THE INESSENTIAL QUASI-INDEXICAL

 

As Perry originally formulated things, the primary casualty of the problem of the essential indexical was the analysis of belief as a two-place relation between a subject and a proposition.[1] Strictly speaking, of course, Perry argued that the problem he identified undermined the “doctrine of propositions” which consists of this analysis of belief together with the claims that the truth-values of propositions are independent of contextual parameters (other than worlds) and that propositions are individuated more finely than truth-conditions.[2] But he went on to assert that the only adequate solution to the problem involved the rejection of the “propositional-relation” (let’s call it) analysis of belief.[3] The central aim of this paper is to defend a version of the propositional-relation analysis from Perry’s criticisms.[4]

 

It is worth emphasizing from the get go that the problem of the essential indexical has a semantic dimension as well as a psychological dimension. Corresponding to the propositional relation analysis of belief is an account of belief reports according to which ‘believes’ is a two-place predicate and the contents of “that’-clauses in canonical belief reports are the propositional relata of the belief relation.[5] And Perry’s argumentative strategy involves drawing an inference from the role indexicals play in reports of “locating beliefs” to a conclusion about the nature of these beliefs which is facilitated by this account of belief reports. Let me elaborate.

 

Perry argues that certain behavioural changes can be explained by a subject’s coming to acquire new locating beliefs – “…beliefs about where one is, when it is, and who one is.”[6] Such subjects, however, can express what they come to believe only by means of the use of an indexical expression. But if locating beliefs were relations between subjects and propositions – and these propositions served as the contents of belief report “that”-clauses – they should be capable of expression by means of any belief sentence whose “that”-clause denotes the requisite proposition. As a result, the propositional-relation analysis fails for an important class of beliefs.[7]

 

In his discussion of the semantic issue, Perry focuses on first-person belief reports of the form “I [came to] believe …” which subjects use to express their own beliefs, and the role indexicals play therein. Greater generality can be achieved, however, by considering as well the role Castaneda’s “quasi-indicators” play in reports from third parties.[8] Roughly, a lexically indexical expression, such as ‘she’, is used as a quasi-indexical in a belief (or other attitude) report if it does not express an indexical reference made by the speaker but rather is used to attribute implicit indexical reference to the subject of the report.[9] So, for example, the expression ‘she’ in Fred’s utterance,

Mary believes she is a star,

is used as an indexical if it’s content is determined by features of the context in which Fred utters it (such as the demonstratum of Fred’s demonstration, if he made one). And it is used as a quasi-indexical if it attributes to Mary a belief she would express as,

I believe I am a star.

Following Castaneda, I will use a superscripted asterix – as in ‘she*’ – to indicate that an expression is being used as a quasi-indexical. Note: exactly the same case can be made for the thesis that quasi-indexicals are essential to third-person reports of locating beliefs as can be made for the claim that indexicals are essential to first-person reports.[10]

 

Finally, when discussing the contents of belief reports, the notion of content at issue is what might be termed “explanatory content” – the content of a belief reports which enables it to play the role it does in explanations of behaviour. The motivation for focusing on explanatory content is that Perry’s argument for the essential role of indexicals relies on the observation that replacing them with co-referential expressions in locating-belief reports undermines the explanatory power of such reports.[11] For present purposes, I am neutral on the question as to whether the explanatory content of a belief report should be identified with its semantic content,[12] its illocutionary content, [13]or its pragmatic content.[14] 

 

This paper consists of four parts. First, I develop Perry’s argument against the propositional-relational analysis of belief in more detail and consider a number of responses to it. Second, I offer a diagnosis of the phenomenon of the putatively essential indexical. Third, I develop a version of the propositional-relational analysis which is immune to the difficulties Perry presents. And finally, I consider a number of potential objections to this view.

 

I: The Argument

 

For simplicity I will focus only on Perry’s second example, the details of which are of follows. Professor Jones, who desires to attend a department meeting which she believes begins at noon, spends the morning seated at her desk. Suddenly at noon, she gets up from her desk, exits her office, and sets off down the hall. She believed all morning that the meeting began at noon and yet did not get up from her desk, so this belief cannot explain her behaviour. Instead her behaviour needs to be explained by appeal to a newly acquired locating belief – about when it is – which she would (at the time) express as,

I believe the meeting starts now[15]

(and which a third party might report as,

Professor Jones believed that the meeting started then*.)

                            

Now according to the doctrine of propositions, this newly acquired belief, like all beliefs, is a relation in which the subject stands to a proposition. The difficulty which locating beliefs pose for this doctrine is that the propositions to which subjects are allegedly doxastically related when they have such beliefs cannot be identified. Since the propositional relata of belief are supposed to be “true or false absolutely,”[16] they must be specifiable without using indexical expressions – because ‘the meeting starts now” is true in some contexts but not others, it does not uniquely specify a proposition. But any belief report which lacks an (appropriate) indexical will fail to explain the behaviour at issue. 

 

Consider again Professor Jones and her noontime department meeting. If the belief whose acquisition explains her behaviour is a relation to a proposition, there should be a non-indexical expression A such that

Professor Jones believed that the meeting started at A,

reports or expresses this belief.[17] The trouble is that the candidates for A yield belief reports which do not explain why Jones left her office when she did and hence which do not report the locating belief that does explain her behaviour. If, for example, A is ‘noon’, the resulting belief sentence,

Professor Jones believed the meeting started at noon,

reports a belief that she held all morning long, which, therefore, cannot by itself explain the change in his behaviour which occurred at noon. Similarly, if A is a description, such  as ‘the time the Dean has lunch’, which expresses a concept she has in mind which fits noon uniquely, the resulting belief report, 

Professor Jones believed the meeting started at the time the Dean has lunch,

again does not explain his behaviour. After all, at noon she might not have realized that the time the Dean has lunch was then* (even if she realized that the time the Dean had lunch was noon).[18]

 

Of course, if Jones had (at the time) a belief she could have expressed using a sentence of the form,

I believe that A is now,

then a report of form,

Professor Jones believed that the meeting started at A,

would explain her behaviour. For example,

Professor Jones believed the meeting started at the time the Dean has lunch,

and

Professor Jones believed the time then Dean has lunch was then*

do jointly explain Jones’ behaviour, even though the former report does not by itself do so. But not only does this presuppose that there is a locating belief – distinct from the original target of the analysis – whose specification requires the use of indexicals, it also undercuts the identification of the locating belief Jones expressed as,

I believe the meeting starts now,

with the belief reported by

Professor Jones believed the meeting started at the time the Dean has lunch.

After all, since the latter explains Jones’ behaviour only when combined with the belief reported by

Professor Jones believed the time then Dean has lunch was then*

and the former explains her behaviour by itself, they can hardly be identical.

 

Balaguer[19] and Pruim[20] have recently offered distinct responses to Perry’s argument which are notable because they deploy strategies which Perry himself explicitly considers and rejects. Both defend the propositional-relation analysis of belief by rejecting the requirement of a non-indexical specification of the propositional relatum of belief. But they reject this requirement for very different reasons. Balaguer rejects this requirement because he takes the objects of the belief relation to be “indexical general propositions” (or GPs),[21] or what Perry calls “relativized propositions.”[22] Unlike propositions as traditionally conceived, indexical GPs do not have (or determine) fixed truth conditions but rather can have different truth conditions in different contexts. In effect, they can be viewed as functions from various contextual features or indices to truth conditions. The reason no non-indexical specification of the propositional relata of the belief relation is forthcoming is, on this view, because any such specification involves singling out a specific set of truth conditions.

 

Perry, however, raises serious difficulties for this view. He points out that indexical GPs cannot be believed absolutely but only relative to sets (or, perhaps, ordered n-tuples) of indices. Professor Jones, for example, did not simply believe the indexical GP that the meeting starts now. Instead, she believed it only relative to the time of noon.[23] Moreover, the indices relative to which a subject believes an indexical GP – the “context of evaluation” – cannot simply be identified with features of the context in which the belief is held – the “context of belief.” Jones (or some other subject, for that matter) could believe at some other time, say 1:00 pm, that the meeting starts now was true relative to noon (or at noon that it was true relative to some other time). [24] The problem this poses for Balaguer’s view is that believing indexical GPs relative to various indices does not explain the subjects’ behaviour in the puzzle cases Perry marshals. The fact that Professor Jones believed that the meeting starts now was true relative to noon does not explain why she got up from her desk an left her office at noon; after all, she may well have believed this all morning.[25] As a result, indexical GPs cannot be identified with locating beliefs.[26]

 

In contrast to Balaguer, Pruim concedes that the propositional relata of belief are traditional propositions with fixed truth conditions. Nevertheless he still rejects the requirement of a non-indexical specification of at least those propositions which serve as the relata of locating beliefs. His reason for this rejection, however, is that we have limited access to these propositions. In particular, he argues that grasping certain propositions – that is, believing, doubting, entertaining, etc., them – requires being properly situated.[27] For example, the propositional relatum of Professor Jones’ locating belief can be grasped only at noon on the relevant day. Now subjects who are not properly situated can talk about such propositions and even attribute them to various people as relata of the belief (or some other doxastic) relation. Nevertheless, they cannot express them – the verbal expression of a proposition is reserved for those who grasp it. Instead they can pick them out only via description – ‘the proposition Professor Jones expressed by means of “the meeting starts now,”’ for example.[28] And any adequate descriptive specification of the propositional relatum of a locating belief – that is, one which serves to explain the behaviour of the subject – will have to utilize an appropriate indexical.[29]  

 

There are, again, a number of reasons to balk at this proposal. It is worth noting, however, that Perry’s original objection – that invoking limited accessibility involves abandoning a “common actual world”[30] – is not very compelling. There are lots of perfectly ordinary relations in which people can stand to perfectly ordinary things only if they are properly situated – the relation one stands to a computer keyboard when typing is a perfectly banal example – but such relations pose no threat to a common actual world.

 

Perry has, however, also raised more persuasive considerations against the thesis of limited accessibility.[31] As above, let’s suppose that the following believe report is true:

(1) Professor Jones believes at noon that the meeting starts then*

The proposition to which Jones stands in the belief relation can be represented as,

(2) that the meeting starts at i,

where i is a Fregean sense which uniquely picks out noon (and putatively can only be grasped at noon.). Now suppose, in addition, that at 11:00 am Professor Smith expects that at noon Jones will get up from her desk, exit her office, and set off down the hall, and that this expectation is due to a belief correctly reported as follows:

(3) Smith believes at 11:00 am that Jones believes (or will believe) at noon that the meeting starts then*.

Now in order for the belief reported by (3) to yield this expectation, Smith must believe at 11:00 am the proposition

(4) that i is noon.

Otherwise Smith would expect Jones to engage in said pattern of behaviour at some other time. But if this is right, Smith believes at 11:00 am a proposition which by Pruim’s lights is accessible only at noon. Moreover, if Smith believes at 11:00 am the proposition

(5) that the meeting starts at noon,

we would expect that as a consequence of this and his 11:00 am belief in (4) that he also believes (2) at 11:00 am, again despite the fact that this proposition allegedly accessible only at noon.

 

Now Pruim might respond by pointing out that the belief we are attributing to Smith can be reported only by a sentence such as,

(6) Smith believes at 11:00 am that the meeting starts at the time that can be expressed at noon by ‘now’.

And this suggests that the propositional relatum of his belief is not (2), but rather a general proposition containing a denoting complex whose denotation is i. And if this is right, no reason has been given to suppose that Smith can grasp (2) at any time other than noon. But Jones’ belief too can be reported only by a sentence such as

(7) Jones believes at noon that the meeting starts at the time that can be expressed at noon by ‘now’

(for which,

Jones believes at noon that the meeting starts then*

is elliptical, on the view in question). To nevertheless insist that (7) reports Jones’ doxastic relation to (2) while (6) reports Smith’s relation to a distinct proposition would be ad hoc.

 

Finally, one might worry that propositions to which access is so limited cannot play the role in communication that we expect of them. Propositions are theoretical entities posited to play a certain explanatory role in thought and talk. In particular, they are meant to serve as the informational contents of various thought-states which are conveyed or communicated via speech acts. Insofar as propositions – or a conception thereof – fail to occupy this role, the whole motivation for invoking propositions is to that degree undermined. And this exactly what happens on Pruim’s account. Suppose, for example, at noon Professor Jones leaves a note in her office for Smith which reads,

I couldn’t wait any longer. The meeting is starting now.

According to the view on the table, the proposition which Jones expresses by means of her inscription of the second sentence cannot be communicated to Smith,[32] at least if he fails to arrive just at noon. Pruim does suggest that deviations from the simple model wherein the proposition expressed coincides with the proposition communicated are sometimes tolerable, as when information is conveyed by Gricean pragmatic mechanisms.[33] But given that the pragmatic impartation of information is mediated by the communication of what is literally expressed on the Gricean story,[34] this phenomenon does not adequately prop up the hypothesis of limited accessibility. Neither of the considerations raised here decisively undermines Pruim’s theory, but the cumulative effect does, in my view, diminish its desirability.

 

If Perry’s argument succeeds then the propositional-relation analysis of belief – according to which belief is a two place relation between subjects and propositions – will have to be abandoned. Broadly speaking there are three alternative approaches to the analysis of belief. First, one could take it to be a two-place relation between subjects and something other than propositions. Lewis, for example, takes belief to be a relation between subjects and properties.[35] Second, one could analyze belief as a three place-relation between subjects, propositions, and some third class of entities, such as “characters”[36] or “propositional guises”[37]. And third, one could treat belief as a monadic property of subjects which resists relational analysis.[38] My own view, however, is that Perry’s argument does not ultimately succeed, and that a version of the propositional-relation analysis remains both attractive and defensible. 

 

II: The Diagnosis

 

Before developing my own variant of the propositional-relation analysis, it will prove fruitful to provide some motivation for it. And luckily, from the viewpoint of the unity of this very paper, such motivation is to be found in the proper diagnosis of the phenomenon Perry has identified.[39] In my view, what this phenomenon reveals is a widely made oversimplification in the theoretical reconstruction of folk, or belief-desire, explanation of behaviour.[40] What I have in mind is the assumption that the core of an explanation of an instance of rational behaviour consists solely of an application of the belief-desire-action (or BDA) principle: if a rational subject S desires outcome O and believes that performing action A she can achieve O, then S will perform A. So, for example, using this principle, one might explain the behaviour of a student of Professor Jones – who drops by her office at 11 am to discuss a paper assignment – as follows: the student desired a good grade in Professor Jones’ course and he believed that he could achieve this end by dropping by her office to discuss the paper assignment. Of course, the BDA needs to be elaborated to account for such things as the ordering of preferences, degrees of belief, competing strategies for certain goals, the believed likelihood of various outcomes, cognitive dynamics, etc. Nevertheless, it is typically thought that once a suitably gussied up version of it has been deployed, most of the explanatory work has been done. 

 

Now I do not mean to deny the importance of the BDA to the explanation of behaviour. Rather, what I object to is its overemphasis. In order to explain why a subject engaged in a particular course of behaviour, one needs to explain both (i) why she engaged in the type of behaviour she did and (ii) why she did so in the specific circumstances that she did. For example, a complete explanation of the behaviour of Professor Jones’ student requires an account of why he performed an instance of dropping-by-Jones’-office-and-discussing-a-paper-assignment behaviour but also an account of why he did so at 11 am. The BDA principle, however, is suited only to the former task. This, I take it is the fundamental lesson of Perry’s argument. Let me elaborate.

 

Suppose a subject S engages in behaviour of type B in a context C. One might attempt to explain her behaviour by invoking BDA as follows:

EXPLANANDUM:

S B’s in C

EXPLANANS:

1) S desires O

2) S believes that if S B’s then O

3) Principle BDA

As it stands, however, this explanation is defective in two respects. First, it explains only why she ensures that someone engages in behaviour B, not why she does so herself. After all, S might not realize that she* is S. And second, it best explains only why S engages in behaviour B in some context or other, not why she specifically B’s in context C. One might attempt to respond to this second worry by replacing (2) with,

2*) S believes that if S B’s in C then O.

But S could believe this in various contexts other than C without its culminating in behaviour B – recall Professor Jones who believed all morning that if she exited her office and set off down the hall at noon she would be able to attend the department meeting. And, as we have seen, believing it in C does not explain why S B’s in C either. After all, S might not realize in C that C is there* and then*.

 

Of course, the belief that engaging in a type of behaviour will yield a desired outcome is typically part of the explanation of a subject’s producing a corresponding behavioural token. What is missing from the explanation, however, is the subject’s recognition of an opportunity to engage in this behaviour. That is, what explains S’s B-ing in C is (i) S’s belief that engaging in B would enable her to achieve her desired outcome O and (ii) S’s recognition, while in context C, of an opportunity to engage in B.[41] So, for example, what explains Joe’s behaviour of filling his car up with gasoline at 4:00 pm at the PetroCan station on  6th Ave might be a standing belief that filling his car with gasoline would enable him to achieve certain of his desired ends and, as he was driving along 6th Ave just before 4:00 pm, the recognition of an opportunity to engage in that behaviour.[42] And what explains Professor Jones’ behaviour is the standing belief that exiting her office and setting off down the hall at noon – a highly specific type of behaviour – will enable her to achieve her goal of attending the department meeting and the recognition, at noon, of an opportunity to do so.[43]

 

When recast in this way, two features of the phenomenon we have been discussing become evident. First, it is a much broader phenomenon than Perry originally suggested.[44] On Perry’s view, the recognition of an opportunity to engage in a course of action is identified with the having of a locating belief. Professor Jones’ recognition of the opportunity to exit her office and setting off down the hall at noon, for example, just is her possession of the locating belief that noon was then*. But consider Kent, who has a standing belief that punching Paul in the nose will have the desired effect of avenging some slight that Paul committed, and who punches Paul in the nose at the department Christmas party. The explanation as to why Kent punched Paul at all is, of course, provided by this standing belief. And the explanation of why he did so at the party is provided by his recognition of an opportunity. But this recognition did not consist in the possession of a locating belief – about who he was, where he was, or when it when. Rather it consisted of his experience of Paul at the party, perhaps in a state rendering him particularly vulnerable to nose punches. Moreover, Kent’s state of recognition can be reported and used to explain why he punched Paul at the party (as opposed to doing so in some other contexts) without the use of quasi-indexicals by a sentence such as,

Kent saw Paul at the party.

The lesson here is that the class of states which count as the recognition of opportunities is broader than the class of locating beliefs.[45]  

 

Second, the force of Perry’s charge that a belief report wherein a quasi-indexical ‘A*’ has been replaced by a co-referential (and non-indexical) expression ‘B’ retains the explanatory power of the original only if supplemented by a report of the form

S believes that A* is B

(where S is the subject) has been blunted. Given that a report such as,

Professor Jones believed that the meeting started at noon,

explains only the type of behaviour in which she engaged – exiting her office and setting off down the hall at noon – it is unsurprising and unproblematic that some other psychological state – in this case her belief that noon was then* – was required to explain why she behaved when she did. Moreover, the psychological state reported by,   

Professor Jones believed that the meeting started then*,

does not by itself explain Jones’ behaviour, and also requires supplementation by the belief that noon was then* in order to adequately do so. The type of behaviour Jones all morning intended to engage in was not merely to exit her office and set off down the hall but rather to exit her office and set off down the hall at noon. After all, given her beliefs about when the meeting was scheduled only the latter would be expected to satisfy her desire to attend it. But the belief that the meeting started then* explains only why she engaged in the former type of behaviour – which she could have engaged in at any time during the day – although it does explain why she did so at noon.

 

One might, of course, object that since the behaviour in question is, in fact, an instance of exiting-her-office-and-setting-off-down-the-hall behaviour, there is no further explanatory work to be done; after all both “the” type of behaviour engaged in and the context in which this was done have been explained. The problem with this manoeuvre, however, is that it severs the link between the intention Jones had all morning and the behaviour she ultimately engaged in. Consider, by way of analogy, someone who intends to play tennis with Boris Becker and ends up playing with whoever happens to be at the club that morning. Even if it turns out that, unbeknownst to the subject, the person at the club that morning was in fact Boris Becker (perhaps in disguise), she has nevertheless abandoned her original project of playing tennis with Becker in favour of the more general project of playing tennis with someone.   

 

III: The Solution

 

What remains to be done is to develop a version of the propositional-relation analysis of belief compatible with this two-fold picture of psychological explanation. At the core of the theory I wish to propose is the notion of a cognitive relation. A cognitive relation is a relation in which a thinking subject stands to the objects of thought and talk, in virtue of which she is able to think or speak of them. Paradigmatic cognitive relations include experiential relations, such as seeing or smelling objects, reputational relations, such as hearing or reading about an object, and conceptual relations, thinking about objects via the deployment of concepts. Also important are memory relations wherein a subject remembers experiencing, hearing about, or conceiving of an object.   

 

Conceptual relations come closest to the Fregean notion of sense,[46] but unlike sense they often pick out objects (or, perhaps better, have particular relata) only relative to contextually supplied features of various kinds.[47] So, for example, a subject might be conceptually related to a object by means of a concept expressible by “the tallest person in Turcotte Hall at 3:00 on December 12, 2006,” which picks out an individual only relative to the comparative heights of those in Turcotte Hall at that time, or “the current Canadian Prime Minister,” which picks out an individual only relative to the time at which the concept is deployed. Indexical relations – those which can be expressed only by means of indexicals – are a species of contextually relative conceptual relation. One can stand in a conceptual relation to a given time, for example, by means of deploying the concept expressed by ‘now’, but which time one is thereby cognitive related to depends upon the time at which the concept is deployed; in particular, the time one is cognitively related to by the deployment of this concept just is the time at which the concept is deployed.  

 

Now it is important to note that a subject’s notional objects – the objects she believes to exist – do not correspond or map onto the individual cognitive relations in which she stands. Rather they correspond to collections or “teams”[48] of these relations. Subjects constantly make judgements to the effect that certain of the cognitive relations in which they stand share relata. So, for example, a subject might judge that the person she heard about (or remembers hearing about) yesterday is the person she conceives of as the shortest spy or that the place she was born is her current location. And such shared-relatum judgments carve up the totality of cognitive relations in which the subject stands into collections, corresponding to each of which is an object – understood here in a broad sense to include times and places – which she believes to exist. Moreover, these collections can undergo changes in membership as a subject comes to stand in new cognitive relations to things, as well as when she revises her shared relatum judgements or simply forgets about various experiences she has undergone.

 

What the kind of psychological state characterized above as the recognition of an opportunity amounts to, on this view, is standing in a new cognitive relation to some object – typically an experiential or indexical relation – and judging that it shares a relatum with the members of some pre-existing collection (thereby adding it to said collection). So, for example, Kent’s recognition of the opportunity to punch Paul in the nose consisted in his having a visual experience of someone at the party and judging that person to be Paul – the putatively shared relatum of one of his collections of cognitive relations. And Professor Jones’ recognition of the opportunity to exit her office and set off down the hall at noon consisted of her indexical relation (at noon) to the time noon and her judgement that noon was then*.

 

Thinking subjects stand in cognitive relations not only to objects, but also to properties and relations. And they make shared relatum judgments regarding these relations as well. A subject might, for example, judge that the colour she is currently experiencing is the colour she heard about in her design class yesterday. As a result, we can identify a subject’s notional properties and relations – the properties and relations she believes to exist – with the collections of cognitive relations generated by these judgements. And, on my view, such collections, along with collections object-relations, serve as constituents of the propositional relata of the belief relation. Russellian propositions are abstract structures – roughly, ordered sequences – of objects, properties, and relations. My propositional relata share this abstract structure, but positions in it are occupied by collections of cognitive relations to objects, properties, and relations, rather than the objects, properties, and relations themselves.[49] So, for example, Mary’s belief that Fred stinks can be represented as,

Believes (Mary, < CSTINKS; CFRED>),  

where ‘CFRED’ denotes a collection of cognitive relations in which Mary stands to Fred and ‘CSTINKS’ denotes a collection of cognitive relations in which Mary stands to the property of stinkiness.[50]

 

Consider, again, Kent’s behaviour at the department Christmas party. As above, a complete explanation of a Kent’s behaviour requires an explanation of the type of behaviour he engaged in – punching Paul in the nose – and an explanation of why he engaged in it in the particular circumstances that he did. An explanation of the former is provided by a belief-desire pair such as,

Kent desired that he be avenged against Paul

Kent believed that if he punched Paul in the nose, he would be avenged against Paul.

On the picture on the table, these psychological states can be represented as,

Desired[51] (Kent, <CAVENGES; CKENT, CPAUL>)

Believed (Kent, <<CPUNCHES; CKENT, CPAUL> É < CAVENGES; CKENT, CPAUL>>),[52]

where ‘CKENT’ denotes a collection of cognitive relations in which Kent stands to himself, ‘CPAUL’ denotes a collection of cognitive relations in which he stands to Paul, and ‘CAVENGES’ and ‘CPUNCHES’ denote respectively collections of cognitive relations in which he stands to the relation of being avenged against and the relation of punching. What is important to note, however, is that if CPAUL includes Kent’s experiential relation to Paul at the party (and CKENT includes his indexical relation to himself), then this belief-desire pair also explains why Kent punched Paul in the circumstances in question. After all, the membership of Kent’s experiential relation in CPAUL reflects the fact that Kent judged it to share a relatum with the prior membership of this collection.

 

Similarly, an explanation of the type of behaviour engaged in by Professor Jones is provided by,

Professor Jones desired that she attend the department meeting

Professor Jones believed that if she exited her office and set of down the hall at noon she would attend the department meeting,

which can be analyzed as,

Desired (Jones, <CATTENDS; CJONES, CDM>)

Believed (Jones, <<CEXIT&SETS; CJONES, CNOON> É <CATTENDS; CJONES, CDM>>).

But again, if Jones’ indexical relation (at noon) to noon is a member of CNOON, then this belief-desire pair also explains why she exited her office and set off down the hall when she did. 

 

The upshot of this is that any referring expression that successfully picks out the requisite collection of cognitive relations can be used to report a belief which explains both a type of behaviour and the circumstances in which it occurred. Let’s suppose that a referring expression in the content clause of a belief report picks out one of a subject’s collection of cognitive relations by means of picking out a member of that collection. And let’s suppose that definite descriptions pick out conceptual relations,[53] proper names pick out reputational relations,[54] and that quasi-indexicals pick out indexical relations.[55] Finally, suppose that the reputational relation Professor Jones’ would express using ‘noon’, the conceptual relation she would express using ‘the time the Dean has lunch’, and the indexical relation she would express (at noon) using ‘now’ are all members of CNOON. In such circumstances, the occurrences of ‘noon’, ‘the time the Dean has lunch’ and ‘now’ respectively in,

(i) Professor Jones believed that the meeting started at noon,

(ii) Professor Jones believed that the meeting started at the time the Dean has lunch,

and

(iii) Professor Jones believed that the meeting started then*,

would each pick out CNOON, and the reported belief would in each case be the same:

Believed (Jones, <CSTARTS; CDM, CNOON>).

And as a result, each report would provide an equally good explanation of Jones’ behaviour.

 

Recall, however, that Perry argued that these reports differ in explanatory power because (i) and (ii) explain Jones’ behaviour only when supplemented by a report of the form,

(iv) Jones believed that A was then*

whereas (iii) explains by itself. [56] But as we have seen, reports such as (iv) simply make explicit that the collection of cognitive relations picked out by ‘noon’ or ‘the time the Dean has lunch’ contains a member which counts as the recognition of an opportunity to engage in the (instrumentally) desired pattern of behaviour. And if it were not already explicit, (iii) might similarly need supplementation by a report such as (iv) in order to reveal that the behaviour she engaged in was the behaviour she spent the morning intending to engage in, namely, exiting her office and setting off down the hall at noon.  

 

IV: Objections and Replies

 

There are a number of objections one might have to the theory presented here. I will consider three of them. First, one might worry that, despite their propositional structure, the propositional relata I have identified are not really propositions. And if this is right, my claim to have rescued the propositional relation analysis of belief from Perry’s criticisms will have proven to be specious. The reason one might think this is because propositions are supposed to be the bearers of truth and falsity, but if the members of any of constituent collections of one of my propositional relata fail to share a common relatum, it cannot be assigned a truth-value. Suppose, for example, propositional relatum of the belief reported by,

Mary believes that Steven Harper is the current Canadian Prime Minister

is

<CCDNPM; CHARPER, CDEC06>.

This putative proposition is true only if the individual who is shared relatum of the members of CHARPER has the property which is the shared relatum of the members of CCDNPM at the time which is the shared relatum of the members of CDEC06. But if any of these collections lack a shared relatum, the proposition is neither true nor false, despite the fact that Mary’s belief is, in fact, true. Moreover, this is no mere hypothetical worry. A failure of the membership of some collection to have a common relatum need not reflect any serious confusion on the part of the subject. All it takes is one mistaken shared-relatum judgement, such as Mary mistakenly judging the person behind her in line at an Ottawa supermarket to be Steven Harper.

 

There a couple of things worth noting by way of response to this worry. First, it is not in general true that propositions have truth-values in possible circumstances of evaluation. For example, the Russellian proposition,

<Being the Canadian PM; Steven Harper, December-2006>

is neither true nor false in circumstances in which Steven Harper does not exist. And second, when a subject thinks or talks about one of her notional objects on a given occasion, typically a sub-collection of the corresponding collection of cognitive relations is salient. Mary might think of “Steven Harper” by means of the conceptual relation she expresses as “the leader of the federal Conservatives” or by means of her (memory of her) experiential relation to the person behind her in line at the supermarket. And even if the members of the larger collection lack a shared relatum, the members of the sub-collection may have one. As a result, propositions containing these sub-collections can have truth-values, even if the corresponding propositions containing the larger collections do not.

 

Second, one might worry that my theory runs afoul of the very same difficulty which beset Pruim’s vis-ŕ-vis the explanatory role of propositions in thought and talk. Recall: propositions are entities posited to serve a dual role as the contents of thought-states and the communicated contents of speech acts. The problem is that my propositional relata cannot serve this latter role. The propositional relatum of a belief is in part constituted by collections of cognitive relations in which the subject stands to various items. And since other subjects presumably cannot (or at least typically do not) stand in exactly the very same cognitive relations to those items, they cannot stand in the belief relation to this proposition. As a result, insofar as successful communication requires that the listener come to believe (or entertain or even doubt) what the speaker believes, my propositional relata cannot serve as the communicated contents of speech acts.

 

Even though these propositional relata cannot themselves serve as communicated contents, they fix or determine Russellian propositions which can so serve. Suppose Randa, while in conversation with Michael, asserts the sentence,

Paul is a twit,

and in so doing expresses a belief that can be represented as,

Believes (Randa, <CR-TWIT; CR-PAUL>),

where ‘CR-TWIT’ denotes a collection of Randa’s cognitive relations to the property of being a twit and ‘CR-PAUL’ denotes a collection of Randa’s cognitive relations to Paul. In such circumstances, the communicated content of Randa’s assertion would be a Russellian proposition with the same structure but whose constituents are the property which is the shared relatum of the members of CR-TWIT and the object which is the shared relatum of the members of CR-PAUL,

<Twithood; Paul>.[57]

If Michael were to accept Randa’s assertion and judge that the person Randa was referring to by means if her utterance of ‘Paul’ was the shared relatum of one of his collections of cognitive relations, CM-PAUL, and that the property she attributed to this object by means of her utterance of the predicate ‘is a twit’ was the shared relatum of another of his collections, CM-TWIT, then Michael would acquire the following belief:

Believes (Michael, < CM-TWIT; CM-PAUL >).

Although, strictly speaking, the propositional relata of belief are not the communicated contents of speech acts, they are nevertheless intimately connected to them. As a result, the dual-role is occupied by distinct but tightly connected entities.

 

Third, one might worry that indexical cognitive relations – especially those in which subjects stand to themselves – are epistemologically or metaphysically suspect. In particular, this idea seems to entail a metaphysical picture according to which thinking subjects are non-physical Cartesian egos – occupying essentially subjective spatio-temporal locations – that are epistemically transparent to the subjects themselves and epistemically inaccessible or opaque to others.

 

Such worries, however, are ill-founded. The relata of indexical relations are perfectly objective locations and times and biologically constituted subjects. And as such, they are both inter-subjectively accessible and open to perfectly familiar empirical investigation. Moreover, not only does standing in such relations fail to render their objects transparent to a subject, it yields little or no substantive information regarding them whatsoever; a subject who knows of a certain location only that it is her current location, for example, knows very little about it. Now it is true that only an appropriately situated subject can stand in an indexical relation to a given time or place – a subject located at that time or place – and that only the subject herself can stand in an indexical relation to herself. But being appropriately situated is required for standing in many perfectly ordinary relations – being three feet to the left of some object or being born at the same date as some event, for example. And it is neither surprising nor mysterious that only the subject can be indexically related to herself. After all, to be so related involves deploying a contextually relative concept which denotes whoever deploys it. 

 

V: Conclusion

 

Perry’s “problem of the essential indexical,” although very important, has been widely misunderstood. This significance of this phenomenon is not that it undermines the propositional-relation analysis of belief; as demonstrated above, there is a compelling version of this analysis which is compatible with Perry’s problem. Nor is it that there is a class of psychological states whose expression requires the use of indexicals or quasi-indexicals. Rather, what Perry’s argument reveals is the need for an analysis of belief capable of explaining not only why thinking subjects engage in the types of behaviour that they do, but also why they do so in the circumstances that they do. And by taking the constituents of believed propositions to be collections of cognitive relations – including those which count as the recognition of opportunities – this explanatory need can be met.

 

 

Peter Alward

Department of Philosophy

University of Lethbridge

peter.alward@uleth.ca

 

 

References:

 

Alward, Peter (manuscript), “Cluster Theory: Resurrection.”

 

Balaguer, Mark (2005), “Indexical Propositions and De Re Belief Ascriptions,” Synthese 146: 325-55.

 

Cappelen, Herman and Lepore, Ernie (2005), Insensitive Semantics, Blackwell.

 

Castaneda, Hector-Neri (1967), “Indicators and Quasi-indicators,” American Philosophical Quarterly 4: 85-100.

 

---- (1968), “On the Logic of Attributions of Self-Knowledge to Others,” The Journal of Philosophy 65: 439-56.

 

Frege, Gottlob (1892/ 1997), “On Sinn and Bedeutung,” in The Frege Reader, Michael Beaney, ed., Oxford: Blackwell, pp. 151-71.

 

---- (1919/ 1997), “Thought,” in The Frege Reader, Michael Beaney, ed., Oxford: Blackwell, pp. 325-45. 

 

Grice, Paul (1975), “Logic and Conversation,” in Syntax and Semantics, vol. 3: Speech Acts, P. Cole and J.L. Morgan (eds), New York: Academic Press.

 

Kaplan, David (1980), “Demonstratives,” in Themes from Kaplan, Almog, Joseph, Perry, John, and Wettstein, Howard, eds., Oxford University Press.

 

Kripke, Saul (1980), Naming and Necessity, Cambridge: Harvard University Press.

 

Lewis, David (1979), “Attitudes De Dicto and De Se,” The Philosophical Review 88: 513-43.

 

Perry, John (1979/ 1993), “The Problem of the Essential Indexical,” in The Problem of the Essential and Other Essays, Oxford University Press, pp. 33-52.

 

---- (1983/ 1993), “Castaneda on He and I,” in The Problem of the Essential and Other Essays, Oxford University Press, pp.91-120.

 

Pruim, Peter (1996), “Elusive Thoughts: The Limited Accessibility of Indexical Beliefs,” Philosophical Studies 83: 171-190.

 

Salmon, Nathan (1986), Frege’s Puzzle, Cambridge: The MIT Press.

 

Tiffany, Evan (2000), “What is Essential About Indexicals?” Philosophical Studies 100: 35-50.

 

Quine, Willard Van Orman (1960), Word and Object, Cambridge, Mass.: The MIT Press.


[1] Perry (1979), p. 36.

[2] Note: the latter claim is characteristic of Fregean, in contrast to Millian, accounts of propositions.

[3] Perry (1979), pp. 46 ff.

[4] The view I defend in fact satisfies all the requirements laid down by the doctrine of propositions. My focus here, however, will be only on the analysis of belief. 

[5] Perry (1979), p. 36. The relevant notion of content here is “explanatory content.” More on this below.

[6] Perry (1979), p. 35.

[7] Perry’s argument will be developed in more detail below.

[8] See Castaneda, 1967, 1968.

[9] Castaneda (1968), p. 441.

[10] Arguably, since the speaker and the subject are the same, indexical and quasi-indexical uses of expressions coincide in first-person present-tense belief reports.

[11] Perry (1979), p. 35.

[12] Semantic content is typically glossed as what is literally expressed by an utterance. Fregeans and neo-Fregeans normally identify explanatory content with semantic content.

[13] Cappelen and Lepore (2005) distinguish between semantic content and illocutionary or “speech act” content – what is asserted or requested, etc. by means of an utterance. Although they do not directly address the issue, I take it they would identify explanatory content with illocutionary content rather than semantic content.

[14] Millians often identify explanatory content with pragmatic content – what is pragmatically imparted by means of an utterance – rather than semantic content. See, e.g., Salmon, 1986. 

[15] Presumably this locating belief would arise as a result of the combination of his prior belief that the meeting started at noon and the locating belief he would (at the time) express as,

I believe that now is noon.

[16] Perry (1979), p. 37.

[17] For present purposes I will remain neutral on this issue of semantic innocence – that is, whether or not (referring) expressions have the same contents when they occur within the scope of various sorts of sentential frames as when they do not. I will instead focus on the contents of expression only when they occur within belief report “that”-clauses. 

[18] Moreover, if she erroneously believed that noon uniquely satisfied the concept expressed by ‘the time the Dean has lunch’ then the proposition to which

Professor Jones believed the meeting started at the time the Dean has lunch

reports her to be doxastically related is false even though the allegedly identical locating belief which she expresses as

I believed the meeting starts now

is true. Perry (1979), p. 38.

[19] Balaguer, 2005.

[20] Pruim, 2006.

[21] Belaguer (2005), p. 332.

[22] Perry (1979), pp. 42 ff.

[23] Presumably a subject might believe an indexical GP relative to a range of values for certain indices. For example, Jones might believe the meeting starts now relative to some time between 11:30 am and 12:30 pm.

[24] To deny this would be, in effect, to suggest that indexical GPs are of “limited accessibility.” But as we shall see, to the extent that the thesis of limited accessibility undercuts the Perry’s argument for the essential indexical, it does so independently of any appeal to relativized propositions. It also runs afoul of its own difficulties.

[25] Nor does believing it absolutely – and not relative to any particular time – explain her behaviour at noon. Balaguer could, I suppose, respond by insisting that Jones could believe that the meeting starts now relative to noon only at noon. But again this sounds like limited accessibility, which will be discussed immediately below.

[26] Perry (1979), p. 44. It is worth noting that Balaguer nowhere addresses this objection.

[27] Pruim (1996), pp. 175 ff. Pruim intends his theory to be a generalization of Frege’s (1919) account of ‘I’.

[28] Pruim (1996), pp. 179-81.

[29] On this view, “Professor Jones believed that the meeting started then*” is elliptical for “Professor believed the proposition she expressed by means of ‘the meeting starts now’.”

[30] Perry (1979), p. 46. Unlike Balaguer, Pruim (1996, p. 179) is sensitive to Perry’s objections.

[31] What follows is adapted from Perry (1983), pp. 97-8.

[32] Nor can the proposition expressed by the first sentence. After all, on the view in question sentences containing the word ‘I’ express propositions which can only be grasped by the speaker.

[33] Pruim (1996), p. 181.

[34] Grice, 1975.

[35] Lewis, 1979.

[36] Kaplan, 1980.

[37] Salmon, 1986.

[38] Quine (1960, chapter 6) suggests something along these lines. 

[39] Although it does not establish what he claims for it, I think the phenomenon Perry has identified is very significant.

[40] I do not mean to suggest that philosophers are typically unaware of making this simplification. I do mean to suggest, however, that many philosophers are unaware of the distortions that making it can yield in the analysis of belief and in a proper account of belief reports.

[41] Note: recognizing or judging, in C, that there is an opportunity to engage in B is distinct from judging that there is in C an opportunity to engage in B. In the former case, C is the context in which the judgement is made; in the latter, it is part of the content of the judgement. 

[42] Of course, a more elaborate explanation will be needed if Fred forgoes earlier opportunities to acquire gas for his car prior to filling up at the PetroCan. This might involve appeal to such things as competing desires or optimal vs. less than optimal opportunities.

[43] We do have here a kind of limited accessibility, not to propositions, but to types of behaviour. Sometimes one has only a single opportunity to engage in a course action, and once the opportunity passes the behaviour in question becomes, in effect, inaccessible.

[44] Tiffany (2000, p. 40) makes a distinct but related point.

[45] And narrower too, if certain locating beliefs fail to correspond to the recognition of opportunities.

[46] Frege, 1892. Fregean senses, however, are abstract particulars which mediate subjects’ relations to objects of thought rather than the relations to those objects themselves.

[47] Although thinking subjects are sometimes cognitively related to objects via the deployment of concepts which uniquely denote them without the aid of context features, this is rarer than the literature often has it.

[48] The collections of cognitive relations are more like teams than sets because like the former, and unlike the latter, such collections can survive changes in membership and their membership, even relativized to times, is contingent.

[49] In effect, they are Russellian propositions with actual objects, etc., replaced by a subject’s corresponding notional objects.

[50] It is worth noting that noting that there is no guarantee that all the members of a collection of cognitive relations will share a common relatum, despite being judged by the subject to do so.

[51] Strictly speaking, belief and desire are correctly analyzed as triadic relations between subjects, propositions, and times.

[52] For simplicity, I am assuming that Kent’s belief has the logical form of a material conditional rather than of some kind of counterfactual conditional. 

[53] Descriptions which invoke subjects’ experiences – ‘the person Kent saw as the party’, e.g. – are better thought of as picking out experiential relations.

[54] Kripke’s (1980) causal-historical theory of name reference is in the background here. Systematic names – like names of times, ‘noon’, e.g. – may be better thought of as picking out conceptual cognitive relations. For present purposes, this will make no difference.   

[55] This is, of course, a gross oversimplification made for illustrative purposes. For a developed account of referring to collections of cognitive relations, see Alward (manuscript). 

[56] Or at least if it is presupposed that (a) Jones desired that she attend the department meeting and (b) Jones believed that if she exited her office and set of down the hall at the time of the department meeting, she would attend the meeting.

[57] If the members of either collection do not share a relatum, then the constituents of the Russellian proposition are the shared relata of the salient sub-collections underlying Randa’s assertion. If the members of the salient sub-collections do not share relata, then no proposition has been communicated.