Synthese (2009) 168:81–96

DOI 10.1007/s11229-008-9314-x

Indeterminacy, identity and counterparts: Evans

reconsidered

Elizabeth Barnes

Received: 26 April 2007 / Accepted: 19 February 2008 / Published online: 9 May 2008
© Springer Science+Business Media B.V. 2008

Abstract In this paper I argue that Gareth Evans’ famous proof of the impossibility
of de re indeterminate identity fails on a counterpart-theoretic interpretation of the
determinacy operators. I attempt to motivate a counterpart-theoretic reading of the
determinacy operators and then show that, understood counterpart-theoretically,
Evans’ argument is straightforwardly invalid.

Keywords Vague identity · Ontic vagueness · Indeterminacy · Counterpart theory

Gareth Evans’ famous (1978) argument against the possibility of ontic vagueness is
one of those philosophical problems that just will not die. Weatherson (2003) claims
it’s a lynchpin in the case against de re vagueness, whereas Noonan (2004) remarks
that everyone knows that all the argument shows is that every vague object is deter-
minately distinct from every precise one (and every other vague one). Obviously, it’s
the sort of thing that philosophers tend not to agree about. The Evans argument is thus
still very much a ‘live’ issue in the ontic vagueness debate, one which anyone who
wants to go in for ontic vagueness must have something to say about.
With that in mind, I’d like to propose a solution to the Evans argument for the
would-be defender of ontic vagueness. I certainly do not intend this to be read as a
knock-down objection to Evans’ argument—far from it. But I put this solution forward
because it’s a simple and straightforward solution to Evans’ puzzle that, as far as I
know, has been largely unexplored. My contention will be that, should you be tempted
by an ontology which commits you to vague or indeterminate identities, there is a
reading of the determinacy operators available to you according to which the Evans
argument fails for quite basic reasons.

E. Barnes (B)
University of Leeds, Woodhouse Lane, Leeds, LS2 9JT, UK
e-mail: [email protected]

123
82 Synthese (2009) 168:81–96

My proposal, which I’ll outline below, is that the determinacy operators should be
understood as quantifiers that range over precisifications in the same way that modal
operators are quantifiers that range over possible worlds, and that claims of ‘trans-
precisificational identity’ should be treated much like a counterpart theorist in the
mould of Lewis (1971) treats claims of trans-world identity. On this understanding, I
argue, the Evans argument fails, for quite straightforward reasons.

1 The Evans argument

Gareth Evans, in his seminal paper “Can There Be Vague Objects?” offers a pur-
ported proof against the existence of ontic vagueness. The proof hinges on the idea
that whenever there are vague objects, there will be vague identity.1 But vague iden-
tity (as in genuine vague identity between entities, rather than merely vague identity
statements),2 the argument attempts to show, is incoherent. Thus, Evans concludes,
there cannot be vague objects.
The proof, as Evans presents it, runs as follows:
(1) ∇(a = b)
(2) λ[∇(x = a)]b
(3) ∼ ∇(a = a)
(4) ∼ λ[∇(x = a)]a
(5) ∼ (a = b) [from (2) and (4), by Leibniz’s Law]
This is not a straightforward contradiction, but Evans then reasons that “if ‘Indefi-
nitely’ and its dual ‘Definitely’ (‘’) generate a modal logic as strong as S5, then
(1)–(4) and presumably Leibniz’s Law can each be strengthened with a ‘Definitely’
prefix, enabling us to derive
(5 )  ∼ (a = b)
which is straightforwardly inconsistent with (1).”
Thus, if Evans’ reasoning is correct, we have a logical proof that there cannot be
ontic vagueness.

2 The logic of determinacy

It seems clear, though, that to assess whether or not Evans’ argument is valid, we
must first provide an adequate analysis of ‘determinately’ and its converse ‘indeter-
minately’. Evans’ proof hinges on the use of these operators, and thus any judgment
of its validity will depend on how we interpret them. In the subsequent sections I
will attempt to motivate an interpretation of the determinately/indeterminately opera-
tors—one which I’ll argue is very natural—on which the Evans argument does not go
through.

1 Many philosophers have disputed this claim (see, for example, Williamson 1994), but I will grant Evans
his assumption for the sake of argument.
2 See Lewis (1988).

123
Synthese (2009) 168:81–96 83

2.1 Precisificaitons

It’s a common assumption3 in the literature that the determinacy operators have a
semantics that mimics that of the modal operators. For example, the family of posi-
tions broadly known as supervaluationism maintain that the determinacy operators
should be read as a type of pseudo-modal operator that ranges over precisifications
rather than (as the modal operators do) worlds.4
According to this model, it’s quite easy to provide for the analogy between modal
semantics and determinacy semantics. Familiarly, on a possible worlds semantics ‘nec-
essary’ simply means true in all (accessible) possible worlds (
P iff: P obtains at every
accessible world). Its dual, ‘possibly’, just means not necessarily not; i.e., true in at
least one possible world (♦P iff: P obtains at some accessible world).
A similar understanding of the determinacy operators is then readily available if
we read them as ranging over precisifications. Instead of (accessible) possible worlds,
the operators will be concerned with admissible precisifications (inadmissible preci-
sifications being analogous to impossible worlds). ‘Determinate’ will, on this model,
simply mean true on every admissible precisification (P iff: P obtains in every admis-
sible precisification). The key disanalogy with the modal operators, however, will be
that, while something can obviously be both necessary and possible, something cannot
likewise be both determinate and indeterminate.5 So it does not suffice to characterize
‘indeterminate’ as true in at least one admissible precisification (since this would leave
open the possibility of truth in every precisification, which would amount to deter-
minacy). Instead, ‘indeterminate’ needs to be characterized as true in at least one but
not all admissible precisifications (∇P iff: P obtains at some admissible precisification
and does not obtain at some other admissible precisification).

2.2 Modal vagueness

The view outlined above—that determinacy operators are in various way structurally
analogous to modal operators—is familiar from the vagueness literature. But there’s
an alternative, stronger way of understanding the link between determinacy/indeter-
minacy semantics and modality that I find preferable, at least for the case of ontic
indeterminacy (which is what concerns us here).6

3 Which Evans seems to share, given that he thinks that the operators ‘generate a modal logical at least as
strong as S5.’
4 See Fine (1975) for the classic treatment.
5 The closest modal analogy for ‘indeterminate’ thus seems to be ‘contingent’—something cannot be
both contingent and necessary, contingency being a middle-ground, as it were, between necessarily and
necessarily not. See Parsons and Woodruff (1992).
6 What follows is a very brief exposition, but the full details of the view proposed can be found in Barnes
(manuscript) ‘Ontic vagueness: a guide for the perplexed’ and Barnes and Williams (manuscript) ‘A theory
of metaphysical indeterminacy’; additionally, Williams (2008) outlines a very similar and equally useful
view. The structure (though not the ontological commitments) is analogous to the so-called non-standard
supervaluationism of, e.g., McGee and McLaughlin (1994).

123
84 Synthese (2009) 168:81–96

Rather than claiming that precisifications are like worlds, we can claim precisifica-
tions are worlds. That is, we can understand precisifications as a special subset of the
possible worlds. On this view, determinacy operators are more than just formally anal-
ogous to modal operators: determinacy is just a form of restricted modality (wherein
the operators range only over the subset of possible worlds which count as admissible
precisifications of the actual world).
This modal (as opposed to pseudo-modal) construal of the determinacy operators
relies on the distinction, familiar from Ersatz theories of modality, between the actual
world and the actualized world.7 If the world contains indeterminacy, then there is
more than one world in the space of precisifications—because if the world contains
indeterminacy then the world admits of precisification in various ways. If, for exam-
ple, it’s indeterminate whether p, then there will be at least two worlds in the space
of precisifications—one which represents things as being p, and one which represents
things as being not-p. Both possible worlds in the set of precisifications are fully pre-
cise, but if at the actual world it is indeterminate whether p, it will be indeterminate
which of these two fully precise worlds is actualized.
This indeterminacy arises because, as in standard Ersatz theory, each world rep-
resents itself as being actualized. So if we then apply the basic definitions, given in
the previous section, of determinacy (truth at all precisifications) and indeterminacy
(truth and some precisifications and falsity at others), we get the result that it’s indeter-
minate which world is the actualized world. Determinately, there is one and only one
actualized world. But if there’s indeterminacy in the actual world, then it’s indetermi-
nate which of the possible worlds it is.8 This is just a modelling of the basic thought
that if the world itself is indeterminate, then there’s no unique, determinately correct
representation of it.
There are formal as well as philosophical advantages to using such a picture of ontic
indeterminacy. On the formal side of things, it upholds both bivalence and excluded
middle.9 On the more philosophical front, it gives us a way of understanding ontic inde-
terminacy that fully avoids assigning ‘third-category’ status to indeterminacy (where
p’s being indeterminate is just another way p can be, just like being true or being
false, rather than being some kind of ‘unsettledness’ between the two poles of truth
and falsity). It also gains points for simplicity—why have both precisifications and
worlds, if you can instead account for the former solely in terms of the latter? As has
often been pointed out, the determinacy operators do seem to behave strikingly like

7 This terminology can be somewhat confusing—another option is Stalnaker’s distinction between a world
and its possible states, one of which is actual. See, e.g., Stalnaker (1976).
8 Because if there’s indeterminacy at the actual world there will be more than one world in the space of
precisification, and we have defined indeterminacy with respect to p as disagreement between precisifica-
tions about the truth value of p—thus the precisifications will disagree over the truth value of ‘world w is
actualized’ (since each world represents itself as being actualized), making it indeterminate which world is
actualized.
9 Everything’s precise at the precisifications; indeterminacy arises simply if there’s more than one prec-
isification. So determinately, everything’s precise—but if there’s no determinately actualized world, that
means it’s indeterminate which precise way things are. For any p, it’s determinate that either p or not-p
obtains—those are the only two options; it’s just that it can be indeterminate which obtains. Likewise, it’s
determinate that p is either true or false, because again those are the only two options; but sometimes it can
be ‘unsettled’ which of those two exhaustive and exclusive options in fact obtain.

123
Synthese (2009) 168:81–96 85

the modal operators. With that in mind, it would seem beneficial, if it were tenable, to
simply incorporate treatment of the determinacy operators into a theory of modality.
And, lastly, as mentioned, the modal picture gives full weight to the highly intuitive
thought that, if the world is in fact indeterminate, there’s simply no determinately
correct way of representing it.10
It’s very important to note that this type of modal representation of indeterminacy
is in no way essential to the response to the Evans argument I’m about to give. You
can easily accept a counterpart theoretic analysis of the determinacy operators while
maintaining a standard (non-modal) account of precisification. But the modal frame-
work does make my subsequent position both more plausible and more motivated, and
also helps me to answer some potential objections. I’ll thus incorporate it in several
places to do some work, though I’ll make sure to highlight when I’m doing so.

3 Counterpart theory

Much of what I have said so far is review, and none of it is going to help us out of
the Evans argument. In direct relevance to Evans, here is my main proposal: given
the above interpretation of the determinacy operators (i.e., as either modal or pseudo-
modal operators ranging over precisifications), we should read them counterpart-the-
oretically. And if we construe the determinacy operators according to the counterpart
theory model, the Evans argument will fail.11
Before I continue, let me make some clarificatory remarks on the project. The term
‘counterpart theory’ is somewhat ambiguous. It could be used to mean the Lewis
(1968) formal system, together with his translation manual for taking sentences of
quantified modal logic and turning them into sentences of the language of counterpart
theory. But the term is also used simply to refer to the idea that the truth conditions of
modal predications such as ‘a could have been F’ are to be analysed not in terms
of strict and literal trans-world identity (a is F at some world) but in terms of a relation
of similarity (a has a counterpart—an object sufficiently similar to it—at some world
that is F), together with the idea that which similarity relation is the counterpart relation
is inconstant. These two projects are largely independent. One can obviously hold that
the truth-conditions of modal predications concern distinct but similar individuals,
and hold that the standards of similarity vary depending on what counterpart relation
is invoked, without holding that sentences of QML should be translated into the 1968
formal system as Lewis suggests. Likewise, one can accept the Lewis (1968) formal
system and hold that the truth of a modal predication is not inconstant. These two
doctrines—both of which go by the name ‘counterpart theory’—are simply tangen-
tial. When I use the term I do not mean the 1968 formal system. I am a counterpart
theorist only in the sense that I think the truth conditions of ‘a could have been F’
are not that a is F at some world but that there is something appropriately similar

10 Unless we assign a third-category, separate-but-equal status to indeterminacy, which is of course to be

avoided.
11 The application of counterpart theory to vagueness (though not specifically to Evans’ argument) has
been explored before—see Forbes (1983)—though in that context its application is embedded within the
framework of the degree theory of vagueness.

123
86 Synthese (2009) 168:81–96

to a at some world that is F, and that what counts as ‘appropriately similar’ depends
on what the salient standards of similarity are, and this can varies with a variation in
how we refer to things. In particular, then, I am not going along with Lewis (1968)
treatment of names as definite descriptions or anything like that. If you think that
means my proposed solution does not deserve to be called a ‘counterpart theoretic’
solution, that’s fine—simply sub in an alternative term. The terminology is unimpor-
tant; what matters is that the assumptions I am making render the Evans argument
invalid.

3.1 Motivations

First things first, why on earth would you be tempted to go counterpart-theoretic12

at this stage (and the answer better not be ‘because it solves the Evans argument!’)?
In what follows I’ll lay out the motivations, which I actually think are very strong,
whether or not you’re remotely tempted by counterpart theory for the modal
case.13
Most importantly, counterpart theory lends itself particularly well to the semantic
framework in question because of its unique treatment of transworld identity. Accord-
ing to counterpart theory, individuals are world-bound, but at other worlds they do have
counterparts. These counterparts are world-bound objects in other possible worlds
which are similar in relevant respects to the object in question. Modal claims are true
of an object in the actual world in virtue of what is true of their counterparts. So it’s
true that I might have had red hair (rather than brown) if and only if I have a counter-
part who has red hair. But my red-haired counterpart and I are not, strictly speaking,
identical to one another.
Thus, in counterpart theoretic terms we do not have identity across precisifications
(so you, strictly speaking, do not exist at any precisification of your current situation,
but you do have a counterpart that does). For any x, x is F at a given precisification
if and only if x’s counterpart at that precisification is F. To extend to the case of
determinacy operators, Fx iff: all of x’s counterparts (at every precisification) are
F; ∇Fx iff: some of x’s counterpart’s (at some precisification) are F and some are
not.
To use a paradigm case of vagueness as an example, if it is indeterminate that Simon
is bald, then Simon will have some bald counterparts and some non-bald counterparts
(with ‘counterparts’ understood as ways Simon might be presented under various prec-
isifications; but each ‘counterpart’, as in traditional counterpart theory, is distinct from

12 In the sense outlined above. From now on I’ll drop this qualification.
13 It’s helpful to note that endorsing the modal semantics for ontic indeterminacy outlined in Sect. 2.2 would
also allow for the incorporation of standard motivations for counterpart theory, particularly the inconstancy
of modality de re. Such arguments need not be accepted, however, if a modal semantics is used, because
(as I will discuss later) those sceptical about contingent identity can always claim that contingent identity
obtains only when indeterminate identity obtains (so that a = b entails
a = b). But for those who are
tempted by contingent identity in the first instance, the standard motivations for it can also be brought to
bear for the case of indeterminate identity.

123
Synthese (2009) 168:81–96 87

Simon himself).14 Conversely, if it is determinate that Garfunkel is bald, then—for all

admissible precisifications—all Garfunkel’s counterparts will be bald.
Intuitively, this denial of transworld (or trans-precisification) identity fits neatly
with the semantics for determinacy in question. If we take vagueness as a real (and
perhaps even necessary/ineliminable) feature of the world, then it’s natural to sup-
pose that, for any vague term or object, the precisified versions of it will be similar
in relevant respects, but they will not quite be that term (or object). This copes well
with the common objection to the supervalutionist—that she has merely ‘changed the
subject’ by precisifying. Well, in a sense she has; strictly speaking, she’s talking about
a different word/object when dealing with a precisification. But that does not prevent
her from gaining information about the actual word/object.
‘Bald’ taken to mean ‘has less than 537 hairs on the scalp’ rather than our loose,
common-usage term has certainly lost something in translation. Or, to give a potential
ontic example, vague Tibbles the cat is a fuzzy creature quite distinct from any of
the determinate collections of molecules that might stand in for him at his various
precisifications. Better than adopting strict identity then—identity across precisifica-
tions, which would entail that our loose, vague word and the precisified word are
identical, or that our blurry-bounded cat and some collection of molecules x(1)–x(n)
are identical—seems to be claiming that the extension of ‘less than 537 hairs on the
scalp’ is not identical to the extension of our word ‘bald’, but rather is a precisified
counterpart of ‘bald’ (and likewise for Tibbles and the molecules). In a certain preci-
sification, that is, ‘less than 537 hairs on the scalp’ has the relevant similarity and does
the semantic work that ‘bald’ does in the actual world, and so claims of determinacy
such as, e.g. ‘x is determinately bald’, are made true, in part, by the truth (simpliciter) of
‘counterpart-x has less than 537 hairs on the scalp’.15 Similarly, in a certain precisifi-
cation there is group of molecules that plays the role that Tibbles plays in the actual
world, such that ‘molecule x(3) is determinately part of Tibbles’ is made true at least
in part by the truth simpliciter of ‘counterpart-molecule x(3) is part of the counterparts
of x(1)–x(n)’.
The general picture here is similar to notions of verisimilitude in the sciences—the
idea that our theories make an approximation at truth, a best representation of the sit-
uation, but that they do not quite paint the picture as things really are.16 Analogously,
precisifications give us a helpful model for distinguishing determinacy and indeter-
minacy, but they are not the way things really are—in reality, things (words, objects,

14 What is in fact being precisified in these cases will depend on what you think the source of vagueness
is: if you think vagueness is semantic, then our term ‘bald’ will be precisified; if you think it is ontic, then
something like facts about what is a member of the object playing the Simon role or what properties that
object instantiates, etc., will be precisified; if you think vagueness is a mixture of both, then precisifications
will likewise be a mixture of both. Counterpart theory is particularly useful in the third case, because its
context-sensitivity helps to pick out what sorts of precisifications are in order at what times. Precisifications,
however, should be understood as complete precisifications of the actual world—nothing is precisified in
isolation.
15 The truth of ‘determinately’ statements will, of course, depend on the true application of all admissible
precisifications—that is, that all the counterparts of a vague term can truly be applied to the subject in
question.
16 The analogy is only partial, however, as verisimilitude’s picture is one of getting closer and closer to the
truth, which everyone who is not a degree theorist probably will not be happy with.

123
88 Synthese (2009) 168:81–96

properties, etc.—depends on your theory of vagueness) are vague.17 With this in mind,
then, it’s best not to speak of strict identity across precisifications. Rather, we have
actual (vague) things and then we have their precisified counterparts. We can thus latch
on to truth claims about actual things based on the way things are in precisifications,
while at the same time maintaining that nothing here in the vague world is identical
with anything in the precisifications.
As a further point in its favour, counterpart theory allows that the truth of modal
claims can be context sensitive. Sentences uttered in different contexts may invoke
different counterpart relations, and thus who your counterparts are will depend in part
on the context of utterance. This feature is particularly appealing when its determinacy
analogue is applied to vagueness, an area which has a great deal of context-sensitivity.18
On precisificational accounts of determinacy, if ‘x is determinately F’ changes truth
value depending on the context, this means that some precisifications of F count as
admissible in some contexts but not in others. The precisification-style modelling of
determinacy thus needs to account for why certain precisifications can be sometimes
admissible, sometimes not. Counterpart theory gives us an excellent way of modelling
and explaining the reasons for this shift in precisificational admissibility. ‘x is deter-
minately F’ is true in a context just in case the counterpart relation invoked in that
context only picks out as counterparts of x things that are F. And the same utterance
which is true in one context can be false in another context, simply because a different
counterpart relation is invoked, one which picks out some x-counterparts which are not
F. The counterpart-theoretic notion of the salience of a particular counterpart relation
tells us why there is a shift in what precisifications are admissible.
So, for example, for a man who is 6 2, ‘He is determinately tall’ is true when uttered
at the APA, but false when uttered at an NBA convention.19 There is a shift in which
precisifications we count as admissible. Counterpart theory, with its well-developed
apparatus for modelling context-sensitivity, provides a particularly useful framework
for explaining how this shift in admissibility works.20
It’s important to point out here, however, exactly how counterpart theory realizes
the context-dependence of modal properties, and thus to demonstrate how the analo-
gous treatment of determinacy might work. There’s a common temptation to think that

17 Standard counterpart theory, however, applies only to objects, not to properties (see especially Lewis
(1971)). So the person who wants to adopt counterpart theory to account for vagueness in properties will
have to either give an account of vagueness in properties via objects or, alternatively, adopt the version of
counterpart theory that incorporates counterparts of properties, as developed in Heller (1998).
18 Most vague words can, of course, be context sensitive independent of apparent embeddings in determi-
nacy operators. But the standard supervaluationist-style theories of vagueness equate truth with ‘super-truth’
(determinate truth). So for any simple utterance of, e.g., ‘he is tall’, we need to look at precisifications to
evaluate its truth.
19 It’s tempting to say that the context-sensitivity of a word like ‘tall’ is explained much more simply:
‘tall’ just picks out different properties in different contexts. But the vagueness of tall means that we cannot
determine what property ‘tall’ picks out without precisifying (at least not on a Lewisian-style picture of
properties). And thus we still need to explain the shift in precisificational admissibility.
20 Think of the analogy with the statue and the clay puzzle. In the context of referring to the (single) object
qua statue, we say truly that it could not survive being smashed, but when we switch contexts to refer to
the object qua clay, we say truly that it could. Thus the same object can have different modal properties (by
having different counterparts) depending on what context we refer to it in.

123
Synthese (2009) 168:81–96 89

a counterpart-theoretic analysis makes it the case that which modal properties a thing
has is in some sense mind-dependent (after all, modal properties vary from context to
context and we’re the ones picking out the contexts). Anyone who thought this was the
type of solution to be offered for indeterminacy of identity would be forgiven for not
giving it the time of day; but in fact I think it’s a misreading of counterpart theory. Yes,
counterpart theory does say that when we refer to, e.g., the single clay/statue entity
as the clay, it has the property of being squashable, yet when we refer to the same
entity qua statue it does not have the property of being squashable. So prima facie it
does seem that by the very act of referring to a single object in different way we’re
changing what modal properties that object has.
In fact, counterpart theory claims only that a change in how we refer to objects can
change how we single out properties. For the case of the statue/clay, when we refer
to it qua clay we say truly that it could be squashed, whereas when we refer to it qua
statue we say truly that it could not. But all that has happened, when we refer to it as
‘the statue’, is that we have made it the case that the very same property from when
we called it ‘the clay’ no longer deserves the name ‘squashability’.21
Likewise for determinacy. If it’s determinate that you’re tall at the APA but deter-
minate that you’re not tall at the NBA, it’s not that you have changed any of your
properties. It’s just that the different contexts single out different precisifications for
the word ‘tall’, because different standards of tallness are salient.

3.2 Counterpart theory and Evans’ argument

3.2.1 The basics

And now, at long last, we come to it: the proposed solution to Evans’ famous argument.
Let me be clear about what follows: I’m in no way asserting that indeterminate identity
follows from counterpart theory. What I want to argue is that, within a counterpart-
theoretic semantics, indeterminate identity is perfectly coherent. You do not need to
embrace indeterminate identity if you interpret the determinacy operators counterpart
theoretically—and it will not give you independent motivation for doing so—but if
you already have reason to think indeterminate identity should be tenable, counterpart
theory can let you have it.
Assuming that we’re reading the determinacy operators counterpart theoretically,
Evans’ argument straightforwardly fails because, on a counterpart-theoretic interpre-
tation, the two property abstraction steps equivocate. Because of this equivocation,
the argument is invalid (though the premises are true). Thus a major advantage of a
counterpart theoretic treatment of the argument is that it allows us to say everything

21 In the clay context, the property would have been the set of all individuals with squashed counterparts;
the property is still the very same set (it has the same members), but by changing the context we have made
it the case that not all it’s members have squashed counterparts (because referring to the statue/clay qua
statue invokes a counterpart relation according to which the object has no squashed counterparts); so the
very same property no longer deserves to be called the property ‘squashability’. See Cameron (2008) for
detailed discussion of these issues.

123
90 Synthese (2009) 168:81–96

that Evans thinks we should say in cases of ontically indeterminate identity (∇a = b,
∼ ∇a = a, etc), make all the predications de re, and yet reject Evans’ conclusion.
Familiarly, in counterpart theory it can be the case that:
(a = b)
but:
♦(a = b)
yet:
∼ ♦(a = a) (that is,
(a = a)).
Thus, from the fact that ♦(a = b) but ∼ ♦(a = a), you cannot infer that a and b
have different properties. In the actual world, they are identical (@a = b), and thus
they share all their properties (via the indiscernibility of identicals). Yet the different
names can invoke different counterpart relations, so in some worlds they are not iden-
tical.22 But we cannot conclude from this that a and b are actually distinct; they are
actually identical, but could have been distinct. a and b can in fact satisfy different
modal predicates (since a has some counterpart which is qualitatively unlike any of
b’s counterparts) in the actual world, but never in the same context. When we refer to
a qua identical to b, then necessarily all its counterparts are counterparts of b. When
we refer to the one thing a/b as ‘a’, the counterpart relation invoked is such that it has
different counterparts from the counterparts it would have were we to refer to it as ‘b’.
But from this we cannot, of course, conclude that a and b are distinct; they are one
and the same thing, simply with different counterpart relations singled out depending
on which terms of reference we choose.
If we are to understand the determinacy operators in a modal/pseudo-modal seman-
tics, then we can straightforwardly apply a similar treatment in the indeterminacy of
identity case.
We have:
∇(a = b)
But, of course,
∼ ∇a = a.
From there, Evans infers that a has a property (being determinately identical to a),
which b lacks (since ex hypothesi b is not determinately identical to a). Yet, if we
interpret the indeterminacy operator using counterpart theory, this inference does not
go through. As in the modal case, a and b might have all and only the same properties,
but there be no context in which it is appropriate to both refer to the object as ‘b’ and
refer to one of its properties as ‘being determinately identical to a’. But were this the

22 Again, think of the counterpart-theoretic treatment of the statue and the clay. They are identical, the very
same thing. But they could have been different—the statue could have been made out of bronze. Referring
to the (single) object qua statue invokes a different counterpart relation than referring to it qua clay, so the
same object can satisfy different modal predicates, but—crucially—never in the same context. There’s no
violation of Leibniz’s Law here—there’s one object and one set of properties, but the correct descriptions
of those properties varies from context to context.

123
Synthese (2009) 168:81–96 91

case, the Evans argument would not in fact demonstrate that a and b have different
properties (and thus must be distinct)—it would only demonstrate that some of a/b’s
properties have different descriptions appropriate to them, depending on whether we
are in an ‘a’ context or a ‘b’ context.
Saying that it’s indeterminate that a = b is (at least structurally) equivalent to saying
that it’s possible that a = b—a is identical to b at some precisifications, but not at
others. But, as aforementioned, in counterpart theory you cannot infer from this that
a and b have different properties. It may simply be that the names ‘a’ and ‘b’ invoke
different counterpart relations.
Thus, in a counterpart-theoretic framework, the Evans argument will fail to go
through. The property abstraction reasoning will be equivocal, and thus invalid. Both
(2) and (4) in the argument are each true in some context, but there is no single context
in which both of them are true. The argument can only go through by shifting contexts,
and thus equivocating.
If it is indeterminate that a = b, that means that some of a’s counterparts are coun-
terparts of b and some are not. This is perfectly compatible with it being determinate
that a = a. The determinacy of a = a simply means that all of a’s counterparts are coun-
terparts of a, which is trivially true.23 Yet we cannot infer from this that a has some
property that b lacks, because the counterpart relation invoked in the a = b case is dif-
ferent; we have simply changed the context. The indeterminacy of identity between
a and b simply tells us that not all a’s counterparts are b’s counterparts; that is, at
some precisifications a and b are distinct. But this tells us nothing about how things
are with a and b in the actual, non-precisified, world. Notably, it does not license the
conclusion that a and b have different properties, and thus are non-identical, which
is the move Evans attempts to make. Evans’ argument thus fails, when interpreted
counterpart-theoretically.

3.2.2 Why the argument is equivocal

It’s worth spelling out the details of this, just to be clear about how, exactly, the pro-
posed solution is meant to work. Object a has the property of being determinately
identical to a—counterpart-theoretically, this property amounts to a saying ‘all my a
counterparts are a counterparts’. But of course, b has the same property; b can also
say ‘all my a counterparts are a counterparts’. It’s just that, when we have shifted
the counterpart relation by using the term ‘b’, that property no longer deserves to be
called ‘being determinately identical to a’ On this relation the property that deserves
to be called ‘being determinately identical to a’ is the property of having all one’s b
counterparts be a counterparts. b does not have this property (and neither does a, of

23 Of course, in the analogous modal case
a = a is false, so one might likewise think that a = a should
be false. Were this the case, then so much the worse for the Evans argument, as counterpart theory renders
it unsound as well as invalid. But what we count as precisifications is a very restricted modality, and it’s far
from obvious that anything resembling Lewis (1971) ‘twin cases’, which falsify
a = a, would be present in
admissible precisifications. Indeed, one might just count as a condition on the admissibility of a precisifica-
tion that it satisfies a = a. I do not want to take a stand of this either way. What I’m interested in doing here
is showing that each of the premises of the Evans argument can be true, and yet on a counterpart-theoretic
interpretation the argument still does not work.

123
92 Synthese (2009) 168:81–96

course), and so it is false to say that b is determinately identical to a. Evans’ argument

cannot go through, on a counterpart-theoretic interpretation, without equivocating.
To see precisely why this is the case, it’s helpful to look at an example. Suppose the
only instance of indeterminacy in the world is a case of indeterminate identity between
a and b, such that ∇(a = b). There would be two admissible precisifications of such
a situation—one in which a and b are identical, and one in which they are distinct.
Thus, in counterpart theory, there would be two precisifcations, one in which there is
a single object, c, which is a counterpart of both a and b, and one in which there are
two objects, d and e, where d is a counterpart of a and e is a counterpart of b. So the
situation is as follows:
(i) c is a counterpart of a and a counterpart of b
(ii) d is a counterpart of a and not a counterpart of b
(iii) e is a counterpart of b and not a counterpart of a
The Evans worry is then, in counterpart theoretic terms, that a has some property
that b lacks, in virtue of having a counterpart that b does not have. But from that we
can conclude that a and b are in fact distinct, rather than indeterminately identical.
The indeterminate identity of a and b needs to be compatible with both their identity
and their non-identity, but it looks like it’s straightforwardly inconsistent with their
identity, forcing us to conclude that they’re non-identical. The idea is just the very
basic:
(iv) Suppose a = b
(v) d is a counterpart of b (from (ii) and (iv))
(vi) contradiction (from (v) and (ii))
Yet on closer examination of the counterpart-theoretic interpretation, we can see
that claims (i)–(iii) do not in fact license the reasoning made in (iv)–(vi). This is
because, though each of (i)–(v) are true, they are not all true according to any single
counterpart relation, and thus the contradiction (vi) does not follow. The only way to
generate a contradiction would be by equivocating—by changing counterpart relations
mid-argument—and thus the argument is invalid.
(i) says that according to the counterpart relation invoked by ‘a’ (call it Ca ) c is
a counterpart of a, and according to the counterpart relation invoked by ‘b’ (call it
Cb ) c is a counterpart of b. (ii) says that according to Ca d is a counterpart of a and
according to Cb d is not a counterpart of b. (iii) says that according to Ca e is not a
counterpart of a and according to Cb e is a counterpart of b.
What would be needed for a contradiction, then, is that according to Ca e is a coun-
terpart of b. But that’s not the case. The counterpart relation which tells us that e is
a counterpart of b is Cb , and that’s the appropriate relation to invoke in the context.
According to counterpart relation Ca , though, e is not a counterpart of b. Whenever we
start speaking of b as ‘b’, we switch the context to the use of Cb , and so we correctly
say that e is a counterpart of b. But we cannot then generate a worry from the claim
that e is not a counterpart of a, because in doing so we automatically switch the context
back to Ca , and thus any argument we set up would rest on an equivocation.
Again, it’s helpful to think of the analogy to the statue/clay case. According to coun-
terpart theory, we can truly say that the clay has some counterparts—some squashed

123
Synthese (2009) 168:81–96 93

lumps, maybe—which the statue does not have. But we can only do this by changing
the context from the counterpartstatue relation to the counterpartclay relation.

3.2.3 Contingent and indeterminate identity

This should come as no surprise to the philosopher who endorses the modal frame-
work for ontic indeterminacy sketched out in Sect. 2.2. Determinacy is being treated
as a restricted necessity: what’s determinately true is what’s true not at all worlds
unrestrictedly, but what’s true at all those worlds which count as precisifications of the
actual world. Indeterminate identity simply is contingent identity, where ‘contingency’
here is the contingency defined on the restricted necessity that is determinacy. Thus,
any vindication of the coherence of (absolutely) contingent identity will automatically
yield a vindication of the coherence of indeterminate identity. The coherence of both
stands or falls together. Since counterpart theory allows the contingent identity theo-
rist to resist the Barcan/Kripke proof of the necessity of identity, we should likewise
expect it to allow the indeterminate identity theorist to resist the structurally similar
proof Evans gives for the determinateness of identity.24
It should be noted, however, that accepting this solution to the Evans argument
would still leave room for a position that accommodates many of the intuitions of
those resistant to contingent identity. It could be maintained that identity can only be
contingent when it is indeterminate—so that if a = b then
a = b. So while objects
which are indeterminately identical are contingently identical, objects which are deter-
minately identical may in every case be necessarily identical.25
Moreover, any form of contingent identity is vindicated by the response to Evans
offered here only if the genuinely modal interpretation of precisification is taken on
board. But, as emphasized earlier, acceptance of that broader picture of indeterminacy
is in no way necessary for endorsing a counterpart-theoretic treatment of Evans. One
might easily maintain that modality and indeterminacy are only limitedly analogous,
and thus adopt counterpart theory for the determinacy operators while retaining a
standard treatment of the modal operators. This point is particularly clear when one
considers that there are considerations in favour of a counterpart-theoretic interpre-
tation that seem wholly specific to the case of indeterminacy and/or vagueness, and
features of counterpart theory which might prove much more salient in the case of
indeterminacy than they do in the case of modality. If, as many do, we consider vague-
ness to be ineliminable, then we should expect the denial of trans-precisificational
identity yielded by counterpart theory. The words (or objects) we find in precisifi-
cations are useful tools for doing semantics, but they are in no way identical to the
essentially vague representations (or objects) that we actually have. Counterpart the-
ory thus lets us say something akin to the notion of verisimilitude for our modelling
of vagueness—it’s a useful and helpful representation, but do not mistake it for the
real thing. This kind of picture is particularly attractive for the supervaluationist, who

24 The analogy between these two proofs has been highlighted by E. J. Lowe in his (2005).
25 My only point here is that such a position would be perfectly consistent, given everything I have said so
far, and thus might be amenable to those who would welcome a solution to Evans’ puzzle but find contingent
identity largely unacceptable.

123
94 Synthese (2009) 168:81–96

is often accused of changing the subject when she precisifies. Counterpart theory can
allow her to embrace this—she does change the subject, but that is exactly what she
should do. These considerations26 are pressing in the case of vagueness, though they
might have no weight (or, indeed, be explicitly contradicted) in the modal case.
What is most significant for the purposes here, though, is that counterpart theory
gives us a way of understanding indeterminate identity which is both fully coherent
and compatible with both identity and non-identity. This compatibility is crucial to
avoid lapsing into a third-category analysis of indeterminacy. If two things are inde-
terminately identical, it’s not that they have some special status, on equal footing with
being identical or being non-identical. Rather, there are only two options—either they
are identical or they are distinct. But what counterpart theory (and added help from
modal semantics) allows is the idea that sometimes things can be ‘unsettled’ between
these two poles.

4 Objections

One worry about going counterpart-theoretic in a treatment of the Evans argument,

however, is that the entrenched context-dependence of counterpart theory might point
toward a strictly semantic, rather than ontic, interpretation of vagueness. Thus using
counterpart theory to undermine Evans’ proof of the impossibility of ontic vagueness
turns out to be self-defeating, because it shows that the vagueness in question is a
semantic matter after all.
However, though context sensitivity is a key feature of counterpart theory (indeed,
the key feature in the analysis of Evans), I do not think this pushes in the direc-
tion of a ‘semantic only’ view of the indeterminacy in question. Indeed, I think the
counterpart-theoretic interpretation is entirely neutral as to the source of indetermi-
nacy; that question will have to be settled by theoretical considerations elsewhere.
But counterpart theory, while in no way motivating ontically indeterminate identity,
does vindicate its consistency, despite its inherent reliance on context-sensitivity. The
defender of ontic indeterminacy could easily maintain that context determines what
an object’s counterparts are while still maintaining that the indeterminacy in question
is ontic. The position would be that what we take as an acceptable precisification of an
object will be a fact about us—our context, our language, our minds—but the bare fact
that the object itself admits of precisification is a fact about the world, independent
of what we think about it.27 So, in short, how we precifisy is a fact about us, but that
precisification is applicable in the first place need not be.

26 In addition to the usefulness of counterpart theory in explaining shifts in precificational admissibility,

as the context-dependence of vagueness is far more widely accepted and less controversial than that of
modality.
27 Such a position would need to maintain that there is no referential vagueness in the terms on either
side of the identity claim—but I take it that this is Evans’ original intention; they are Kripke-style names.
Counterpart theory does not, of course, interfere with this construal of the names in question (‘a’ and ‘b’)
as rigid designators, because in a counterpart-theoretic framework such names always refer back to their
actual world referent (so ‘a’ always refers to a, rather than to any of a’s counterparts).

123
Synthese (2009) 168:81–96 95

This line of thought is particularly suited to the modal representation of

indeterminacy mentioned previously. The counterpart theory semantics is, of course,
like any semantics, just conceptual machinery—a way of setting out truth conditions,
of representing the way things are. But what, exactly, that formal structure represents
is a further question. On the modal approach to indeterminacy, how we single out
counterpart relations is a fact about us, but the fact that there is more than one world in
the space of precisifications could at least conceivably be determined by how things
are in and of themselves.28 If indeterminacy is representational, then what is being
represented by the formal structure are facts about our representations (what we know,
how we speak, what we mean, etc.). But if, ex hypothesi, indeterminacy were ontic,
then what would be represented by the structure are facts about mind-independent
reality, about how the world is in and of itself.29 That is, if the world itself (rather than
just our representations of it) were to admit of precisification, that would be a fully
ontological, mind-independent fact.
Nothing like commitment to ontological indeterminacy falls out of (or is motivated
by) the counterpart-theoretic treatment of determinacy operators or, more broadly, any
kind of modal approach to indeterminacy like that outlined above. It would be incum-
bent upon the defender of ontic indeterminacy to give reasons (ostensibly reasons quite
separate from her semantics for the determinacy operators) why she thinks a particular
case of indeterminacy, and more specifically indeterminate identity, is a genuine fact
about the world itself, rather than a fact about how we describe it or what we know
about it. What sorts or reasons she might give, or whether any such are available, is
outside the remit of this paper. The purpose of the arguments given here is simply to
show that the basic idea of ontically indeterminate identity is coherent, whether or not
it is in fact motivated.

5 Conclusion

In summary, then, I have attempted to motivate the idea that given a construal of the
determinacy operators as either modal or pseudo-modal operators ranging over prec-
isifications, we ought to interpret them counterpart-theoretically. On the counterpart
theory model, however, the Evans argument against ontic vagueness is straightfor-
wardly invalid. This obviously is not meant as a knock down objection to the Evans
argument, as there is clearly much of the above that can be resisted. But for those who
like the idea of ontic vagueness independently or are tempted by the type of ontolog-
ical commitments that tend to lead to it (three-dimensionalism, restricted composi-

28 See Barnes (manuscript) ‘Ontic vagueness: a guide for the perplexed’ and Barnes and Williams
(manuscript) ‘A theory of metaphysical indeterminacy’.
29 A helpful analogy here is the debate between the conventionalist and realist about modality. Both mo-
dalizers can have the same formal semantics for modal operators, but they disagree about what is being
represented by that formal structure. The conventionalist thinks the truth conditions for
P tell us something
interesting about our beliefs, our conceptual apparatus, how we use ‘P’, etc., whereas the realist thinks they
tell us some deep metaphysical fact about how the world is. The question is what the formal structure is
representing (facts about our minds, or facts about the world). But nothing about the formal structure itself
will decide this either way.

123
96 Synthese (2009) 168:81–96

tion, etc.) it’s something to keep in mind. Evans’ argument is formidable enough that
anyone who goes in for ontic vagueness has to have something to say about it; I hope
to add counterpart theory to the list of options.

Acknowledgement Many thanks to Katherine Hawley, Carrie Jenkins, Daniel Nolan, Robbie Williams,
participants in the Arché Vagueness Seminar, two anonymous referees, and most especially to Ross
Cameron.

References

Cameron, R. (2008). What’s metaphysical about metaphysical necessity? Philosophy and Phenomeno-
logical Research, 77 (forthcoming).
Evans, G. (1978). Can there be vague objects? Analysis, 38, 208.
Fine, K. (1975). Vagueness, truth, and logic. Synthese, 30, 265–300.
Forbes, G. (1983). Thisness and vagueness. Synthese, 54, 235–259.
Heller, M. (1998). Property counterparts in Ersatz worlds. Journal of Philosophy, 95(6), 293–316.
Lewis, D. (1968). Counterpart theory and quantified modal logic. Journal of Philosophy, 65, 113–126.
Lewis, D. (1971). Counterparts of persons and their bodies. Journal of Philosophy, 68(7), 203–211.
Lewis, D. (1988). Vague identity: Evans misunderstood. Analysis, 48, 128–130.
Lowe, E. J. (2005). Identity, vagueness, and modality. In J. L. Bermudez (Ed.), Thought, language, and
experience: themes from the philosophy of Gareth Evans. Oxford: OUP.
McGee, V., & McLaughlin, B. (1994). Distinctions without a difference. The Southern Journal of
Philosophy, 33, 203–253.
Noonan, H. (2004). Are there vague objects? Analysis, 64, 131–134.
Parsons, T., & Woodruff, P. (1995). Worldly indeterminacy of identity. Proceedings of the Aristotelian
Society, 95, 171–191.
Stalnaker, R. (1976). Possible worlds. Nous, 10, 65–75.
Weatherson, B. (2003). Many, many problems. Philosophical Quarterly, 53, 481–501.
Williams, J. R. G. (2008). Multiple actualities and ontically vague identity. The Philosophical Quarterly,
58, 134–154.
Williamson, T. (1994). Vagueness. Oxford: OUP.

123