01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 1
Introduction
JOSÉ LUIS BERMÚDEZ
The framework for the papers in this volume is set by the wideranging philosophical contributions of Gareth Evans, who died in
1980 at the age of 34. In the papers gathered together in the
posthumously published Collected Papers (CP) and in The Varieties
of Reference (VR) (prepared for publication from drafts and lecture
notes by John McDowell after Evans’s death) Evans made a number of important contributions to the philosophy of language, the
philosophy of thought, and philosophical logic. Some of these contributions have been extensively discussed and assimilated into
ongoing debates and discussion, but considerable areas of his
thought remain relatively unknown. I hope that this volume will
contribute to Evans being widely recognized as one of the most
original philosophers of the second half of the twentieth century.
This introduction is in two parts. Evans was a far more systematic
thinker than is usual in contemporary analytical philosophy,
and the first part gives an overview of his thinking. The second part
of the introduction summarizes the papers and provides any
necessary background.
EVANS: INFLUENCES AND OVERVIEW
Evans’s philosophical work is best viewed against the backdrop of
four powerful and in many ways competing currents in the philosophy of the 1960s and 1970s. Two of these currents were firmly
I am grateful to Mark Sainsbury and Christopher Peacocke for written comments on
an earlier draft of this introduction.
01-Luis-Intro.qxd
2
22:02:05
9:16 AM
Page 2
José Luis Bermúdez
rooted in Oxford, Evans’s intellectual home throughout his career,
while two originated in the very different philosophical climate of
North America.
Inevitably for an Oxford philosopher of his generation, Evans
was exposed to the very different philosophical concerns and styles
of Michael Dummett and Peter Strawson. Dummett’s Frege-inspired
philosophy of language was a powerful influence throughout
Evans’s career, as was Strawson’s neo-Kantian project of using the
techniques of conceptual analysis and transcendental arguments to
plot the limits and structure of our conceptual scheme.1 At the same
time Evans was very open to, and informed about, developments in
philosophy on the other side of the Atlantic. As a philosopher
deeply versed in philosophical logic, Evans was influenced by the
powerful cluster of ideas about modality and designation that came
in the wake of the semantics for quantified modal logic proposed
by Saul Kripke, Ruth Marcus, and others. A final important influence on his philosophical framework was Donald Davidson’s proposal to develop a theory of meaning for a regimented version of
natural language in terms of a broadly Tarskian theory of truth—
the so-called the Davidsonian program in semantics.
Evans shared Dummett’s powerfully developed conviction that
the best approach to the philosophy of language is broadly
Fregean, incorporating a truth theory at the level of reference and a
theory of understanding at the level of sense, supplemented by a
theory of force that would explain how language is used to perform
different types of linguistic act.2 Like Dummett, Evans held that the
most problematic notion in the theory of meaning is the notion of
sense, and also like Dummett, Evans was troubled by Frege’s lack of
concern with the crucial notion of what it is to grasp a sense. Evans’s
most systematic and substantial work, The Varieties of Reference,
is, among other things, an essay on the relation between a theory of
sense and an account of linguistic understanding. Yet the emphasis
there is rather different from that of Dummett. Dummett’s principal
1
The most comprehensive statement of Dummett’s philosophical outlook remains
Frege: Philosophy of Language (1973), but the essays in Truth and Other Enigmas
(1981) provide a more accessible introduction, as does Dummett’s well-known essay
“What is a theory of meaning? (II)” (1976), originally published in Evans and
McDowell 1976. The best single source for Strawson is Individuals (1959).
2
The overall picture is sketched out in Dummett 1976: 74. Dummett’s most
sustained exploration of the theory of force is Dummett 1973: ch. 13.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 3
Introduction
3
concern is with what it is to understand a sentence (to grasp the
sense of a sentence), and, as is well known, this leads him away
from treating truth as the central notion in a theory of sense. We
cannot, Dummett argues, give a satisfactory account of what it is to
grasp the truth conditions of a sentence that is not effectively decidable, and hence the notion of truth cannot be at the core of a theory
of understanding in the way that it is standardly taken to be. Evans
does not directly engage with this aspect of Dummett’s thought
(although Evans’s writings suggest a robust realism that would not be
sympathetic to Dummett’s anti-realist conclusions). His concern is
more directly with the theory of understanding at the sub-sentential
level—and in particular with what it is to understand those linguistic
expressions that have the job of picking out individuals (the class of
referring expressions). Evans holds that an account of what it is to
understand referring expressions should be formulated in the context
of a more general account of what it is to think about individuals.
Our language contains referring expressions because we think about
objects in certain ways, and those expressions work the way they
do because of how we think about the objects that they pick out. In
this sense, then, he thinks that the philosophy of language should in
the last analysis be answerable to the philosophy of thought.
Evans’s views about the direction of explanation in our thinking
about language mark a significant divergence from Dummett (noted
and discussed by Dummett in his 1991b). Dummett has consistently
argued that the philosophical analysis of thought can proceed only
via the philosophical analysis of language. In fact, he takes this
principle (which he attributes to Frege) to be the defining feature of
analytical philosophy (Dummett 1994). Evans, in contrast, holds
that we can elucidate the nature of thought independently of the
nature of language. The idea that the direction of explanation goes
from thought to language is not new to Evans. It is clearly built into
the Gricean program in semantics, for example, since that program
aims to elucidate linguistic meaning through speakers’ communicative
intentions. In contrast to Grice and his supporters, however, Evans
confronts the obvious challenge that this throws up of giving a
substantive and independent account of the nature of thought. Part II
of VR is a sustained investigation of the different types of what
Evans calls singular thoughts. Singular thoughts are, roughly,
thoughts about specific and identifiable individuals (as opposed, for
example, to quantificational thoughts that are not generally targeted
01-Luis-Intro.qxd
4
22:02:05
9:16 AM
Page 4
José Luis Bermúdez
at a specific individual). Paradigm singular thoughts, in addition to
thoughts that might be expressed using proper names, are perceptual
beliefs of the sort that might be communicated through demonstrative
expressions such as “this” or “that” and beliefs about oneself or
the present moment that might be expressed through token-reflexive
expressions such as “I” or “now” (see also “Understanding demonstratives” (Evans 1981b)). Both Dummett and Evans think that
thoughts are the senses of sentences, and hence that to explore
what is distinctive of, say, “I”-thoughts is to explore the sense of the
first person pronoun. But Evans, unlike Dummett, holds that we
can investigate “I”-thoughts without proceeding via the sentences
that express them.
Evans’s understanding of the category of singular thoughts
marks a further significant divergence from Dummett. Singular
thoughts, for Evans, are Russellian. They involve ways of thinking
about objects that require the existence of the object being thought
about. A singular thought requires the existence of the object being
thought about, because thinking it depends upon the ability to pick
out, or otherwise keep track of, the object in question.3 In developing a Russellian account of singular thoughts, Evans took himself
to be following Frege. Despite Frege’s apparent concession that
sentences with empty referring terms can have a sense, and hence
that there can be thoughts with non-denoting senses, Evans maintains
that a proper understanding of Frege’s notion of sense shows this to
be impossible. The best way to understand Frege’s notion of sense,
Evans maintains, is through the metaphor of a mode of presentation, and, he argues, there cannot be a mode of presentation of
something that does not exist. This way of understanding sentences
with empty referring terms and thoughts with non-denoting senses
is diametrically opposed to Dummett’s own theory of sense (and
indeed to Dummett’s interpretation of Frege). John McDowell’s
contribution to this volume (Chapter 1) assesses Evans’s interpretation of Frege.
Although Evans’s ideas about the Russellian nature of singular
thoughts are diametrically opposed to Dummett’s, they have a certain affinity with well-known views of Peter Strawson’s (thus bringing us to the second of the four influences in Evans’s philosophical
3
The case for the Russellian status of singular thoughts is made in the first three
chapters of VR and worked out for specific classes of singular thoughts in part II of
that work.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 5
Introduction
5
framework). In “On referring” (Strawson 1950), his well-known
essay on Russell’s theory of descriptions, Strawson argued that particular uses of sentences with non-referring singular terms can fail
to make statements that are true or false. There is, according to
Strawson, a logical relation between a sentence expressing a singular
thought and a sentence asserting that the object being thought
about exists, such that the first sentence can only be true or false
when the second sentence is true. When this relation of presupposition fails to hold, no statement is made. Strawson’s theory of presupposition is clearly situated at the level of language. Evans can be
seen as extending Strawson’s account to the level of thought—and,
indeed, of explaining why the relation of presupposition holds. If it
is indeed the case that a thinker cannot entertain a singular thought
without being appropriately connected with the relevant object,
then it is fully to be expected that no such thought can be expressed
linguistically if the object in question does not exist.
Developing this line of thought of course requires clarifying what
is to count as being appropriately connected with the object.
Evans’s thinking here exemplifies one of his many points of difference from Kripke and other philosophers drawing philosophical
consequences from the semantics of quantified modal logic. As part
of his repudiation of those theories of sense that understand the
sense of a proper name in terms of definite descriptions holding
true of the bearer of that name, Kripke developed a very austere
account of the type of connection with an object required in order
to refer to it successfully (Kripke 1972, 1980). According to
Kripke, an utterance of a name refers to an object just in case there
is a series of reference-preserving causal links going back from the
utterance in question to the initial occasion or process where the
name was bestowed. A reference-preserving link is one where a
speaker intends to use the name to pick out the same object as the
person from whom she learnt the name. This type of account is
deeply antithetical to Evans’s understanding of thought and language.
In his first publication, “The causal theory of names” (1973), Evans
offers a nuanced critique of Kripke’s account. He maintains that
the information that speakers associate with a name is indeed
crucial to fixing reference, but concedes that information is individuated by causal origin (what matters is not the object that best
fits the information that speakers possess, but rather the object
from which that information originates). In VR, however, Evans
01-Luis-Intro.qxd
6
22:02:05
9:16 AM
Page 6
José Luis Bermúdez
puts more distance between himself and those theories of the
content of thought loosely based on Kripke’s account of names
(what in VR §§3.3 and 3.4 he collectively terms the Photograph
Model). The fundamental emphasis in VR is on working through
the implications of the principle (Russell’s Principle) that thinking
about an object requires discriminating knowledge of that object
(see particularly VR ch. 4). Most of the cases of discriminating
knowledge that Evans discusses clearly involve some sort of causal
link with the object being thought about, but what is important is
the thinker’s capacity to exploit that link in order to distinguish the
object from all other things.
This concern with discriminating knowledge signals a further
connection between Evans and Strawson. In Individuals (1959)
Strawson undertakes the neo-Kantian project of mapping the structure of our conceptual scheme. For Strawson our conceptual
scheme is firmly anchored in our ability to think about individual
objects (what he terms particulars). Our ability to think about
objects is inextricably tied to our ability to identify and reidentify
them, and it is through our ability to identify and reidentify objects
that we are able to identify and reidentify places and hence get a
grip on the spatial structure of the world.4 The idea that there is an
intimate connection between thinking about objects and thinking
about space is very much to the fore in Evans. It is a prominent theme
in his discussion of demonstrative identification in VR chapters 6
and 7 and is discussed with particular reference to Strawson’s
Individuals in “Things without the mind” (1980c): see further
Cassam’s contribution to this volume (Chapter 8). Thinking about
an object that one is currently perceiving requires being able both to
locate that object relative to oneself and to integrate one’s personal,
egocentric space with one’s understanding of the objective layout of
the environment (with what psychologists term a cognitive map).
In at least one respect Evans is more Kantian than Strawson. One
of the principal themes of Kant’s Critique of Pure Reason is the
relation between self-consciousness and the objectivity of the world.
This theme is not directly explored in Strawson’s Individuals, but is
4
Strawson and Evans adopt the same three-way classification of types of discriminating knowledge (as Evans notes at VR 89 n. 2, referring to Strawson 1971).
One has discriminating knowledge of an object if one is currently perceiving it; if
one could recognize it were one to be presented with it; or if one knows distinguishing facts about it.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 7
Introduction
7
very much to the fore in Evans’s discussion of self-identification in
VR chapter 7. Thinking about ourselves requires discriminating
knowledge no less than thought about any other type of object, and
Evans analyzes the discriminating knowledge involved here in
terms of the thinker’s ability to locate herself within a cognitive
map of the environment. These themes are discussed in Bermúdez’s
contribution to this volume (Chapter 5).
It is because Evans takes thinking about objects to be such a
highly sophisticated cognitive ability that he allows for so many
ways in which our thoughts can fail to connect to the relevant
objects, and correspondingly many ways in which we can have the
illusion of a thought without thinking a genuine thought. Yet, not
all thoughts about objects can fail in this manner. It is only singular
thoughts that run the risk of giving “the illusion of thought”.5 This
raises the question of how to delineate the class of singular
thoughts. Which ways of thinking about objects are Russellian, and
which not? Here Evans’s analysis of thought proceeds hand in hand
with his analysis of language. The category of singular thoughts is
delineated through the referring expressions that appear in the sentences that express them. A singular thought is a thought whose
canonical linguistic expression would involve a genuine referring
expression. But what is a referring expression?
Here we see once again the influence of Kripke and possible
worlds semantics. In VR §2.4 Evans addresses the question of why
definite descriptions should not be counted as referring expressions.
After rejecting the arguments Russell provided in “On denoting”,
Evans offers an argument based on the behavior of definite descriptions in modal contexts. If we treat definite descriptions as referring
expressions, we will need to relativize the relation of reference to
a possible world. Only thus will we be able to capture the much
discussed difference between wide and narrow scope readings of
definite descriptions in modal contexts (the difference between, on
the one hand, saying that the object that satisfies a definite description in this possible world has a particular property F in another
possible world and, on the other, saying that the object that satisfies
5
Evans’s concern in VR is almost exclusively with the “illusions of thought” that
arise when one fails to be appropriately connected to the object one is thinking
about. But he would no doubt have been sympathetic to the idea that thoughts can
fail with respect to their predicative component—when, for example, concepts fail
to be clearly specified.
01-Luis-Intro.qxd
8
22:02:05
9:16 AM
Page 8
José Luis Bermúdez
that same definite description in another possible world is F in that
world). But no such relativization of the reference relation is necessary in the case of “standard” referring expressions. As Evans puts
it, “As a consequence of this change we ascribe to names, pronouns,
and demonstratives semantical properties of a type which would
allow them to get up to tricks they never in fact get up to; since their
reference never varies from world to world, this semantic power is
never exploited” (VR 56). In essence, Evans identifies referring
expressions as a semantic kind in virtue of how they can be most
illuminatingly and economically treated within a semantic theory—
and his preferred semantic theory is one that treats referring
expressions as rigid designators in the way that Kripke suggested.
In one sense Evans’s characterization of referring expressions is
rather narrow. He holds, for example, that all referring expressions
are singular, thus excluding by fiat the possibility of plural terms
counting as referring expressions—an exclusion with which Rumfitt
takes issue in his contribution to this volume (Chapter 3). In
another sense, however, he drastically expands the scope of referring expressions. Some of Evans’s most interesting (and least well
known) contributions to semantic theory come in his explorations
of the semantics of pronouns in his two-part “Pronouns, quantifiers,
and relative clauses” (1977) and in the more synoptic “Pronouns”
(1980b). Evans is particularly concerned with pronouns that have
quantifier expressions as antecedents—pronouns such as the “her”
in “every woman loves her mother”. Bound pronouns are frequently thought to be the natural language analogues of the logician’s bound variable (the “x” in “∀x Fx”). On this view, most
emphatically developed by Geach (1962), there is no sense in which
pronouns with quantifier antecedents can be treated as referring
expressions. In opposition to this, Evans develops what he terms a
coreferential treatment of bound pronouns, showing how to provide a semantics for bound pronouns that trades on the fact that
any sentence with a quantifier as antecedent can be paired with a
sentence (or set of sentences) that has a singular term as antecedent.
Evans suggests that by evaluating sentences with quantifier
antecedents in terms of appropriately paired sentences with singular term antecedents we can understand bound pronouns in referential terms, and hence as semantically similar to pronouns with
singular antecedents. Just as the “her” in “Alexandra beats her
donkey” corefers with “Alexandra”, so too does the “her” in
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 9
Introduction
9
“Some woman beats her donkey” corefer with the proper name in
the sentence or sentence whose truth underwrites the truth of the
existentially quantified sentence (in this case, of course,
“Alexandra”). Some of the implications of Evans’s account for linguistics are explored in Safir’s essay in this volume (Chapter 4).
Evans devoted considerable attention to a type of referring
expression that he termed descriptive names (VR §§1.7, 1.8, 2.3).
Descriptive names are names whose reference is fixed by description
in such a way that they will always take wide scope in modal
contexts. Evans’s favorite example of such a reference-fixing stipulation is
(1) Let us call whoever invented the zip “Julius”
but in fact there are a number of descriptive names in common
currency. “African Eve”, generally taken to refer to a specific individual who lived 200,000 years ago and is our earliest common
ancestor, is a good example. Descriptive names are interesting for
two reasons. First, and in opposition to much of the post-Naming
and Necessity discussion of proper names and definite descriptions,
they are both descriptive and rigid (that is, they refer to the same
object in all possible worlds in which they refer at all). Second, they
are not Russellian. A sentence with a non-referring descriptive name
can still express a genuine thought (“African Eve did not exist”
might be a true sentence). The implications for semantic theory of
recognizing the category of descriptive names is discussed in
Sainsbury’s contribution to this volume (Chapter 2).
Insofar as descriptive names are rigid designators that clearly
express a descriptive Fregean sense, recognizing the existence of
descriptive names is already compromising the insistence of Kripke
and other direct reference theorists that referring expressions cannot have a Fregean sense. As far as Evans is concerned, however,
this is just the tip of the iceberg. Like Dummett, Evans thinks that
Frege’s notion of sense is an indispensable tool for understanding
language, as well as being the thread that links together the analysis
of language and the analysis of thought. VR is a sustained attempt
to show that, despite the widely held view to the contrary, there is
no incompatibility between a referring expression functioning as a
rigid designator and having a Fregean sense (see VR §2.5 for an
emphatic statement of this point). An important element in the case
he makes is pointing out, surely correctly, that descriptive names,
01-Luis-Intro.qxd
10
22:02:05
9:16 AM
Page 10
José Luis Bermúdez
and definite descriptions in general, are a poor model for Frege’s
notion of sense.6 Evans’s reading of Frege’s notion of sense begins
with a relatively neutral characterization (VR §§1.4, 1.5). We cannot, Evans holds, think about an object unless we think about it in
a particular way. The sense of a referring expression is a particular
way of thinking about the object it picks out, such that anyone who
is to understand a sentence featuring that expression must think
about its referent in that way. It is clear from Kripke’s extended discussion in Naming and Necessity that no definite description or set
of definite descriptions can satisfy this requirement in the case of a
proper name. But it should be equally clear that there is no reason,
even in the case of a proper name, to think that the sense of a referring expression should take a descriptive form—still less so for
those referring expressions that are not proper names (other than
descriptive names, of course). Thinking about an object in a particular
way requires having discriminating knowledge of that object. But,
as pointed out earlier, thinking of an object as the unique satisfier of
a definite description is only one way of possessing distinguishing
information about it—and distinguishing information is only one
form of discriminating knowledge.
In VR Evans explains how the discriminating knowledge requirement can be met for each of the principal categories of referring
expressions. In the case of what he calls “one-off referential devices”
such as demonstratives and personal pronouns, perception and
recognition are the fundamental forms of discrimination. Things are
more complex for proper names (VR ch. 11). Here Evans emphasizes
the existence of name-using practices and distinguishes between the
producers and the consumers of such practices. The producers of a
practice have dealings with the bearer of a name, and their use of
the name is bound up with their abilities to recognize the bearer of
the name in a way that clearly satisfies the discriminating knowledge requirement. Consumers of the name, in contrast, use names
with the intention of participating in specific name-using practices,
with the particular practice fixed by information deriving from the
bearer of the name (information that might be partially or even
wholly inaccurate). Here the discriminating knowledge requirement is not directly met (as Evans notes: VR 387 n. 13).
6
The point has been made many times (Dummett 1973: 110–11; Bell 1979). It is
astonishing that it is not more widely accepted.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 11
Introduction
11
One can see Evans’s discussion of proper names in “The causal
theory of names” (1973) and VR chapter 11 as aiming simultaneously to do justice to Kripke’s modality-inspired insights, while
blunting the more drastic conclusions that might be drawn from
them. In “Reference and contingency” (1979) Evans applies a similar strategy to Kripke’s arguments for the existence of contingent
a priori truths, arguing (against Dummett and others) that there is
nothing paradoxical about the existence of truths that are a priori
and contingent, and (against Kripke) that there is nothing particularly interesting in the fact that such truths exist. One very interesting
feature of this paper is his attempt to undercut Kripke’s trademark
modal arguments, all of which draw conclusions about the content
of a sentence from its behavior in modal contexts.
In §III of “The causal theory of names” Evans distinguishes
between the proposition a sentence expresses and its content.
Propositions are understood in the standard way as functions from
possible worlds to truth-values, while the notion of content is picked
out by Frege’s intuitive criterion for sameness of content (viz. that
two sentences have the same content just if it is impossible for someone to understand both while believing one and disbelieving the
other). It is widely accepted that two sentences can have different
contents while expressing the same proposition (in virtue of being
true in exactly the same possible worlds). Evans, however, argues
that sentences can be epistemically equivalent (have the same content)
while expressing different propositions, which means of course that
it is illegitimate to make inferences about content on the basis of
behavior in modal contexts. Consider “Julius” once again. “Julius”
and “the inventor of the zip” embed differently in modal contexts,
since “the inventor of the zip” can take narrow scope within modal
operators in a way that “Julius” cannot. Yet, argues Evans,
I cannot imagine how the belief that Julius is F might be characterized
which is not simultaneously a characterization of the belief that the inventor
of the zip is F, i.e. that one and only one man invented the zip. Belief states
are individuated by the evidence that gives rise to them, the expectations,
behavior, and further beliefs that may be based upon them, and in all
of these respects the belief states associated with the two sentences are
indistinguishable. (CP 202)
This bold argumentative move may fail to convince. One might
wonder, for example, why the two sentences “Julius is F” and
01-Luis-Intro.qxd
12
22:02:05
9:16 AM
Page 12
José Luis Bermúdez
“the inventor of the zip is F” should behave differently in modal
contexts if they are synonymous in the way that Evans suggests.
There seem to be no features of a sentence, other than its content,
that could explain how a sentence behaves in modal contexts.
But those who find this line of argument persuasive will still have to
answer Evans’s challenge to explain what difference there is
between the belief that Julius is F and the belief that the inventor of
the zip is F. Evans is surely correct that philosophers have played
too fast and loose with modal arguments about content.7
Let us turn now to the fourth and final influence on Evans, the
Davidsonian program in semantics. In addition to the influential
collection Truth and Meaning: Essays in Semantics, which Evans
edited with John McDowell in 1976, Evans engages with Davidson’s
project of using a Tarskian theory of truth as a theory of meaning
in his papers “Semantic structure and logical form” (1976) and
“Semantic theory and tacit knowledge” (1981a) and in VR §1.8. In
the Introduction to Truth and Meaning Evans and McDowell
strongly align themselves with one aspect of Davidson’s conception
of the form a theory of meaning should take. This is its divergence
from what they term translational semantics. Translational semantic
theories offer mappings between sentences of the object language
and semantic representations of those sentences—what are, in effect,
translation rules. Suppose that the meaning of object language
sentence S is given by the metalanguage sentence S. A translational
semantics would give a rule stating the relation between S and S—a
rule of the form “ ‘S’ means ‘S’ ”, as one might have a French–
German translation rule stating that “ ‘Il pleut’ means ‘es regnet’ ”.
The key objection that Evans and McDowell make to translational
semantics is that it fails to provide a theory of understanding.
What we ought to be doing is stating what the sentences of the language
mean, stating something such that, if someone knew it, he would be able to
speak and understand the language . . . But there is no escaping the fact that
one could have a competence based upon the mapping relation, and yet
not know what a single sentence of the language meant. A speaker-hearer
would know that only if he knew what sentences of the theory’s language
meant; but this is knowledge of precisely the kind that was to be accounted
for in the first place. (Evans and McDowell 1976: p. ix)
7
It is worth noting, though, that some neo-Kripkean arguments in this area
are not dependent upon modal contexts in ways that make them straightforwardly
susceptible to Evans’s challenge. See Soames 2002.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 13
Introduction
13
A theory of meaning, then, must be a theory of understanding. This
gives a necessary condition for a theory of meaning: no theory
can qualify if a speaker could know it without understanding the
language for which the theory is being given. This condition is
clearly satisfied by Davidson’s theory, which is designed to yield a
T-sentence of the form “ ‘p’ is true iff p” for every sentence of the
language. Davidson avoids the translational fallacy because his
recursive theory of meaning yields theorems that use the meaningspecifying sentence rather than mention it. No one could grasp a
theorem such as
(2) “Madrid is the capital of Spain” is true iff Madrid is the capital
of Spain without understanding the sentence “Madrid is the
capital of Spain”.
Nonetheless, Evans does not follow Davidson in the claims he
makes on behalf of a recursive truth theory. Evans’s divergence from
Davidson comes across very clearly in “Semantic structure and logical form” (1976), where he takes issue with Davidson’s claim that
a recursive truth theory will provide a satisfying explanation of
what makes an inference valid. Davidson holds that a satisfying
account of validity will have to show that valid inferences are valid
in virtue of the structure of their premises and conclusion, so that
formal validity is more fundamental than material validity. Evans
glosses the requirement that this imposes upon a recursive truth
theory as follows:
Let us say that the conditional ‘If S1 is true, . . . , and Sn1 is true, then Sn is
true’ is the validating conditional for that inference S1 . . . Sn1|Sn. I think
Davidson’s idea is that an inference is structurally valid according to a
theory T if and only if its validating conditional is a semantic consequence
of the theory’s recursive clauses. (CP 53)
Two features of a Tarskian truth theory explain how the requirement
is met in individual cases. The first is that the truth theory picks out
a class of expressions as logical constants. Davidson, like many
others before him, thinks that we should look to the logical constants to explain what secures structurally valid inferences. Second,
and slightly less obviously, structurally valid inferences depend
upon premises and conclusion having common elements. These
must be semantically common elements, and it is the job of a theory of truth to display the logical structure of sentences in a way
01-Luis-Intro.qxd
14
22:02:05
9:16 AM
Page 14
José Luis Bermúdez
that will make manifest what these semantically common elements
are and how inferences can trade on them.
Attractive though this picture is (and of course this conception of
what makes an inference structurally valid is very widely held),
Evans finds it wanting. There are important classes of structurally
valid inferences whose validity cannot, without considerable
contortion, be traced to the logical structure of their constituent
sentences. The non-logical parts of a sentence have a structure that
can validate certain inferences, so that structural validity can, Evans
thinks, be a function of semantic structure no less than of logical
structure. The semantic structure of a sentence is a function of
the semantic kinds to which its constituent expressions belong.
These semantic kinds determine the inferential properties of expressions falling under them as a function of the type of entity that can
serve as their semantic values. Consider, for example, the sentence
“Clare is a skinny ballet dancer”. We are interested in explaining the
structural validity of the inference from “Clare is a skinny ballet
dancer” to “Clare is a ballet dancer”. According to Evans, this follows
from a correct account of the word “skinny”, which is functioning
here as an attributive adjective. Attributive adjectives take as their
semantic values functions from sets to subsets of those sets. In this
case the function will be from the set of ballet dancers to the subset
of just those ballet dancers who are skinny. Clearly, then, any skinny
ballet dancer will be a ballet dancer, and the inference from “Clare
is a skinny ballet dancer” to “Clare is a ballet dancer” will be valid
in virtue of its semantic structure.
In stressing the need to discern semantic structure by what he
calls interpretational semantics (VR 33; CP 61), Evans diverges from
Davidson, since a Tarskian truth theory of the sort that Davidson
thinks can serve as a theory of meaning does not assign semantic
values to any expressions other than proper names. Evans’s
interpretational semantics is quite clearly Fregean in inspiration.
Nonetheless, as he makes clear at VR 34, Evans is proposing interpretational semantics as a validation for a Tarski-style truth theory,
rather than as a replacement for it. At the very least, he thinks, any
adequate truth theory must be in harmony with a coherent interpretational semantic theory. And in fact it may be the case that one
can read off from what a truth theory explicitly says about a type
of expression the class of semantic values that an interpretational
semantics would assign to it.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 15
Introduction
15
In addition to the shift to interpretational semantics, Evans
proposes a further extension of Davidson’s approach to meaning.
Evans’s guiding thought is that a theory of meaning must be a theory
of understanding. One implication of this is that we must be able to
explain speakers’ understanding of a language (and in particular
their ability to understand novel sentences of that language) in terms
of their knowledge of a theory of meaning. This is not something
that can be achieved merely by spelling out the form that the theory
of meaning must take (presumably the conjunction of a coherent
interpretational semantics and a suitable truth theory). We need to
know how to understand what it is to have knowledge of such a
theory, which is something that Davidson does not address and
takes himself not to need to address.
Some philosophers have expressed skepticism about the prospects
for an informative and explanatory account of what it would be to
know a theory of meaning. Plainly, the only knowledge of a theory
of meaning that speakers can be claimed to have is tacit or implicit
knowledge, and it has been argued that no account of tacit knowledge can be anything other than a redescription of a speaker’s
competence (Quine 1972; Wright 1981). The putative problem is
that the meaning theorist has no way of identifying which of a
number of extensionally equivalent theories of meaning a speaker
can properly be described as tacitly knowing. Evans confronts this
line of argument in ‘Semantic theory and tacit knowledge’ (1981a).
He begins from the thought that tacit knowledge of a theory of meaning involves having a disposition corresponding to each axiom of
the theory. These dispositions, he argues, should be interpreted in a
full-blooded sense, as requiring causally efficacious categorical bases.
When we ascribe to something the disposition to V in circumstances C, we
are claiming that there is a state S which, when taken together with C,
provides a causal explanation of all the episodes of the subject’s V-ing
(in C) . . . Understood in this way, the ascription of tacit knowledge does
not merely report upon the regularity in the way in which the subject
reacts to sentences containing a given expression (for this regularity can be
observed in the linguistic behavior of someone for whom the sentence is
unstructured). It involves the claim that there is a single state of the subject
which figures in a causal explanation of why he reacts in this regular way
to all the sentences containing the expression. (CP 329–30)
This way of thinking about the causal basis of comprehension
offers a way of distinguishing between extensionally equivalent
01-Luis-Intro.qxd
16
22:02:05
9:16 AM
Page 16
José Luis Bermúdez
meaning theories, and hence of responding to the skeptical challenge.
It also explains how a hearer can understand a previously unencountered sentence S. The hearer understands S in virtue of a complex
set of dispositions corresponding to the constituents of S and
derived from exposure to those constituents in different sentences
and different contexts.
Evans’s discussion of tacit knowledge deploys a theoretical
notion that marks him off from all four of the philosophers we have
been considering—the notion of nonconceptual content (although
he only uses the expression “nonconceptual content” in VR). Evans
is happy to concede that the causally effective inner states in terms
of which he understands tacit knowledge are very different from
“ordinary” states of knowledge or belief. Although there are overlaps
on both the output side (in that someone with the tacit knowledge
that p will be disposed to do many of the same things as someone
who believes that p in the ordinary sense of belief) and the input
side (in that someone will typically acquire the tacit knowledge
that p in circumstances that might well induce the ordinary belief
that p), Evans maintains that there is a fundamental difference in
kind between tacit knowledge and the propositional attitudes.
Although he does not himself put it in these terms, states of tacit
knowledge are essentially one-track rather than multi-track dispositions. Whereas tacitly knowing a principle of a semantic theory
manifests itself in a narrowly defined disposition to react to sentences and to use words in certain ways, the dispositions associated
with beliefs have a significantly wider application and are at the
service of a far wider range of projects.
One who possesses a belief will typically be sensitive to a wide variety of
ways in which it can be established. (what it can be inferred from), and a
wide variety of different ways in which it can be used (what can be inferred
from it)—if we think of plans for intentional action as being generated
from beliefs by the same kind of rational inferential process as yields
further beliefs from beliefs. To have a belief requires one to appreciate its
location in a network of beliefs. (CP 337)
Evans’s discussion of the distinctiveness of beliefs (and other propositional attitudes) returns to the important connection between
semantic structure and inference. It is because propositional attitudes
have structured contents that they can feature in such a complicated
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 17
Introduction
17
web of inferential connections. A thinker’s appreciation of the
inferential power of any given belief reflects her ability to appreciate
the inferential power of other beliefs that have constituent concepts
in common with it. As Evans puts it, “behind the idea of a system
of beliefs lies that of a system of concepts, the structure of which
determines the inferential properties which thoughts involving an
exercise of the various component concepts of the system are
treated as possessing” (CP 338).
Tacit knowledge states, in contrast, are single-track dispositions
because they lack this structure. This makes them a class of what in
VR Evans terms nonconceptual information states (VR §§5.2
and 6.3). In VR Evans’s emphasis is on nonconceptual perceptual
states and the role that they play in demonstrative thought. The
discriminating knowledge required for demonstrative thought
about an occurrently perceived object takes the form of an ability
to keep track of it over time. Thinkers are able to do this when they
stand in the appropriate information links to the object. These
information links are nonconceptual, issuing in practical capacities
to locate the object relative to oneself and to other objects—practical
capacities that can be viewed as one-track dispositions in just the
same way as the tacitly known principles of a semantic theory.
Although similar ideas had been mooted by other theorists (two
very different examples can be found in Stich 1981, which distinguishes propositional attitudes from what Stich terms subdoxastic
states, and in the distinction between analog and digital content
made in Dretske’s 1981 discussion of the content of perception),
Evans offers an original and sustained development of the idea that
there might be content-bearing states that do not involve the exercise
of concepts and that are fundamentally different from conceptual
states.8 Campbell’s contribution to this volume (Chapter 6) explores
Evans’s distinction between conceptual and nonconceptual content
in the context of Evans’s discussion of Molyneux’s question (the
question, discussed by Locke, of whether a congenitally blind person able to discriminate certain shapes by touch would be able,
were they suddenly to acquire the capacity to see, to discriminate
the same shapes using vision alone).
8
Evans’s discussion of nonconceptual content in VR inspired a number of theorists to explore the notion further. A number of the contributions to the ensuing
debate are collected in Gunther 2003.
01-Luis-Intro.qxd
18
22:02:05
9:16 AM
Page 18
José Luis Bermúdez
PAPERS
A central theme in VR is the project of reconciling two ideas frequently treated as incompatible—the idea that singular terms are
Russellian, on the one hand, and the idea that they have Fregean
senses, on the other. As Evans freely acknowledges (CP 291 n.), it
was John McDowell who originally proposed to ascribe Fregean
sense to Russellian singular terms (McDowell 1977). Fittingly,
McDowell’s contribution to this volume (Chapter 1) assesses
Evans’s claim in the first chapter of VR that this was in fact Frege’s
considered view of singular terms, despite his apparent willingness
to ascribe sense to empty names.
McDowell agrees with Evans that the Russellianism discernible
in the early (pre “On sense and reference” (1892)) Frege remains
“submerged” in Frege’s later thought, rather than being definitively
discarded. It is true that Frege does seem explicitly to countenance
the possibility of sentences with empty names expressing genuine
thoughts, suggesting that we can understand what is going on
here by analogy with fictional discourse. But McDowell agrees with
Evans that this is fundamentally incompatible with how Frege
understands thoughts. A thought, for Frege, is a postulate to the
effect that the world is a certain way. The way the world is postulated to be is the truth condition of the thought—the way the world
has to be for the thought to be true. To judge the thought is to
judge that the truth condition holds. But, McDowell argues,
we cannot apply the notion of a truth condition to the putative
thoughts expressed by sentences with empty names. If the putative
thought does indeed have such a truth condition, then there seems
every reason to hold that the condition is not satisfied if the sentence expressing the thought contains an empty singular term—in
which case the putative thought would come out as false, rather
than as lacking in semantic value. This means that the putative
thoughts expressed by sentences with empty names cannot have
truth conditions—and the very sketchy comments that Frege makes
about fiction give us no handle on what thoughts without truth
conditions would be, or what it would be to judge them.
In addition to this line of argument (which develops points made
by Evans at 1982: 24–5), McDowell identifies the stress in Evans’s
Frege on the object dependence of singular thoughts as part of a
broader intellectual project—the project of “integrating our rational
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 19
Introduction
19
powers with our natural situatedness in the world”. These stem from
McDowell’s distinctive way of understanding Frege’s introduction
of the notion of sense. For McDowell, Frege’s prime motivation for
introducing the notion of sense is to accommodate the normative
requirements of rationality.
The controlling aim of Frege’s introduction of Sinn is to provide a conception of thoughts—possible contents of propositional attitudes and speech
acts—that, in conjunction with a repertoire of concepts of kinds of propositional attitudes partly explained in terms of rational relations between
them (for instance, that rationality precludes believing and disbelieving, or
withholding judgment with respect to, the same thing), yields descriptions
of ways in which minds are laid out such that the descriptions put a subject’s rationality in question only when the subject’s rationality is indeed
open to question. (p. 00 below)
A thinker can rationally judge an object both to be F and not to be
F if the thinker is thinking about the object in a different way
(under a different mode of presentation) on each occasion. But this
does not imply that we should characterize those modes of presentation in completely object-independent terms. Quite the contrary:
to entertain a thought is already to be directed at the world, and in
the case of a singular thought that directedness involves contextual
relations that can only hold if the object exists. Traditional interpretations of Frege have not paid sufficient attention to these contextual factors (thereby losing touch with our “natural situatedness
in the world”), while revisionist accounts of reference and content
have lost touch with the idea that thinking involves the exercise of
rationality. Evans’s Frege tries to do justice to both constraints.
As we saw in the first part of the introduction, Evans does not
think that all referring expressions are Russellian. Descriptive
names (such as “Julius”, introduced by the stipulation that “Julius”
refers to the inventor of the zip) are referring expressions, since they
satisfy the criterion of rigidity, and yet can plainly have a sense in
the absence of a bearer. The implications that this has for the theory
of truth are the subject of Mark Sainsbury’s “Names in Free
Logical Truth Theory” (Chapter 2). As Evans himself notes, the
existence of descriptive names in a language means that a truth
theory for that language cannot have a classical logic. The axiom
for “Julius” takes the form:
(3) For all x (“Julius” refers to x iff x Julius)
01-Luis-Intro.qxd
20
22:02:05
9:16 AM
Page 20
José Luis Bermúdez
In classical logic (3) entails that Julius exists, contravening our
assumption that descriptive names are not Russellian. Evans proposes a truth theory formulated in a negative free logic (a free logic
is a logic in which the constants do not carry existential import, so
that it is not the case that, for a given constant a, ∃x (x a) ). Within
a negative free logic, even though (3) does not entail the existence
of Julius, the existence of a non-descriptive and hence Russellian
name such as “Hesperus” is nonetheless entailed by an axiom
such as
(4) “Hesperus” refers to Hesperus,
since (4) is false if Hesperus does not exist. The significant difference between (3) and (4) is that (4) is a simple sentence, whereas
(3) is a complex sentence (a simple sentence is one in which an
n-place predicate is concatenated with n simple names). In negative
free logic it remains the case that every simple sentence with an
empty name is false.
Sainsbury’s chapter explores a tension between Evans’s need for a
free logical framework, on the one hand, and his emphasis on
truth theories as theories of interpretation, on the other. In using
a truth theory as a theory of interpretation, we use the T-theorems of
the theory rather than its axioms. That is, in interpreting a speaker’s
utterance of, say, “Hesperus is large”, we deploy the T-theorem
(5) “Hesperus is large” is true iff Hesperus is large
without proceeding via an axiom such as (4). Even though the
interpreter typically uses such an axiom in deriving the T-theorem,
the interpretation will exploit the theorem regardless of how it
is derived. The problem is that the distinction between Russellian
and non-Russellian names cannot be captured at the level of the
T-theorems in a free logical framework. In the case of (5) the righthand side of the biconditional is a simple sentence, and hence false
in the (counterfactual) case where Hesperus does not exist. Since in
such a situation “Hesperus is large” will also be false, the biconditional will itself be true. As Sainsbury puts it, this obscures the distinction between Russellian and non-Russellian names in a way that
seems to “emasculate” Evans’s insistence on the object dependence
of non-descriptive names, given his fundamental claim that an
utterance with an empty singular term fails to express a genuine
thought.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 21
Introduction
21
The resolution of the tension that Sainsbury offers on Evans’s
behalf exploits a principled modification of the standard truth theories that he (Sainsbury) has developed in previous writings in
order to accommodate demonstratives and indexicals (see the papers
in Sainsbury 2002). According to Sainsbury, the T-theorems for
sentences involving demonstratives such as “that man” contain two
parts. The first part essentially sets the scene for the utterance in
question, by specifying the person whom the speaker is picking out,
the place and time, and so on, while the second part gives the truth
condition. So, for example, a T-theorem for Hermione’s utterance of
the sentence “That man is disgraced banker” might take the form:
(6) Speaking of Bob on 1 January 2004, Hermione said that
he was a disgraced banker, and her utterance is true iff Bob is
a disgraced banker whose fall from grace took place prior to
1 January 2004.
Since this idea of a two-part T-sentence is independently motivated,
Sainsbury feels able to apply it to mark the distinction between
Russellian names and descriptive names. The basic idea is that
the object dependence or object independence of any given object
language name will be given by the first, scene-setting part of the
relevant T-theorem, in virtue of indications of the scope of the corresponding metalanguage names. Sainsbury follows Evans in using
Russell’s own scope-indicating device of square brackets. Prefixing
an expression with a constant in square brackets indicates that the
expression is false unless the constant bound by the square brackets
has a bearer. So, for example, “[a] ~ Fa” will be false if “a” fails to
refer, while “~ [a] Fa” will be true in the same circumstance.
Clearly, Russellian names take the widest possible scope, and this
will be clearly indicated in the appropriate T-theorems. A specimen
such T-theorem is
(7) [Hesperus] “Hesperus is large” is true iff Hesperus is large.
The scope indication would be different for descriptive names, as
follows:
(8) “Julius did not invent the vacuum cleaner” is true iff [ Julius]
Julius did not invent the vacuum cleaner.
(7) comes out as false if “Hesperus” fails to refer, while (8) is true
if “Julius” fails to refer, thus giving the right result. As Sainsbury
01-Luis-Intro.qxd
22
22:02:05
9:16 AM
Page 22
José Luis Bermúdez
recognizes, his emendation imposes modifications elsewhere in
Evans’s semantics, but he argues that these remain well within the
spirit of Evans’s project. The final section of his paper explores
Evans’s motivations for thinking that natural language names must
be either Russellian or descriptive.
Although Evans is studiously attentive to distinctions within the
class of referring expressions (such as the distinction between
Russellian names and descriptive names) and to the differences
between referring expressions and other types of singular term,
such as definite descriptions, he is curiously silent on the question
of whether there can be such a thing as plural reference. He is
emphatic that not all singular terms are referring expressions, but
seems unquestioningly to accept that all referring expressions must
be singular terms. Ian Rumfitt’s contribution “Plural Terms: Another
Variety of Reference?” explores the case for what he terms referentialism about plurals. The plurals in question include compound
names, such as “William and Mary”, collective names such as “the
Aleutian islands”, compound expressions such as “my teammates
and I”, and plural indexicals and demonstratives (“these flowers”
and “we”). (Rumfitt leaves to one side plural descriptions such as
“the men who independently discovered the calculus”.) There are
two ways of defending referentialism about plurals. The first way
is to treat the referent of a plural term as semantically singular, so
that plural terms come out as singular terms that refer to a distinctive type of object. The most obvious candidate objects are sets (e.g.
the set of flowers to which I am currently pointing), mereological
fusions (e.g. the fusion of the members of my cycling team), and
aggregates (where aggregates, unlike sets, are physical objects with
spatio-temporal locations). Rumfit rejects all three versions of
singular referentialism in favor of a version of referentialism characterized by the following three theses:
(i) Semantic predicates, such as “designate” and “satisfy”, remain
constant in sense whether they attach to a plural subject or a
singular subject.
(ii) In the principles that state the designations of plural terms the
expressions that follow the verb “designates” (or “refers to”)
are themselves plural expressions.
(iii) The designation relation does not distribute, so there is no
entailment from “␣ refers to x & y & z” to “␣ refers to x”.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 23
Introduction
23
In defending this type of referentialism, Rumfitt engages (in §4)
with Dummett’s neo-Fregean argument that plural phrases should
be understood predicatively rather than referentially, and provides
a limited defence (in §§5 and 6) of Boolos’s proposal to regiment
arguments involving plurals by using second-order quantifiers and
predicate variables.
In the final section of his paper Rumfitt deploys Boolos’s secondorder account of the logic of plurals to explore how plurals behave
in modal contexts. In the context of Evans’s understanding of referring expressions, this is a very important question. As we saw earlier,
Evans holds that genuine referring expressions must be rigid. The
thesis has some plausibility for singular terms, but, as Rumfitt notes,
the admissibility of plural referring expressions has a potentially
disruptive effect. If there are non-rigid plural referring expressions,
then they can only be accommodated by a semantic theory that relativizes reference to a possible world in precisely the way that
Evans tries to rule out. Rumfitt himself argues that some varieties
of plurals are rigid, but he is agnostic about others. What would it
be for a plural term to be rigid? Rumfitt holds that the rigidity of
singular terms is given by appropriate instances of the following
two schemata:
(R1a) x a → ⵧ (Ex → x a)
(R1b) ◊ (x a) → x a
The first schema is effectively the principle of the necessity of identity
(with “Ex” to be read as “x exists”, while the second schema states
that no object could possibly be identical to a unless it is actually
identical to a. There are, Rumfitt holds, plural analogues of these
principles. In the following “Tx” is to be read as “x is one of the
T-things” and “E2 (T)” as “the T-things exist”.
(R2a) Tx → ⵧ (E2 (T) → Tx)
(R2b) ◊ Tx → Tx
In English, (R2a) states that if something is one of the T-things,
then, necessarily, if the T-things exist, x is one of them, while (R2b)
states that it is possible for something to be a T-thing only if that
thing is actually a T-thing.
The simplest case comes with compound names formed by
concatenating two or more rigid singular terms (as in “William and
Mary”). Here, Rumfitt argues, the compound name inherits the
01-Luis-Intro.qxd
24
22:02:05
9:16 AM
Page 24
José Luis Bermúdez
rigidity of its singular constituents, and he offers a derivation to
show how two singular terms each satisfying (R1a) will concatenate
to form a compound term that satisfies (R2a). There is a certain
intuitive plausibility in the claim that, if William is one of William
and Mary, then he will be one of William and Mary in any possible
world in which William and Mary exist. But other varieties of
plural terms are less straightforward. Consider collective names
such as “The Aleutian islands”, for example. Is it really the case
that, if a given island x is one of the Aleutian islands, then in
any possible world in which the Aleutian islands exist, x must be
one of them? Are there really no possible worlds in which there are
more Aleutian islands than there are in this possible world? Plural
demonstratives are also problematic. Let “these people” pick out
the people in the line at the fast food restaurant. It seems plausible
to say that had Martha not decided that the line was too long,
Martha might have been one of them. But this is incompatible with
(R2b). As Rumfitt notes, however, arguments in this area need to
be formulated carefully. It is important not to prejudge the issue by
tacitly assimilating plural terms to plural descriptions, for example.
If “the Aleutian islands” is read as “the group of islands off
the west coast of Alaska”, then it seems hard to deny that there are
possible worlds in which an island that is actually one of the
Aleutian islands is not in the appropriate archipelago. And we also
need to make sure that we are not tacitly importing additional uses
of the demonstrative, imagining an utterance of “these people” made
to a different group of people in line for the fast food restaurant—
a group that on this occasion includes Martha. Rumfitt is surely
correct when he says that these issues will need to be investigated
further before we can make a full assessment of Evans’s account of
referring expressions.
Although Evans does not discuss pronouns in VR, his extended
discussion of pronouns in “Pronouns, quantifiers, and relative
clauses” (1977) and “Pronouns” (1980b) is continuous with his
discussion of proper names, demonstratives, and indexicals. The
second paper was written in an interdisciplinary spirit, aiming to
build bridges between the philosophically motivated semantics he
proposed for pronouns and the extensive discussion of pronouns
and anaphora in linguistics. As Kenneth Safir brings out in his
paper “Abandoning Coreference” (Chapter 4), Evans’s treatment
of pronouns has had considerable influence within linguistics.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 25
Introduction
25
Evans made two fundamental contributions to the study of pronouns.
The first (discussed briefly above) is his Frege-inspired development
of a co-referential treatment of bound pronouns, which has the theoretical advantage of allowing both bound pronouns and pronouns
with a singular term as antecedent to be understood in referential
terms. To take an example that Safir also discusses, Evans’s semantics allow us to give a uniform account of the pronoun “his” in the
two sentences “John loves his mother” and “Everyone loves his
mother”. Evans’s second contribution was to identify a class of pronouns that do not fit this model. These are what he termed E-type
pronouns. These are pronouns with quantifier antecedents that
nonetheless do not appear to operate as bound variables. Evans
gives the following example.
(9) John owns some sheep, and Henry vaccinates them.
We cannot, Evans argues, understand (9) as saying of some sheep
that they are such that John owns them and Henry vaccinates them.
This would leave open the possibility, which (9) rules out, that John
owns some sheep that Henry does not vaccinate. Evans proposes a
different semantics for E-type pronouns, according to which they
are singular terms whose denotation is fixed by a description
recoverable from their quantifier antecedent. So, the denotation of
“them” in (9) is fixed by the description “sheep that John owns”.
In “Pronouns” (1980b), Evans uses his semantics of pronouns to
argue against an influential treatment of pronouns by Howard
Lasnik (1976). Lasnik argued that relations of coreference between
pronouns and their antecedents were not fixed by linguistic rule, but
instead by pragmatic, extralinguistic factors. The paradigm case of
pronominal coreference for Lasnik would be a sentence such as
(10) I’m glad he’s gone
said of someone who has just left the room, where the person who
has just left the room is suitably salient to both speaker and hearer.
Of course, coreference cannot be completely unconstrained, and
Lasnik proposes the following rule excluding certain forms of
coreference;
(11) A name cannot be c-commanded by a coreferent noun phrase.
The technical notion of c-command is explained below (pp. 000–00),
but the basic idea is that the position of pronouns in the hierarchical
AQ: Please
update page
number.
01-Luis-Intro.qxd
26
22:02:05
9:16 AM
Page 26
José Luis Bermúdez
architecture of a sentence places restrictions on the noun phrases
with which they can corefer. Evans makes a number of objections
to Lasnik’s proposal. He is concerned, for example, that it completely obscures the commonalities between pronouns with singular antecedents and pronouns with quantifier antecedents, arguing
that these two types of pronoun form a semantic natural kind completely distinct from pragmatic uses of pronouns. In this context,
Evans makes an important distinction (emphasized by Safir)
between dependent reference and intended coreference. A pronoun
is referentially dependent upon a noun phrase when it corefers with
that noun phrase and picks up its referent from it. But there can be
intended coreference between a noun phrase and a pronoun without dependent reference. A pronoun can corefer with a noun
phrase even though it does not derive its reference from that noun
phrase. Evans gives the following discourse as an example:
(12) What do you mean John loves no one? He loves John.
It seems clear (and it follows from Lasnik’s non coreference rule) that
“He” cannot be referentially dependent upon the second occurrence
of “John”. Nonetheless, there can be intended coreference between
the two (because the pronoun is referentially dependent upon the
first occurrence of “John”, which stands in the relation of intentional coreference to the second occurrence of “John”).
In opposition to Lasnik, therefore, Evans proposes that all
cases of coreference can be understood in terms of a linguistic rule
enforcing a dependency relation and a principle blocking dependent reference in certain circumstances. Safir takes issue with Evans’s
central claim that coreference is enforced by linguistic rule, and (as
the title of his paper suggests) argues that the notion of coreference
is not the correct notion to use in thinking about pronouns. On the
general picture that Safir favors, a number of linguistic principles
constrain the possibilities of coconstrual (the notion he prefers to
coreference, for reasons that will become apparent shortly), but
there are no rules mandating the coconstrual of a pronoun with an
antecedent (or following) noun phrase. One of these constraining
principles is effectively Evans’s reformulation of Lasnik’s non coreference rule in terms of dependent reference. Safir terms this the
Independence Principle (IP):
(13) IP: If X c-commands Y, then X cannot depend upon Y.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 27
Introduction
27
Yet, Safir argues, the Independence Principle must be supplemented.
He notes that those contexts, such as (12), where there is intended
coreference without referential dependence are all contexts where
the coreference is surprising—where there is, as he puts it, an
expectation of non covaluation. The Independence Principle does
not explain where there should be such an expectation. We can
obtain such an explanation, however, if we take the failure of
dependence in such cases to be generated by what he terms the
Form-to-Interpretation Principle (FTIP):
(14) FTIP: If X c-commands Y and Z is not the most dependent
form available in position Y with respect to X, then Y cannot
be directly identity dependent on X.
Safir proposes that expectations of non covaluation arise in just
those cases where the failure of referential dependence is generated
by FTIP—this is what he terms Pragmatic Obviation (although it
should be stressed that the FTIP itself is not a pragmatic principle
in the sense made familiar by Grice, but rather an algorithm for
computing dependency relations between formal representations).
One of the key claims of Safir’s chapter is that Evans was fundamentally mistaken in thinking about the behavior of pronouns in
terms of coreference, where two terms are coreferential just if they
both refer to the same object/person. Safir argues that a number of
linguistic phenomena show that there can be co construal without
coreference. These are cases where a pronoun is to be understood in
terms of the noun phrase on which it is referentially dependent, even
though the pronoun cannot strictly speaking be described as referring
to the referent of that noun phrase. One such linguistic phenomenon
is effected by what Safir terms proxy readings of pronouns, where the
pronoun refers to an object that in some sense stands proxy for the
object picked out by its antecedent. Here are two examples:
(15) As they strolled through the wax museum, Fidel could not
help thinking that he would have looked better in a uniform,
and Marlene could not help thinking that she would have
looked better without one.
(16) Patton realized that he would be vulnerable to a flanking
movement.
In (15) “Fidel” and “Marlene” refer to the relevant individuals,
while the pronouns refer to their respective waxwork models.
01-Luis-Intro.qxd
28
22:02:05
9:16 AM
Page 28
José Luis Bermúdez
In (16) it is Patton’s army, which is commanding from a distance,
rather than Patton himself, that is vulnerable. The pronouns are
referentially dependent upon their proper name antecedents, even
though they do not refer to those antecedents.
Most of Evans’s discussion of referring expressions in VR is
taken up with demonstratives (such as “that” and “this”, whose
reference is fixed in part by an accompanying ostension) and indexicals (such as “now”, “here”, and “I”, where reference is fixed as a
function of the context of utterance), a topic that he also takes up
in “Understanding demonstratives” (1981b). José Luis Bermúdez’s
contribution to the volume (Chapter 5) focuses on Evans’s discussion of the first person pronoun. The overall aim of Evans’s treatment of indexical expressions is to show, in opposition to direct
reference theorists and almost all contemporary philosophers of
language, that we can and should apply the Fregean model of sense
to indexicals and demonstratives. Evans’s position is particularly
significant, since demonstratives and indexicals are frequently
taken as prime examples of referring expressions that do not have a
Fregean sense. For Evans, the need for the notion of sense arises
because the successful use of indexicals and demonstratives to pick
out objects is subject to the constraints imposed by Russell’s principle.
The requirements of Russell’s principle are met in different ways by
different types of indexical and demonstrative expression, but
Evans’s treatment of “I” incorporates many of the points that
emerged in discussing other types of expression, such as “here” and
“this”. Evans takes a very metaphysical approach to explaining the
sense of the first person, treating the task as requiring an exploration of the nature of self-consciousness.
Evans’s account of the discriminating knowledge of oneself
required to use and understand the first person pronoun has three
components. We can understand the first as a way of fleshing out
Frege’s well-known remark that “everyone is presented to himself
in a special and primitive way in which he is presented to no one
else” (Frege 1918: 25). According to Evans, everyone who uses
the first person pronoun with understanding does so in virtue of
ways they have of thinking about themselves that are both primitive and not available to anyone else. What makes this possible is
that self-conscious thinkers are suitably receptive to certain types of
information—information that is distinctive because it feeds into
judgment without requiring the thinker to identify a particular
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 29
Introduction
29
object as the source of that information. In the case of information
about oneself that is derived from autobiographical memory, somatic
proprioception, or introspection, for example, there can be no
question but that one is oneself the source of the relevant information. There is no gap between judging on the basis of the appropriate information channels that there is F-ness, and judging that one
is oneself F. These sources of information are immune to error
through misidentification relative to the first person pronoun, to
use the terminology that Evans adopts from Shoemaker (1968).
The second component in Evans’s picture has to do with the
“output” side of those first-person judgments for which information
that is immune to error through misidentification serves as input.
As Perry and others have emphasized, first-person thoughts have
particular and direct implications for action. This is particularly
evident when we are dealing with judgments based on information
that is immune to error through misidentification, as with Perry’s
own well-known example of my seeing that a bear is coming
towards me. There is no sense in which I can judge that the bear is
coming towards someone and wonder whether I am that person.
By the same token, nothing further is needed to rationalize an
immediate reaction.
Since we are capable of entertaining thoughts about ourselves
that involve possibilities to which we have no informational
links, the sense of “I” cannot be exhausted by our sensitivity to
information that is immune to error through misidentification in the
required sense. Whereas judgments based on information sources
that are immune to error through misidentification do not involve
identifying a particular person as oneself, these more complex
judgments do involve discriminating knowledge of oneself, and
hence bring us back to Russell’s principle. Evans’s account of how
Russell’s principle is met for the first person hinges on the subject’s
ability to locate himself within the objective spatio-temporal world.
As he puts it, “to know what it is for [␦ I] to be true, for arbitrary ␦,
is to know what is involved in locating oneself in a spatio-temporal
map of the world (Evans 1982: 211). This knowledge, which Evans
analyzes in terms of the ability to bring egocentric and objective
ways of thinking about space into harmony, allows a thinker to
determine the truth condition of any thought about himself
(whether that thought involves predicates that are susceptible to
error through misidentification, or immune to such error).
01-Luis-Intro.qxd
30
22:02:05
9:16 AM
Page 30
José Luis Bermúdez
Evans’s stress on sensitivity to information that is immune to
error through misidentification leads him to affirm the privacy of
“I”-thoughts. Since the sense of the first-person pronoun depends
upon the ability to exploit sources of information that concern
only the thinker and are available only to the thinker, it follows
that “I”-thoughts cannot be shared. Whereas Evans claims that the
privacy of “I”-thoughts does not compromise their objectivity,
Bermúdez argues that the shareability of “I”-thoughts is required
both by the requirements of communication and by the need for
objective thoughts to exist and have a truth-value independently of
anyone thinking them. This leads him to propose what he terms the
symmetry constraint upon a satisfactory account of “I”-thoughts.
The symmetry constraint requires the token-synonymy (in a given
context) of an utterance that I would express with the first-person
pronoun and an appropriately related utterance that you might
express with the second-person pronoun, so that your utterance
of “You are F” comes out as synonymous with my utterance that
“I am F”, in a way that allows you to utter the negation of what
I utter in the relevant context.
Although Evans’s own account of the sense of “I” fails to meet
the symmetry constraint, Bermúdez shows how it can be modified
to meet it. He argues that Evans is wrong to include the information component in the sense of “I”. Consider a first-person judgment of the form “I am F”, where the information that one is F is
derived from sources that are immune to error through misidentification. An account of the thought expressed will have to reflect the
fact that it is derived from information that is appropriately
immune to error through misidentification. But Bermúdez suggests
that it is part of the sense of the predicate, rather than of the first
person pronoun. Where the property of F-ness is such that there is
no gap between judging on the appropriate grounds that F-ness is
present and judging that one is oneself F, an account of what it is to
possess the corresponding concept (the concept of F-ness, or the
sense of “F”) will include a first-person clause to that effect.
Shifting the information component away from the sense of the
first-person pronoun in this way means that privacy is not an
immediate consequence of Evans’s account of “I”, but it does not
yet show how it might meet the symmetry constraint. Bermúdez
argues that the location component of the sense of “I” allows the
symmetry constraint to be met. When you describe me as being F,
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 31
Introduction
31
your use of the second-person pronoun depends upon your knowledge of what it would be to locate me on a spatio-temporal map
of the world. This knowledge, which is part of the sense of the
second-person pronoun “you”, is the very same knowledge that
underpins my use of “I”. Your use of a particular token of “you” in
a given context and my use of a particular token of “I” in that same
context both exploit the same practical ability.
Evans’s innovative use of the concept of information to illuminate
the nature of content is at the forefront of John Campbell’s contribution, “Information Processing, Phenomenal Consciousness, and
Molyneux’s Question” (Chapter 6). Campbell addresses some of
the central issues raised by Evans’s paper “Molyneux’s question”
(1985b), exploring how Evans’s approach to Molyneux’s question
is informed by his account of the content of conscious perceptual
experience, as developed in VR. Campbell takes issue with Evans’s
account of the content of perceptual experience, and in particular
with what he sees as its conflation of subpersonal informationprocessing content and personal-level experiential content.
The underlying problem that motivates Evans in “Molyneux’s
question” has to do with the concepts of shape. It seems clear that
our concepts of shape are derived from our experience of shape,
and that we can have this experience through either the modality of
touch or the modality of vision. Even though our tactile experience
of shape is very different from our visual experience of shape, we
appear to have unitary concepts of shape, rather than (as Bishop
Berkeley suggested) specific concepts of shape tied to specific sensory modalities. But the fundamental differences in the phenomenal
character of tactile and visual experience mean that there is a genuine and important question as to how the shape concepts acquired
and used on the basis of vision can be the same as those acquired
and used on the basis of touch. This is the question that Molyneux
originally posed by asking whether whether someone born
blind who has learned to discriminate certain shapes by touch
would be able, were they suddenly to acquire the capacity to see, to
discriminate the same shapes using vision alone.
Campbell presents Evans as arguing from the concept of egocentric
space to the unitary nature of our shape concepts (and hence to an
affirmative answer to Molyneux’s question). Our shape concepts
have the content they do in virtue of their connections to our egocentric ways of thinking about space. These egocentric ways of
01-Luis-Intro.qxd
32
22:02:05
9:16 AM
Page 32
José Luis Bermúdez
thinking about space are themselves a function of how we behave
within our immediate environment and, as Evans puts it, “there is
only one egocentric space, because there is only one behavioral
space” (1982: 160). The content of shape concepts is a subspecies
of egocentric spatial content, and since egocentric spatial content
has to be understood through the ways we act upon the environment, the unitary nature of our behavioral repertoire secures the
unitary nature of our shape concepts. We do not have distinct
shape concepts for the tactile and visual modalities, because we do
not have distinct behavioral repertoires associated with touch and
vision. As Campbell brings out, however, there are two ways of
interpreting the relation between egocentric spatial content and
behavior. We can effectively assimilate the two, so that the egocentric spatial content of perception is actually constituted by its
implications for action. On this view, perceiving an object at a particular egocentric location is to perceive it as something that might
be reached by a certain set of movements, or that might be acted
upon in certain ways were one to make the appropriate set of movements. Or we can interpret the distinction so that the egocentric
spatial content of perception is the categorical ground of actions
derived from perception, so that we perceive an object as something
that can be reached by an appropriate set of movements precisely
because we perceive it as having a particular egocentric location.
Campbell notes that Evans’s response to Molyneux’s question
works only on the first reading, but argues that the second reading is
preferable.
Putting this objection to one side, however, Campbell finds a
deeper problem with Evans’s argument. The issue here is the extent
to which sameness of egocentric spatial content can properly be
expected to be transparent to the perceiving subject. Evans’s argument from egocentric spatial content will yield an affirmative
answer to Molyneux’s question only if the perceiving subject is
himself able to identify the sameness of egocentric spatial content
that Evans claims to exist. Even if we grant that visual experience
of shape and tactile experience of shape ultimately have the same
content, there is a question whether this sameness of content is
manifest to the perceiver. When Evans discusses conceptual contents (the type of contents that can be the objects of propositional
attitudes and the meanings of sentences), he holds them to be constrained by what he calls the Intuitive Criterion of Difference,
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 33
Introduction
33
which entails that if two sentences express the same thought, this
must be manifest to the thinker. But as Evans stresses, the content
of perceptual experience is not conceptual content; rather, it is nonconceptual content. As we saw in the first part of the Introduction,
nonconceptual content is not individuated by the holistic considerations of inferential role associated with conceptual thought. The
question arises, therefore, whether nonconceptual content has
the transparency property. At this point Campbell finds serious
tensions in Evans’s discussion. Evans has a unitary notion of nonconceptual content applicable both to subpersonal computational
states (such as those involved, for example, in the subpersonal
computation of the distance and direction of the source of a sound)
and to conscious perceptual experiences (such as one’s hearing a
sound as coming from an object located at a certain distance and in
a certain direction). The picture that Evans develops is one on which
subpersonal information-processing contents become conscious
when they become available to a “thinking, concept-applying,
and reasoning system” (1982: 158). Nonconscious subpersonal
information-processing contents clearly lack the transparency
property. One might plausibly expect the specialized mechanisms
responsible for processing tactile information to be insulated from the
mechanisms that process visual information. Yet Evans thinks that
once they become conscious, their sameness or difference is manifest
to the thinking and perceiving subject. Campbell argues that this is a
mistake—a mistake that reflects a more general problem with Evans’s
discussion of non conceptual content. The mistake comes from
assimilating the content of perceptual experience to the content of
subpersonal information-processing modules. Campbell argues that
we need to recognize that the content of perceptual experience is
fundamentally different both from the content of subpersonal modules
responsible for processing sensory inputs and from the content of
beliefs and other propositional attitudes. These three types of content
are related in complex ways that have not yet been adequately
explored. The first step in clarifying these relations is making sure that
they are kept carefully distinguished.
Whereas Campbell’s contribution is focused on the nature of
perceptual experience and the particular type of content that it
has, Christopher Peacocke’s chapter, “ ‘Another I’: Representing
Conscious States, Perception, and Others” (Chapter 7), explores
how we think about conscious perceptual experience, both our
01-Luis-Intro.qxd
34
22:02:05
9:16 AM
Page 34
José Luis Bermúdez
own and that of others. This investigation continues Peacocke’s
long-standing project (Peacocke 1992, 1999, 2004) of explicating
the possession conditions of concepts. As suggested by the title of
his chapter, which alludes to Zeno’s characterization of a friend
as “another I”, one of the themes of the paper is the relation
between the first-person and third-person aspects of our perceptual
concept (where a perceptual concept is the concept associated with
perception through a particular sensory modality, such as the concept seeing). We can see Peacocke’s overall project as affirming the
priority of the first-person dimension of perceptual concepts, while
avoiding the epistemological and other problems that tend to go
with a first-person/third-person asymmetry. According to Peacocke,
perceptual concepts are what he calls deeply first-personal, because
one’s knowledge of what it is for any arbitrary person to be a
perceiver will depend upon how one knows that one is oneself a
perceiver.
Peacocke’s starting point is what he terms the Core Rule, which
he initially discusses in the context of the concept of seeing. The
Core Rule for vision is that when a thinker sees that it is the case
that p, he is entitled to judge “I am seeing that p”. One important
feature of the Core Rule is that it places the justificatory weight of
a perceptual self-ascription upon the episode of perception, rather
than upon any quasi-perceptual awareness of one’s own perceptual
experience. It is, in Peacocke’s phrase, an “Outside-In” rule. Nor is
this characteristic of the Core Rule a function of the factivity of seeing (of the fact that one cannot see that p unless p is actually the
case). Peacocke offers an Extended Core Rule to cover cases where
a perceiver is entitled to self-ascribe a visual experience even
though that experience is illusory (and may be known to be illusory). The Extended Core Rule states that if a thinker is in a state
that is subjectively as if she is seeing that p, then she is entitled to
judge “I am having a visual experience as of p”.
As Peacocke notes, the Core Rule and the Extended Core Rule
respect Evans’s observation in VR that a subject’s “internal state
cannot in any sense become an object to him. (He is in it.)” (1982:
227). We learn about our own experiences through our experience
of the world, not by taking those experiences as objects.
Nonetheless, the Core Rule differs significantly from Evans’s own
account of perceptual self-ascription. According to Evans, a thinker
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 35
Introduction
35
works backwards from perceptual experience to perceptual selfascription in the following way:
He goes through exactly the same procedure as he would go through if he
were trying to make a judgment about how it is at this place now, but
excluding any knowledge of an extraneous kind. (That is, he seeks to
determine what he would judge if he did not have such extraneous information.) The result will necessarily be correlated with the content of the
informational state which he is in at that time. Now he may prefix this
result with the operator ‘It seems to me as though . . .’. This is a way of
producing in himself, and giving expression to, a cognitive state whose
content is systematically dependent upon the content of the informational
state, and the systematic dependence is a basis for him to claim knowledge
of the informational state. But in no sense has that state become an object
to him. (1982: 227–8)
Evans’s proposal in effect calls for an act of the recreative imagination, expecting the thinker imaginatively to undergo an act of trying to determine how things are that abstracts away from the sort
of background knowledge that thinkers would typically employ to
determine whether the world really is as it perceptually appears to
be. As Peacocke notes, the type of abstraction that Evans describes
seems in many cases to stand in the way of accurate self-ascription
of perceptual experiences. There are many cases where how we
experience the world is directly a function of our background
knowledge. Phonetically there is no difference between the English
sentence “Peter leaped” and the German sentence “Pieter liebt”.
Yet one’s experience differs depending upon how one hears the relevant phonemes, which in turn seems to depend upon information
that would count as extraneous for Evans (such as one’s knowledge
that one is in Heidelberg, Mississippi, for example, rather than
Heidelberg, Germany).
Peacocke shares with Evans (and a number of other contributors
to this volume) a sense that philosophical accounts of concepts and
conceptual abilities must respect psychological facts about how the
concepts in question are acquired and deployed. Perceptual concepts
have been extensively studied in the context of children’s acquisition
of what has come to be known as a “theory of mind”, and Peacocke’s
discussion of the deeply first-personal nature of perceptual contexts
is developed in the context of developmental data on young children’s understanding of perception. The data he considers include,
01-Luis-Intro.qxd
36
22:02:05
9:16 AM
Page 36
José Luis Bermúdez
for example, the fact that 2-year-old toddlers do not seem to be able
to appreciate that they can see something that cannot be seen by
someone on the other side of an opaque screen, or that older children
seem not to understand that someone in a different location who sees
the same object is nonetheless likely to see a different part of that
object. According to Peacocke, these and other phenomena are to be
explained by young children’s naïve use of the Core Rule in otherascriptions. Their ascriptions of perceptual experiences to others are
based on straightforward extrapolation from their own case, whereas
thinkers with a full grasp of perceptual concepts will appreciate that
there are significant constraints and conditions upon applying the
Core Rule in all cases where other perceivers are not in the same
location and under the same conditions. Moreover, full mastery of
the practice of ascribing perceptual states to others requires not
simply working outwards from the application of the Core Rule in
one’s own case, but also appreciating that the other person is herself
capable of deploying the Core Rule to make perceptual selfascriptions. In this respect, as Peacocke notes, the deeply first-personal
nature of perceptual concepts fits well with simulationist approaches
to understanding other minds—both because it suggests a development in terms of imagining what it would be like to perceive the
world from another person’s perspective and because it depends upon
reasoning in accordance with the Core Rule rather than (as a more
theoretical approach would require) explicitly formulating and
applying the Core Rule.
Evans was interested not just in how we should explain what
experience is and how we think about it, but also in the more
Kantian question of whether there are any necessary conditions
upon our having the type of experience that we have. He addresses
this question in “Things without the mind” (1980c), which
addresses Strawson’s discussion of the relation between space and
objective experience in chapter 3 of Individuals (1959) and was
originally published in a Festschrift for Peter Strawson. Evans and
Strawson are both discussing what makes it possible for us to
experience and think about an objective world, where an objective
world is a world that can exist whether or not we are perceiving it.
Both, although in different ways and on different grounds, defend
the Kantian thesis that space is a necessary condition for objective
experience and hence that we cannot think about or experience an
objective world that is not a spatial world.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 37
Introduction
37
As Quassim Cassam points out in “Space and Objective
Experience” (Chapter 8), there are a number of different ways of
developing what he terms the Spatiality Thesis. One might, for
example, argue that the existence of space is a necessary condition
for objective experience. This line of argument, which Cassam
labels STe, attempts to derive metaphysical conclusions from epistemic conditions upon experience. Alternatively, one might argue
that the idea of space (STi) or the perception of space (STp) are
epistemic conditions of objective experience. These different arguments are not completely independent of each other. Since perception is factive, an argument for STp might turn out to be an
argument for STe, although it might equally turn out that what
really matters for objective experience is not the perception of
space (in the factive sense), but rather the experience as of space.
In any case, Evans’s arguments in “Things without the mind” are
focused primarily on STi, whereas Kant’s own arguments in the
Critique of Pure Reason seem to be intended to support STp. (It is
not altogether clear how to interpret Strawson’s arguments in
Individuals, but they seem to be directed at STi.) Cassam’s paper
argues that Evans’s two principal arguments for STp are unsuccessful, but offers in their place a more directly Kantian argument
for STp.
Strawson sets up the issues by asking us to consider a No-Space
world composed only of sounds and inquiring whether it would
be possible for a thinker in such a world to think of it as an objective
world. The conclusion he reaches is that objective experience would
be available to such a thinker only if the auditory universe effectively had built into it an analogue of space (what Strawson calls the
Master-sound). In “Things without the mind”, Evans considers and
rejects Strawson’s arguments, for two reasons. The first is that
the Master-sound is not as quasi-spatial as Strawson claims, since
there is no need for it to be ordered in the way that space must be
ordered. The second is that the Master-sound is not really required,
since what is really important is that the thinker’s experience
should exhibit the type of regularity that would make it appropriate
to think of its course as due simultaneously to how the world is laid
out and to the subject’s movement through it. This second criticism
of Strawson introduces one of Evans’s key ideas in this area. This
is the idea that our perception of the world is bound up with
what he calls “a simple theory of perception”, by which he means
01-Luis-Intro.qxd
38
22:02:05
9:16 AM
Page 38
José Luis Bermúdez
a quasi-theoretical understanding of what make it possible for us to
have the experiences that we do.
Evans’s first argument for STi, which Cassam terms the
Argument from Perception, is based on this idea that a simple theory of perception is an indispensable basis for the idea of existence
unperceived. Such a theory can contain, Evans argues, only two
possible explanations for why something perceptible should not be
being perceived at a given moment. It might appeal to deficiencies
in the perceiver or to the absence of factors causally necessary for
perception. Evans argues that deficiencies in the perceiver alone
cannot yield the idea of existence unperceived, because the subject
has no independent understanding of what it would be to be suitably receptive other than perceiving the relevant object, which
makes citing a lack of receptivity to explain the absence of perception completely uninformative. In contrast, incorporating spatial
notions into the simple theory of perception gives a clear explanation of why a perceptible phenomenon might not be perceived (one
might be at the wrong location, for example). As Cassam points
out, however, Evans gives a convincing account of how a simple
theory of perception incorporating spatial concepts would allow
a thinker to make sense of existence unperceived, but does not
make the case that this is the only way of making sense of existence
unperceived. He finds a similar difficulty for Evans’s second argument for STi (the Substance Argument), which is based on the idea
that we need to think of things in terms of primary properties in
order to think of them as persisting in a way that would allow them
to exist unperceived. It may well be the case, Cassam argues, that
persistence and existence in space go together in our conceptual
scheme, but Evans is trying to establish the stronger conclusion
that existence in space is a necessary feature of any conceptual
scheme that allows for persisting objects that can exist unperceived.
After rejecting Evans’s attempts to establish STi Cassam turns to
a more directly Kantian argument for STp—for a necessity claim
about the perception, rather than the concept, of space. The central
claim here is that the perception of space is necessary for perceiving
objective particulars, because one cannot perceive an objective particular without perceiving it as having spatial properties such as
shape, extension, and location. Cassam recognizes that the thesis
needs to be circumscribed to deal with objective particulars, such
as sounds, that might be thought not to have spatial properties, and
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 39
Introduction
39
accordingly proposes STpb, according to which the perception of
space is necessary for objective experience insofar as it is necessary
for the perception of bodies. It is obviously not the case that a perceiver must perceive all of an object’s spatial properties. In fact,
Cassam argues, it is only the perception of location that can plausibly be argued to be strictly necessary for the perception of bodies.
Cassam suggests that STpb is far more promising than the version
of STi defended by Evans, not least because it offers the possibility
of explaining what might ground the necessity claim. Whereas Evans
follows Strawson in seeing necessities in this area as essentially
conceptual necessities, secured by the structure of our conceptual
scheme, Cassam sees the possibility of an empirical and/or metaphysical basis for the necessity claim. He points to two possible
argumentative strategies. The first highlights the role that information about spatial location has been suggested to play in solving the
so-called binding problem, while the second derives from the fact
that material objects are individuated according to their location.
In each case it appears that we have the material for an explanation
of why the Spatiality Thesis holds.
Turning now to a very different aspect of Evans’s thought,
E. J. Lowe’s “Identity, Vagueness, and Modality” (Chapter 9) evaluates Evans’s very influential one-page article “Can there be vague
objects?” (1980a). It is a familiar idea that our language is vague;
that there are terms for which it is indeterminate whether they
apply to certain objects. There are borderline cases of hairlessness
for which there is no fact of the matter whether they count as bald
or not. Far more contentious is the idea that reality itself might be
vague—that there might be, as Lowe puts it, ontic vagueness in
addition to semantic vagueness. Evans’s article contains an attempted
reductio ad absurdum of the idea that there might be vagueness
actually in the world, in addition to in our description of the world.
The argument is as follows. (I follow Lowe in using the more perspicuous lambda notation over Evans’s own notation. The expression “x[Fx] a” is to be read as “a has the property of being F” and
“∇ϕ” as “It is not determinately the case that ϕ”, with ∆ϕ” as “It is
determinately the case that ϕ”).
(i) ∇ (a b)
(ii) x [∇ (x a)] b
(iii) ¬ ∇ (a a)
01-Luis-Intro.qxd
22:02:05
9:16 AM
40
Page 40
José Luis Bermúdez
(iv) ¬ x[∇ (x a)] a
(v) ¬ (a b)
(vi) ¬ (a b)
Evans’s point is that assuming the indeterminate identity of a and b
means that there is a property that b has (ii) but a does not (iv).
This is the property of being indeterminately identical to a. We are
assuming that it is indeterminate whether a is identical to b, but
there is clearly no indeterminacy about a’s identity to itself. Hence,
an application of the principle of the non-identity of discernibles
yields the conclusion that a is not identical to b (v). Since this
conclusion is reached by a series of truth-preserving steps from
premises that are being assumed to be true, it follows that it is
determinately true, and hence that there is, contradicting (i), a fact
of the matter about whether a is identical to b (vi).
Lowe suggests that Evans’s argument is subtly question-begging.
He grants that there is no indeterminacy about a’s identity to itself,
but notes that Evans’s application of the principle of the non-identity
of discernibles rests upon taking it to be equally determinate that
there is a property that b has but a lacks, viz. the property of being
indeterminately identical to a. Lowe claims, however, that this is
not warranted by the original assumption. If it is indeterminate
whether a is identical to b, then it must also be indeterminate
whether b has the property of being indeterminately identical to a.
One can only assume, as Evans does, that it is fact that there is a
property that a lacks but b has if one tacitly assumes that it is
determinately true that a is not equal to b, which is of course the
conclusion of the argument.
Lowe’s principal charge against Evans’s argument, though, is not
that it is question-begging, but rather that it is invalid. He draws an
instructive comparison with the well-known family of arguments
for the non-contingency of identity inspired by Ruth Barcan and
Saul Kripke. As he notes, Evans’s argument against ontic vagueness
can be easily transformed into an argument for the non-contingency
of identity by reading Evans’s indeterminacy operator (“∇”) as “It is
contingently the case that”. Construed in this manner, moreover,
the argument would not be question-begging in the manner
just identified. If a is contingently identical to b, then it seems
clear that the property of being contingently identical to a must be
different from the property of being contingently identical to b.
01-Luis-Intro.qxd
22:02:05
9:16 AM
Page 41
Introduction
41
Yet there is a fallacy of equivocation between the step from (iii) to
(iv) and the step from (iv) to (v). Continuing to read “∇” as “It is
contingently the case that”, the step from (iii) to (iv) is the move
from “It is necessarily the case that a = a” to “It is not the case
that a has the property of being contingently identical to a”.
Presumably what makes it true that it is not the case that a has the
property of being contingently identical to a is the fact that a has
the property of being necessarily identical to a.
It is here that the putative fallacy of equivocation arises, since
there are two very different properties that might be in play here.
The first is the property, which a shares with b and every other
object, of being necessarily identical to itself. The second is the
property, peculiar to a, of being necessarily identical to a. It is
uncontentious, according to Lowe, that “It is necessarily the case
that a a” entails “a has the property of being necessarily identical to itself”, which validates the move from (iii) to (iv). But this
will not secure Evans’s conclusion, since b also has this property,
thus blocking the application of the principle of the non-identity of
discernibles in the move from (iv) to (v). The step from (iv) requires
a to have the property of being necessarily identical to a. But if this
is the property in question, Lowe argues, the step from (iii) to (iv) no
longer seems plausible. The thesis that a is necessarily identical to
itself is a substitution instance of a law of logic (the law of the
reflexivity of identity). But the thesis that a is necessarily identical
to a is a controversial metaphysical claim (and in fact a close relative
of the controversial metaphysical claim that the argument is trying
to establish). Lowe maintains that Evans’s argument commits an
analogous fallacy.