The Problem of Essential Anaphora

Abstract

This paper represents a development of arguments first deployed in two papers from 2015, namely, “Attitudes de se and de me” and “Consciousness, Self-Consciousness and Essential Indexicality.” (Both these earlier papers are also available at this website.) This paper improves on the previous ones by including a more extensive and improved discussion of what various authors mean by, and what could be meant by, indexical, more precisely, first-person belief. It also includes a discussion of Perry’s later concept of belief, which appeals to the notion of reflexive truth conditions—see § 4.
I would like to thank Al Hàjek, Brian Garrett and Daniel Stoljar for very useful conversations and comments.

Introduction

Ever since the publication of John Perry’s article “The Problem of the Essential Indexical,” notions of essential indexicality and self-locating belief have been widely endorsed in the philosophy of language, mind and action. Their close relatives, attitudes inelimably de se, have also been received well, thanks to David Lewis’s article “Attitudes de se and de dicto,” which was published in the same year. Recently, however, some strident opposition has emerged.1 In their book The Inessential Indexical—On the Philosophical Insignificance of Perspective and the First-Person Herman Cappelen and Josh Dever argue that the widespread acceptance of essential indexicality, attitudes ineliminably de se and self-locating belief is far too comfortable and uncritical because these notions are seriously under-elaborated, hence unclear and insufficiently justified. Cappelen and Dever do not, however, simply point out how unclear and poorly justified these notions are. They maintain that the phenomena themselves are spurious: there are no such things as essential indexicals, attitudes ineliminably de se and self-locating beliefs in any philosophically significant sense.

For this reason, their critique is not motivated by any reconstructive ambition. Their goal is not to argue that essential indexicality, irreducibly de se attitudes, self-locating belief or whatever else one might like to call these phenomena

… should be explained in ways different from how, e.g., Lewis and Perry explained them. Our goal is to show that the entire topic is an illusion—there’s nothing there. … It isn’t that [Lewis’s and Perry’s] theories fail to adequately explain the phenomenon they identified. The problem is deeper; there’s nothing to explain or account for. … Perry and Lewis, and their followers have observed fragments of various different phenomena. They have then made the mistake of thinking that there’s something—call it essential indexicality—that ties all of these fragments together, when in fact there’s no unified phenomenon at all. It’s just a bunch of different phenomena with differing explanations. (Cappelen and Dever, p.3)

This deflationary, even dismissive stance raises an obvious question: how can so many have been so wrong for so long? In this paper I shall argue that while Cappelen and Dever are right in their observations about the under-elaborated, unclear and insufficiently justified character of claims made for the essentiality of indexicality, of attitudes de se and self-locating belief, the majority opinion has not been as wrong as they allege. Something deep has been seen with a glass but darkly and the goal of this paper is to bring it out—this via a critique of Cappelen and Dever’s argument that the entire topic is illusory.

Evidently, this goal requires that one first clarify how essential indexicality and the like have come to be understood in the literature. Let us look at four writers, Colin McGinn, Hugh Price, Cara Spencer and Carol Chen, all four of whom are more or less representative of a broad consensus that there are such things as essential indexicals, attitudes ineliminably de se and self-locating beliefs. In 1983 we find Colin McGinn claiming that

… indexical concepts are ineliminable because without them agency would be impossible: when I imagine myself divested of indexical thoughts, employing only centreless mental representations, I eo ipso imagine myself deprived of the power to act. (McGinn 1983, p.104)

Almost 30 years later Huw Price claims something similar:

In his well-known paper ‘The Problem of the Essential Indexical’ … John Perry points out that action depends on indexical beliefs. In addition to “impersonal” or “third-personal” information about her environment, an agent needs “situated” or “first-personal” information—information about the “I, here and now” of her situation. (Price 2007, p.1)

One could adduce passages from many other authors, all making similar claims. As Cappelen and Dever point out, little effort is made either to clarify these claims or to justify them. Typically, the reader is simply directed to Perry or Lewis.

This is surprising. What, after all, are the indexical thoughts, beliefs or information which McGinn and Price are claiming to be essential to agency? No doubt when John Perry moves from looking for the shopper who is making a mess to re-arranging the items in his shopping trolley, he does so because he has suddenly and quite explicitly realised, “I am the shopper making a mess,” “I am making a mess,” etc. But it is not generally true that when acting agents undergo such episodes, in which first person thoughts literally run through the mind. Nor can McGinn and Price be claiming that the capacity for such explicit indexical, in particular, first-personal thinking is essential. No mere capacity for indexical and specifically first-personal thinking could be essential for the exercise of agency since no mere capacity for anything could be essential for the exercise of another capacity.

McGinn, Price and the many others who make similar claims would no doubt protest that they do not understand by first-person thought or belief anything as literal as this. But what then do they mean by indexical, and in particular, first-person thought and belief? Perhaps McGinn, for example, really means something like the following:

(I)ndexical concepts are ineliminable because without them agency would be impossible: when I imagine myself divested of …. thoughts [whose content an agent with these thoughts and capable of expressing its thoughts linguistically, must express in indexical, indeed first-personal terms if it is to express this content perspicuously. In other words, when I imagine myself employing only centreless mental representations [,which are precisely representations whose content a being capable of expressing its thoughts would have to express non-indexically and in particular non-first personally], I eo ipso imagine myself deprived of the power to act. (McGinn 1983, p.104)

And similarly Price the following:

In his well-known paper ‘The Problem of the Essential Indexical’ … John Perry points out that action depends on … beliefs [whose contents an agent with these beliefs, insofar as it is capable of expressing its beliefs linguistically at all, must express using indexicals if it is to express these contents perspicuously]. In addition to “impersonal” or “third-personal” information about her environment, an agent needs “situated” … information—information [which the agent, given the capacity to express her beliefs linguistically, must express as information] about the “I, here and now” of her situation. (Price 2007, p.1)

And so on, mutatis mutandis, for the many others who make similar claims.

Clearly such convoluted reformulations avoid the falsity or incoherence of literal interpretation. Moreover, there are precedents for them in the literature. For instance, Spencer claims that the problem of the essential indexical “concerns indexical belief, that is, belief expressible with a sentence containing an indexical pronoun such as ‘I’, ‘here’, or ‘now’,2 or a demonstrative pronoun such as ‘this’ or ‘that’.” (Spencer 2007, p.179; italics added) According to Chen, in his paper “Perry makes a persuasive case for the idea that we must use indexicals to characterize certain beliefs that are required to give a rational explanation of a subject’s bodily actions.” (Chen 2002, p.13)

Of course, these precedents are all little more than remarks in passing. They are also very imprecise; they fail, for example, to distinguish clearly between the expression or attribution of belief and the expression or attribution of a content of belief—between what Perry (Perry 2006, p.204) eventually calls reporting a belief, i.e., attributing the belief itself; and expressing it, i.e., stating what is believed, the content of belief. Yet the thesis of essential indexicality is primarily a claim about the contents of belief: the change in Perry’s behaviour in the supermarket is due to his acquiring beliefs with a different content and so it seems that beliefs with a certain kind of content, a content which Perry himself would express using indexicals, specifically, the first person, are required for his agency.

Moreover, these passing remarks are, when interpreted literally, false. Thus, Spencer’s characterisation is too weak: Beliefs and contents of belief are expressible with all sorts of different sentences, indexical or non-indexical in varying ways and to different degrees. My friend Mary might believe, and I might know her to believe, that John has lost his wallet at the supermarket. If I now went to the supermarket to look for the wallet on John’s behalf, I could quite correctly say to the manager, “My friend Mary believes that John has lost his wallet here” or, if John is with me, “My friend Mary believes that this man here has lost his wallet in the supermarket.” Or perhaps John mutters to me while we are standing in the checkout queue of the supermarket that the man in front of us was very rude to the cashier. Unfortunately, the rude man overhears him. This individual could very naturally say to the cashier, “The man behind me believes that I was very rude to you.” So indexical and in particular first person belief or thought is not simply belief or thought expressible with a sentence containing an indexical pronoun such as ‘I’, ‘here’, or ‘now’ since many non-indexical beliefs are expressible with indexical sentences.

By contrast, Chen’s characterisation is too strong: when we explain Perry’s change of behaviour in the supermarket, we typically say that Perry realised that he was the shopper making a mess, that he was making a mess, etc. Now on a standard understanding of indexicality it is non-linguistic rather than linguistic context which determines the reference of an indexically functioning singular term; it does not borrow its reference from that of another term in some relevant stretch of discourse. Yet in the explanation of Perry’s behaviour this is precisely what occurs; the pronoun ‘he’ is functioning anaphorically rather than indexically. The pronoun ‘he’ can, of course, be used indexically. I might, for example, point to an individual and say, “He is a millionaire.” But in this case, the pronoun is functioning in demonstratively indexical fashion, not anaphorically.

Of course, the fact that it is functioning in such demonstratively indexical fashion might lead one to regard such reference as a species of indexicality alongside which one might then place anaphora as another species of indexicality. Perry himself gestures towards this—see Perry 2012, pp.68-73, esp. Table 1 on p.69, which lists anaphora as a type of indexicality. Such an understanding of anaphora would make what Chen says true. Even so, one would still have to distinguish anaphoric indexicality from the kind associated with notions of essential indexicality and self-locating belief since the latter kind concerns expressions such as ‘I’, ‘here’ and ‘now’. These expressions refer quite differently and distinctively from anaphorically used pronouns: non-linguistic context determines their reference and so they never borrow it from other terms in the linguistic context. For this reason, Perry characterises the kind of indexical reference they enable as narrow and explicitly distinguishes it from the wide type enabled by anaphora.3 Given the need to make this distinction, characterisation of the anaphoric use of third-personal pronouns in the attribution of belief as indexical merely requires one to reformulate notions of essential indexicality, of self-locating belief and even of attitudes ineliminably de se, as well as whatever problems they raise, in terms of narrow indexicality.

In any case, there is a prior and more fundamental problem confronted by efforts to characterise a kind of belief or thought as indexical, indeed first-personal because indexicals or the first person are used in the linguistic expression or attribution of such belief. The fact that Spencer’s characterisation is too weak shows that the attempt to characterise any kind of belief or thought in terms of its linguistic expression must appeal to a particular kind of expression or attribution, hence to a particular kind of sentence—that kind of expression or attribution which seeks to capture the belief’s inherent intentional structure, the structure which explains why it must be characterised by such and such a sentence and no other insofar as it is characterised perspicuously.4 For as we have seen, in ordinary conversational contexts one can correctly express and attribute belief in all sorts of ways, using all kinds of sentence. In short, one must embrace the notion of perspicuous expression or attribution, hence the notion of that sentence which depicts how the belief in question really or inherently is. But this raises the question of how one knows that one is expressing or attributing perspicuously, how one knows that such and such a sentence accomplishes perspicuous expression or attribution. Circularity now threatens, for surely one can only know what sentence such and such a belief disposes one to utter in order to express or attribute it perspicuously by knowing what the belief is in its inherent intentional structure.

As we shall see in section three, Perry himself, across three influential papers from 1977, 1979 and in particular 1980, works towards a characterisation of belief states in terms of the sentences they dispose one to utter insofar as one seeks to express or attribute them perspicuously. This yields in turn an account of indexical and first-personal belief that avoids the superficial difficulties to which Spencer’s and Chen’s merely passing remarks about indexical belief are subject. At the same time, he cannot avoid the circularity just intimated. As we shall also see, the conception of a belief state he develops across these three papers tacitly presupposes what it seeks to explain. So in principle one cannot explicate indexical, in particular, first-personal belief in terms of the sentences someone willing and able to describe the belief and its content perspicuously would have to use.

Let us, however, defer further consideration of this until section three. For the moment, two general observations about the literature on essential indexicality suffice. Firstly, this review of what four representative authors take the thesis of essential indexicality or, alternatively, of self-locating beliefs or attitudes ineliminably de se, to be is, just as Cappelen and Dever allege, radically unclear. Secondly, it is so at least in part because the notion of so-called indexical, in particular, of first person belief is left unclarified. These authors take it as obvious or at least unproblematic what such belief or thought is. Any attempt, therefore, to determine what claims of essential indexicality really amount to and why they present a problem for traditional conceptions of belief and its contents must acknowledge and address this lack of clarity.

Instructively, Cappelen and Dever make no such attempt. They, after all, have no desire to reconstruct the claims and arguments of those who embrace notions of essential indexicality, self-locating belief and attitudes ineliminably de se. Rather, they proceed directly to their argument against essential indexicality, an argument which purports to show that indexical and in particular first-personal belief is essential neither to the explanation or justification nor to the generation of action. I will present this argument in the next section (§ 1) in order then to show (§ 2) that their lack of concern for reconstruction is a mistake. Pace Cappelen and Dever, there is something deep and philosophically important towards which those who endorse notions of essential indexicality, self-locating belief and attitudes ineliminably de se are inchoately gesturing. Moreover, one can get at this through a critique of their own argument. Specifically, this argument can be used as a basis for reconstructing out of the intuitions of those who endorse essential indexicality and the like a counter-argument which reveals precisely both where the problem Perry was glimpsing emerges; and where the concept of indexical and specifically first-personal belief intervenes to generate the thesis of essential indexicality. In so doing, it shows that the problem emerges prior to the intervention of concepts of indexicality and the first person. So while there is indeed a genuine problem concerning the explanation, justification and generation of action for traditional conceptions of belief and its contents, it does not really or primarily concern indexicality. The problem is not one of the essential indexical; in this sense (and this sense only) Cappelen and Dever are right to say that there is no distinctive problem of the essential indexical.

This of course raises the question of why the concept of indexical and specifically first-personal belief should ever have intervened at all. Precisely here it becomes crucial not to leave this notion unexplicated, as do Cappelen and Dever. For once one appreciates the role played by the account Perry gives of belief generally, hence of indexical and in particular first-personal belief specifically, one can see why he puts the problem he uncovers in his paper of 1979 as a problem of indexicality. And this is crucial for understanding both what the problem is of what Perry calls the essential indexical; and why in characterising in this way he misdescribes it.

§ 1: Cappelen and Dever’s Critique of the Essential Indexical

Having pointed out how unclear and insufficiently justified the thesis of essential indexicality is, Cappelen and Dever set out to clarify it to the point where they have a clear target to aim at. They rightly claim that “attempts to articulate the essence of Perry and Lewis’s insight tend to come back to something or other about a distinctive connection with agency.”5 (Cappelen and Dever, p.18) This connection is, for example, clearly evident in the passages quoted above from McGinn, Price and Chen. Consequently, Cappelen and Dever orient their critique around a consideration of the putative role of indexicality in agency in the broad sense of purposive behaviour. Specifically, they begin with an account of the explanation and justification or, as they call it, the rationalisation of such behaviour.

Consider the following explanation and rationalisation of François’ clearly purposive behaviour, hence in this broad sense action, of ducking under a table in order to avoid being shot:

  • Belief: François is about to be shot.

  • Desire: François not be shot.

  • Belief: If François ducks under the table, François will not be shot.

  • Action: François ducks under the table. (Adapted from Cappelen and Dever, p. 36)

Cappelen and Dever call this an impersonal action explanation or impersonal action rationalisation. They do so because the content of the beliefs and desires involved in it are not indexical, hence not first-personal. These beliefs and desires are thus completely impersonal: they can be had by anyone. Precisely because they can be had by anyone, François, although obviously the subject of these beliefs and desires, is not mentioned as their subject since their being his is not constitutive of their identity. This notion of impersonal belief—of a content which can be “grasped”, hence of a belief with this content which can be had, by anyone—needs to be borne in mind. For it plays, as we shall see, a key role in identifying the core phenomenon underpinning talk of essential indexicality and the like.

The notion of impersonal action explanation or rationalisation permits Cappelen and Dever to formulate what they take to be the core thesis of essential indexicality. This presumed essential connection Cappelen and Dever formulate as the Impersonal Incompleteness Claim or IIC:

Impersonal action rationalisations and explanations “… are necessarily incomplete because of a missing indexical component.” (Cappelen and Dever, p.37)

Note a distinctive feature of the IIC. McGinn and Price construe indexicality as essential in the generation or causation of action. McGinn, for example, claims that indexical concepts, hence beliefs and desires involving these concepts, are essential to agency, not to the explanation or rationalisation thereof. Yet the IIC speaks only of the explanation or rationalisation of agency. Clarifying why Cappelen and Dever formulate the IIC in terms of the explanation or rationalisation of action rather than its generation is in fact crucial for understanding both the claim Perry is making when he uses various examples in order to demonstrate what he describes as essential indexicality; and the basis for the deflationary response Cappelen and Dever give to this claim. We must therefore consider why they put things this way.

i. Why is the Impersonal Incompleteness Claim about the Explanation or Rationalisation of Action rather than the Generation Thereof?

On the face of it, explaining why Cappelen and Dever formulate the IIC in terms of the explanation or rationalisation of action is easy: this better reflects how Perry proceeds in identifying the problem he calls that of the essential indexical. As Perry puts things in a later paper, this time with regard to his action of leaving for campus because he has a class to teach,

… my use of the indexical “I” in expressing and reporting my belief that I had a class to teach was essential to the expression or report of belief explaining the fact that I was leaving for campus. I didn’t think [in either this later paper or the one from 1979] that there was some proposition [as opposed to some belief!] that could only be expressed with the indexical “I”. (Perry 2006, p.213)

For this reason, and writing now of the well-known example of Perry’s making a mess in the supermarket, Cappelen and Dever believe that the claim Perry is making concerns explanation contexts, hence the explanation or rationalisation rather than the generation of action.

Now the fact that Perry, in formulating his claim, does so with regard to the the expression or reporting of belief, typically for purposes of explanation or rationalisation, suggests to Cappelen and Dever their deflationary response to Perry’s argument. Specifically, they argue that although McGinn, Price and indeed Perry himself understand the claim and problem of essential indexicality to concern the very nature and structure of belief itself, in reality Perry merely demonstrates something linguistic, more precisely, something concerning the logical behaviour of indexical singular terms. “The claim Perry makes about [the supermarket case] and the related scenarios is,” they say, “most naturally read as a claim about the opacity of explanation contexts.” (p.32) In explanations generally co-referential singular terms often cannot be substituted salva explanatione even when they can be substituted salva veritate. And in explanations specifically of action they cannot be substituted even salva veritate since such explanations involve psychologically intensional contexts. In consequence, in explanations of action they can never be substituted salva explanatione because they cannot be substituted salva veritate.

None of this is, of course, new or surprising. Nor indeed does it concern indexical singular terms specifically. Understandably, then, Cappelen and Dever describe it as “surprising to find Perry summarizing his view in [this] way … ” (p.34), namely, as a point about the behaviour of (a certain kind of) singular terms in (a certain kind of) linguistic context. Proponents of essential indexicality agree that “the allegedly distinctive feature of indexicals goes beyond not being substitutable in explanation contexts.” (p.34) And Perry himself thinks he is making a more interesting and important point than this. Although he formulates matters in terms of what Cappelen and Dever calls the explanation or rationalisation of action, he does take himself to be showing something about how belief generates or causes action. To put things in the language of Cappelen and Dever: he takes himself to be showing that an impersonal action explanation or rationalisation is incomplete because it fails to mention certain key states of belief involved in the generation of action which are in some sense indexical, indeed sometimes first-personal. These belief states, he believes, cannot be adequately characterised by the traditional conception of belief, thanks to its reliance on a traditional conception of the contentfulness of belief

Cappelen and Dever fail to see, however, or at least do not sufficiently appreciate, how important the qualification “in some sense indexical, indeed sometimes first-personal” is. For Perry the claim that indexical belief is essential to the generation or causation of action (and only for this reason to the explanation or rationalisation thereof) is bound up with a particular conception of belief which operates in the background of Perry’s all too short and under-elaborated argument. When in his paper of 1979 Perry speaks of the belief states causally involved in the scenarios he considers as essentially involving indexicality, he is presupposing that conception of belief which is already implicit in Perry 1977 and more fully elaborated in Perry 1980. As was pointed out above, this conception of belief constitutes a sophisticated version of the kind of thing encountered in Spencer and Chen. In particular, it attempts to individuate the contents of belief, and thus belief states themselves, in terms of the sentences one would use to express these contents if one were in the belief states oneself and capable of such linguistic expression and attribution.

Now this conception of belief determines why in his paper of 1979 Perry can describe the problem and phenomenon he uses various examples to reveal as a problem and phenomenon of indexical belief. For it yields the notion of specifically indexical belief operative in the background of his argument. This notion allows Perry to avoid the absurdity to which one can reduce the claim that agency requires indexical belief if one interprets the notion of indexical belief as the term quite literally suggests, namely, as a matter of having some indexical sentence run through one’s head in an act of mental judgement. Because Cappelen and Dever fail to see how Perry’s understanding of indexical belief shapes the very way Perry understands the problem he claims to have found, namely, as one of essential indexicality, they can see nothing in Perry’s examples and his use of them other than his appeal to the opacity of indexical singular terms in action explanation contexts. Everything now depends on showing what point Perry was really trying to make—this by showing in much more detail than Perry does how he makes it. Instructively, this can be done by appeal to Cappelen and Dever themselves, specifically by turning their own argument against them.

ii. Cappelen and Dever’s Argument against the Impersonal Incompleteness Claim

It is crucial to understand precisely what Cappelen and Dever claim to show in their argument for the inessentiality of indexicality. They claim to show that adequate explanation of action requires appeal only to non-indexical beliefs and desires of the traditional kind, i.e., beliefs and desires that p, whose content is strictly propositional, hence can be expressed by full sentences. Or, to put things in terms of the generation or causation rather than explanation or rationalisation of action, they claim to show that action only requires beliefs and desires of this traditional kind. One can, they maintain, imagine such beliefs and desires interfacing with various sub-personal mechanisms to generate action in a completely plausible way, without any mediation from any other kind of belief and desire. In particular, one needs no mediation by indexical beliefs and desires.

In order to see how exactly they attempt to show this, let us return to the structure of belief and desire illustrated above:

  • Belief: François is about to be shot.

  • Desire: François not be shot.

  • Belief: If François ducks under the table, François will not be shot.

According to Cappelen and Dever nothing more needs to be added by way of belief and desire in order to arrive at the explanandum or justificandum:

  • Action: François ducks under the table.

Thus far, this is mere assertion with which many would presumably disagree. So Cappelen and Dever set out to explain away the considerations which would lead to disagreement, i.e., to the belief that impersonal action explanations and rationalisations are incomplete. One might well wonder, they point out, “why it is the case that “when you and I both apprehend the thought that I am about to be attacked by a bear, we behave differently” (Perry 1977, p.494).” (p.49) At this point the following two-step argument appears compelling to many:

Premise 1 : Two agents can have all the same impersonal beliefs and desires yet act differently.

Premise 2 : In such cases there must be a difference in beliefs and/or desires between the two agents which explains the difference in behaviour.

Lemma : This explanatory difference must therefore be a difference in beliefs and/or desires which are not impersonal, that is, are personal, since ex hypothesi the two agents have the same impersonal beliefs and desires.

Premise 3 : The only kind of personal belief the agents could have, at least in the kind of case represented by François and Dilip, is first-personal, hence indexical.

Conclusion : In this kind of case a truly complete action explanation or rationalisation must attribute first-personal, hence indexical belief. Merely impersonal action explanations or rationalisations are incomplete because they lack this missing indexical component.

Clearly, the conclusion of this argument is the IIC, at least in its application to the kind of case represented by François and Dilip.

Evidently, those attracted by this kind of argument are assuming that if two agents display different behaviours, then they have different beliefs or desires. Or, as Cappelen and Dever put it, they are assuming that “when two agents have the same belief-desire sets, they perform the same action.” (p.52) More accurately, they are assuming that when two agents have the same belief-desire sets, they perform the same action ceteris paribus since for some external reason an agent might fail to perform an action towards which he or she is psychologically disposed, e.g., just as the agent is about to act, the agent is struck dead by a bolt of lightning. Cappelen and Dever rightly accept this assumption—see p.52. They claim, however, that this assumption does not entitle one to conclude that when agents with the same impersonal belief/desire sets behave differently, the difference in behaviour must be due to a difference in belief and desire, hence a difference in either first-personal or more generally indexical belief and desire. There can be, they maintain, some other difference between agents which explains the difference in behaviour, so the second premise of the argument above is false.

But what could this explanatory difference be if not one of belief and desire? Consider the following action explanation or rationalisation:

  • Belief: François is about to be attacked by a bear.

  • Desire: François not be attacked by the bear.

  • Belief: If François climbs into the tree and Dilip calls for help, François will not be attacked by the bear.

  • Action: François climbs into the tree and Dilip calls for help.

Evidently, François and Dilip could and in this case actually do have these impersonal beliefs and desires. Yet the behaviour of each is different: François climbs into the tree and Dilip calls for help, thereby together saving François from the bear. What could the difference be between them which explains the behavioural difference if not a difference in further belief?

Cappelen and Dever open their attempt to answer this question with the following move: in cases like that of François and Dilip,

actions will differ but not what we are rationally motivated to do. Someone other than François, e.g., Dilip, can be rationally motivated to perform the action that François climb into the tree. (p.52)

At first sight, talk of Dilip’s being rationally motivated to perform the action that François climb into the tree seems odd. Cappelen and Dever appear, however, to believe that to perform an action is always simply to make it the case that p, i.e., to so affect the world that the proposition that p becomes true. This makes the claim that someone other than François, precisely Dilip, might be rationally motivated to perform the action that François climb into the tree less odd. Cappelen and Dever now go on to claim that even though Dilip might be no less rationally motivated to make it the case that François climb into the tree,

Dilip just can’t perform the action. François can perform it, so he does. François is, like Dilip, rationally motivated to perform the action that Dilip call for help. But François can’t perform that act. So François climbs into a tree and Dilip calls for help. (p.52)

Why, one will ask, can François not perform the action Dilip performs, namely make it the case that Dilip calls for help? And why can Dilip not make it the case that François climbs into a tree. Each, say Cappelen and Dever, is subject to “various physical or psychological constraints” which limit the range of actions he can perform. More generally,

a given agent has an “action inventory”: a range of actions that he can perform. An agent constantly seeks to match his intentions with his action inventory, and when he finds a match, action occurs. When there’s no match, the intention idles, and doesn’t motivate or rationalize action. … (O)nly certain actions result because only certain actions were available in the first place. (p.50)

It seems, then, that although the beliefs and desires of each agent cause an intention to do what the other does, because this intention does not match up with a doable action—what Cappelen and Dever call an actionable content—, it neither causes nor makes rational any action.

Cappelen and Dever make clear that this process of matching is not to be understood as something done consciously. Rather, it occurs at the sub-personal level:

(N)o one has to “look” to see if there’s a match, … hence no one has to “think about available actions” in any way, let alone a first-person way. The actions have already been “thought about” by the time they emerge as potential actions (and thought about just as … third-person [actions of the form] “that p”); all that remains is to see if the actions are among the things that can be done. (p.51)

Obviously, if the agent had to think about available actions in a first-person way, then we would have reinstated essential indexicality. But right from the outset this is ruled out because, properly understood, the matching process involves no thinking at all. Clearly, if this story suffices as an account of the explanatory difference between François and Dilip, then Cappelen and Dever can maintain that their completely impersonal action explanation and rationalisation is complete. But is their argument really satisfactory?

§ 2: The Incompleteness of Impersonal Action Explanations or Rationalisations

There are two kinds of flaw in the way Cappelen and Dever seek to show that impersonal action explanations or rationalisations are complete, hence adequate. Firstly, their account of the explanatory difference between François and Dilip is arguably internally incoherent. Secondly, it involves a category mistake in the sense that it does not really speak to the question which needs to be addressed.

In at least four respects, the account seems internally incoherent:

  1. Cappelen and Dever have claimed that in the case of François and Dilip (and, for that matter, anyone else who happens to hold the beliefs these two agents hold) the one can be rationally motivated to do what the other does without actually doing it. Yet the story they tell of action-inventories, matching and the like entails that the intention to do what the other does finds no matching actionable content and so, as they themselves say, does not motivate anything—see p.50, cited above. So doing what the other does is not rationally motivated after all. Is this not an outright contradiction which one can only avoid by maintaining that while beliefs and desires can motivate an action, the intention they cause can fail to do so?

  2. Having an intention surely entails having a motivation of which distinctively self-conscious beings like François and Dilip could be aware. It is, however, phenomenologically and intuitively clear that François and Dilip have no motivation to do what the other does in any sense which would involve either actual or potential awareness of themselves as having this motivation.

  3. According to Cappelen and Dever belief and desire combine to generate intentions to act which then “hit the “action center,” … a big switchboard with a bunch of available actions” in order to determine whether the potential actions which the intentions constitute “are among the things that can be done.” (Cappelen and Dever, p.51) Clearly, the actions available in the action inventory or action centre cannot themselves determined by belief and desire since they are that with which the intentions to act are compared to determine whether they can be realised. If, however, this is so, then for any action that a be F in the action centre, if b = a, then the action that b be F is also in the action centre. Moreover, these must be different actions since according to Cappelen and Dever both to act is to make it the case that such and such a proposition is true and the proposition that a is F is not the proposition that b is F. Trivially, however, there are infinitely many co-referential terms of the form “The object at <x, y>” since there are infinitely many frames of reference. So there must be infinitely many actions in the action inventory or action centre. But surely there can be no process of sorting through a chronically infinite number of possible actions in order to determine whether an intention to act generated by belief and desire is “among the things that can be done.” (p.51)

  4. What hits “the “action center” and is compared with “a bunch of available actions” in order to determine whether it can be done is an intention to act. So on this picture of them, intentions are formed prior to determining whether they can be realised. Surely, however, one only forms an intention if one believes that one can act upon it or at least if one does not believe that one cannot act upon it. There is no further question to be resolved once intentions have been formed; it remains only to act. Naturally, one could discover that one cannot in fact do what one intended to do. But one discovers this by doing and failing. And this cannot be what Cappelen and Dever mean by the matching process since the matching process cannot be identical with behaviour if different matching processes are to explain different behaviours.

But these four kinds of potential internal coherence are not the truly fundamental problem with the attempt Cappelen and Dever make to find an explanatory difference between François and Dilip which is not a matter of belief. For in fact this exercise in speculative cognitive psychology answers the wrong question. At best, the account constitutes a hypothesis about how the explanatory difference between François and Dilip is sub-personally realised or implemented but it does not tell us what this difference is. In this sense, the account involves a kind of category mistake. When Cappelen and Dever claim that although “Dilip can be rationally motivated to perform the action that François climb into the tree … “, “he just can’t perform the action” (p.52), what sense of ‘cannot’, hence of ‘can’, are they operating with? The action that François climb into the tree is not in Dilip’s action centre. That is, it is not an actionable content for him. But

(w)hat constitutes actionable content for an agent will depend on the kinds of sensory inputs the agent has—for example, whether he moves around the world only with smell and taste or using sonar and sounds. It will depend on the agent’s physical setup: whether he has fingers, claws, a trunk, wings, feet, fins, etc. It will depend on whether the agent’s skills include steering cars out of skids or balancing plates on poles, and whether the agent’s casual (sic.) reach includes having employees publish Kripke’s manuscripts. (Cappelen and Dever, p.51)

So an actionable content is simply a function of what an agent can physically do. In other words, the sense of ‘can’ and ‘cannot’ with which Cappelen and Dever operate is that of bodily or physical ability and inability. But this is not the right sense of ‘can’ and ‘cannot’, the one actually operative in the case of François and Dilip. After all, the action that François climb into the tree may well be in Dilip’s action centre in a purely bodily or physical sense. That is, Dilip may well be able to move François to climb into the tree. Similarly, François may well be able to move Dilip to call for help. But neither do these things because they cannot do them rationally: from the perspective of ensuring that François not be attacked by a bear, it is better that Dilip call for help rather than encourage François to climb into the tree and that François climb into the tree rather than encourage Dilip to call for help.

If, however, difference in behaviour is due to what different agents can and cannot rationally do, then we must explain it in terms of differences in belief. These differing beliefs need not be beliefs about what each can rationally do because perhaps the agents in question are not sophisticated enough to have such beliefs. But each must at least have beliefs about its particular situation which determine to what degree what it can physically do it can do rationally. In other words, in addition to whatever impersonal beliefs each has and possibly shares with the other, it must have beliefs about its particular situation which so to speak select amongst what it can physically do for what it can rationally do.6 In principle, impersonal beliefs can only cause action insofar as they are mediated by beliefs which constitute awareness of the particular situation.

Note that this mediation must be accomplished by beliefs about the particular situation; it cannot be accomplished by the particular situation itself, in all its distinctiveness and difference from the situation of another agent with the same impersonal beliefs. If this were not so, then whatever beliefs about its particular situation the agent had would be causally irrelevant. But that such beliefs are not causally irrelevant is shown by the fact that when the agent has beliefs about its particular situation which diverge from this situation itself, i.e., when these beliefs are erroneous, its behaviour diverges from what the situation objectively determines as appropriate. (And of course even when this behaviour diverges from what the situation objectively determines as appropriate, it can still be perfectly rational.) Causal mediation between impersonal belief and desire on the one hand and action on the other can only be accomplished by intervening beliefs about the agent’s particular situation. It is indeed constitutive of impersonal belief and desire precisely as impersonal that it operate causally through belief (and, for that matter, desire) concerning the agent in its particular situation.7 Evidently, the belief (and desire) which mediates between impersonal belief and desire on the one hand and action on the other cannot itself be impersonal. It must be non-impersonal and therefore personal.

Yet as we shall now see, this does not entitle one to conclude that the belief (and desire) which mediates between impersonal belief (and desire) on the one hand and action on the other is first-personal, hence indexical. Consider once again François’ action of ducking under a table in order not to be shot. Someone who does not identify with François might explain or rationalise François’ action in the following third-personal terms:

  • François believes: He is about to be shot.

  • François desires: He not be shot.

  • François believes: If he ducks under the table, he will not be shot.

  • Action: François ducks under the table.

The attributions of belief and desire contained as components in this action explanation or rationalisation are obviously not impersonal in the sense defined above. That is, they are not beliefs (and desires) which anyone could have. Precisely for this reason, the subject of belief must now be specified in these attributions since this ensures that the pronoun ‘he’ refers to the right object, namely, François, in the right way, namely, anaphorically, hence (third-)personally.8 Only if this is ensured will the right content be attributed. Now insofar as this explanation or rationalisation is complete, the beliefs (and, for that matter, the desire) attributed are not impersonal;9 they are therefore personal. But this is not yet to claim that they are first-personal. Before one can claim this, one must claim something further, namely, the third premise in the argument given above. For only this premise entitles one to maintain that the beliefs (and desire) in question are in reality first-personal even though they are being attributed in third-personal fashion, presumably by someone who does not identify with François.

But is the third premise true? Are the beliefs attributed in this action explanation or rationalisation in reality first-personal, hence indexical? The answer depends on what one means by first-personal belief. In this regard, there are, as we have already intimated, at least two possibilities. By first-personal belief one might mean what is suggested by talk of ‘first person thought’, ‘belief about the ‘I’, ‘here’ or ‘now’ of one’s situation’, etc: thinking quite literally in the first person, either an actual episode of thinking to oneself, “Φ(I/me),”10 or the disposition thereto. Alternatively, one might mean something counterfactually dispositional: a first-personal belief is a belief state such that a subject in this belief state, capable of English and both willing and able to express the content of this belief state perspicuously, would use a sentence of the form “Φ(I/me)” to do so. If we embrace the first of these two options, then we render the third premise false since as we have seen in commenting on McGinn and Price, neither agency requires explicit, first-personal thinking nor does complete explanation or rationalisation of action require attribution of such thinking. So only the second option remains and clearly it makes the third premise true.

But is the second option acceptable? Already a simple consideration suggests that it is not: it in effect characterises indexical and in particular first-personal belief by appeal to the sentence a subject capable of English and prepared to express perspicuously the content of its belief would be disposed by the belief to utter. So what is it about the belief, more precisely, about its content, which explains why the belief disposes to utter this sentence and no other? Obviously, it will not do to say that this is the character of the belief as having a content expressible by such and such a sentence of the form “Φ(I/me).” This simply re-states the explanandum; it does not state an explanans. It seems, then, that some structural, non-dispositional characterisation of the belief, in particular, of its content, is needed if one is to explain the dispositional property in terms of which the character of the belief as indexical, specifically, first-personal, has been defined.

We will, however, defer more thorough consideration of the second option until the next section. For as indicated above, Perry himself initially adopts a counterfactually dispositional conception of belief and its contents which yields a sophisticated version of this counterfactually dispositional characterisation of indexical and first-personal belief. Understanding this conception of belief is crucial to appreciating why Perry construes the problem he outlines in his paper of 1979 as a problem of essential indexicality. Moreover, we must also consider the possibility of options for explicating the concept of indexical, and in particular, first-personal belief beyond the two thus far considered. This is all the more necessary as Perry, in his later work, in particular, in Crimmins and Perry 1989 and in the first edition of Perry 2012, published in 2001, himself pursues a third option for explicating the concept of belief in response to the problems his original account encounters. This third option is examined in section four.

For the moment, then, let the following preliminary diagnosis suffice: whatever may or may not be wrong with Perry’s conception of belief and the account of indexical, specifically first-personal belief it yields, the argument thus far has already identified the problem he identifies in his paper of 1979. Moreover, it has done so without recourse to notions of indexical or distinctively first-personal belief. The fundamental problem created by the role of belief in agency for the traditional conception of belief lies prior to the operation of the third premise, the premise which yields essential indexicality and the IIC. For it lies in the lemma of the argument rather than its conclusion. According to this lemma, impersonal belief alone is insufficient for agency because distinctively personal belief is needed. It is left entirely open whether personal belief is eo ipso indexical and in particular first-personal belief, however one understands these latter notions. The real point of Perry’s examples, the real problem they intimate, may be put as the claim that impersonal action explanations or rationalisations are necessarily incomplete because of missing personal component.

An example of a personal belief is simply François’ belief that he is about to be shot. An example of a personal component missing from a strictly impersonal action explanation or rationalisation is the statement that François believes that he is about to be shot. The problem of the essential indexical can be put without recourse to indexicality or the first-person at all. One might call it the problem of essential anaphora: agency requires belief whose content must be characterised in terms involving anaphoric reference to the subject of belief and so the complete explanation or rationalisation of agency requires attribution of belief in which such anaphoric reference is involved. Clearly, this creates just the same problem for the traditional conception of belief and its contents as would the essential (first-personal) indexical. Of course, if the second way of understanding indexical and specifically first-personal belief is acceptable, the third premise therefore true, then the phenomenon of essential anaphora just is the phenomenon of essential indexicality. Impersonal action explanations or rationalisations are incomplete. And the phenomenon required to render action explanations or rationalisations complete does engender a significant philosophical problem, namely, for the traditional conception of belief and its contents.

But now a crucial question arises: if the problem towards which Perry was gesturing can be put as a problem of essential anaphora and still be just as much of a problem for the traditional conception of belief, why does Perry not put things this way? Why the emphasis on first-personal, hence indexical rather than third-personal, hence anaphoric attribution, given that the latter would suffice to make the point about the problem the generation, explanation and rationalisation of agency raise for the traditional concept of belief and the contents of belief? Here it is crucial to determine the role played in Perry’s argument by his conception of belief and its contents, which, as we have seen, yields sophisticated version of the second option available for construing indexical and in particular first-personal belief. As already noted, if this conception is acceptable, then the third premise of our reconstruction of his argument is true and the problem he uncovers is indeed a problem of essential indexicality.

More, however, follows from the acceptability of Perry’s initial conception of belief generally, hence of indexical belief specifically, than this. In particular, it follows that first-personal, indexical characterisation, hence attribution, of belief is prior to third-personal anaphorical characterisation—prior in the sense that it captures perspicuously what the belief characterised, hence attributed is. In other words, third-personal anaphorical characterisation and attribution, in which content is individuated so to speak oratio oliqua, by using a sentence (“François believes that he is about to be shot”) is merely a convenient façon de parler. By contrast, first-personal indexical characterisation and attribution, in which content is individuated so to speak oratio recta, by quoting a sentence (François has a belief which would dispose a sincere English speaker to assert, “I am about to be shot”), gets at how things really are. This is a crucial result. For if first-personal indexical characterisation and attribution is primary in the indicated sense, then the content of at least first-personally indexical belief can be individuated by means of a sentence which, however much it might rely on non-linguistic context for the determination of its references, is free-standing in the sense that it does not rely on linguistic context, in particular, on any psychologically intensional embedding, for the determination of its references. This means in turn that one can maintain even for first-personally indexical belief that it conforms to the traditional conception of belief as a relation to an object of belief. In short, one can give an account of such belief which both conforms to the traditional model and accounts for the causal role of such belief in the generation or causation of action. This is fundamentally why Perry speaks of essential indexicality and specifically, for the case of his making mess, of first-personality. And so this is why the third premise must be included in any reconstruction of his argument.

What, however, if Perry’s initial conception of belief should fail? What indeed if third-personal anaphoric characterisation and attribution were not in fact posterior but rather prior to first-personal indexical attribution—prior because it perspicuously captures the intentional structure of the belief attributed in the sense that an account of first-personal, hence indexical self-ascription of belief draws upon, and so itself presupposes, third-personal anaphoric characterisation? Then one would have to conclude that Perry was wrong to speak of essential indexicality and the first person. And one could, indeed would have to, drop the third premise from the reconstruction of his argument. For now one would have to say that what Perry was really getting at was essential anaphora. In other words, in his paper of 1979 Perry was obscurely glimpsing the following:

There are forms of belief such that

  1. Someone S* who does not identify with the subject S of these forms of intentionality only truly individuates them as the particular intentional states or experiences they are if S* uses a sentence of the form “S CV Φ(PN)”, in which

    (a) ‘CV’ is some cognitive or epistemic verb such as ‘believes’, ‘realises’, ‘sees’, etc.;

    (b) ‘PN’ is a third personal pronoun, whether in the nominative or some other case, which refers anaphorically to S; and

    (c) ‘Φ’ is whatever linguistic construction—sentential function, gerundive function, etc.—is required for forming, together with ‘PN’ and, as the case may be, also the subordinating conjunction ‘that’, an appropriate complement to the verb ‘CV’.

  2. These forms of belief cannot plausibly be interpreted as cases of first-person thought or belief under any understanding of what it is to be a first-person thought or belief.11

  3. These forms of belief are necessarily involved in the exercise of agency, hence are, in this sense, essential to agency.

Finally, one would have to concede that the traditional conception of belief as a relation of a subject to some object of belief (and of a content of belief as precisely such an object) fails. For if there are forms of belief which can only be perspicuously characterised in third-personal anaphoric fashion, then there are forms of belief whose content is not an object of belief, at least not when this latter notion is understood to be something which could be individuated by a free-standing sentence, whether indexical or not.

§ 3: Belief and Acceptance—Perry’s Early Conception of Belief

As already indicated, in three papers from 1977, 1979 and in particular 1980 Perry makes a first attempt to develop a conception of belief which permits him to preserve the traditional idea that belief is a relation to an object of belief even for the case of (what he and others in the literature call) indexical belief. Perry begins by claiming that “(w)e use sentences with indexicals or relativized propositions to individuate [the ostensibly indexical belief states essential for action], for the purposes of classifying believers in ways useful for explanation and prediction.”12 (Perry 2000, p.39) Note that, prima facie, this claim is false because as we have also seen the ostensibly first-personal belief states essential for agency—what we have called here simply personal beliefs—and the contents of such belief states may be characterised in third-personally anaphoric terms. At no point in the explanations or rationalisations given above of François’ action of ducking under a table were sentences containing indexicals used to characterise the beliefs which led him to perform this action. It sufficed simply to judge or assert, “François believes that he is about to be shot.”

It seems, however, that Perry does not intend to be taken so very literally. He goes on to say,

(B)elief states individuated in this [indexical] way enter into our common sense theory about human behavior and more sophisticated theories emerging from it. We expect all good-hearted people in that state which leads them to say “I am making a mess” to examine their grocery carts, no matter what belief they have in virtue of being in that state. That we individuate belief states in this way doubtless has something to do with the fact that one criterion for being in the states we postulate, at least for articulate sincere adults, is being disposed to utter the indexical sentence in question. (Perry 2000, pp.39-40)

So Perry is not denying that one can individuate François’ belief in the completely third-personal, impersonal manner indicated. Rather, he is saying that this belief is perspicuously depicted as a belief state which, were it had by an articulate sincere adult, would dispose this latter to utter the corresponding indexical, indeed first-personal sentence, “I am about to be shot.” Evidence for this lies in the fact that we place such weight upon what people themselves say about their beliefs and other psychological states when we seek to explain or justify their actions. If, however, this is so, then anaphoric individuation is merely a convenient device for attributing something whose real structure is perspicuously captured at least as well by appeal to the notion of an indexical sentence and a relativised proposition.

Clearly, waiting in the wings here is a version of the second option considered above and intimated by Spencer and Chen: indexical and in particular first-personal belief is belief which, should it be had by an articulate sincere adult, would dispose this latter to utter some corresponding indexical, indeed first-personal sentence. This conception of indexical and in particular first-personal belief has two important consequences: Firstly, it allows for the possibility that non-linguistic beings such as dogs and cats, or pre-linguistic beings such as children, can find themselves in at least similar belief states to the one in which François find himself. As a matter of fact, François is an articulate sincere adult capable of English, so François is actually disposed by his belief state to assert sincerely, “I am about to be shot.” This need not, however, be the case. Of course, non- and pre-linguistic beings might not be able to have beliefs as complex as ones whose truth entails the shooting death of someone. Even so, we can easily imagine simpler beliefs expressible in the form “Φ(I/me)” which non- and pre-linguistic beings could have. Secondly, this conception of indexical and specifically first-personal belief makes the third premise of the reconstruction of Perry’s argument given in the previous section true, indeed almost trivially so. It thus directly yields the IIC, indeed it determines how this is to be understood. For it determines that the missing indexical component which according to the IIC impersonal action explanations and rationalisations lack consists in attribution of a belief state such that, if one had this belief state oneself and were a sincere and competent speaker of English seeking to explain or justify one’s own action, this belief state would dispose one to utter an appropriate indexical, indeed first-personal sentence by way of characterisation of the state’s content.

Now this conception of indexical and in particular first-personal belief enables Perry to avoid the threat to the traditional conception of belief presented by the essentiality of non-impersonal, hence in this sense personal belief to agency. This can be seen by considering how matters would stand were we to regard the third-personal anaphoric characterisation of such belief as perspicuously capturing its intentional structure. To do this would be to admit forms of belief whose contents cannot be individuated by means of a full sentence, not even when the sentence contains indexical singular terms. In consequence, we would not be able to regard the beliefs themselves as relations to objects of belief since we would have lost grip on what it could mean for a content of belief to be an object of belief to which a subject could stand in the relation which is belief. If, however, personal belief can be explicated as indexical and in particular first-personal in Perry’s counterfactually dispositional sense, then we leave room for the traditional relational conception. Such belief would not, of course, be an actual relation to a sentence since no sentence is actually involved or implicated in belief. But one could construe it as an actual relation to a proposition on the basis that it is inherently such as to dispose, possibly only counterfactually, anyone in this belief state and willing and able to characterise the state perspicuously to utter a sentence which expresses a proposition of the same type, at least in the context of utterance, i.e., relative to certain entities—in the case of a first-personal sentence, to the utterer itself.13

True, this conception of indexical and specifically first-personal belief makes for a more complicated conception of belief as relational.14 More importantly, it requires further elaboration, specifically, a well-worked out account of how sentence and belief relate one another. Providing such an account is one of Perry’s concerns in his paper of 1980, “Belief and Acceptance.” Here he seeks to provide an account of belief in general which construes it, not as itself a triadic relation of the believer to a proposition through a sentence, but as imposing such a relation on a self-explainer or –rationaliser. The key to this account is Perry’s doctrine of acceptance:15

To accept a sentence S is to be in a belief state that would distinguish such speakers who would think and utter S from those who would not. Thus my conception allows an animal or preverbal child to be meaningfully said to accept a sentence. (Perry 2000, p.45)

Acceptance in this sense

… is not belief and not analyzable in terms of belief; rather, it is an important component of belief. … One has a belief by accepting a sentence. (Perry 2000, p.45)

So acceptance is part of belief because and in the sense that it is that by means of which we believe. At the same time, the sentence accepted is not literally part of belief. It would be more accurate to say that the sentence accepted is something which the belief state one has by accepting it, in virtue of what this belief inherently is, imposes on a self-attributor. Such self-attribution is essentially and actually a triadic relation of a subject to a proposition through a sentence. This is intimated by the following passage from Perry: “(C)onfusing acceptance has,” he thinks,

wreaked havoc in the philosophy of belief, in the philosophy of mind, and in metaphysics generally. It requires that we see what is believed, and so what is true and false, on the model of what is accepted; belief is thus treated as a relation to a sentence or sentence-like entity. … We want what is believed to classify belief states for purposes of explaining thought and action—the proper role of what is accepted—while at the same time being objectively true or false, the common objects of belief for different persons at different times. (Perry 2000, p.46)

In other words, we should not do what has been traditionally done, namely, use the concept of what is believed to account for both the causal or functional, or indeed inferential role of belief and the character of belief as having a content which is either true or false, a content which different subjects can believe at different times. Belief is an intentional state, mental judgement an occurrent mental event, in which the subject accepts a sentence, therein finding itself in a state or event which has a propositional content and disposes the subject to certain behaviours, including, of course, the utterance of the sentence accepted (in appropriate circumstances).

Clearly, by distinguishing what is believed from what is accepted whilst construing belief as achieved through (imposition upon a self-attributor of) what is accepted, Perry construes the relation to a proposition in which belief consists as mediated by sentences. The advantage of this is that what is believed need no longer be conceived as isomorphic in its structure to what is accepted. In consequence, it need no longer be a proposition in Frege’s sense. What is believed, says Perry,

is not a sentence, nor a sentence meaning, nor one of Frege’s thoughts—an abstract object with a sentencelike structure. It is rather, as Russell thought, a complex of objects and properties16—objects and properties which are part of the world, not part of the mind (except in rare instances). (Perry 2000, p.45)

Meanwhile responsibility for the causal or functional, indeed in part inferential role of belief is assigned to the dimension of acceptance. What is accepted thus determines what Stalnaker calls informational content,17 i.e., what in deference to Frege one might term its cognitive and practical value. In summary, we may say that to believe that Φ(a) is to be in a belief state whose propositional content is true if and only if Φ(a)18 by accepting the sentence “Φ(a).” Because the second occurrence of the singular term ‘a’ is completely extensional, the directedness of the belief state out into the world is Russellian rather than Fregean. That is, what is believed can be conceived in Russellian rather than Fregean terms; it is a Russellian proposition.

Now the one reason why Perry recommends this account of belief is that, in its application to first-personal belief, it avoids construal of what is believed as a ‘private thought’, i.e., a proposition accessible only to the believer. But for current purposes its primary virtue is more specific. If belief can be explicated as a matter of a believer accepting a sentence, thereby determining a certain proposition as what it takes for the belief to be true, then third-personal anaphoric characterisation of indexical, in particular, first-personal belief need not be taken as capturing the intentional structure of belief perspicuously. Such characterisation, which is typically appropriate for attributors who do take themselves to be identical with the subject of belief, turns out to be merely a convenient device for what is more clumsily but also more accurately put in counterfactually dispositional terms.

Yet there is something wrong with Perry’s effort to provide a counterfactually dispositional characterisation of belief, in which a sentence is quoted in order to individuate content. The problem is intimated by the condition that an individual who accepts a sentence must be sincere,19 i.e., concerned to articulate the content of the belief state which, given sincerity and the capacity to speak the relevant language, disposes the individual to utter the sentence accepted. This condition is essential since without it the individual could conceivably not be disposed to utter the sentence which articulates what it believes. Indeed, if this condition is not fulfilled, an individual might be disposed to assert something which entails the falsity of what it believes. The concern, however, to articulate or, as the case may be, to conceal the content of belief clearly presupposes knowledge of what this content is. Knowledge of this content of belief cannot consist in or be analysed as knowing what sentence the belief state with this content disposes one to utter since, as the need to incorporate sincerity into the putative analysis shows, knowing the latter presupposes knowing the belief state and its content. The putative analysis thus presupposes that of which it would be an analysis. So little, then, does counterfactually dispositional characterisation of belief have ontological priority over non-dispositional characterisation that things are in fact the other way around: what comes first is non-dispositional characterisation, in which a sentence is used to individuate content, hence belief characterised simply as belief that such and such is the case. Perry has not succeeded in giving priority to a counterfactually dispositional concept of belief over a non-dispositional one.

§ 4: Reflexive Truth-Conditionality—Perry’s Later Conception of Belief

The failure of his attempt to give priority to a counterfactually dispositional concept of belief over a non-dispositional structural one constitutes an initial failure of Perry’s long-standing effort to correct Frege’s account of thought and sense, first undertaken in his paper of 1977 “Frege on Demonstratives.” There Perry had argued that in order to solve the problem which indexicals constitute for Frege and all those in the philosophy of language and mind who follow him, one must break with Frege’s identification of the ‘thought’ (Gedanke) and the ‘sense’ (Sinn) of a sentence with one another. The former, understood as the proposition expressed by a sentence, constitutes the truth-conditional dimension of statement or belief. It is, says Perry (Perry 2006, p.216), what is stated or believed. The latter constitutes, as so-called cognitive or practical value, the causal or functional role of statement or belief—how what is stated or believed is stated or believed. (ibid.) These two dimensions must, thinks Perry, be kept separate in the sense that they cannot be understood as accomplished by the same entity to which staters or believers are related when they state or believe.

At the same time, the failure of his first attempt teaches a lesson which Perry acknowledges20 to have been implicit in a criticism made of him by Stalnaker as early as 1981. In his article “Indexical Belief” Stalnaker argues that the distinction Perry makes in his early articles between the content of belief, which is, as a result of this distinction, now a Russellian rather than Fregean proposition, and the belief state, which consists in accepting a sentence expressive of this proposition, does not adequately provide for the notion of informational content. In other words, it does not really capture cognitive or practical value. In consequence, Perry cannot provide an adequate account of the causal or functional role of belief. As Stalnaker puts it,

(b)elief states are too subjective to represent informational content … . But belief objects or contents, in Perry’s sense, are … too extensional to represent information conveyed in an act of communication. We need an intermediate entity …. . (Stalnaker 1981, p.148)

Stalnaker’s point seems to be precisely that implicit in the critique given above of Perry’s counterfactually dispositional account of belief in general and of indexical belief in particular: the dispositional properties in which causal or functional role consists must be founded in content, hence cannot be identified with a part or aspect of it. If, therefore, we must regard what is believed or, mutatis mutandis, what is stated in the extensional manner Perry recommends, that is, as a Russellian proposition, then between it and causal role we must insert a structural intermediary which accounts for the relevant dispositional properties. There may indeed be a distinction to be made between what is believed or stated and how it is believed or stated. The latter must, however, be characterised structurally and yet account for causal role.

In his later writings, particularly Crimmins and Perry 1989 and the first edition of Perry 2012, published in 2001, Perry makes another attempt to characterise statement and in particular belief in a manner which preserves the traditional conception of belief as a relation of a subject to an object of belief even for the case of so-called indexical, in particular first-personal belief. It would be inaccurate to describe this second attempt as completely different from, or unrelated to, the first. According to Perry himself, a concern to ensure “the distinction between what is believed and how it is believed … ” (Perry 2006, p.216) drives the second attempt as much as it does the first. The second attempt does not replace the distinction between what is believed or stated and how it is believed; rather, it enables one to understand it better (ibid.)—this, it seems, by providing a genuinely structural account of how what is stated or believed is stated or believed.

This second attempt consists in construing how what is stated or believed is stated or believed in terms of additional higher-order truth conditions which Perry calls reflexive. Traditionally, in both the philosophy of language and the philosophy of mind one has focussed upon the subject matter of statement and belief, i.e., upon the reality which makes the statement or belief true. Indeed, so much has this been so, thinks Perry, that the philosophy of language and mind has fallen prey to what he calls the subject-matter fallacy, the supposition, namely, “that the content of a statement or belief is wholly constituted by the conditions its truth puts on the subject of the statement or belief; that is, the conditions it puts on the objects … words designate or … ideas are of.” (Perry 2012, p.58) Evidently, to make this supposition is to assume that statements and beliefs have content in one single sense: there is only one (theoretically interesting) biconditional determining truth conditions for the statement or belief and the right hand side of this biconditional refers to neither the utterance or the belief itself but merely that independent reality which makes the statement or belief true. Thus, the content of the statement or belief that Lingens is lost in the Stanford library is given by one biconditional only and this biconditional make no mention on its right hand side of either statement or belief but merely articulates the objective state of affairs necessary and sufficient for truth:

The statement or belief that Lingens is lost in the Stanford library is true
if and only if
Lingens is lost in the Stanford library

According to Perry, however, statements and beliefs have layers of interlocking content, hence layers of interlocking truth conditions. All hands can acknowledge that the statement or belief that Lingens is lost in the Stanford library has more than one truth condition. The statement, for example, is true not just if and only if Lingens is lost in the Stanford library but also if and only if there is an the entity determined as referent by the name ‘Lingens’ as uttered on this occasion, in this utterance u, this entity is Lingens and it belongs to the extension of the relational property determined by the meaning of the predicate ‘is lost in the Stanford library’ once again, as uttered on this occasion, in this utterance u. The qualification ‘as uttered on this occasion’ is crucial since the name ‘Lingens’ might well be ambiguous; there might, after all, be two individuals named Lingens. Similarly, the predicate ‘is lost in the Stanford library’ might on other occasions of its utterance refer to a different library or even express a different property, e.g., being lost in the sense of having lost oneself in the wealth of texts and information contained in the Stanford library.

Now Perry claims that this latter truth condition, which is reflexive because it involves reference on its right hand side to the utterance u with which the statement is made (or, mutatis mutandis, to the belief b), is not a theoretically uninteresting consequence of content but rather a crucial part or aspect of thereof. In particular, it is that aspect of content which captures the causal or functional role of statement (or, mutatis mutandis, belief). One can see this by considering how this conception of reflexive content and reflexive truth conditions applies to indexical statements. Let us imagine Lingens is an amnesiac who has forgotten that he is Rudolf Lingens. Lingens makes an utterance u of the English sentence “I am lost in the Stanford library,” therein stating that he is lost in the Stanford library. Thereby he produces an utterance u of this sentence, in other words, a token of it, which is true if and only if Lingens is lost in the Stanford library. At the same time, the fact that the utterance u contains a sub-utterance of the first-person pronoun ‘I’ shows that this statement is subject to a further, higher-order truth condition:

Lingens’ utterance u of the sentence “I am lost in the Stanford library” is true
if and only if
there is an x such that x is the referent determined by the meaning of the indexical ‘I’ as contained in u, x is Lingens and x belongs to the extension of the relational property determined by the meaning of the predicate predicated of x in u, namely, ‘am lost in the Stanford library’

More succinctly put:

Lingens’ utterance u of the sentence “I am lost in the Stanford library” is true
if and only if
the speaker of u belongs to the extension of the relational property determined by the predicate ‘am lost in the Stanford library’ contained in u

And most succinctly put:21

Lingens’ utterance u of the sentence “I am lost in the Stanford library” is true
if and only if
the speaker of u is lost in the Stanford library

Clearly, this reflexive truth condition enables a sufficiently competent hearer of it to ascertain what is stated, namely, that Lingens is lost (if the hearer knows that the speaker is Lingens); or that the man speaking to the hearer is lost in Stanford library (if the hearer knows Lingens simply as the man is speaking to the hearer); or that the strange man wandering around muttering to himself is lost in the Stanford library (if the hearer knows Lingens simply as the strange man he overhears muttering in the library); and so on, ad indefinitum, for all the ways in which a hearer might be aware of Lingens as claiming to be lost in the Stanford library. Two important things follow from this: firstly, the reflexive truth condition, because it is what any hearer must grasp in order to identify what is stated, hence to be causally affected by the statement, constitutes the causal or functional role of the statement. Secondly, what is stated is so to speak indifferent to how it is characterised, provided only that different characterisations are co-referential, either with respect to referent, with respect to extension, or with respect to truth-value. In other words, it shows that what is stated is indeed, just Perry wants to maintain, a Russellian rather than Fregean proposition.

Thus far the account has only been formulated for the linguistic case. It must, however, be made to apply also to the psychological case of belief. Here Perry suggests (something like) the following. In the case of the belief that Lingens is lost in the Stanford library, what is believed will be the (Russellian) proposition that Lingens is lost in the Stanford library. This will be the ordered triple <Lingens, x is lost in y, Stanford library>. The belief has also (something like) the following reflexive truth condition:

The belief b that Lingens is lost in the Stanford library is true
if and only if
there is an entity x such that b is about x, x is Lingens and x satisfies the relational property determined by the content of b, namely, being lost in the Stanford library

This, too, can obviously be given a maximally succinct formulation:

The belief b that Lingens is lost in the Stanford library is true
if and only if
Lingens is lost in the Stanford library

In the case of Lingen’s first-personal belief, “I am lost in the Stanford library,” i.e., that belief state which disposes Lingen, a sincere and competent speaker of English, to assert the sentence, “I am lost in the Stanford library,” the ground-floor subject-matter truth condition—what Perry calls the referential truth condition22—will be the same as in the case of anyone’s belief that Lingens is lost in the Stanford library. The reflexive truth condition of the first-personal belief will, however, be different:

The belief b which Lingens would articulate by stating, “I am lost in the Stanford library,” is true if and only if the possessor of b is lost in the Stanford library

Note a crucial assumption of this account: Like statements, “(b)eliefs are concrete cognitive structures: they are particulars that belong to an agent, come into existence, endure and go out of existence.” (Perry 2000, p.209) Why does Perry maintain this? Perry explicitly provides three reasons—see Perry 2000, p.210—which need not be mentioned here. A deeper reason should, however, be noted, one embedded in the very nature of Perry’s undertaking which he himself does not mention. Of statements we may ask what makes it the case that they are utterances subject to different kinds and level of truth condition. Perry cannot answer this question by saying that it is because they have such and such a propositional content, as if propositional content determined truth conditions. The whole point of the account is, after all, to show how possession of different kinds and level of truth condition constitutes the contentfulness of the statement made. Indeed, to maintain that truth conditions were determined by propositional content (which propositional content was determined by the meaning of the sentence uttered) would take us back to Frege and his identification of the dimensions of truth-conditionality (Gedanke) and causal or functional role (Sinn). So according to Perry the meaning of the sentence uttered, more precisely, the meanings of its various semantic components in their combination with one another, must first directly determine the different kinds and level of truth condition and only thereby propositional contentfulness.

Now statements are clearly utterances and as such concrete particulars in space and time. If, however, the same strategy of explication is to work for belief, then one must similarly regard beliefs as concrete particulars. Indeed, one must regard them precisely as particular cognitive structures in a manner analogous to the way statements, in virtue of the linguistic structures used in their making, are such structures. Just as the expressions uttered in the making of a statement combine to determine that in the context the particular statement has such and such truth conditions, hence expresses such and such a proposition, so, too, a particular belief has an analogous composition. For this reason Perry speaks of beliefs as composed of notions and ideas, the former corresponding to singular terms, the latter to predicate expressions. Consequently, just as beliefs and desires are particulars in a subject’s head or at least its mind—see Perry 2006, p.217—, so, too, are notions and ideas. In particular,

(b)eliefs are structured entities that contain ideas and notions as constituents. Ideas and notions, like beliefs, are … concrete cognitive particulars. So there is no such thing as agents have the same idea or notion, but only similar ones. (Perry 2000, p.211)

Unsurprisingly, Perry goes on to allow that “notions and ideas might be … particular words in a language of thought, physical objects like file folders, or things with more of a dispositional character, like the process underlying the disposition of an agent to have a specific “pattern of neural activation” in certain circumstances.” (Perry 2000, pp.215-216)

The distinction between higher-order reflexive and ground-floor referential or subject-matter truth conditions23 is meant to implement the distinction between what is stated or believed and how it is stated or believed. In this way Perry hopes to rend asunder what Frege bound together, namely, the propositional dimension of the thought (Gedanke) and the causal dimension of sense (Sinn). According to Perry, the aspect of contentfulness constituted by reflexive truth conditions corresponds to the dimension of sense and is thus that structural feature of both statement and belief which accounts for (the dispositional properties constitutive of) causal role. Thus Perry says, “We should expect a belief to motivate actions that will serve the goals of its owner, given its reflexive truth conditions, and the reflexive truth conditions of his other beliefs that are involved in motivating the actions.” (Perry 2006, p.217)

Now it is clear how reflexive truth conditions might enable causal role in the case of at least certain kinds of statement. The first and primary effect of a statement, in its character as a statement, consists in its being understood by a hearer. So the causal role of a statement qua statement lies primarily in its capacity to cause understanding. And a grasp of reflexive truth conditions does seem to be essential24 to understanding in the case of most interest to Perry, namely, statements made using indexical sentences. In these cases, in order to identify what is stated, the hearer must first grasp those identifying conditions for the referent which are specified by the linguistic conventions governing indexical expressions. So a hearer can only understand by recognising their utterance as subject to such conventions. Only by recognising that Lingen’s utterance of the sentence “I am lost in the Stanford library” contains a token of the word ‘I’, hence involves reference to whoever is uttering this token, can a hearer ascertain who Lingens is talking about, hence what Lingens is stating. Since this identifying condition is contained in the reflexive truth conditions, one may indeed maintain that a grasp of these is essential to understanding, hence causal role. And if this is so, one may also maintain that at least in the case of statements made with sentences containing indexicals, specifically, the first person pronoun, a grasp of reflexive truth conditions is similarly essential.

Precisely here, however, in the case of statements made with sentences in the first person, we encounter the Achilles’ heel of Perry’s account, not just for such statements, indeed, not just for statements in general, but for belief as well. As true as it is that in the case of first-personal statements, reflexive truth conditions are an essential component of their understandability, hence their distinctive causal role, it is equally true that such truth conditions only accomplish this under presuppositions which Perry is committed to denying. Imagine a language in which there were no dedicated first-personal pronoun but this role were taken by some other term, for example, the expression ‘George’,25 which could also be used as a proper name. Evidently, speakers must now decide which of two competing linguistic conventions is to be employed on any given occasion.

But how must this choice be formulated? It is not hard to see that it must be formulated as a choice between whether the speaker refers (according to the one convention) to an individual named George or whether the speaker refers (according to the other convention) to him- or herself. One can see this by considering what one encounters when one attempts to formulate the choice speakers face between competing conventions in a manner which employs, not third-personal anaphora but Reichenbachian reflexivity. For now one must say something like the following: the speaker must choose between referring in accordance with the one convention to an individual named George; and referring in accordance with the other to the speaker of that utterance of (a sentence containing) ‘George’ which exists if the speaker chooses to refer in accordance with this convention to the speaker of that utterance of (a sentence containing) ‘George’ which exists if the speaker chooses to refer in accordance with this convention … . This is an infinite regress and it is clearly vicious—vicious because ungrounded.

It is clear why the regress arises: in general, a convention can only be characterised in terms which appeal to that for the sake of which it is applied—what the defining point of the convention is. In the case of a first-personal pronoun, this defining point can only be put as reference to oneself (as oneself). When one denies oneself, as Perry surely must, this obvious way of putting the convention-defining point, the inherent reflexivity of the convention—the fact that it enables a speaker to create an utterance u which refers to the speaker of u in virtue of the speaker’s applying this convention here and now in order to achieve its defining point—generates a vicious infinite regress as soon as one asks what the defining point is. And of course speakers must ask this question if they are to wield the convention and the types of expression it governs.

The regress does not, of course, arise because in the case imagined two conventions governed the expression at issue. Imagining a situation in which two conventions of reference, one first-personal, the other nominal, governed the one expression served merely to render visible the constitutive role of the notion of application, hence of there being, as part of the very identity of the relevant convention, that for the sake of which the convention is applied: its defining point. Even in those cases in which there is just one linguistic convention, in other words, in which there is a dedicated first-personal pronoun, speakers must still apply a convention. So even in such cases speaker must still choose a convention constitutive of an expression as a first-personal pronoun, and each must do so in order to refer and each must do so in order to refer to him- or herself. This is the point of the convention and this point is part of what the convention is.

But is it true, as we are tacitly assuming, that Perry must deny himself use of reflexive pronouns in characterising the defining point of the convention governing first-personal pronouns, hence in characterising this convention itself? Indeed it is! Statements certainly possess reflexive truth conditions and, at least in the case of first-personal statements, hearers certainly resort to them, or at least to the reflexive reference they contain, in order to identify the referent and thereby what is stated. But all this is true only because in such statements speakers are referring to themselves, and they are referring to themselves only because and insofar as they are stating, hence desiring, intending and believing that they state, that they are thus and so or that such and such is the case with them. In other words, the reflexive pronoun essential to characterising the defining point of the convention governing a first-personal pronoun entails the possession by speakers of intentions and indeed beliefs which require for their perspicuous characterisation an anaphorically-used third-personal pronoun.26 Since Perry denies the latter, he must also deny the former.

The upshot of all this is therefore the following: third-personal anaphoric characterisation of the statement and the reference it contains is prior to Perry’s of course equally third-personal reflexively truth conditional characterisation. Problems of regress only arise for the case of first-personal statement when one attempts to invert this priority by construing the reflexive referentiality and truth conditionality of statement as how things really are when speakers make statements—as if the anaphoric forms one might use in the characterisation of the underlying desires, intentions and beliefs, as well as in the characterisation of the statement itself if this latter is first-personal, were merely conveniences which, for all their convenience, mask the underlying reality of reflexive referentiality and truth conditionality. All in all, characterisation in essentially anaphoric third-personal terms is prior to the reflexively truth-conditional characterisation of which first-personal statements and the beliefs, desires and intentions driving all forms of statement are undoubtedly capable. Once this is conceded, there is no reason for thinking that essentially anaphoric, third-personal characterisation of all forms of utterance or intentionality, from statements and beliefs through commands and desires to promises, intentions and indeed perceptual experiences, is merely a façon de parler for items only perspicuously characterised in reflexively referential or -truth conditional terms.

Instructively, Perry’s appeal to reflexive truth conditions fails for a higher-order version of the very reason that undid Reichenbach’s original account, which sought to build reflexivity into (what Perry calls) referential truth conditions, i.e., into meaning.27 Just as first-personal reference cannot be identified with that of the phrase “the speaker of this utterance,” so, too, the first-personally referential character and therefore the content of first-personal statements cannot be analysed as a combination of two separate dimensions, viz., the referentially truth conditional dimension of what is stated and the reflexively truth conditional dimension of Reichenbachian reflexive truth conditions. For in both cases the character of the statement as involving first-personal reference and having the kind of content associated with such reference accounts for, hence is presupposed by, possession of truth conditions involving reflexivity, whether at the ground-floor referential level or at the higher-order reflexive level. Perry, however, no less than Reichenbach, wants to have things the other way around and this is not possible.

For current purposes, however, the most relevant consequence of the failure of Perry’s later attempt to provide an account of statement and belief in general and of first-personally indexical statement and belief in particular is the following: the personal belief required for agency is not perspicuously characterised as it is in itself as first-personally indexical. Rather, it is characterisable as it is in itself only in third-personally anaphoric terms, the kind of characterisation typically given by someone who does not take themselves to be the subject of belief. In other words, as least with regard to belief we may assert that there is such a thing as essential anaphora—forms of belief which satisfy the three conditions given at the end of section two. We may also assert that this is the phenomenon Perry was glimpsing but also misdescribing as essential (first-personal) indexicality.

Note a crucial implication of this: if François’s belief that he is about to be shot cannot be explicated as a belief state with a certain combination of referential and reflexive truth conditions, if, in other words, the third-personally anaphoric characterisation is ineluctable (because the former in fact presupposes the latter), then one has de facto saved the marriage Perry sought to dissolve. Frege was following a sound instinct when he grounded the truth-conditional and causal dimensions of statement and belief in the one thing. But does this not also bring back all the problems which Perry had originally and rightly found in Frege’s account? No, it does not because it does not bring back Frege’s account. Quite the opposite: recognition that there are forms of belief which can only be perspicuously characterised in third-personally anaphoric terms entails that there are forms of belief which are not in any way relations to Fregean thoughts. They can, of course, be relations to Russellian propositions and they can, of course, have reflexive truth conditions—but only because they have a content which resists characterisation as an object of belief when this is understood as something which could be individuated by a free-standing sentence, whether indexical or not.

None of this entails, of course, that there is literally no such a thing as first personally indexical belief. Quite the contrary: it prepares us for understanding just what this is, an issue we return to in conclusion. For it reveals the following basic insight from which all attempts to explicate what such belief is must proceed: either such belief is explicit, episodic first-personal thinking, as when the thought, “I am the shopper making a mess,” runs through one’s head; or it is simply the capacity for such explicit, episodic thinking. There is nothing else for it to be. This does not preclude, of course, that genuinely first-personal thinking might be causally involved in agency. The case of Perry in the supermarket provides an excellent illustration of how it can be. Perry writes of how it dawned on him (Perry 2000, p.27) that he was the shopper he was trying to catch, i.e., the shopper making a mess. Clearly, Perry is describing a moment of epiphany, in which he explicitly judges, “I am the shopper making a mess.” Equally clearly, this causes him to acquire the belief that he is the shopping making a mess, which latter causes him in turn to rearrange the items in his shopping trolley.28 In other words, in this particular case a subject of belief acquires the behaviour-generating, third-personal, anaphoric belief through undergoing an episode of explicit first-personal thinking (mental judgement).

Lastly, the claim that the personal belief required for agency is essentially anaphoric rather than (first-personally) indexical in no way makes self-attribution of belief impossible. The worry here is that if one insists that there is personal belief which is not indexically, in particular, first-personally characterisable, then one cannot make sense of what it is for a subject ascribe such belief to itself—as when Perry thinks or says to himself, “I believe that I am making a mess.” For if the claim were right, then, or so one might argue, when Perry ascribes this belief to himself he would surely have to say, “John Perry believes that he is making a mess”—therein, of course, meaning it precisely not in the manner of de Gaulle. This worry rests, however, on a misunderstanding of the claim just made. To characterise a belief perspicuously is to characterise it as it ‘truly’ is, that is, as it is solely in itself. This is not inconsistent with the claim that when a personal belief is had by an agent with the additional capacity for self-attribution of belief, this agent can and indeed must attribute the belief to itself in indexical, indeed first-personal terms, by thinking to itself or indeed asserting, “I believe that Φ(I/me).” In other words, when the agent is sophisticated enough to be capable of self-attribution of belief, third-personally anaphoric belief can and must be rendered explicit in such episodic, first-personally indexical thinking and stating. And as we have already suggested, sometimes such episodic, first-personally indexical thinking and stating can engender third-personally anaphoric belief.

In fact, so little does the claim that the personal belief required for agency is in itself only third-personally and anaphorically characterisable render self-attribution impossible that it actually permits a clarification of the capacity for explicit self-ascription of belief and, mutatis mutandis, of other properties and relations. Given that a personal belief is, considered purely and simply in itself, third-personal and anaphoric in structure, a subject which ascribes such a belief to itself by thinking or asserting, “I believe that Φ(I/me),” must also be prepared to ascribe this belief to itself by thinking or asserting, “I, S, believe that Φ(I/me),” where ‘S’ is, of course, some non-first-personally-indexical, at least partially descriptive singular term. In other words, first-personal self-ascription of belief and indeed of all properties and relations presupposes that the subject believes that it is S, or again, that ‘S’ refers to it itself, and that it possesses the ability to think to itself, “I am S.” The personal belief that one is S is unique in that it goes hand in hand with the ability to think to oneself, “I am S”; it entails the capacity of its subject to make it explicit in an act of first-personal self-ascription. This personal belief is thus a constitutive, defining feature of self-consciousness. To put the matter in Kantian terms, awareness of self as subject goes hand in hand with awareness of self as object, that is, as an empirical self.

Of course, implicit in this is the idea that certain forms of personal belief do not entail the capacity for self-ascription of them. Now it has already in effect been shown, at least with regard to belief, that there is essential anaphora, that is, that there are beliefs which satisfy the three conditions given at the end of § 2. If, however, it is not generally true that one can have such forms of belief only if one can ascribe them to oneself (as oneself), then there can be beings possessed of such beliefs, or at least of other forms of essentially anaphoric cognitive intentionality, which lack all capacity for such self-ascription or for explicit, first-personal thinking generally. In other words, it must be possible for non-self-conscious beings to possess forms of essentially anaphoric cognitive intentionality. This would be an important result since it would constitute a break with a defining assumption of much philosophy since Descartes, namely, that, to put things crudely, consciousness entails self-consciousness.

§ 5: Essential Anaphora without the First Person

So can there be cases of agency in which third-personally anaphoric forms of intentionality are involved in which the agent is in no way capable of first-personally indexical forms of intentionality? To put the point in Lewis’s terminology,29 can there be attitudes de se possessed by a subject incapable of attitudes de me, that is, incapable either of ascribing these forms to itself or of thinking their contents in episodes of first-personal thinking?

Consider the following example:

  • S sees: A large, aggressive dog is charging towards it.

  • S desires: It not be attacked.

  • S believes: If it runs into the house, it will not be attacked.30

  • Action: S runs into the house.

We are now assuming the subject of belief S to be another dog or some other kind of animal incapable of attitudes de me. That there are such animals, on the one hand capable of a diverse range of intentional states and experiences, on the other unable to think to themselves, “I am thus and so,” is surely a plausible assumption. Yet the anaphoric construction is just as essential here as it is in the examples considered above and it is so for precisely the same reasons. How so?

Any truly complete and adequate explanation of S’s behaviour, namely, its running into the house, must involve a characterisation of S’s perceptual experience which accounts for why S engages in this behaviour and not another. If, however, instead of using the pronoun ‘it’ anaphorically, we use a non-indexical singular term to refer to S in our characterisation of the content of its perceptual experience, then we have characterised the intentional content of S’s experience as if the experience were an impersonal one, hence an experience with a content anyone could have. Consequently, we have not fully characterised or captured the causal role and efficacy of S’s perceptual experience, to the point where we can explain why S runs into the house rather than doing something else.

But nor may we use an indexical singular term. The first person pronoun is in any case already ruled out because ex hypothesi S is not capable of first-personal thinking. More importantly, the first person could only be used without referring to the wrong individual if the content of S’s perceptual experience were characterisable oratio recta, in some perceptual analogue of direct speech, for example, “S sees, “A large, aggressive dog is charging towards me”.” But this characterisation is surely only intelligible as a perverse way of putting what is better put by using some indirect form, e.g., “S sees a large, aggressive dog charging towards it” or “S sees that a large, aggressive dog is charging towards it.”

Finally, we cannot characterise the content of S’s perceptual experience by using some kind of demonstratively indexical phrase. In other words, we cannot characterise it as a matter of S’s seeing a large, aggressive dog charging towards this dog, the dog here, this dog here, and the like. For one thing, this kind of characterisation only enables full explanation of S’s behaviour in conjunction with the further claim that this dog (or the dog here, this dog here, etc.) is S. And explainer can, however, surely only know this further claim by understanding S’s behaviour, which involves identifying the perceptual experience involved in generating it. In short, the claim “S sees a large, aggressive dog charging towards this dog and this dog is S” presupposes knowledge of S as seeing a large, aggressive dog is charging towards it. For another and more important thing, such non-first-personal indexical characterisations, no less than any non-indexical characterisation, fail to distinguish S’s experience from a perceptual experience anyone or anything could have. Precisely for this reason one needs to add to the explanation the identity claim that this dog (or the dog here, this dog here, etc.) is S. So all indexical characterisations not in the first person must also fail to capture fully the causal role and efficacy of S’s experience. The problem with indexical singular terms which are not first-personal is in fact no different from the problem with non-indexical singular terms. The only indexical singular term which avoids this problem is the first-personal pronoun. But this attributes to S a capacity for first-personal awareness of which S is not capable.

So neither non-indexical nor indexical singular terms will do. This underscores a point made in the argument given above against Cappelen and Dever: the reason why impersonal action explanations and rationalisations are incomplete is that they fail to mention any non-impersonal, hence in this sense personal beliefs. The defining feature of such beliefs is precisely that they cannot be had by anyone, not that they are indexical, however the notion of indexical belief might ulimately be conceived. The character of belief as personal (in the sense defined) is the feature needed for complete and adequate explanation or rationalisation since belief with this feature presents an agent’s awareness of its particular situation, the kind of awareness an agent needs in order to bring its physical capacities and whatever impersonal beliefs it has to bear in objectively intelligible fashion. As we have seen, in order to conceive of beliefs with this feature, we do not have to resort to indexicality or the first person.

All this makes clear that the problem of accurately capturing the causal role and efficacy of the psychological states and experiences driving behaviour actually has nothing to do with the limitations of non-indexical singular terms as opposed to indexical ones, or of impersonal singular terms as opposed to first-personal ones. The problem has to do with singular terms as such. Take any conception one likes of how a singular term refers. Perhaps it refers to something x by standing in the right causal relations to x. Perhaps it refers to x in virtue of x’s satisfying some sufficiently individuating identifying condition or definite description.31 Finally, perhaps it refers to x in virtue of some general rule of the kind characteristic of indexicals, e.g., “‘I’ refers to x iff x is its speaker,” “‘here’ refers to x iff x is a region close to its speaker in referring to which the speaker has an interest,” etc. Even so, a singular term possesses a linguistic meaning which permits it refer to something off its own bat, without borrowing its referential properties from any other singular term occurring in the context of utterance. In the sense defined by the enumeration just given of how different singular terms refer (and in this sense only) we may say that a singular term refers to something by representing it. We may then formulate the problem to which the example of S the dog is pointing as follows: any characterisation of the content of S’s experience which refers to S by representing S will either underdescribe S’s experience by failing to capture its distinctive causal role; or misdescribe it as something it neither is nor can be, namely, first-personal. To put things in a slogan: We must allow that there can be reference without representation, i.e., a kind of reference essentially expressible only in anaphoric terms. For only then will we be able perspicuously to characterise and attribute the kind of belief essential to agency.

Of course, there is a price to pay for this. The thesis of essential anaphora contradicts the traditional view that an intentional state or experience is (always) an attitude towards, hence a genuine relation of the subject to, an intentional object which constitutes the content of the state or experience.32 This view rests on the assumption that there is no reference without representation and that therefore the third-personally anaphoric construction is merely a device for attributing to another what are in reality first-personally indexical attitudes. For then the content of all forms of belief and intentionality generally can be articulated, hence individuated, by means of a free-standing sentence. This permits one in turn to construe a state or experience with this content as a relation to what the free-standing sentence articulates. Of course, sometimes this sentence must be indexical, indeed first personal and what it articulates or expresses must be what Perry calls a relativised proposition.33 Even so, it is a free-standing sentence and so one can retain the idea that all intentional states and experiences are attitudes towards, hence relations to, something. But the price of this is commitment to the IIC and Cappelen and Dever are right in their claim that essential indexicality in the sense defined by the IIC is at best false, at worst incoherent.

Conclusion: Implications and Speculations

In conclusion, let us draw out the possible significance of giving up the idea that all belief, hence all intentional states and experiences are relations of a subject to some object of belief or, more generally put, to some entity which constitutes intentional content. In particular, let us speculate on some of the advantages potentially to be had by giving up this idea and accepting that there are intentional states and experience whose structure is perspicuously captured by anaphoric constructions of the form “S believes that Φ(it)” or “S perceives Φ(it), etc.—most generally, of the form “S-cognitive verb-Φ(pronoun).” (In what follows, I will restrict myself to the case of nominative belief but the points made apply just as much to non-nominative belief and indeed to non-doxastic forms of intentionality.)

One important advantage lies in the possibility of giving an account of what it is to have, in the clearly graphic and explicit way in which Perry realises that he is the shopper making a mess, an episode of first-personal thinking. It is indeed remarkable that although the first person has been much discussed, no analysis or explication has been given of what it is explicitly to think, “I am thus and so.” Perhaps this is because, as will now be suggested, any such analysis or explication must take the anaphoric construction seriously, as something which is not simply a device for attributing belief which if not (as it sometimes can be) de dicto is only ever indexical and first-personal. One thing is clear: whatever one says about the anaphoric construction, one must acknowledge it to have an intimate conceptual connection with explicit first-personal thinking. So if one wants to maintain that there is such a thing as belief whose structure is perspicuously captured by the anaphoric construction, it is incumbent upon one to reveal this intimate connection. After all, in some sense, S’s believing or thinking (in the sense of mentally judging), “I am Φ,” implies or implicates S’s believing that it is Φ. We need to explain this, in other words, to set belief or perhaps rather thinking de me in proper conceptual relation to belief de se.

So what is happening to Perry when, in the supermarket, he suddenly realises, “I am the shopper who is making a mess”, thereupon acquiring the belief that he is making a mess and in consequence shifting to re-arranging items in his own shopping trolley? There are at least two conditions of adequacy on any putative analysis or explication of this: firstly, it must accurately describe the causal role of Perry’s explicit, first-personal thinking since this episode of thinking is intimately involved in his change of behaviour. Secondly and relatedly, it must plausibly account for how Perry acquires the standing belief that he is the shopper making a mess through this realisation. In particular, we must give an account of explicit, first-personal thinking which does not entail that the subject of such thinking is already in such standing belief-states, as if its explicit first-personal thinking were merely, as it sometimes can be, the making-explicit of belief states it is already in.

As the core component of an account of explicit, first-personal thinking in general I suggest the following hypothesis: the episode of explicit, first-personal thinking which Perry undergoes in the supermarket is a matter of his imagining that he is asserting that he is the shopper who is making a mess (or that he is making a mess, etc.). In general, to undergo an episode of explicit, first-personal thinking (in the sense of mentally judging), “I am Φ,” is to imagine that one is asserting that one is Φ—mutatis mutandis, of course, for non-nominative cases. Obviously, by imagining is not meant that strictly discursive sense in which one constructs a hypothetical or fictional story, as when Hilary Putnam calls upon his readers to imagine that in some very distant part of the universe there were a planet Twin Earth exactly like Earth except that what English speakers there call water is in fact XYZ. I mean the kind of imagining in which one imagines what it would be or even feel like for such and such to be the case.34 That these are two different senses is shown by the fact that I can imagine in the first sense that I am dead only to discover that I cannot imagine my being dead in the second sense.

Clearly, an account of explicit, first-personal thinking based on this core component presupposes that certain forms of belief and, mutatis mutandis, of intentionality generally are essentially anaphoric, that is, such that they are perspicuously captured in their intentional structure in third-personally anaphoric terms. If this were not so, then the account would obviously be viciously circular. Furthermore, the account directly entails a dependence of explicit, first-personal thinking on linguistic competence. This is plausible since there is surely a very strong intuition that non-linguistic beings such as dogs and cats cannot literally think to themselves, “I am Φ.” As such, the account can explain the phenomenological observation that thinking to oneself takes place in some language or another, that indeed it is, as Plato and others have noted, the dialogue of the soul with itself.

Going hand in hand with this explanatory capacity is another: the account creates a conceptual link between the discursive and the perceptual since it appeals to imagination not in the sense of conceiving such and such to be the case but in a sense which implicates what it would be or even feel like for such and such to be the case. The notion of imagination to which the account appeals thus implicates a subject’s imagining itself perceiving. In consequence, the account reveals itself to be decidedly anti-Cartesian since it is essential to Descartes’ conception of the self-conscious self that perception and therefore also imagination be capacities the self merely contingently possesses. Moreover, this anti-Cartesianism is also an anti-computationalism since the Cartesian and the computationalist, for all their differences, agree that a purely discursive intelligence is possible.

Now it would be wrong to object to the idea that an episode of explicit, first-personal thinking (in the sense of mentally judging), “I am Φ,” involves imagining oneself asserting that one is Φ on the grounds that to imagine is to fantasise, hence cannot imply or implicate actually believing that one is Φ. For it is just false that to imagine is always only to fantasise. In other words, there can be ‘serious’ imagining. Yet although imagining can sometimes be ‘serious’, it is obviously not necessarily so. We must therefore explain what it is for such imagining to be ‘serious’ and we must do so in a way which, as indicated above, explains how there can be an intimate connection between explicit first-personal thinking and a subject’s believing that it is thus and so or that things are thus and so with it. Only if this account can explain this will it be able to set belief or perhaps rather thinking de me in clear conceptual relation to belief de se—this by showing how, in what sense, the former is founded in the latter.

How, then, can imagining oneself asserting be construed as ‘serious’, such that a sufficiently strong inferential relation is preserved from a subject’s thinking, “I am Φ,” to its believing that it is Φ? One way in which the imagining at issue might be ‘serious’ would be when the episode of imagining oneself asserting is caused by the relevant belief. It is perfectly possible that a subject might imagine itself asserting such and such in virtue of believing such and such. So a case of explicitly thinking to oneself, “I am Φ,” in a sense which entails that the subject believes that it is Φ, might be one in which the subject imagines itself asserting that it is Φ in virtue of believing that it is Φ. In this kind of case, the act of explicit, first-personal thinking makes a prior standing belief explicit.

This will not deal, however, with the episode of explicit first personal thinking Perry undergoes in the supermarket. Perry realises that he is the shopper who is making a mess, that he is making a mess, etc. In other words, he acquires the belief as a result of the episode of explicit, first personal thinking rather than the other way around. How is this to be understood? Perry writes, “I once followed a trail of sugar on a supermarket floor, pushing my cart down the aisle on one side of a tall counter and back the aisle on the other, seeking the shopper with the torn sack to tell him he was making a mess.” (Perry 2000, p.27) Now one can only seek another shopper with a torn sack in order to tell him that he is making a mess. So Perry’s behaviour prior to his realisation is based on his positively believing that he is not the shopper making a mess and not merely on his not believing that he is the shopper making a mess. It is clear why Perry positively possesses the negative belief rather than simply not possessing the positive one: when he first sees the trail of sugar extending out before him, he knows that there is a shopper making a mess. He also knows that the shopper has gone before him along the aisle precisely because the trail of sugar extends out before him. Since, however, he has either forgotten or at least not taken into consideration that he has himself gone before along the aisle, he assumes that someone else must be making the mess.

The behaviour Perry bases upon this assumption is, however, unsuccessful. Moreover, it eventually shows itself as such, that is, as unsuccessful. Thus, his realising that he is the shopper making a mess does not merely have a prospective character, in that it causes him to acquire new beliefs which then lead to new behaviour, specifically, to his adjusting items in his own shopping trolley. It also has a retrospective character in that it allows him to make sense of his current situation and immediate past. Note that, strictly speaking, what causes him to adjust items in his shopping trolley is not the belief he directly forms as a result of his explicitly realising, “I am the shopper making a mess.” What motivates, hence explains and rationalises this behaviour is the belief that he is making a mess, a belief Perry could acquire and possess without there being any prior search for that other shopper who is making a mess, hence without that identity belief arising that he is the shopper making a mess which corrects his prior behaviour. Perry’s identity belief and his belief that he is making a mess are therefore distinct existences such that the former gives rise to the latter. At the same time, the identity belief does not just brutely cause the belief driving Perry’s new behaviour; rather, it does so rationally. That is, it gives rise to the causally relevant belief by making sense of his current and immediately past behaviour: Perry comes to see that this behaviour is doomed to failure because its underlying assumption, namely, that he is not the shopper making a mess, is false. Perry’s episode of realisation thus primarily causes rational insight—on the one hand, into the failure of his present and immediately past behaviour, on the other, into what rationality requires of immediately future behaviour, given a goal governing his immediate past, his present and his immediate future.

This suggests the following pattern of explication for the kind of case in which one thinks explicitly to oneself, “I am Φ,” in the sense of realising that one is Φ: to undergo such an episode of first-personal thinking is to be caused by one’s objective situation to imagine oneself asserting, “I am Φ,” therein grasping the content of this assertion as yielding rational insight into one’s situation and on this basis forming the appropriate standing beliefs which guide subsequent behaviour. This aptly characterises the episode of realisation Perry undergoes. Note, however, that it also fits even the situation of Lingens the amnesiac who, as a result of pursuing various clues as to his identity, suddenly realises that he is Lingens. Thus, he might learn of the truth of various impersonal propositions about a man called Lingens and conclude on this basis that this individual is likely to be where he is, having such beliefs and perceptual experiences as he has now, etc. For this reason, a decisive step towards recalling who he is consists in learning that Lingens finds himself in the very same situation now as he does, i.e., has suffered amnesia and is now lost in the Stanford library. Note that this gives us a further reason for thinking that animals such as dogs and cats do not undergo episodes of explicit, first-personal thinking. For they surely cannot realise anything in this explicit, first-personal sense, a sense which implicates insight into the nature of their current situation and what this says about the rationality of their behaviour in it. Episodes of explicit, first-personal realisation are events in and through which one self-consciously and rationally orients and in particular re-orients oneself in the world. Dogs and cats are surely not capable of this.

At this point, however, a serious problem with the whole idea of distinctively anaphoric belief (and of anaphoric intentionality generally) might seem to arise. The problem lies in the way the account just given of what it is to realise that one is Φ appeals to the notion of content. For it seems to suggest that the subject of any such episode of explicit, first-personal realisation relates itself to the content of its realisation, as if this content were an object of belief. This is phenomenologically confirmed by the fact that when Perry undergoes his episode of realisation, he could certainly think or say, as a result of his realisation, “This explains why I cannot catch up with the shopper making a mess.” In so doing, he would seem to be so to speak ostending something with the demonstrative pronoun ‘this’—ostending it precisely as that which makes sense of his situation, in particular, his inability to catch up with the shopper making a mess. Does this not imply that the content of his realisation a genuine object of belief, hence a proposition? The crucial point here is not just that in our account of the phenomenon of explicit, first-personal realisation we have re-admitted objects of belief. Rather, it seems that the idea of essentially anaphoric belief makes it impossible to account for that aspect of explicit, first-personal thinking which consists in awareness of a content of belief as one’s own. As such, it makes it impossible to account for self-conscious awareness of one’s own inner mental life—clearly, an unacceptable result.

The answer to this objection is, however, easy: the core component of the notion of an object of statement or belief is the idea that the content of statement or belief can be individuated by a free-standing sentence. In this sense, the content of the assertion one imagines oneself making in thinking to oneself, “I am Φ,” is clearly an object of belief. It is precisely what is captured by the sentence “I am Φ” (if the subject uttering this sentence is the subject of the episode of first-personal thinking) or by some non-first-personal sentence, e.g., “S is Φ” (if the subject uttering this sentence is not, or at least does not take itself to be, the subject of the episode of first-personal thinking). Of course, the reason why the answer is easy is that this account of explicit, first-personal thinking, in particular, of that kind in which a subject realises something, ties the concept of such thinking to that of linguistic assertion. At the same time, this account is consistent with there being, indeed presupposes that there are, forms of belief and, mutatis mutandis, other forms of intentionality which are third-personally anaphoric in (the linguistic characterisation of) their structure. The contents of these forms of belief (and of intentionality generally) need not be objects of belief in the sense indicated even as these forms of belief (and of intentionality generally) can only be made explicit in episodes of first-personal thinking in which the content of the assertion imagined is an object of belief.

It would seem, then, that this account of explicit, first personal thinking can accommodate the case of Perry in the supermarket, hence meet the conditions of adequacy outlined above. Specifically, it gives a relatively clear and flexible account of thinking de me which presupposes belief de se and indeed, mutatis mutandis, other attitudes de se because it founds the former in the latter. This success suggests, however, that it might be possible to generalise this founding relation. On the account of explicit, first personal thinking just given, S’s believing that it is Φ is possible even when S is non- or pre-linguistic whereas S’s believing or thinking, “I am Φ,” is not. Perhaps, then, we can see in belief and, more generally, intentionality de se a model for thinking about the intentionality of non-linguistic beings. There would be numerous issues to resolve here, namely, whether non-first-personal belief whose content could be individuated by a free-standing sentence—belief that a is Φ—were not equally subject to the requirement of linguistic competence. If this were so, then one would have to see the first-person belief, “I am Φ,” and the standardly propositional belief that a is Φ as complementary forms of sophisticated intentionality which require linguistic competence and founding in diverse forms of intentionality de se which do not require such competence.

Such issues cannot be resolved here. Suffice, then, to conclude by pointing out what this conception of the primordial significance of belief and, more generally, intentionality de se would mean for a very long-standing issue within the philosophy of language. On the one hand, there are those who regard psychological intentionality as not just genetically but also conceptually prior to linguistic intentionality. Examples would be Grice, Bennett, Schiffer, Lewis and the later Searle. On the other hand, there are those who, like Davidson, Dennett and Brandom, see the priority as running the other way around, so much so, indeed, that non-linguistic beings end up having intentional states and experiences in some less than fully literal sense. Evidently, the claim that there are forms of psychological intentionality, namely, intentionality de se, whose intentional contents cannot be expressed by free-standing sentences is clearly consistent with the former position and surely inconsistent with the latter. Yet the account of explicit, first-personal thinking made possible by this claim ties such thinking to language in a manner congenial to the latter position. In particular, it explains why one is instinctively wary of the enthusiasm with which the former camp, in their accounts of the rise of linguistic convention, ascribe to non-linguistic beings forms of intentionality so complex that they suggest a capacity for explicit first-personal thinking and reasoning. The claim that there are forms of intentionality ineluctably and intrinsically de se therefore appears to provide a basis for mediating between and moderating these two extreme positions.

But are there any disadvantages to embracing the claim? In other words, are the any reasons, apart from mere considerations of simplicity, why one should retain the assumption that forms of belief and, mutatis mutandis, of intentionality generally are relations of a subject to an intentional object which constitutes the content of the intentional state or experience in question? One such reason might be the following: linguistic intentional content is a function of the linguistic meanings of component linguistic expressions and their order—sometimes, of course, in the case of indexicals, only together with a contribution from context. In other words, linguistic intentional content is compositionally structured. This compositionality permits one to develop formally semantic accounts of language and one would presumably like to think that psychological intentional content offers nothing to contradict this. For then whatever a formal semantic account of language tells us about the nature of linguistic intentional content it will presumably also tell us about psychological intentional content. If, however, there are forms of belief and of intentionality generally which are ineliminably de se, then one might have to renounce this hope.

References

Altham, J. (1980) “Indirect Reflexives and Indirect Speech,” in Intention and Intentionality: Essays for G. E. M. Anscombe, edited by Cora Diamond and Jenny Teichmann, Cornell University Press, pp.25-37

Bennet, J. (1976) Linguistic Behaviour, Cambridge: Cambridge University Press

Brandom, R. (1998) Making It Explicit, Cambridge, Mass.: Harvard University Press

Cappelen, H., and Dever, J. (2013) The Inessential Indexical, Oxford: Oxford University Press

Cappelen, H., and Lepore, E. (2007) “The Myth of Unarticulated Constituents,” in O’Rourke and Washington 2007, pp.199-214

Castañeda, H-N. (1966) “‘He’—A Study in the Logic of Self-Consciousness,” in Castañeda 1999, pp.35-60

Castañeda, H-N. (1968) “The Phenomeno-Logic of the I,” in Castañeda 1999, pp.89-95

Castañeda, H-N. (1999) The Phenomeno-Logic of the I: Essays on Self-Consciousness, edited by James G. Hart and Tomis Kapitan, Indiana University Press, first edition

Christensen, C. (1997) “Meaning Things and Meaning Others,” in Philosophy and Phenomenological Research, Vol. 57, No. 3, pp.495-522

Christensen, C. (2007) “Nichts Neues unter der Sonne: Bewußtsein und Selbstbewußtsein bei Paul Natorp,” in Kant-Studien, Vol. 98, No. 3, pp.372-398

Christensen, C. (2008) Self and World: From Analytic Philosophy to Phenomenology, Walter de Gruyter, Berlin

Christensen, C. (2013) “The Horizonal Structure of Perceptual Experience,” in Logical Analysis and History of Philosophy/Philosophiegeschichte und Logische Analyse, Vol. 16, pp.109-141

Christensen, C. (2015) “Consciousness, Self-Consciousness and Essential Indexicality,” available at http://cbchristensen.net/

Christensen, C. (2015) “Attitudes de se and de me,” available at http://cbchristensen.net/

Davidson, D. (2001) Inquiries into Truth and Interpretation, 2nd edition, Oxford: Oxford University Press

Dennett, D. (1971) “Intentional Systems,” in The Journal of Philosophy, Vol.68, No. 4, pp.87-106

Dennett, D. (1996), The Intentional Stance, 6th. printing, Cambridge: Cambridge University Press

Evans, G. (1984) The Varieties of Reference, edited by J. McDowell, Oxford University Press, Oxford

Geach, P. (1957) “On Beliefs about Oneself,” in Analysis, Vol. 18, No. 1, pp.23-24

Grice, H.-P., (1957) “Meaning,” in The Philosophical Review, Vol. 66, pp.377–88

Heck, Jr., R. (2002) “Do Demonstratives have Senses?,” in Philosophers’ Imprint, Vol.2, No. 2; http://quod.lib.umich.edu/p/phimp/3521354.0002.002/1, accessed on January 31st. , 2016

Kapitan, T. (2014) “Review of Cappelen and Dever, The Inessential Indexical: On the Philosophical Insignificance of Perspective and the First Person,” in Notre Dame Philosophical Reviews, http://ndpr.nd.edu/news/50266-the-inessential-indexical-on-the-philosophical-insignifi/; accessed on April 25th. , 2015

Lewis, D. (1979) “Attitudes De Dicto and De Se,” in The Philosophical Review, Vol. 88, No. 4, pp.513-543

McGinn, C. (1983) The Subjective View: Secondary Qualities and Indexical Thought, Oxford: Oxford University Press

Nagel, T. (1974) “What is it Like to be a Bat,” in The Philosophical Review, Vol. 83, No. 4, pp.435-450

O’Rourke, M., and Washington, C. (2007) Situating Semantics: Essay on the Philosophy of John Perry, The MIT Press, Cambridge, Massachusetts

Peacocke, C. (1992) A Theory of Concepts, The MIT Press, Cambridge, Massachusetts

Perry, J. (1977) “Frege on Demonstratives,” in Perry 2000, pp.1-26

Perry, J. (1979) “The Essential Indexical,” in Perry 2000, pp.27-44

Perry, J. (1983) “Castañeda on He and I,” in Perry 2000, pp.77-100

Perry, J. (1986a) “Thought without Representation,” in Perry 2000, pp.171-188

Perry, J. (1998) “Myself and I”, in Perry 2000, pp.325-339

Perry, J. (1999) “Self-Notions,” in Logos: Philosophical Issues in Christian Perspective, Vol. 11, pp.17-31

Perry, J. (2000) The Problem of the Essential Indexical and Other Essays, expanded edition, CSLI Publications, Stanford, California

Perry, J. (2006) “Stalnaker and Indexical Belief,” in Content and Modality: Themes from the Philosophy of Robert Stalnaker, edited by J.J. Thompson and Alex Byrne, Oxford University Press, Oxford, pp.204-221

Perry, J. (2012) Reference and Reflexivity, 2nd. Edition, CSLI Publications, Stanford, California, first published in 2001

Price, H. (2007) “The Effective Indexical,” unpublished ms., Centre for Time, University of Sydney, http://philsci-archive.pitt.edu/4487/

Reichenbach, H. (1947) Elements of Symbolic Logic, The Free Press, New York,

Schiffer, S. (1972) Meaning, Oxford: Oxford University Press

Searle, J. (1983) Intentionality—An Essay in the Philosophy of Mind, Cambridge University Press, Cambridge

Spencer, C. (2007) “Is there a Problem of the Essential Indexical?,” in O’Rourke and Washington 2007, pp.179-197

Stalnaker, R. (1981) “Indexical Belief,” in Synthese, Vol. 49, No. 1, pp.129-151

Tiffany, E. (2000) “What Is Essential about Indexicals?,” in Philosophical Studies, Vol. 100, No.1, pp. 35-50

Notes

  1. There have, of course, been earlier critics, e.g., Ruth Millikan and Evans Tiffany.

  2. In fact, ‘here’ and ‘now’, like ‘there’ are demonstrative adverbs, not pronouns.

  3. Even the notion of attitudes ineliminably de se are understood to concern indexicality of the narrow kind Perry associates with ‘I’, ‘here’ and ‘now’ since it is universally assumed that the de se construction is a third-personal means of reporting what is standardly called indexical, because specifically first-personal belief.

  4. Perry speaks at one point in these kinds of term: “When Lingens believes that he is in the library, it may perhaps be useful for certain purposes to say that he believes the singular proposition consisting of the individual Lingens and the property of being in the library. But that doesn’t get at the real underlying structure of his belief.” (Perry 2006, p.209) Admittedly, Perry is writing here in criticism of a position he does not hold.

  5. Elsewhere they write, “In talking to philosophers about these issues over many years, we find that attempts to articulate the essence of Perry and Lewis’s insight tend to come back to something or other about a distinctive connection to agency.” (p.18)

  6. Note that this does not entail that the agent selects—as if it had beliefs about what it can do, either physically or rationally.

  7. The ultimate reason for this is clear enough: action is essentially subject to standards of rationality. Psychological states do not merely cause action, they do so more or less intelligibly, such that it is possible to evaluate action in the light of the beliefs and desires which have caused it as more or less appropriate to the objective situation. Now the situations of different agents with the very same impersonal beliefs and desires may differ in ways which determine that different courses of action are rational for them. So these impersonal beliefs can only cause action rationally if they are mediated in their causal role by non-impersonal, hence beliefs which constitute awareness of the particular situation.

  8. This points, incidentally, to the fundamental mistake of assuming that to be third-personal is to be impersonal.

  9. Anaphora is not being used here in that fashion which permits one, for example, to say that François believes that he is about to be shot but does not know it (because François in fact believes that the man he sees in the mirror is about to be shot but does not appreciate that he is the man he sees in the mirror).

  10. This covers the case in which the first-person pronoun does not occur nominatively. Incidentally, the need to cover this case drives home the obvious point that to think to oneself, “Φ(I/me),” is to think to oneself a sentence of the form “Φ(I/me)” as true under its literal interpretation, i.e., as expressing something true just in case the speaker of the particular token of the sentence is Φ.

  11. Indeed, as is intimated below, I would argue that one could and indeed must extend this claim to cover non-cognitive forms of intentionality such as desire or intention.

  12. Obviously Perry does not mean that we only use sentences with indexicals, which express relativised propositions, in order to pick out belief states. We also use non-indexical sentences to pick out non-indexical contents, such as the proposition that E = mc2.

  13. The proposition expressed by the sentence need not be exactly the same proposition as the one believed; precisely in the case of indexical beliefs it will only be of the same type.

  14. One might regard it as construing belief to be a dual relation, both to a sentence and to a proposition, provided one not understand this as the claim that a sentence is actually involved or contained in the belief which consists in its acceptance. The sentence is not an intermediary entity actually involved in the belief; rather it is a means used by an attributor of belief for “getting at an inner state involved in belief, in virtue of an effect that state would have, in certain circumstances.” (Perry 2006, p.216)

  15. As Perry makes clear in the first footnote to his paper, what he understands by acceptance does not have much to do with the technical meanings given the term ‘acceptance’ by other authors—see Perry 2000, p.45, note 15. He thus does not mean what, e.g., L. Jonathan Cohen means by the term—see Cohen 1992.

  16. In the earlier essay “Frege on Demonstratives” Perry characterises the propositional contentfulness of an asserted sentence S as the assertion’s determining an ordered n-tuple containing the referents of the singular terms in S and the incomplete sense expressed by the sentential function in S—in effect S stripped of its singular terms. Now Frege, unlike, say, Husserl, does not identify the sense of a sentential function with the property or relation which applies to the referent or referents just in case the assertion of S expresses a true proposition; for Frege, this is rather the referent of the sentential function. Since in his earlier essay Perry is clearly using the notion of sense in Frege’s manner, it follows that he is also using a slightly different notion of (Russellian) proposition. But presumably the later notion is more accurate since, after all, the second occurrence of the sentential function ‘Φ(…)’ is just as extensional as the second occurrence of the singular term ‘a’.

  17. See Stalnaker 1981, p.147. Stalnaker’s introduction of an intermediary between propositional content in Perry’s sense, i.e., what is believed, and the belief state is consistent with, indeed completes Perry’s conception of belief and its contents. Just as Perry sees propositional content in his sense as accruing to the belief state, so, too, one can see the intermediary as accruing to it in the same way: through accepting a sentence. Of course, in introducing this further kind of proposition Stalnaker thinks he is making a critical point against Perry. But this is arguably incorrect, a result of wrongly taking what Perry calls a belief state to be one of the two objects of belief into which Perry differentiates the one traditional Fregean proposition (Gedanke). In fact, the belief state is not itself an object of belief but rather has an object of belief. More precisely, it has or incorporates two interconnected ones: the proposition in Perry’s sense and the sentence through accepting which the proposition is, as Frege would put it, ‘grasped’. More precisely still, of course, the belief state is a relation of a believer to a proposition via (acceptance of) a sentence. There is no competition between the belief state and the intermediary for the status of object of belief. As a result of this mistake, Stalnaker appears to make another: he claims that for Perry the belief state is “a mental analogue of[what Kaplan calls] a character.” (Stalnaker 1981, p.147) This is not so. The belief state has a mental analogue of a character, this in virtue of its having as a component the acceptance of something which does have a character in the sense intended by Kaplan, namely, a sentence. Importantly, if this is how we understand Stalnaker’s intermediary kind of proposition as accruing to belief, then his account is implicitly dependent on Perry’s conception of belief. It will thus be subject to the same weakness.

  18. Is Perry sometimes tempted to claim something stronger than this? At one point he says, “When we come across some ineluctably ordinary belief—a belief that some object has some property—we invent a special name for it (“de re belief”) and wonder how it is possible.” (Perry 2000, p.46) This remark perhaps suggests that Perry is at least contemplating the possibility that belief de dicto might be analysable in terms of belief de re and acceptance: to believe that Φ(a) is to believe of a that Φ(it) by accepting the sentence “Φ(a).” The obvious problem here is, of course, that a subject might believe that Φ(a) when there is no a. It should be noted, however, that in the same essay Perry says that because his goals are limited, “de dicto belief will not be totally banished.” (Perry 2000, p.46) Specifically, it will be banished only for those beliefs whose dimension of acceptance involves context-dependent, hence indexical sentences. It seems, then, that only belief indexical in this sense will be shown to analysable in terms of belief de re and acceptance.

  19. Perry acknowledges the need to make sincerity a condition only in his paper of 1979—see Perry 2000, p.40. He does not mention it in the paper “Belief and Acceptance,” nor in the earlier paper “Frege on Demonstratives” from 1977, which already contains the notion of belief-by-acceptance, admittedly expressed in the rather different terms of apprehending-a-thought-by-entertaining-a-sense—see Perry 2000, p.17ff.

  20. See Perry 2006, p.216.

  21. Incidentally, this last and most succinct way of formulating the reflexive truth conditions of the statement that Lingens is lost in the Stanford library reveals Perry’s debt to Reichenbach’s token-reflexive account of indexicals. Unlike Reichenbach, however, Perry rightly refuses to regard the token-reflexive expression ‘speaker of this utterance of ‘I’’ as constituting the meaning of ‘I’—see Perry 2012, pp.10-11 and pp.87-88.

  22. See Perry 2012, p.135.

  23. See Perry 2006, p.217.

  24. It is obviously not sufficient since knowledge of the reflexive truth conditions of a statement does not in and of itself constitute knowledge of the referential or subject-matter truth conditions and it is primarily in the knowledge of these that understanding consists.

  25. Obviously, the first-person function would still exist in such a language even though there were no dedicated expression for it. There are indeed languages which have no such dedicated expression: “W. von Humboldt has alluded to languages which express the ‘I’ by ‘here’, the ‘thou’ by ‘there’, the ‘he’ by ‘yonder’, thus rendering the personal pronouns by locative adverbs … .” (Heidegger 1984, § 26, H 119) The text to which Heidegger is referring is “Über die Verwandschaft der Ortsadverbien mit dem Pronomen in einigen Sprachen” from 1829, in Gesammelte Schriften, edited by the Prussian Academy of the Sciences, Vol. VI, Division 1, pp.304-330.

  26. Two ancillary observations are worth making about this argument: Firstly, it intimates how important it is to distinguish between reflexively used pronouns (which in English, unlike, say, German, are always expressions exclusively dedicated to this task) and anaphorically used pronouns. Unfortunately, the literature persistently fails to make this distinction—see in this connection, note 29 below. Secondly, in the situation envisaged in the argument even when speakers uttered the expression ‘George’ in order to refer nominally rather than first-personally, they would only refer and state as they did because and insofar as they were stating, hence desiring, intending and believing that they state, that George is thus and so. And this suffices to show the priority of third-personal anaphoric characterisation over Perry’s reflexively truth conditional characterisation. It does so, however, rather less dramatically: it appeals to what we intuitively recognise to be the kind of desire and intention to act, and the kind of belief about one’s own actions, which such speakers must have, rather than to the absurd consequences of denying the priority of third-personal anaphoric characterisation over reflexively truth conditional characterisation.

  27. Reichenbach writes, “The word ‘I’ … means the same as ‘the person who utters this token’ … .” (Reichenbach 1947, p.284; quoted in Perry 2012, p.88)

  28. It seems plausible to maintain that an episode of explicitly and first-personally judging, “I am Φ,” only causes behaviour via causing the standing belief that one is Φ. After all, it is perfectly possible for such an episode to be behaviourally epiphenomenal, a mere fleeting moment in (self-)consciousness. Thus, immediately upon thinking to oneself, “I am Φ,” one might be suddenly distracted by something else. In consequence, no standing belief that one is Φ precipitates out and so no behaviour reflective of the belief that one is Φ emerges.

  29. For grammatical reasons this terminology is misleading. Se is the Latin reflexive pronoun, comparable to the German sich and the English ‘himself’, ‘herself’, ‘oneself’, and ‘themselves’. Thus, one has vidit se in speculo (as opposed to Brutus eum, Caesarem, in speculo vidit, i.e., Brutus saw him, Caesar, in the mirror), Er sah sich im Spiegel and “He saw himself in the mirror.” As reflexive pronouns, there is nothing in intrinsically indexical about these Latin, German and English expressions at all, in contrast to the demonstrative use to which personal pronouns generally can be put, as when a speaker points at someone and says, “He is a millionaire.” In English the reflexive pronoun has a first-person form, in Latin and German, the ordinary personal pronoun is used in non-third-person cases, but their person, first, second or third, is determined by the pronoun or noun which governs the reflexive pronoun. So in their capacity simply as reflexive there is also nothing intrinsically first- (or, for that matter, second- or third-) personal about reflexive pronouns at all.

    In fact, it seems that the only reason why English-speaking philosophers have associated reflexive pronouns in English with the first person is the fact that in English the reflexive pronoun can be used, in conjunction with the corresponding personal pronoun, for extra emphasis—sometimes indeed to emphasise that the subject has some kind of first-personal awareness. But when the reflexive pronoun is being used in this way, it is not functioning as a reflexive pronoun! That it is not thus functioning is shown by the fact that German accomplishes this kind of emphasis by using ‘selbst’ or ‘selber’, which is not a reflexive pronoun at all.

    As already indicated, of the three languages Latin, German and English, only the latter has a dedicated first-personal reflexive pronoun: “I saw myself in the mirror” must be translated as in speculo me viderem and Ich sah mich im Spiegel. Me and ‘mich’ are simple first-personal pronouns in the accusative. If one wanted to emphasise the reflexive (and indeed possibly the first-personal, self-conscious) character of these claims in Latin and German, one would say, me ipsum in speculo viderem and Ich sah mich selbst im Spiegel. Curiously, in order to do this in English, one would go in the opposite direction, namely, introduce the simple accusative pronoun: “I saw me myself in the mirror.”

    It is thus wrong to take, as so many appear to do, the grammar of the sentence “S believes that it is Φ” as a reason for describing it either as attributing de se a belief or as attributing a belief de se. For there is nothing reflexive at all about this sentence; the pronoun in it is not a reflexive but a (nominative) personal pronoun. This is not, however, necessarily to say that Lewis, who introduced this terminology, was wrong. It may be that he was being characteristically inexplicit and enigmatic. Perhaps he was saying—his discussion of Perry’s example of Hume and Heimson seems to suggest this—that Rudolf Lingen’s believing that he is Rudolf Lingens is perspicuously characterised as his believing himself to be Rudolf Lingens. If this is so, then Lingen’s belief is de se since ‘himself’ is a reflexive pronoun functioning as a reflexive pronoun. Believing that p would then be a matter of believing oneself to inhabit a world at which it is true that p. In this sense, perhaps, “any kind of self-locating belief should be understood as self-ascription of properties.” (Lewis 1979, p.522) Lewis’s thesis is thus that “the de se subsumes the de dicto, but not vice versa … ,” (Lewis 1979, p.521) that is, that all belief is de se.

  30. I do not wish to insist that this belief is necessary; perhaps S just responds to the perception of the large aggressive dog charging towards it by running inside. In this case, S would be structurally programmed, either by evolution or by habituation (association), to behave according to the general behavioural conditional “If one is being attacked, run into the house.”

  31. One need not insist on absolute uniqueness, but merely uniqueness sufficient for a hearer to determine what entity the speaker is referring to. Thus, if I say, “The man standing in the corner is my brother,” I can succeed in referring to the man I know to be my brother even if there is, indeed, even if I know there is, another man standing in the corner. It all depends on what I believe to be sufficient for my hearer to distinguish whom I am referring to from all others. One might say that the uniqueness condition of definite descriptions as used in actual speech contains a suppressed component: “The only man standing in the corner who I can be rationally interpreted as intending to refer to is my brother.”

  32. One could regard the object of belief as a sentence.

  33. See Perry 2000, pp.35-37.

  34. My thanks to Brian Garrett for making me aware of the need to distinguish the kind of imagining I intend from imagining in the sense of conceiving.