1. The Entitlement Principle

In his recent paper “Descartes defended” (2012), Peacocke offers the following reconstruction of Descartes's cogito:

• First, Descartes engages in

  1. (1) a particular conscious thinking;

• second, he moves from this conscious event to the judgement

  1. (2) I am thinking;

• third, from (2) he infers

  1. (3) I exist

  (Peacocke Reference Peacocke2012: 110)

In what follows, my focus will be primarily on the first step in the transition (1)–(3). I do not think that the second step is entirely unproblematic, but I take it that the Lichtenbergian view gives us no special reason to resist it. Indeed, Lichtenberg's claim that we should avoid formulating the starting point of the cogito using the first person would be hard to understand if the truth of ‘I am thinking’ failed to require that ‘I’ refers, which is what the step from (2) to (3) makes explicit.

Peacocke points out that the step from (1) to (2) should not be understood as an inference whereby Descartes, having endorsed the proposition ‘I am engaging in a particular conscious thinking’ deduces from it the proposition ‘I am thinking'. Rather, (1) is supposed to describe a mental state of which Descartes is introspectively aware and (2) the content of a judgement that Descartes forms on the basis of this awareness. The crucial question for our purposes is whether Descartes's judgement can be justified by that awareness and, consequently, whether Descartes can know that he's thinking on the pure basis of introspection.Footnote ³

In principle, the answer to these questions could be made to depend on one's views about introspective awareness – ‘yes’ if it is assumed that in being introspectively aware of one's own thinking one is also aware of oneself as the author of the thinking, ‘no’ if introspective awareness is taken to exclude the self. The problem is that it's hard to see how we could assume either of these two views at the start without invoking phenomenological intuitions that at least some participants in a debate about the validity of the cogito will find dubious. Peacocke's approach allows us to bypass this difficulty. He takes the starting point of the cogito to be a kind of awareness that neither encompasses nor excludes the self – not awareness of oneself as thinking, nor awareness of some kind of subjectless thinking, but simply awareness of one's thinking.Footnote ⁴ He then invokes two principles that together explain how such awareness can secure knowledge of the fact that one is thinking. The first principle asserts that there is a tight connection between the first-person concept and the concept thinks, in that:

1. (a) [A]ny plausible account of the nature of the concept thinks entails that to possess it involves taking one's own conscious thinking as conclusive reason for self-applying the concept of thinking in the present tense. (Peacocke Reference Peacocke2012: 120–1)

The second principle asserts that

1. (b) [I]f someone judges a content on grounds that are, as a matter of the nature of one or more concepts in the content, conclusive grounds for judging the content, then the content so judged is knowledge. (Peacocke Reference Peacocke2012: 120–1)

Since nothing in the Lichtenbergian view conflicts specifically with (b), I will concentrate on (a). (a) says that it is built into the nature of the concept thinks and the first-person concept that, whenever a subject competent with these concepts is introspectively aware of what is, in fact, his or her conscious thinking, he or she has conclusive reason to judge ‘I am thinking'. The idea is that one's competence with the relevant concepts will allow one to ‘bridge’ whatever epistemic gap there may be between the introspective awareness of a certain mental activity (e.g., a particular conscious thinking) and the self-ascription of that activity (e.g., a judgement with the content ‘I am thinking').

In his response to Peacocke, Campbell (Reference Campbell2012) seems to reject this idea in favour of a more ‘Lichtenbergian’ approach. He suggests that, if one is “genuinely open-minded about [one's] own existence” and is struck by a conscious thought, one may fail to regard oneself as having the right to move from that to ‘I am thinking'. Instead, one “might say something like ‘I am thinking, if there is any such person’. Or better, one might use the kind of formulation proposed by Lichtenberg: ‘There is thinking’” (Campbell Reference Campbell2012: 366–7).

But these suggestions only appear to contradict (a). Note that (a) speaks of a conclusive reason for self-applying the concept thinks in the present tense. Given (b), a conclusive reason must be one that speaks so strongly in favour of judging a certain content as to be able to support knowledge of the content. Now, Peacocke doesn't say whether anything else beyond the capacity to support knowledge is required of a reason in order for it to qualify as conclusive. But, elsewhere, he suggests that even a conclusive reason can in principle interact with other reasons in such a way as to leave the subject without all-things-considered, doxastic justification for judging the relevant content.Footnote ⁵ I shall put this by saying that a reason can be conclusive without being indefeasible. Given this distinction, the fact that one could be (as Campbell puts it) ‘open-minded’ about his or her own existence is neither here nor there with respect to the truth of (a). For even if one's conceptual competence and introspective awareness provide one with conclusive reason for believing that one exists, as (a) says, there remain all sorts of defeating grounds one might be given for being so open minded.

In effect, Campbell himself makes it clear that he is not interested in questioning the legitimacy of the first step in the transition (1)–(3). What interests him is the idea that “possession of [the first-person concept] may already require that one knows of one's own existence” (Campbell Reference Campbell2012: 366) and that, since the cogito does not explain our possession of that concept (but, rather, presupposes it) the source of such knowledge must lie “somewhere quite outside the range of cogito-style reasoning” (Campbell Reference Campbell2012: 365). Note that none of this contradicts the claim that the first step in the transition (1)–(3) results in knowledge of (2) whenever the subject possesses the first-person concept and the concept thinks. The point is only that for the first step in the transition (1)–(3) to result in knowledge of (2) Descartes must have knowledge of (3) already.

As it turns out, then, both Peacocke and Campbell can be said to agree on a common point, namely that:

I speak of entitlement to signal that, for all the principle says, the subject might not be able to appreciate the fact that he or she is aware of a conscious thinking as his or her reason for judging ‘I am thinking'. If the subject had to be able to do this, we would presumably need some account of how one can know or justifiedly judge ‘I am introspectively aware of a conscious thinking', and it's hard to see how such an account could be put forward without triggering a vicious regress. Instead, the fact that one is introspectively aware of one's own conscious thinking should be regarded as a condition whose obtaining conclusively justifies one to judge ‘I am thinking’ if one possesses the relevant concepts and whether or not one actually grasps that condition.Footnote ⁷ In a context in which all parties agree to a policy of phenomenological neutrality – avoiding any assumption about the self-involving vs. self-excluding nature of conscious experience – this seems the most plausible line to take for the defender of the Cartesian view. So, for the purpose of assessing the prospects of that view, it is useful to set aside the differences between Peacocke and Campbell, and focus on the Entitlement Principle itself. The thought-experiment I will introduce in the next section is meant to bring into focus the reasons why I find this principle problematic.

2. The monoselfists and the poliselfists

Imagine two communities of rational thinkers with radically different views on the metaphysics of the self – the monoselfists and the poliselfists.

The monoselfists believe that each normally functioning human brain promotes the existence of a single self (I use the term ‘promotes’ without any metaphysical connotation – some monoselfists regard the self as wholly distinct and independent from the brain; others treat it as a broadly physical entity). For the monoselfists, any instance of thinking is an instance of the thinking of a particular self. So, based on their competence with the first-person concept and the concept thinks, whenever the monoselfists engage in conscious thinking, they feel justified to form a judgement with the content ‘I am thinking'. In other words, they regard the kind of transition exemplified by the move from (1) to (2) as epistemically warranted.

The poliselfists disagree. They believe that each normally functioning human brain promotes the existence of several selves – a sensitive self, a deliberative self, a desiderative self, etc. (again, ‘promotes’ is meant to be neutral on the exact relation between the selves and the brain; some poliselfists are dualists, others are not). The poliselfists regard thinking as a cooperative activity carried out by the brain's many selves. So, based on his or her competence with the first-person concept and the concept thinks, whenever a member of the poliselfist community engages in conscious thinking, he or she does not feel justified to judge, on this basis, ‘I am thinking', but, at most, ‘We are thinking'. In other words, the poliselfists regard the kind of transition exemplified by the move from (1) to (2) as epistemically unwarranted.

It is important to get the poliselfist ideology right. We sometimes apply a predicate to several things collectively. For example, we may say of certain premises that they ‘imply’ a certain conclusion, meaning that they jointly imply that conclusion, not that the conclusion is implied by each of them. (We may put this by saying that the conclusion is implied by the set of the premises, but that is a sloppy way of talking, for certainly sets don't imply things.) The poliselfists make a similar use of ‘thinking’ and its cognates. For example, they say of certain selves that they ‘believe’ a certain proposition, meaning that they jointly believe that proposition, not that the proposition is believed by each of them. (They may put this by saying that the proposition is believed by the set of the selves, but that is also a sloppy way of talking, for certainly sets don't believe things.) It is because they take the thinking of which they are introspectively aware to be the collective activity of several selves that the poliselfists refrain from forming judgements with the content ‘I am thinking'. That is the source of their disagreement with the monoselfists.

We can imagine various reasons why the poliselfists conceive of human thinking in the particular way they do. One possibility is that they arrived at their view through a particularly radical interpretation of the thesis that our mind comprises several distinct and mutually independent ‘modules'. Another is that, having been impressed by the existence of certain kinds of dissociative disorders, they found it convenient to regard the condition of a normal human being as one in which several selves think ‘harmoniously’ or ‘in unison'. How exactly we fill in these details does not matter much; the essential point is that, however extravagant and far-fetched it may sound, Poliselfism is not the kind of doctrine which is altogether impossible for a rational subject to embrace. In particular, it is not impossible for a rational subject to embrace Poliselfism with full strictness – refusing to ever replace claims such as ‘Those selves are thinking that p’ with ‘That group of those selves is thinking that p', or other singular paraphrases.Footnote ⁸

I want to suggest that reflection on the disagreement between the monoselfists and the poliselfists can be used to put pressure on the Entitlement Principle. To this end, let me make three further points of clarification.

The first point concerns conceptual competence. I said that the poliselfists can be competent with both the first-person concept and the concept thinks. Given that they refrain from judging ‘I am thinking’ even at times when they are introspectively aware of a conscious thinking, some might be tempted to deny these claims. I will defend the view that a poliselfist can possess the concept thinks in the next section. Here, let me focus on the first-person concept. Suppose we adopt Peacocke's own characterization of this concept, according to which:

Then a good case can be made that the poliselfists do not lack the first-person concept. To see why, imagine we were to ask a poliselfist why she refrains from judging ‘I am thinking’ even at times when she is introspectively aware of a conscious thinking. It seems plausible to think that she would answer our question by criticizing the conception of thinking it presupposes. She would tell us that, contrary to what we are accustomed to assuming, wherever there is a human thought featuring the concept ‘I', there isn't any such thing as the subject who produces the thinking. The thinking is produced collectively by several selves, hence there isn't any particular self (or entity of any kind) for the concept ‘I’ to refer to. Now, it seems to me that an answer of this sort, far from showing that our interlocutor lacks the first-person concept, would be a firm indication that she understands and respects its constitutive condition of application, as specified by Peacocke in the quote above. It is true that – in at least some contexts – our interlocutor doesn't use the concept as we do. But, intuitively, this is because her unusual views about thinking stand in the way of her doing so, not because of an imperfect grasp of how ‘I’ works.

It is also useful to see that, while the poliselfist's use of the concept would differ from ours in certain contexts, it would be exactly like ours in others. For example, it is perfectly possible for a poliselfist to believe that, though human thinking is the cooperative activity of several selves, divine thinking is the activity of a single, unified self – so that, unlike us, God can justifiedly and truthfully judge ‘I am thinking'. If someone were to ascribe to God statements in the first-person, the poliselfist would understand and report such statements in the same way we do. That would be further evidence of her competence with the concept I.Footnote ⁹

Second point. I said that, whenever they engage in conscious thinking, the poliselfists feel unjustified to judge ‘I am thinking'. I now submit that, in imagining the epistemic situation of the poliselfists, we can coherently suppose this feeling to be correct. As we've just seen, a believer in Poliselfism has reasons to think that the ‘I’ in ‘I am thinking’ fails to refer whenever the judgement is formed by a normal human being. If one believes in Monoselfism, one may well find these reasons weak and unconvincing. But it is not difficult to imagine that the poliselfists might have accumulated so much empirical and testimonial evidence in favour of their view that it would be irrational of them to give such reasons less weight than they do, and go on to believe ‘I am thinking’ in spite of them (the evidence in question may of course be misleading, but that's beside the point).

Note that the resulting supposition will be that the poliselfists lack all-things-considered justification to judge ‘I am thinking’ at times when they are introspectively aware of a conscious thinking. This supposition does not beg the question against the Entitlement Principle, which speaks of a conclusive entitlement to judge ‘I am thinking'. As was pointed out in the last section, a conclusive entitlement need not be an indefeasible entitlement. It is possible that the poliselfists lack all-things-considered justification, not because they do not have any conclusive entitlement (contrary to what the Entitlement Principle would predict), but because their entitlement was defeated by the reasons they have for not using ‘I’ in introspective reports. In a moment, I will look at these interpretative options in more detail.

Third and last point. I said nothing about ‘what it is like’ to be a poliselfist, but I submit we are free to stipulate that (modulo their different beliefs) their phenomenology does not differ significantly from that of the monoselfists.Footnote ¹⁰ The point of this stipulation is to rule out one possible explanation of the epistemic difference between the monoselfists and the poliselfists: if the two communities share the same phenomenology, we can safely assume that the reason why they are in different epistemic positions vis-à-vis the transition from (1) to (2) must lie either in their competence with thinks and the first-person concept, or else in their diverging beliefs involving those concepts.

I note in passing that the first of these three points gives us good reason to resist Campbell's suggestion that possession of the first-person concept “may already require that one knows of one's own existence” (Campbell Reference Campbell2012: 366). If possession of the first-person concept required knowledge of one's own existence, the poliselfists would have to be said to lack that concept – for knowledge requires belief, and no poliselfist believes ‘I exist'.

But let us go back to the Entitlement Principle on which both Peacocke and Campbell agree. Recall that, according to that principle, whenever a rational subject, competent with the first-person concept and the concept thinks, is introspectively aware of his or her conscious thinking he or she has a conclusive entitlement to judge ‘I am thinking'. Given the three points I have just made, the poliselfists are rational subjects who, despite being competent with the first-person concept, and being introspectively aware of their own conscious thinking do not have all-things-considered justification to judge ‘I am thinking'. This leaves us with three possible ways of explaining their epistemic condition:

1. (A) They lack the concept thinks, so the fact that, when they engage in conscious thinking, they do not have all-things-considered justification for judging ‘I am thinking’ shows nothing about the truth or falsity of the Entitlement Principle.

2. (B) They possess the concept thinks and the Entitlement Principle is true. However, their belief in Poliselfism acts as a defeater of a conclusive entitlement they have for judging ‘I am thinking’ whenever they engage in conscious thinking.

3. (C) They possess the concept thinks and the Entitlement Principle is false. However, the belief in Monoselfism provides monoselfists with the auxiliary evidence needed for their cogito-judgements to be justified.

Options (A), (B) and (C) are mutually exclusive and jointly exhaustive. For either the poliselfists possess the concept thinks or they don't. And if they do, either the Entitlement Principle holds or it doesn't. If it does, we need an explanation of why the poliselfists do not have all-things-considered justification to judge ‘I am thinking’ (and, given the stipulation that the two communities share the same phenomenology, the explanation will have to do with how their belief in Poliselfism counteracts any epistemic entitlement they share with the monoselfists). If it doesn't, we need an explanation of why the monoselfists are all-things-considered justified to judge ‘I am thinking’ (and, given the stipulation of phenomenological parity, the explanation will have to do with how their belief in Monoselfism supplements whatever epistemic warrants they share with the poliselfists). In the next section, I will make a case for (C) and against the Entitlement Principle.

3. The conceptual possibility of collective thinking

Consider the following central tenet of Monoselfism:

It seems clear that, for option (B) to be true, (T1) must be something that we have conclusive a priori justification to believe. For (B) requires the truth of the Entitlement Principle, which says that possession of the first-person concept and the concept thinks is sufficient to have a conclusive entitlement to judge ‘I am thinking’ whenever one is aware of one's conscious thinking. Now, it is hard to see how possession of the two concepts could secure a conclusive entitlement to judge ‘I am thinking’ whenever one is aware of one's conscious thinking if possession of the concept thinks does not even secure conclusive justification to believe that, when there is thinking going on, there is something (or someone) who is the author of the thinking. And to say that possession of the concept thinks secures such justification just is to say that we have a priori justification to believe (T1).

But the apriority of (T1) is also required by the truth of (A). For according to (A), the poliselfists lack the concept thinks. And the only plausible reason to believe that the poliselfists lack that concept has to do with their seemingly deviant view that thinking can be the cooperative activity of several selves. Only if this view is a priori false (because (T1) is a priori true) is there plausible ground for denying the poliselfists’ competence with thinks.

Thus, both (A) and (B) require that (T1) be a priori. But is (T1) a priori? In particular, does the concept thinks preclude the possibility of collective thinking? In order to get ourselves in a position to answer these questions, let us make two assumptions. First, let us assume that we can treat the concept thinks as a stand-in for any of the concepts expressed by ordinary psychological predicates, including ‘believe', ‘judge', ‘be aware’ and ‘know'. This seems appropriate if we agree that Descartes “could equally have formulated his argument in terms of other conscious properties” (Peacocke Reference Peacocke2012: 111). Second, let us assume that our ordinary practice of assertion involving those predicates is a good (if defeasible) guide to what possibilities the corresponding concepts do or do not preclude. I want to argue that, given these assumptions, our willingness to make collective use of psychological predicates such as ‘think', ‘believe', ‘judge', ‘be aware’ and ‘know’ provides good (if defeasible) evidence that (T1) is not a priori. Let me explain.

When I first introduced their view, I said that the poliselfists conceive of each instance of human thinking as the cooperative activity of several selves. It is because they conceive of human thinking in these terms that the poliselfists make collective use of predicates such as ‘think', ‘believe', ‘judge', ‘be aware’ and ‘know'. They apply such predicates jointly to several selves. Clearly, this is not how we ordinarily use psychological predicates when describing the mental activities of which we are introspectively aware. But there are plenty of other occasions in which we too use psychological predicates in a collective fashion:

Gilbert points out that the examples mentioned in this passage can be classified into two classes, depending on whether or not the relevant psychological predicate is ascribed to several individuals which are also presumed to constitute a single group or social entity of some kind (e.g., a union, a court, a family). But, as Gilbert also observes, group talk can easily be paraphrased into plural talk. For example, “with respect to belief, one might say either, ‘Population P believes that p’, or, equivalently, ‘The members of population P (collectively) believe that p.’” (Gilbert Reference Gilbert2004: 96).Footnote ¹¹ This means that each of the examples above can in principle be seen as one in which we are saying of several individuals that they carry out a certain mental activity (for example, ‘believing that management is being unreasonable'), meaning that they jointly carry out that activity, not that the activity is carried out by each of them. Variations on this theme are rife in the recent literature on collective epistemology: a jury may believe that the defendant is innocent, even if not all jurors would endorse this proposition (Lackey Reference Lackey2015: 3); the FBI may have had knowledge of a cluster of facts highly pertinent to 9/11 even if none of its members did (Goldman Reference Goldman2004: 17); the crew of the USS Palau may know that it is heading NNW at 14.2 knots even if such knowledge is not possessed by each and every member of the crew (Bird Reference Bird and Lackey2014: 46).

There may be an initial temptation to regard these examples as distracting, saying that they involve what is best regarded as loose or metaphorical talk. However, several arguments have been made that plural psychological ascriptions should be taken seriously (see Gilbert Reference Gilbert1989; Nelson Reference Nelson, Alcoff and Potter1993; Searle Reference Searle1995; Tuomela Reference Tuomela1995; Bratman Reference Bratman1999). If we take these arguments to establish the minimal point that examples like the ones above can be regarded as literally true – i.e. that their literal truth cannot be ruled out a priori – how should an advocate of options (A) or (B) deal with such examples? I can think of three replies, and none of them strikes me as fully convincing.

One reply would be that the examples under consideration fail to establish the conceptual possibility of collective thinking without individual thinking. Bill and Jane, the components of a court of justice, the agents of the FBI, etc. are all thinking individuals. The fact that they can satisfy certain psychological predicates collectively without the predicate being satisfied by each of them doesn't show that there can be thinking without at least some individual x who thinks.

But this reply misses the point. (T1) doesn't just say that, wherever there is thinking, there is some x who thinks. (T1) says that, wherever there is thinking, there is some x who is the author of that thinking. To appreciate the difference between these formulations, consider a ‘homuncular’ version of Poliselfism according to which the thoughts of which we are introspectively aware can be ascribed, collectively, to several selves, each of which thinks, in turn, his or her own thoughts. The holder of such a view may well concede that there cannot be thinking without some x who thinks. Nevertheless, she will insist that we are not justified to judge ‘I am thinking’ at times when we are introspectively aware of a particular conscious thinking. If any thinking of which we are introspectively aware is performed collectively by several selves, there won't be any single individual for ‘I’ to pick out as the author of such thinking – no matter whether the selves in question also think their own thoughts. Reflection on this point shows that, if it is to support the Entitlement Principle, (T1) must be understood as saying that every instance of thinking is an instance of the thinking of a particular individual, not just that every instance of thinking presupposes individual thinking. The proposed reply doesn't help to make a claim along these lines any plausible.Footnote ¹²

A second reply points to the existence of a deep difference between the kind of mental activities we are willing to ascribe collectively to several individuals, and the kind of mental activities of which we are introspectively aware. The starting point of the cogito is one's awareness of one's own conscious thinking, and it may be said that none of the examples I mentioned involves a type of collective thinking that is also conscious. So it may be suggested that, even if (T1) is not a priori, we could uphold the Entitlement Principle – and vindicate options (A) or (B) – by appeal to the apriority of:

One basic problem with this reply has to do with its last step – the claim that (T2) is a priori. I agree that we do not ordinarily conceive of consciousness – especially phenomenal consciousness – as the kind of property that can be collectively instantiated by several individuals. But I am not sure that this observation alone suffices to show that, unlike the concept thinks, the concept consciously thinks excludes the possibility of collective application.

Consider how consciousness is discussed in certain scientific contexts. In particular, consider the influential hypothesis that consciousness correlates with the capacity of a system to process and integrate information (Tononi and Koch Reference Tononi and Koch2015) and the related suggestion, recently discussed by List (Reference List2018: 313) and endorsed by Schwitzgebel (Reference Schwitzgebel2015), that very large and functionally integrated groups, perhaps designed on the model of a neural network, can be collectively conscious.Footnote ¹³ No doubt these suggestions are hugely controversial. But precisely the fact that we regard them as controversial, and not as outright incoherent seems to me to indicate that (T2) is not a priori. If (T2) were a priori true we would have to be in a position to rule out ‘from the armchair’ what, to all appearances, are best seen as empirical hypotheses.

Granted, it is not part of our ordinary practice of assertion involving predicates like ‘consciously think’ to apply such predicates collectively to many individuals. But there is a difference between saying that collective uses of ‘consciously think’ are not part of ordinary practice (something which is entirely neutral on whether (T2) is a priori) and saying that it is part of ordinary practice to prohibit such uses (something which would support the apriority of (T2)).Footnote ¹⁴ I don't see any clear evidence of the existence of such a prohibition. On the contrary, I think one could plausibly argue that, if the hypotheses I have just described turned out to be correct, the idea of collective consciousness would become common currency among ordinary speakers. This would not be a ‘change of topic’ on our part, so much as the realization of a possibility left open by the concept of consciousness we actually employ.

The third reply points to a difference, not between two different kinds of mental activities – those we are willing to ascribe to several individuals collectively and those of which we are introspectively aware – but between different concepts of mental activities in general. It may be said that collective uses of psychological predicates involve ‘external’ concepts of psychological conditions – concepts that are applied ‘from the third-person perspective', in the context of ordinary language or scientific practice – whereas the cogito is concerned with an ‘internal’ concept of thinking – a type concept that each of us is supposed to apply ‘from the first-person perspective', in the context of introspecting his or her own thinking.

Now, consider a concept thinks⁺ which resembles thinks in every respect, except that it excludes the possibility of collective application. If thinks⁺ could be shown to be our ‘internal’ concept of thinking, the Entitlement Principle could be vindicated by appeal to the apriority of:

The fact that neither (T1) nor (T2) are a priori would be beside the point.

The success of this line of response depends on whether its proponents can show that our internal concept of thinking is, indeed, thinks⁺, as opposed to a more liberal concept which allows for collective application. Note that it won't be enough to show that, as a matter of fact, our introspective awareness targets thinking+ (the property picked out by thinks⁺) rather than thinking (the property picked out by the ordinary concept thinks). One will also have to show that the property of thinking+ wears the impossibility of its collective instantiation on its sleeves, in the sense that this aspect of its nature is somehow ‘transparent’ to any subject introspectively aware of it. (Compare: the mere fact that the substance of which we are perceptually aware in being aware of water is H₂O does not imply that, in forming judgements on the basis of that awareness, we are entitled to draw on the a priori rules governing a concept as demanding as H₂O – water does not wear its chemical composition on its sleeves).

Here our constraint of phenomenological neutrality becomes once again relevant. Both the claim that the non-collective nature of our thinking is introspectively ‘transparent’ to us and the claim that we have direct introspective evidence for the existence of a self are likely to be endorsed by some and rejected by others on the grounds of phenomenological intuitions that we have little or no independent basis to assess. Thus, it seems we should not be moved by the strategy of appealing to the apriority of (T3) in a context in which it was agreed to shun appeal to such intuitions. The question whether conceptual competence can guarantee a justified transition from (1) to (2) should be settled by reference to the psychological concepts we actually employ – not to those we could employ given a controversial construal of the phenomenological data.

Where does all this leave us with respect to the dispute between Cartesians and Lichtenbergians?

4. Poliselfism and the cogito

In the last section, I argued that (T1) – the principle that wherever there is thinking, there must be some x who is the author of that thinking – is not a priori true. This implies two things. First, that we have no ground to deny that the poliselfists possess the concept thinks. There is nothing conceptually wrong in the idea that thinking might be the cooperative activity of several selves. Second, that the right explanation of the epistemic difference between the monoselfists and the poliselfists is not the one offered by advocates of the Entitlement Principle. We shouldn't say that the belief in Poliselfism acts as a defeater of an epistemic entitlement the poliselfists have for judging ‘I am thinking’ whenever they are introspectively aware of their own conscious thinking. Rather, we should say that there isn't any such entitlement in the first place.

This result, if correct, poses a serious challenge to the Cartesian view. Recall that the Entitlement Principle was supposed to explain, in phenomenologically neutral terms, how a rational subject can justifiedly move from his or her introspective awareness of a particular conscious thinking to the self-ascription of that thinking (§1). If the principle is rejected, an epistemic gap remains open and some auxiliary evidence seems to be called for in order to close it – evidence coming, not from introspection or a priori reflection, but from other a posteriori sources.

It is interesting to note that what are usually described as ‘rationalist’ interpretations of the cogito – i.e., interpretations according to which Descartes's knowledge of his own existence is not in any way based on his introspective awareness of a particular conscious thinking – are no less affected by this conclusion than the ‘empiricist’ one we have been focusing on so far.Footnote ¹⁵ Consider, for instance, Shoemaker's (Reference Shoemaker1996) ‘rationalist’ take on how one can know that one exists:

I don't think we should be fully convinced by what Shoemaker says in this passage – at least, not if we take into proper account the possibility that ‘I think’ or ‘I exist’ might be collectively asserted or entertained in thought. Of course, it is a logically necessary condition of my being deceived that I think. But this is not sufficient to make ‘I think’ indubitable. If collective thinking is a logically viable option, it is logically possible that there be deception without a single deceived subject – hence without anything for the ‘I’ in ‘I am being deceived’ (or ‘my’ in ‘my being deceived') to refer to. This means that it is not a logically necessary condition of ‘I think’ being asserted, or even entertained in thought, that it be true. I can judge ‘I think’ and, based on the possibility of Poliselfism, bring myself to doubt the truth of my judgement.

To put the point more clearly, let us say that a proposition p is a priori self-verifying when it is a priori (or, in Shoemaker's terminology, ‘logically necessary’) that any thought with p as its content must be correct. If we have no a priori grounds to rule out the truth of Poliselfism – or more in general, the possibility of collective thinking – it will not be a priori that any thought with the content ‘I am thinking’ must be correct. A collective thought with the content ‘I am thinking’ would not be correct – at least, not if a necessary condition for its correctness is that its subject term succeeds in referring to something. Hence, the proposition ‘I am thinking’ is not a priori self-verifying, contrary to what this and other ‘rationalist’ readings of the cogito have been assuming (cf. Williams Reference Williams1978: 59; Burge Reference Burge1996: 93).

Some may be inclined to dismiss these observations as irrelevant to the validity of cogito. After all, compatibly with everything I have argued so far, one could accept a weaker version of the Entitlement Principle according to which:

It may be said that, since ‘Either I am thinking or we are thinking’ entails ‘Either I exist or we exist', and ‘We exist’ entails ‘I exist', the Entitlement Principle* gets us back to what Descartes wanted, independently of whether the first step in the transition (1)–(3) is valid or not.

But this reasoning contains a fallacious step. To say that ‘We exist’ entails ‘I exist’ is to say that every context in which ‘We exist’ is true is a context in which ‘I exist’ is also true. Now, ‘We exist’ is true in a context if and only if the agents of the context exist, whereas ‘I exist’ is true in a context if and only if the agent of the context exists. If there are contexts in which certain individuals qualify as the agents of the context (plural) without any of them qualifying as the agent of the context (singular), we will have a counterexample to the entailment – a case in which ‘We exist’ is true while ‘I exist’ fails to be true because there is no single entity for ‘I’ to pick out. The contexts created by collective thoughts and assertions are precisely contexts of this sort. Suppose a court of three judges collectively asserts the sentence ‘We declare the defendant guilty, but I disagree'. The reason why we would be at a loss to interpret this sentence in the context created by the assertion is that, relative to such context, the three judges qualify as the agents (plural) without any of the three qualifying as the agent (singular). This means that ‘I’ cannot be assigned any referent, making it impossible for us to evaluate ‘I disagree’ (or any other continuation of the sentence in the first-person singular). A believer in Poliselfism will see the context relevant to the assessment of any of the thoughts of which we are introspectively aware in the same way. By her lights, a thought with the content ‘I exist’ cannot be true if ‘I’ needs to refer to the agent that produces the thinking (singular) and, by hypothesis, there isn't any such agent.

If these considerations are on the right track, the Entitlement Principle* gives us strictly less than what Descartes wanted. The truth of the principle would leave us with a position according to which introspection and a priori reflection put one in a position to judge, not that one is thinking, but only that one or more things, indexically referable through ‘I’ or ‘we', are doing the thinking of which one is introspectively aware.Footnote ¹⁶ The introspecting subject would be stuck with an irreducibly disjunctive judgement, from which he or she could infer, at most, ‘Either I exist or we exist'. Whatever the merits of this position, it is not the Cartesian one.