Who's Afraid of Substitutivity? (Pre-Print)

Who’s Afraid of Substitutivity?

Stefano Predelli

The discussion of substitutivity of co-referential names

has occupied a central position in the current debate on

the semantics of belief reports. In particular, Millian

analyses of such reports are commonly classified according

to their position with respect to this issue. On the one

hand, one is often told, positions like Mark Crimmins’ and

John Perry’s strive to give semantic recognition to our

pre-theoretic intuition that substitutivity is invalid. One

the other hand, so the traditional description of the

current debate continues, the view typically associated

with Nathan Salmon and Scott Soames insists that our

inclinations with respect to certain reports are

unreliable, and conclude that substitution of co-

referential names in attitude reports is truth-preserving.

In this essay, I challenge the foregoing summary of the

relationships between substitutivity and the Millian

1

analyses of belief reports which I mentioned above. In

section one, I summarize the central tenets of the semantic

analysis of belief reports provided by Salmon and Soames on

one hand, and by Crimmins and Perry on the other. I label

the resulting views as, respectively, the Radical View and

the Contextual View. In section two, I explain why both

views are equally committed to substitutivity, taken as a

logical rule of inference. I then introduce a notion

importantly distinct from logical validity, cogency, which

is prima facie able to account for the difference between the

positions under analysis, with respect to the issue of

substitutivity. However, in section three, I argue that,

not unlike the Contextual View, the Radical View is

consistent with the rejection of the cogency of

substitution moves.

1. Two Analyses of Belief Reports

It is an undisputed datum that, in certain situations, an

utterance of, say,

2

(1) Lois believes that Superman can fly

may be correct, but that an utterance of

(2) Lois believes that Kent can fly

may not provide an adequate picture of Lois’ cognitive

states.1 This prima facie evidence has traditionally been

interpreted as problematic for a certain view of proper

names, the so-called Millian view.

According to Millianism, an utterance of a simple

sentence, such as ‘Superman can fly’ is semantically

correlated to a complex and structured entity, its semantic

content. The content expressed by such an utterance depends

(possibly among other things) on the semantic contributions

provided by the expressions occurring in the uttered

sentence. For a Millian, the sole semantic contribution of

a proper name, such as ‘Superman,’ is the name’s referent.

So, taking certain relatively harmless assumptions for

granted, it follows from Millianism that utterances of

‘Kent can fly’ must express the same content as utterances

of ‘Superman can fly’, given that the names ‘Superman’ and

‘Kent’ refer to the same individual.1 For simplicity’s sake, I ignore the complications generated by thefact that the tale of Superman is fictitious.

3

As I summarized it, the Millian view is silent with

respect to the semantic behavior of more complex sentences,

such as (1) and (2). The following, however, are two

plausible theses concerning belief reports. (i) An

utterance of a sentence of the form ‘A believes that S’

expresses a certain cognitive relation, denoted by the

attitude-verb ‘believes,’ involving A and the semantic

contribution provided by the that-clause, ‘that S.’ (ii)

The semantic contribution of a that-clause ‘that S’ is the

semantic content of the embedded sentence, S. These tenets

seem to commit one to the following conclusion: (iii) an

utterance of (1) is true (with respect to a given

circumstance w) iff the pair consisting of Lois and of the

content expressed by ‘Superman can fly’ belongs to the

extension (in w) of the dyadic predicate ‘believes’. Now,

as we have seen, Millians are committed to the claim that

both ‘Superman can fly’ and ‘Kent can fly’ express the same

content. Thus, it seems to follow from the conjunction of

(iii) and the Millian theory of names that (1) and (2) must

have the same truth-conditions, and that, more generally,

4

substitutions of co-referential names must be truth-

preserving.

How can the intuitions described at the beginning of this

section be accommodated within a Millian framework?

According to the view typically attributed to Salmon and

Soames, our intuitions concerning the semantic behavior of

(1) and (2) are deceptive and unreliable (see Salmon 1986,

Soames 1987a, and Soames 1987b). Salmon and Soames concede

that there may be conversational situations in which an

utterance of (1) is appropriate, but an utterance of (2) is

misleading or unsuitable. But, so they argue, it is a

mistake to interpret such a discrepancy between uses of (1)

and of (2) in terms of a difference in their truth-value.

Such a mistake is allegedly grounded on the confusion

between what these sentences say, namely that Lois is

appropriately related to the Millian content that

Superman/Kent can fly, and a different kind of information,

such as the information that Lois is in the relevant

cognitive relation to that content, when it is presented to

her in a certain manner.2

2 According to a particular version of this position, the source of theconfusion lies in the widespread inability to distinguish between the

5

This strategy appeals to a widely accepted analysis of

psychological attitudes, such as beliefs, hopes, and

desires. According to this analysis, which may be labeled

the Representational View of the Mind, psychological attitudes are

three-place relationships, between a cognitive agent, a

content, and a representation of that content. A rational

agent such as Lois may be belief-related to a certain

content, for instance the content K that Superman/Kent can

fly, with respect to representations of that content of a

certain type, say those prompting Lois’ assent to the

sentence ‘Superman can fly,’ and yet fail to be thus

related to K with respect to representations of another

kind, such as those suitably related to the sentence ‘Kent

can fly.’ Salmon refers to the triadic cognitive relation

at issue as BEL. The mistake which Soames and Salmon

attribute to our intuitive approach to a sentence such as

(2) is the confusion between the information it

semantically conveys, namely that Lois is appropriately

content semantically encoded in a utterance, and the information itpragmatically imparts. The relationships between this version of theview under discussion and Nathan Salmon’s own position is more complexthan it is customarily recognized; on this, see Salmon 1986, __, andSalmon 1989, __.

6

related to K, and the misinformation that Lois is in the

BEL relation with that content and a representation of it

of a particular kind, typically of the kind suitably

related to the embedded sentence ‘Kent can fly.’ According

to the view under analysis, then, the relation semantically

denoted by the English predicate ‘believes’ is a two-place

relation, obtained from the three-place cognitive relation

BEL by existential closure of its third argument place,

i.e., the relation xy[z(BEL(x,y,z))] between agents and

contents (see Salmon 1986, 103-118).

Let me refer to the analysis of belief reports emerging

from this approach as the Radical View. More precisely, the

Radical View consists of the following central claims:

(RV1) the content expressed by a simple sentence, such as

‘Superman can fly’, is the Millian content K that

Superman/Kent can fly;

(RV2) the semantic value of the English predicate

‘believes’ is the dyadic relation obtained from BEL by

existential closure, i.e., the relation holding

between an agent and a content iff the former is in

7

the BEL relation to the latter with respect to some

representation of that content for that agent.

(RV3) the semantic value (in a context c) of a clause ‘that

S’ is the semantic content (in c) of the embedded

sentence, S.

(RV4) an utterance (in a context c) of a sentence ‘A

believes that S’ expresses a content made up exactly

of the semantic contributions (in c) of A, of ‘that S’

and of ‘believes.’

Crimmins and Perry have objected to the foregoing

position, on the grounds that accounts in the spirit of the

Radical View ‘deny the accuracy of our strong intuitions

about truth and falsity’ (Crimmins and Perry 1989, 686).3

In opposition to the Radical View, they aim at providing an

analysis of belief reports which ‘honors the intuition that

[both (1) and the negation of (2)] are true,’ i.e., which

is consistent with our intuitive rejection of the rule of

substitutivity. Crimmins and Perry accept both the Millian

analysis of simple sentences such as ‘Superman can fly’,3. They also write: ‘... we depart from a recent trend to explain theapparent truth of [certain attributions] as an illusion generated bypragmatic features of such claims’ (Crimmins and Perry 1989, 685; seealso 698, footnote 16).

8

and the account of belief reports which I summarized in (i)

and (ii), some paragraphs ago. Thus, not unlike the Radical

View, they agree that a sentence ‘A believes that S’

expresses a relation involving A and the content expressed

by S. But they argue that this tenet does not commit them

to the thesis that such a sentence expresses an information

pertaining only to these two components. Hence, so they

claim, the conclusion in (iii) above need not follow, i.e.,

it need not be the case that an utterance of ‘A believes

that S’ is true iff A is in the appropriate dyadic relation

to the content of S. According to Crimmins and Perry, the

English predicate ‘believes’ stands for a triadic relation,

i.e., for the relation holding between an agent A, a

content K, and a representation-type R, iff for some

representation r of type R, the BEL relation holds between

A, K, and r.4 For them, an utterance of ‘A believes that S’

in a context c expresses the information that A is in the

4 According to the picture of the mind defended by Crimmins, thetriadic relation at issue does not involve representation-types, butconcrete cognitive particulars (see Crimmins 1992 and Crimmins 199_).As Crimmins himself recognizes (see Crimmins 19__, __), thisidiosyncratic feature of his position is independent from the semanticanalysis of belief reports he defends, and will be ignored in whatfollows.

9

BEL relation with the content expressed by S and with a

representation of that content of a type salient in c.

Since such a representation-type enters the content of the

utterance, but is not the semantic value of any expression

occurring in the uttered belief report, it is called by

Crimmins and Perry an unarticulated constituent of that

content (Crimmins and Perry 1989, and Crimmins 1992).

I refer to the view summarized in the foregoing

paragraphs as the Contextual View. It holds the following:

(CV1) the content expressed by a simple sentence, such as

‘Superman can fly’, is the Millian content K that

Superman/Kent can fly (as in RV1)

(CV2) the semantic value of the English predicate

‘believes’ is the aforementioned triadic relation

between agents, contents, and representation-types.

(CV3) the semantic value (in a context c) of a clause ‘that

S’ is the semantic content (in c) of the embedded

sentence, S (as in RV3)

(CV4) an utterance (in a context c) of a sentence ‘A

believes that S’ expresses a content consisting of the

semantic contributions (in c) of A, of ‘that S’, and

10

of ‘believes,’ and of a representation-type salient in

c.

Crimmins and Perry enrich the semantic account provided

by the Contextual View with a relatively complex

explanation of the pragmatic mechanisms that may render a

certain representation-type salient in a context. One

feature of such an explanation is important for the topic

of this essay. For Crimmins and Perry, a representation-

type may become salient in a context in virtue of being

triggered by the wording of the embedded sentence. For

instance, in what Crimmins calls de dicto contexts, an

utterance of (1) renders prominent the name ‘Superman,’ and

hence raises to salience a representation-type suitably

related to the property of being the bearer of that name.

On the assumption of this thesis about contextual salience,

it follows from the Contextual View that such an utterance

of (1) says (truly) that Lois is appropriately related to

K, the content that Superman/Kent can fly, with respect to

some representation of a kind appropriately related to the

sentence ‘Superman can fly’. On the other hand, an

utterance of (2) in a de dicto context is alleged to

11

expresses the false content that Lois is related to K the

respect to a representation of the ‘Kent’-type, i.e., for

instance, a representation which Lane associates to the

sentence ‘Kent can fly’. It thus apparently follows that

the Contextual View, equipped with the thesis about

salience-raising under discussion, is able to assign

contrasting truth-values to utterances of (1) and (2),

consistently with ‘our strong intuitions about truth and

falsity.’

In what follows, I shall refer to the claim that an

utterance of a belief report may raise to salience

representation-types importantly related to certain lexical

items occurring in that report, as the Verbal Salience Thesis.

The Verbal Salience Thesis is not a view in the semantics

of belief reports, it is a hypothesis on the pragmatic

mechanisms governing contextual salience. In particular, it

may be defended independently of the Contextual View of

belief reports.5 It may well be a hypothesis even a

defender of the Radical View could accept for independent

5 Note in particular that views similar to the Verbal Salience Thesishave been defended by philosophers that do not subscribe to theContextual View; see for instance Forbes 19__ and 19__.

12

reasons, even though it seems that no commitment to one or

another view about contextual salience would affect her

account of belief-reports. After all, so one may argue, ,

belief reports are not context-sensitive in the Radical

View, so that no feature in the context of utterance may

affect their semantic properties.

According to the summary presented thus far, opposite

attitudes with respect to substitutivity are among the

driving forces for the competing approaches to the

semantics of belief reports I summarized. Both the Radical

View and the Contextual View agree on a Millian analysis of

simple sentences (see RV1 and CV1), and on the

identification of the semantic value of a that-clause with

the content of the embedded sentence (see RV3 and CV3).

However, they sharply disagree on the proper analysis of

the predicate ‘believes’, which the Radical View associates

to a dyadic relation between agents and contents (see RV2),

in contrast to the Contextual View’s suggestion to evaluate

it in terms of a triadic relation involving representation-

types (see CV2). This disagreement seemingly trickles down

to a difference in the truth-conditions assigned to

13

utterances of attributions: one theory, but not the other,

apparently allows for the existence of non-equivalent

belief reports differing only for occurrences of co-

referential names. It may thus appear that the dyadic

analysis of ‘believes’ proposed by the Radical View must

appeal only to those willing to deny semantic importance to

our intuitions regarding failure of substitutivity, and

that the triadic account defended by the Contextualist View

is at least partly motivated by the conviction that an

adequate semantic theory must invalidate substitution

moves. This interpretation, however, deserves closer

scrutiny.

2. Validity and Cogency

Substitutivity is an inference rule, which sanctions the

validity of arguments containing a premise of the form ‘A

believes that S’, and a conclusion of the form ‘A believes

that S*’, where S* can be obtained from S by replacing

occurrences of proper names in S with occurrences of co-

14

referential proper names. According to the traditional

definition of validity, an argument is valid iff whenever

the premises are evaluated as true with respect to a given

context, the conclusion is also evaluated as true with

respect to that context (see Kaplan 1977). The requirement

that the validity of an argument be assessed by evaluating

premises and conclusion with respect to a unique context is

motivated by the desire that our formal definition of

validity conform to our intuitive understanding of logical

relations. For if the definition of validity were deprived

of that requirement, even instances of the obvious

inference rule of repetition would be invalidated, whenever

indexical expressions are at issue. For example, the

premise of the following argument is true with respect to a

context where Bill Clinton is speaking, but its conclusion

is false with respect to a different context, containing,

say, Boris Yeltsin in its agent co-ordinate:

I am an American. Therefore, I am an American.

Similar considerations hold with respect to cases involving

unarticulated constituents. For example, according to

Crimmins and Perry, a typical utterance of ‘it is raining’

15

expresses a content containing among its constituents the

location of utterance, or a certain contextually salient

location (Crimmins and Perry, ). Improper context shifts

would then invalidate trivially valid arguments, such as

It rains. Therefore it rains.

After all, the premise in this argument may well be true

with respect to a context providing Seattle as its salient

location, and its premise may be false when evaluated at a

context where Los Angeles is prominent.6

Given the foregoing definition of validity, it follows

that, appearances notwithstanding, the Contextual View of

belief reports does not invalidate the inference rule of

substitutivity. For, given any sentence of the form ‘A

believes that S’, and any context c, the Contextual View

evaluates such a sentence with respect to c as true iff A

is in the triadic cognitive relation at issue with the

content expressed by S in c, and with some representation

of the kind salient in c. Identical truth-conditions are

assigned by the Contextual View to ‘A believes that S*’

6 For Crimmins’ own discussion of validity in cases involving indexicalexpressions and unarticulated constituents, see Crimmins 1995.

16

(where S* derives from S by substitutions of co-referential

names), with respect to c.

This conclusion is hardly surprising if one keeps in mind

that the Contextual View agrees with the Radical View on

one crucial tenet: the sole semantic contribution of the

that clause in an attitude report consists in the semantic

content expressed by the embedded sentence (see RV3 and

CV3). Thus, according to either view, tampering with the

layout of a belief report may not produce any interesting

difference as to the truth-conditions of that report (with

respect to a given context), so long as the resulting

alterations do not affect the embedded sentence’s content.7

Once a Millian analysis of the embedded sentences in

examples such as (1) and (2) is taken for granted, it

follows that replacements of proper names with co-

referential names are instances of such content-preserving

alterations. Thus, both views are committed to the

conclusion that utterances of (1) and (2), and of similar

7 In this respect, both positions oppose sententialist analyses of thethat-clause’s semantic behavior, such as Mark Richard’s, i.e.,analyses according to which the semantic value of a that-clausedepends, possibly among other things, on the syntactic layout of theembedded clause (see Richard 1990).

17

pairs, must express the same content with respect to any

given context c.

Strictly speaking, then, the Contextual View, not unlike

the Radical View, does not give a semantic account of our

intuitions regarding failure of substitutivity. However,

the summary I have proposed thus far seems to miss some

sort of fundamental difference between the two views under

analysis. Take once again the examples cited at the

beginning of this essay, and repeated here:

(1) Lois believes that Superman can fly

(2) Lois believes that Kent can fly.

According to the Radical View, given Lois’ predicament, an

utterance of (2) is on a par with an utterance of (1), from

the semantical point of view: if one is true, the other

must be true as well. But, so one may point out, this is

not so for the Contextualist View, if the Verbal Salience

Thesis is viable. For, in this case, the Contextualist View

recognizes the possibility of situations in which an

utterance of (1) encodes the true information that Lois is

in the appropriate relation to the content at issue with

respect to ‘Superman’ representations of Superman/Kent, but

18

an utterance of (2) encodes the misinformation that Lois

stands in that relationship with respect to ‘Kent’

representations. Given that (1) and (2) differ only for the

occurrence of co-referential names, it seems that the

theories under discussion do after all give contrasting

verdicts with respect to substitutivity.

This is not, as we have seen, a disagreement having to do

with the validity of substitutivity as a rule of inference.

But another notion, distinct from that of logical validity,

may be helpful in the attempt of clarifying the suggestion

of the foregoing paragraph. It has been introduced, in a

different context, by Nathan Salmon:

… one may define a complementary notion for the

assessment of arguments, one that looks at such

phenomena as the shifting of contexts that

occurs, or may occur, in the actual utterance of

an argument. … Let us call this speech-act

centered notion pragmatic cogency, to distinguish

it from semantic validity. It is not the proper

19

notion of logical validity, but it is not a

useless notion. (Salmon 1995, 14).

Consider now my utterance u1 of (1), and my utterance u2 of

(2). According to the Radical View, so one may point out,

u1 is a true utterance iff Lois is in the appropriate

relation to a certain content K regarding Superman/Kent,

i.e., iff u2 is a true utterance as well. So, it seems that

for the Radical View the move from (1) to (2) is not only

logically valid, but also cogent. In other words, according

to this view, it is not only the case that the sentences

(1) and (2) must agree in truth-value, whenever they are

evaluated with respect to a given context. It must

apparently also be the case that, whenever an utterance of

the former sentence is true, an utterance of the latter may

not be false. Now, what would the Contextual View say on

this issue? As I have summarized it, the Contextual View is

not in itself committed to any particular result concerning

the cogency of u1 and u2. In particular, given that nothing

in (CV1)-(CV4) precludes that a unique representational

type be salient for both utterances, we may not rest

20

assured that, at least in some kinds of contexts, only the

former may be evaluated as true. Such assurance may however

be obtained from the account of contextual prominence

suggested by the Verbal Salience Thesis. For it follows

from the Contextual View, equipped with such an account of

salience, that (at least in contexts of a certain kind) u1

is true iff Lois is related to K with respect to a

‘Superman’ representation, while u2 is true iff she is in

the BEL relation to K and a ‘Kent’ representation. So,

although the uttered sentences may not differ in truth-

value with respect to a fixed unique context, actual

utterances of them may be assigned contrasting truth-

values, thereby rendering the move from one to the other

not cogent.

It finally seems that the contrast between the views

under analysis vis á vis substitutivity has been identified:

unlike the Radical View, the Contextual View, when coupled

with a certain account of contextual salience, may resist

the cogency of substitutions of co-referential names. Note

that the alleged contrast between the two rival positions

has been obtained by comparing the Radical View with an

21

enrichment of the Contextual View, namely with its

conjunction with the Verbal Salience Thesis. So, the

foregoing summary grants that Contextual View is by itself

in no better position than the Radical View. However, so

one may point out, the Contextual View may reach the

desired result by appealing to the resources of a certain

claim about contextual salience, while no such appeal may

be of any help for the defender of the Radical View.

3. Quantifiers and Contextuality

As we have seen, according to the Radical View, the English

predicate ‘believes’ denotes the existential closure of the

triadic BEL relation, namely the dyadic relation that holds

between an agent A and a content K iff A stands in the BEL

relation to K with respect to some representation of K for

A. Thus, for the Radical View, an utterance of a belief

report ‘A believes that S’ is true iff x(BEL(A, K, x)),

where K is the content expressed by S, i.e., roughly, iff

the following is true

22

(3) A is BEL-related to the content expressed by S with

respect to some representation of that content for A.

But the adequate interpretation of sentences such as (3)

depends, among other things, on an appropriate

understanding of the quantified phrase ‘some

representation.’ And quantified phrases, as is well known,

display a semantic behavior which is importantly dependent

on certain features of the situation in which they are

employed.8 For instance, it is a widely discussed feature

of the expression ‘everything’ that it typically concerns

only items within a relevant, contextually determined range

of objects, rather than anything whatsoever in the

universe. A typical utterance of

(4) the thieves took everything

is not intuitively falsified by the existence of objects,

such as the Eiffel Tower or Bill Clinton’s desk, which were

not taken by the thieves. It is rather intuitively

interpreted as asserting that the thieves took (roughly)

all transportable, valuable items in the apartment at

issue. Similarly, the truth-conditions for an utterance of8 For recent discussion of this topic, see Recanati 1996 and Reimer1998.

23

(5) some students did not come to the party

demand more than the sheer existence of individuals who are

students and who did not show up at the party. What is

required for the truth of (5) is that at least one

individual among the relevant students (say, one of the

students in the utterer’s class) failed to perform the

action denoted by the predicate.

There are two traditional approaches to the issue

sketched in the foregoing paragraph. According to what

Stephen Neale calls the implicit approach, certain contextual

restrictions on the class of relevant items are operative,

which constrain the universe of discourse.9 So, a typical

context for (4) supplies as salient only transportable

valuable items in the apartment, and a typical context for

(5) allows the quantified phrase ‘some student’ to range

only over a class of relevant students. According to the

explicit approach, on the other hand, quantified phrases are

elliptical, and ‘the descriptive content is ‘completed’ by

the context.’10 So, in this approach, a typical utterance

of (4) is interpreted as elliptical for an utterance of9 See Neale 1990; see also Reimer 1998.10 Neale 1990, 95, cited in Reimer 1998, 96.

24

(6) the thieves took everything that was valuable and

transportable in the apartment.

The choice between these two approaches to quantified

phrases is not important for my purpose. In what follows, I

shall assume the implicit approach, but my point may be

easily adapted to alternative analyses of quantified

phrases.

According to the Verbal Salience View, an utterance of a

belief-report in appropriate circumstances renders a

certain representation-type salient in the context of

utterance. So, in this view, an utterance of (2) raises to

prominence representations of the content K that

Superman/Kent can fly interestingly related to the sentence

‘Kent can fly’. It is only representations of this kind

that are of interest for the participants in the

conversational exchange at issue, just as only removable

and valuable items in the apartment were of relevance in

our exchange about the burglary. So, given this suggestion,

the existential quantifier in the analysis of (2) provided

by the Radical View may range only over representations of

the relevant kind. Since none among Lois’ representation of

25

K are of the ‘Kent’ type, it follows that an utterance of

(2) in appropriate circumstances may be evaluated as false,

unlike utterances of (1) in similar circumstances. Hence,

in contrast to the account presented in section two, the

Radical View, when coupled with the Verbal Salience Thesis,

is able to yield a verdict of non-cogency for substitution

moves.

Of course, the development of the Radical View I sketched

in this section is neither Soames’ nor Salmon’s. According

to them, the truth-conditions for an attribution ‘A

believes that S’ may well be expressible in terms of a

quantified sentence along the lines of (3), so long as the

quantifiers in it are stipulated to be unrestricted. In other

words, their point is not that of suggesting an equivalence

between different sentences, such as English sentences of

the form ‘A believes that S’ on the one hand, and the

quasi-English sentences obtainable from (3) on the other,

while remaining open to alternative suggestions as to their

common semantic behavior. However, my point here is

neither historical nor exegetical. The issue under

discussion pertains to the relationships between the Radical

26

View and the pragmatic cogency of substitution moves.

Regardless of the version of that view defended by its

founding fathers, it is theoretically important to notice

that its fundamental tenets do not inevitably entail the

position vis á vis substitutivity which is traditionally

associated to it.

Let us go back to the starting point of our discussion,

i.e., points (i), (ii), and (iii). The Contextual View

grants that, one (iii) is obtained, our intuitions must be

abandoned. But this is wrong. One may accept (iii) and

Millianism, and have substitutivity non-cogent.

4. Conclusion

When it comes to logical validity, the Radical View and the

CP-View agree that substitutivity is valid. When it comes

to pragmatic cogency, they may well agree that it is not

cogent.

These comments about substitutivity may also be adapted

to another aspect of the view defended by Crimmins and

27

Perry, i.e., what one may call conversational sensitivity.

According Crimmins and Perry, an account of belief reports

consistent with our intuitions must allow that a certain

sentence be a true description of an agent’s cognitive

situation when uttered in a certain context, but false when

uttered in another (without assuming cognitive changes in

the agent). For instance, it seems that an utterance of

Pierre believes that London is pretty

ought to be true in a conversation …, but false … According

to Crimmins and Perry, such different settings introduce

distinct contextually salient representations (what they

call the beliefs normal in the context). Let me call it the

Conversational Salience Thesis. But by the same token …

What is the morale of the foregoing considerations on

substitutivity and conversational sensitivity? I think that

they show that the case in favor of the Contextual View is

importantly weakened. For suppose that one has the

inclination not to regard our alleged intuitions as

decisive in these cases; then, the main support for the

Contextual View vanishes. But if one decides to take them

seriously, then there seems to be no aspect where the

28

Contextual View is in a principled advantage over the

Radical View.

References

Crimmins, Mark 1992. Talk About Beliefs. The MIT Press.

Crimmins, Mark 1995. Contextuality, Reflexivity, Iteration,

Logic. Philosophical Perspectives 9, AI, Connectionism, and

Philosophical Psychology: 381-399.

Crimmins, Mark, and John Perry 1989. The Prince and the

Phone Booth: Reporting Puzzling Beliefs. The Journal of

Philosophy, 86: 685-711.

Crimmins, Mark. Notional Specificity.

Forbes

Kaplan, David 1997. Demonstratives. Reprinted in

Neale, Stephen 1990. Descriptions. The MIT Press.

Recanati, Francois 1996. Domains of Discourse. Linguistics and

Philosophy 19: 445-475.

Reimer, Marga 1998. Quantification and Context. Linguistics and

Philosophy 21: 95-115.

29

Richard, Mark 1990. Propositional Attitudes: An Essay on Thoughts and

How we Ascribe Them. Cambridge University Press.

Salmon, Nathan 1986. Frege’s Puzzle. The MIT Press.

Salmon, Nathan 1989. Illogical Belief. Philosophical

Perspectives, 3.

Salmon, Nathan 1995. Being of Two Minds: Belief with Doubt.

Nous 29: 1-20.

Soames, Scott 1987a. Direct Reference, Propositional

Attitudes, and Semantic Content. In Philosophical Topics

15: 47-87. Reprinted in Nathan Salmon and Scott Soames

(eds.), Propositions and Attitudes, Oxford University Press

1988.

Soames, Scott 1987b. Substitutivity. In J. J. Thomson

(ed.), On Being and Saying: Essays for Richard Cartwright, The

MIT Press, 99-132.

30