), rather than denoting its intension, in fact stands for a
particular type of entity, an entity with variable manifestations at different times (such as different people). The
verbs would then be extensional predicates with a derived meaning applying to entities with variable
manifestations.


 38




b. Everything John asked for exists.

c. What John saw exists.

There is evidence that the subject in (11a-c) in fact does not stand for the intension of an NP,
but rather for a kind, that is, what a bare plural or mass noun stands for. Thus, while (12b) as a
continuation of (12a) is perfectly fine, (12c) is not. It is just as bad as (12d), whereas (12b) is
just as good as (12e):

(12) a. John is looking for a tiger.

b. What John is looking for exists.
c ?? What John is looking for exists, namely a tiger.
d. ?? A tiger exists.
e. Tigers exist.

A singular indefinite is much less acceptable with exist than the bare plural. As Carlson
(1977) argued, when an intensional verb takes a bare plural or mass noun as complement, then
in fact the verb does not take an intension as argument, but rather a kind, as denoted by the
bare plural or mass noun. Thus, in (11a) what John needs better stands for a kind rather than
an intension, and so it is the kind and not the intension that exist applies to. Thus, exist in
(12b) applies with the meaning with which it applies in (12e), in an existence claim about a
kind.
There is another type of relative clause with intensional verbs that one might think requires
exist to apply to an individual concept. In this case, I will argue, exist in fact applies to the
corresponding intentional object, not an individual concept. The construction in question
consists in definite descriptions formed with a relative clause containing either a copula verb
as in (13a) (Grosu / Krifka 2008) or an intensional verb, as in (13b) (Moltmann 2008):

(13) a. The gifted mathematician John claims to be could solve this problem in five minutes.
b. The assistant John needs must speak French.

As Grosu/ Krifka (2008) argue for (13a) and Moltmann (2008) for (13b), these definite
descriptions are best considered definite descriptions referring to individual concepts, partial
functions from possible worlds or situations to individuals. In (13a), this partial function maps
a situation s onto John if John is a gifted mathematician in s, and it is undefined otherwise. In
 39




(13b), it maps a situation s onto a person d such that d is an assistant of John’s and s is a
satisfaction situation of John’s needs. Given that the individual concepts are partial functions,
possibly undefined for the actual circumstance, they require a suitable modal attributing a
property to an object in a circumstance for which the function is defined and that is accessible
for the modal. This modal compatibility requirement is satisfied by the presence of could in
(13a) and must in (13b) (note that the sentences would be unacceptable without the modal).
Definite NPs with intensional relative clauses as in (13a) and (13b) might seem good
candidates for cases of reference to and predication of individual concepts. But again, it
appears that exist is not truly applicable to such individual concepts, and when exist seems to
apply to them, we do not in fact have reference to an individual concept. To see this, let us
contrast the definite descriptions in (14) to another type of definite description that can be
formed with an intensional verb. This type of definite description, crucially, is not subject to
the modal compatibility requirement:

(14) a. The ideal woman John wants to marry is tall, blonde, and intelligent.
b. The mountain John is looking for is golden.

There is also a difference in how (13a,b) and (14a,b) are derived; (13a,b) are derived from
(15a,b), with an indefinite as complement:

(15) a. John claims to be an x-much gifted mathematician.

b. John needs an assistant that speaks French.

By contrast (14a) and (14b) are derived from sentences in which the intensional verb,
exceptionally, takes a definite NP as complement:

(16) a. John wants to marry the ideal woman.

b. John is looking for the golden mountain.

Thus, such NPs are not subject to the restriction to weak NPs imposed by intensional verbs.
This indicates that the denotations of the definite NPs in (14a) and (14b) are fundamentally
different from those of the definite NPs in (13), namely that they are intentional objects, rather
than individual concepts. Intentional objects depend on particular intentional acts of attempted
 40




or pretended reference, whereas individual concepts, like possible objects, are mind-
independent.
The observation that is crucial in the present context is that exist applies much better to the
definite NPs in (17) than to those in (18):

(17) a. The mountain John is looking for does not exist.

b. The ideal woman does not exist.
(18) a. ? The gifted mathematician John claims to be does not exist.
b. ?? The assistant John needs does not exist.

The explanation of this difference is that exist can apply to intentional objects but not to
individual concepts.
Obviously there are conditions on when the denotation of a definite NP can be an
intentional object: either some particular agent has conceived of the object in a particular way
(rather than having a generic attitude satisfiable by anything fulfilling a particular condition),
or the object has been established as an intentional object via a historical chain. In the latter
case, the intentional object in fact corresponds to a relatively well-established mythical object.
The latter may exist, the former does not. Definite ‘non-referring’ NPs thus allow for a
systematic switch from standing for an intentional object to standing for a corresponding
fictional or mythical object.

References

Austin, J. L. (1979): ‘Unfair to Facts’. Philosophical Papers. Oxford University Press, New
York.
Carlson, G. (1977): ‘A Unified Analysis of the English Bare Plural’. Linguistics and
Philosophy, 413-457.
Carlson, G. / J. Pelletier (1995): The Generic Book. Chicago UP, Chicago.
Chierchia, G. (1998): ‘Reference to Kinds across Languages’. Natural Language Semantics 6,
339-405.
Fine, Kit (2009): ‘The Question of Ontology’. In D. Chalmers et al. (eds.): Meta-Metaphysics,
OUP.
Frege, G.: Grundlagen der Arithmetik.
Geach, P. (1968): ‘What Actually Exists’. Proceedings of the Aristotelian Society, Supp. Vol.
 41




42, 7-16.
Grosu, A. / M. Krifka (2007): ‘The Gifted Mathematician that You Claim to Be’. Linguistics
and Philosophy 30, 445-485.
Higginbotham, J. (1987): 'Indefiniteness and Predication'. In E. J. Reuland / A. ter Meulen
(eds.): The Representation of (In)definiteness. MIT Press, Cambride, 43-70.
van Inwagen, P. (1998): 'Meta-ontology'. Erkenntnis 48, 233-250.
------------------ (2008): ‘McGinn on Existence’. Philosophical Quarterly 58, 36-58.
Katz, G. (2003): ‘Event Arguments, Adverb Selection, and the Stative Adverb Gap’. In E.
Lang / C. Maienborn / C. Fabricius-Hansen (eds.): Modifying Adjuncts. De Gruyter,
455-474.
Kratzer, A. (1995): ’Stage-Level and Individual-Level Predicates‘. In Carlson / Pelletier
(eds.), 125-176.
Krifka, M. et al. (1995): ‘Genericity: An Introduction‘. In Carlson / Pelletier (eds.).
McGinn, C. (2000): Logical Properties. Oxford UP, Oxford.
Maienborn, C. (2001): ‘On the Position and Interpretation of Locative Modifiers’. Natural
Language Semantics 9.2, 191-240.
----------------- (2007): ‘On Davidsonain and Kimian States’. In I. Comorovski / K. von
Heusinger (eds.): Existence: Semantics and Syntax. Springer, 107-130.
McDaniel, K. (to appear a): ‘A Return to the Analogy of Being’. Philosophy and
Phenomenological Research.
----------------- (to appear b): ‘Being and Almost Nothingness’. Nous.
Miller, B. (1975):’In Defense of the Predicate Exist’. Mind 84, 338-354.
------------ (1986): ‘Exists’ and Existence’. The Review of Metaphysics 40, 237-270.
------------ (2002): ‘Existence’. Stanford Encyclopedia of Phhilosophy.
Moltmann, F. (1997): ‘Intensional Verbs and Quantifiers’. Natural Language Semantics 5.1.,
1-52.
---------------- (2008): ‘Intensional Verbs and their Intentional Objects’. Natural Language
Semantics 16.3., 257-281.
Parsons, T. (1980): Nonexistent Objects. Yale UP, New Haven.
Priest, G. (2005): Towards Nonbeing. Oxford UP, Oxford.
Russell, B. (1905): ‘On Denoting’. Reprinted in R.C. Marsh (ed.): Logic and Knowledge.
Allen & Unwin, London.
Salmon, N. (1987): ‘Existence’. Philosophical Perspectives 1, 49-108.
------------- (1998): ‘Nonexistence’. Nous 32.3., 277-319.
 42




Sider, T. (2001): Four-Dimensionalism. Clarendon Press, Oxford.

Strawson, P. (1950): ‘Truth’, Proceedings of the Aristotelian Society, reprinted in Strawson
1971, Logico-Linguistic Papers, Methuen, London, 1971.
-------------- (1959): Individuals. An Essay in Descriptive Metaphysics. Methuen, London.
Szabo, Z. (2003): ‘Believing in Things’. Philosophy and Phenomenological Research 66,
586-611.
Thomasson, A. (1999): Fiction and Metaphysics. Cambridge UP, Cambridge.
Williams, E. (1984): ‘There-Insertion’. Linguistic Inquiry 15, 131-153.
Zamparelli, E. (1992): ‘Definite and Bare Kind-Denoting Noun Phrases’. In C. Beyssade et al.
(eds.): Romance Languages and Linguistic Theory 2000, Benjamins, Amsterdam, 305-42.
 43



 44