Stopping points: ‘I’, immunity and the real guarantee

Abstract The aim of the paper is to bring out exactly what makes first-personal (and more generally indexical and demonstrative) contents special, by showing that they perform a distinctive cognitive function. Namely, they are stopping points of inquiry. First, I articulate this idea and then I use it to clear the ground from a troublesome conflation. That is, the conflation of this particular function all first-person thoughts have with the property of immunity to error through misidentification, which only some I-thoughts enjoy. Afterward, I show the implications of this idea for a theory of first-person content and of immunity to error though misidentification. I then make some comparisons with Pryor’s notion of wh-misidentification and immunity thereof and with Cappelen and Dever’s position on immunity to error through misidentification and show why they are defective.

propositional content, which displays itself in the failure in intensional contexts of the principle of substitutivity salva veritate of co-referential terms.
I beg to disagree. I do think there are irreducibly first-personal contents. They are not entirely unique in what makes them special, though. They share what makes them distinctive with other indexical contents -particularly with here and now contents -as well as with genuinely demonstrative contents. The aim of this paper is to bring out exactly what makes them special, by showing that they perform a distinctive cognitive function. Namely, they are stopping points of inquiry. First, I articulate this idea (Section 1) and then I use it to clear the ground from a troublesome conflation (Section 2). That is, the conflation of this particular function all first-person thoughts have with the property of immunity to error through misidentification, which only some I-thoughts enjoy. Afterward (Section 3), I show the implications of all that for a theory of first-person content and of immunity to error though misidentification. In particular, I claim that error and immunity to error through misidentification are epistemic phenomena, having simply to do with the structure of the epistemic grounds of the judgment. In particular, I claim that it is necessary and sufficient for an I-judgment to be immune to error through misidentification that it is not based on any identification component as part of its epistemic grounds and I show why this rules out some alternative explanations. I then make some comparisons with Pryor's notion of wh-misidentification and immunity thereof and with Cappelen and Dever's position (Section 4). I claim that Pryor's notions are spurious and that while Cappelen and Dever are right in holding that indexicality, in particular I-indexicality, does not suffice for immunity to error through misidentification, they are wrong in thinking it is not necessary for it. In particular, they have no case for the view that also proper names may be used in judgements, which are so immune. Their case for claiming that proprioception does not contain indexical contents is scant, and shows, at most, that at least demonstrative contents are needed in order to have judgements which are immune to error through misidentification. This is no news, however, since all parties agree that immunity is not the property only of (some) I-contents, but also of (at least some) here, now, or that-contents.

Stopping points: 'I' and the real guarantee
In my view, indexical and in particular first-personal contents possess what, elsewhere (Coliva 2003(Coliva , 2012, I have called 'the real guarantee' . The name may sound cumbersome and probably is. Nevertheless, here is the simple idea: once you entertain an I-content like 'I am F' (and I will focus on propositional contents for convenience in the rest of this paper, if not explicitly indicated otherwise), there is no room for a certain kind of question. Namely, assuming the proposition ascribes a certain property to the individual, it makes no sense to ask 'Who does (or seems to) have that property?'; or, equivalently, 'Which person does (or seems to) have that property?' . To put it otherwise, once you entertain a first-person content, you have reached a stopping point of inquiry. If you are entertaining a genuinely first-personal content, you automatically know that it is you that have (or seem to have) the property in question and there is no room left for an inquiry concerning who has that property. Hence, to put it emphatically, what distinguishes first-person contents from impersonal ones is their 'luminosity' in one respect. Namely, they are luminous regarding the individuation of the subject who has (or seems to have) the property which gets ascribed to her in the proposition. There is, therefore, no more work that could possibly be done to individuate that subject.
If I am right, then, while we can agree with Cappelen and Dever that many cases usually presented to support the view that there are irreducibly first-personal contents are variations on Frege's Hesperus-Phosphorus case, and therefore may be interpreted as showing the opacity of propositional content, the intended reading of all those cases -meaning the one which is in fact relevant to the discussion on indexicality -is actually different. In particular, if I am right, this should bring into sharper relief the idea that those cases actually aim at showing that once you entertain a first-person content, you can no longer be ignorant regarding who has the property ascribed to her in the relevant proposition. Conversely, until you home in an I-content, you may be so ignorant. Hence, while when I entertain the proposition 'AC is F' , I may devise scenarios in which I go on sensibly to ask 'Ok, AC is F, but which person is AC?' , once I home in an I-content, then I can no longer sensibly ask 'OK, I am F, but which person is I?' . If I asked that question, that would show either that I do not have the first-person concept and that, in its turn, will prevent me from entertaining an I-thought at all; or else that I do have the concept but are (perhaps only momentarily) insane -and therefore irrational -not just ignorant.
I cannot possibly survey all cases presented in the literature on the first person, which, to my mind, aim to make this point. However, here is a telling handful of them.
Perry-1: 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. With each trip around the counter, the trail became thicker. But I seemed unable to catch up. Finally it dawned on me. I was the shopper I was trying to catch … I believed at the outset that the shopper with a torn sack was making a mess. And I was right. But I didn't believe that I was making a mess. That seems to be something I came to believe. And when I came to believe that, I stopped following the trail around the counter, and rearranged the torn sack in my cart (1979,3).
Perry-2: An amnesiac, Rudolf Lingens, is lost in the Stanford Library. He reads a number of things in the library, including a biography of himself, and a detailed account of the library in which he is lost … He still won't know who he is, and where he is, no matter how much knowledge he piles up, until that moment when he is ready to say, 'This place is aisle five, floor six, of Main Library, Stanford. I am Rudolf Lingens ' (1977, 492).
Lewis: Consider the case of the two gods. They inhabit a certain possible world, and they know exactly which world it is. Therefore, they know every proposition that is true at their world. Insofar as knowledge is a propositional attitude, they are omniscient. Still I can imagine them to suffer ignorance: neither one knows which of the two he is. (…) The trouble might perhaps be that they have an equally perfect view of every part of their world, and hence cannot identify the perspective from which they view it (1979,520).
To repeat, what these samples collectively show, if we focus on the passages in italics, is that homing in a first-person thought (or indeed a demonstrative one as well, like in Perry-2) is the stopping point of one's inquiry. For it resolves, once and for all, one's previous ignorance regarding who was supposed to have the properties ascribed by a given (list of ) predicative concept(s) in the relevant (list of ) proposition(s). If that is right, then no wonder that first-personal, or more generally indexical and genuinely demonstrative contents, have been taken to play a fundamental role in our cognitive lives. For they actually allow us to locate ourselves spatially, temporally, and with respect to other objects, even when we cannot locate ourselves and our spatio-temporal location in objective -that is to say, non-perspectival -ways. 1 Conversely, the passages reported show that, just by itself, the objective individuation of ourselves and our whereabouts is not sufficient to secure that we know which person, place and time we are thinking about.
To see that this guarantee does not hold with respect to one's proper name, even when one knows it is one's own name, and so to see that indexicals are indispensable, it is enough to recall Perry's own words: 1 E.g. As AC, writing this sentence at 10.05 am Italian time november 23 2015, in her study room in her apartment in Bologna, at such-and-such an address.
Perry-3: Suppose I had said, in the manner of de Gaulle, 'I came to believe that John Perry is making a mess': I would no longer have explained why I stopped and looked in my own cart. To explain that I would have to add, 'and I believe that I am John Perry' , bringing in the indexical again (1979,4).
To unpack the reasoning a bit, consider that to know one's name just amounts to knowing 'I am called NN' (or else, 'I am (identical to the person named) NN'). Now, it is only on the background of such knowledge that the sentence à la de Gaulle, which does not overtly contain the first person in the embedded content, can explain why Perry stopped searching for the messy shopper once he realized that John Perry was making a mess. Take that assumption away -namely, take away the assumption 'I am called John Perry' (or 'I am (identical to the person named) NN') -, and so interpret the phrase as having simply to do with the person named 'John Perry' , clearly the inquiry would not be over. For the shopper would have to look for that guy.
Of course, the reply might be that a similar case can be construed with two different proper names for the same person, like 'Superman' and 'Clarke Kent' . Hence, consider: Cappelen-Dever 1: Pushing my cart down the aisle I was looking for CK to tell him he was making a mess. I kept passing by Superman, but couldn't find CK. Finally, I realized, Superman was CK. I believed at the outset that CK was making a mess. And I was right. But I didn't believe that Superman was making a mess. That seems to be something I came to believe. And when I came to believe that, I stopped looking around and I told Superman to clean up after himself (2013, 33).
Still, a moment reflection suffices to show that the stopping point of our subject's search in not really given by her realization that Superman is Clark Kent, but rather, by her realization that that person she kept passing by, presented to her as Superman, was Clark Kent. Had she not realized that, she would have still gone around looking for Superman/Clark Kent, even once in possession of the revealing piece of news that Superman is Clark Kent.
To sum up: the real guarantee is distinctive of I-contents and is not possessed by other concepts, which pick out the same referent. Only I-contents (and more generally indexical and demonstrative ones) can perform the function of stopping points of inquiry. It is this function, I have claimed, theorists who have defended the indispensability of indexicals have been trying to pinpoint with their various examples. They have not simply brought up more Fregean Hesperous-Phosphorous cases, which -it should be agreed by all parties -simply show the opacity of propositional contents. Hence, there is an interesting and distinctive phenomenon to be explained, contrary to what supporters of the dispensability of indexicality and of the de se, in particular, have been arguing for recently.

The real guarantee vs. immunity to error through misidentification
The real guarantee, enjoyed by first-person contents, or indeed, mutatis mutandis, by indexical and genuinely demonstrative ones, has often been conflated with immunity to error through misidentification. Elsewhere (Coliva 2003), I have tried to disentangle these two issues. Here, I will not go into the details, but I will summarize the gist of that discussion to show how the real guarantee and immunity to error through misidentification are different phenomena, which should be carefully kept apart.
Consider a standard case of error through misidentification, involving the first-person. You look at a glass window and form the thought 'My hair is blowing in the wind' . Unbeknownst to you, it is someone else's hair that gets blown by the wind. Whether or not your hair is blowing in the wind, it is not your hair you saw reflected in the glass window. So you are mistaken about whose hair is blowing in the wind. You have mistaken someone else's hair for yours. You are wrong about the following identity: (1) I = the person whose hair I am now seeing reflected in the glass window Notice, however, that even when you mistakenly think that it is your hair that is blowing in the wind, and you are entertaining such a thought because of a mistaken identification like (1), and are therefore guilty of error through misidentification relative to the first person, there is something that, just in virtue of being entertaining that thought, you cannot be mistaken, or even be ignorant about. Namely which person you are thinking about when you entertain that ill-grounded and possibly false thought -that is to say, yourself.
Thus, you can be wrong about 'My hair is blowing in the wind' and be wrong about it because you wrongly believe (1). But in exactly the same scenario, if possessed of the first-personal concept and rational, you cannot wonder: 'My hair is blowing in the wind, but which is the person whose hair is blowing in the wind?' . To repeat, you may be wrong in thinking that the person whose hair is blowing in the wind is you, but once you have that belief you cannot be ignorant with respect to which person you (mistakenly as it happens) think has her hair blown by the wind. To put it otherwise, if you believe 'My hair is blowing in the wind' , no matter how mistaken you are, and even if you are guilty of an error of misidentification, there is no possibility for you to go on sensibly to inquire which person has her hair blown by the wind. Of course, later on, you could be appraised of your mistake, but at that stage, you would revise by judging 'Well then, someone else had her hair blown by the wind. Let us find out who that person was' and not 'Well then, I didn't know whom I was thinking about when I thought my hair was blowing in the wind. Let us find out who that person is' .
Compare with a different scenario, one in which you believe, because you hear a conversation among by-passers, that Elena Cocò's hair is blowing in the wind. Suppose, moreover, that it is the case that Elena Cocò's hair is blowing in the wind and that the people who said so had correctly identified her and had therefore formed their judgment on a correct epistemological basis. (This is something that any theory of propositional content and of testimonial knowledge/justification should allow for. So we can stay clear of all details for present purposes.) Hence, you correctly believe 'Elena Cocò's hair is blowing in the wind' . Your judgment, though not immune to error through misidentification, is actually not affected by such a mistake. Now you can sensibly wonder 'Ok, Elena Coco's hair is blowing in the wind, but which is this person (i.e. the one whose hair is blowing in the wind)?' . Indeed, you may set out to find out who, in the surroundings of those people whose conversation you heard, has her hair blown by the wind and to figure out, if there is more than one person (but this is strictly irrelevant), which one is Elena Cocò. When that particular person whose hair is blowing in the wind tells you 'I am Elena Cocò' (or else, someone else points out to you that that particular person is Elena Cocò), you have reached your stopping point and cannot sensibly pursue your inquiry any longer (provided no lies were told about her name). In that case, your inquiry is stopped by your having reached a demonstrative content 'That person is Elena Cocò' , which allows you to individuate the relevant person.
To sum up: the real guarantee, which is enjoyed only by the first-person concept (and other indexical or demonstrative concepts), allows the latter to play the role of stopping point of inquiry even when the judgment made using it is affected by error through misidentification. The real guarantee and immunity to error through misidentification are therefore two different phenomena; for the former can occur without the latter. Indeed, the former occurs any time there is an I-content, while the latter does not. Hence, the deep phenomenon to be accounted for is the real guarantee, for it is the distinctive characteristic of I-contents (as well as of other indexical and demonstrative contents). In contrast, as we saw with Elena Cocò's case, you may well make a judgment, which is not affected by error through misidentification and still have not reached a stopping point of inquiry. Once again, that contributes toward showing how indexicality is characterized by the real guarantee, not by the contingent absence of an error through misidentification.

Two morals
I think we can draw two morals from these cases. First, that whatever account you wish to give of error and immunity to error through misidentification, it had better be such that it does not end up entailing that when there is error through misidentification relative to the first person, you have somehow lost the capacity of actually thinking about yourself and would no longer be able to know which person you are thinking about. For all comprehending uses of the first person, even those made in the course of a judgment about oneself that is afflicted by error though misidentification, are such that you are indeed thinking about yourself and therefore know exactly which person you are thinking about. This, in my view, has important consequences. For it rules out two early attempts at explaining -allegedly -immunity to error through misidentification, namely Andrea Christofidou's (1995) and Carol Rovane's (1987Rovane's ( , 1993. It also shows why Elizabeth Anscombe, in her celebrated paper 'The First Person' (1975), was not really addressing immunity to error through misidentification, but grappling with the real guarantee. (For the details, see Coliva 2003).
More generally, however, it also shows why immunity to error through misidentification cannot be explained just by adverting to this or that feature of our first-person concept. For that very concept, with all its distinctive features, is used also in first-personal judgements affected by error through misidentification. Hence, immunity to error through misidentification, if it exists at all, and I believe it does (contrary to Cappelen and Dever), cannot be accounted for simply in semantic terms.
In Coliva (2006Coliva ( , 2012, I have tried to hammer into this point and, in particular, to make a case that error and immunity to error through misidentification are to be understood as epistemic phenomena. That is to say, as having to do with the kind of epistemic grounds on which the relevant first-person judgements are based. People may, and indeed do disagree with the details of my positive epistemic explanation, but I hope the foregoing will convince them of the soundness of the general direction I have taken.
Of course, this leaves open another possibility. Namely, that once we try better to understand the sources of the real guarantee, and therefore dwell deeper into the status and provenance of the first-person concept, it may turn out that possessing that concept depends on being able to think about oneself by exploiting a kind of evidence which is also the one which gives rise to first-personal judgments that are immune to error through misidentification. Indeed, to that kind of I-judgements, which, following Shoemaker (1968), I have claimed (Coliva 2006(Coliva , 2012 are logically, and not merely de facto, immune to error through misidentification. For we need to secure that the first-person concept homes-in the right referent, no matter how deviant a subject's overall set up might be; otherwise, we would have simply failed to individuate the first-person concept at all. If that is the case, then we would witness the following intriguing structure. The possession of the first-person concept is grounded in evidence, which allows one to single out the right referent in all possible circumstances. Hence, it is grounded in evidence which, when exploited to produce first-personal judgements leads to I-contents, which are logically immune to error through misidentification. Yet, the first-person concept can be used to produce I-judgments, which may not be logically immune to error through misidentification (either because they are affected by error through misidentification, or are simply de facto so immune). 2 What that would show, in its turn, is that there are non-conceptual yet first-personal representations, which ground our possession of the first-person concept. Hence, indexicality would run quite deep. It would not just make its appearance in thought, let alone in language. Rather, it would already be present in those representations -either in their very content or in their anchoring point -that make it possible for us to acquire indexical concepts in the first place. I will not pursue the details of this account here. I have partly done so in previous work, and I intend to pursue this complex task in the future. 3 Nevertheless, this seems to me the most promising way of developing Frege's celebrated yet cryptic remark: Frege-1: Now everyone is presented to himself in a particular and primitive way, in which he is presented to no-one else. So, when Dr. Lauben thinks that he has been wounded, he will probably take as a basis this primitive way in which he is presented to himself. And only Dr Lauben himself can grasp thoughts determined in this way (1956,298).
The second moral is that if all this is right, we should be careful in our use of philosophical terminology. In particular, it would be important to introduce two different terms and their cognates and to try to stick to them in the interest of clarity and perspicuity. Namely, 'to individuate' and 'to identify' .

2
There will be more about these in the following. 3 See Coliva and Sacchi (2001). For a possible development that I find congenial, I defer the reader to Peacocke (2014).
The individuation of an object, a person, or indeed a place and a time is achieved by indexicals and genuinely demonstrative concepts. They single out particular objects in one's surroundings, or individuals, or specific places and times. They populate our cognitive lives with contextually bounded occurrences of this, that, self, you, here, now 4 (etc.) and, of the particulars so individuated, we may go on to think about as having this or that property. Furthermore, as we saw in the case of 'John Perry' and 'Elena Cocò' , they serve as anchoring stopping points of other devices we may use to think about particulars, like proper names. We can put the point by saying that in virtue of their being stopping points, indexicals and demonstratives allow us to think about particulars directly, by satisfying what is known in the literature on the topic as Russell Principle. 5 Russell Principle: In order to think of an object one must know which object one's thought is about. Now, quite clearly, if we think of demonstratives, the presence in our minds of an instance of such a kind of concept is underwritten by abilities that are more basic. In the case of perceptual demonstratives, for instance, it seems difficult to deny that the relevant concept depends on being able perceptually to discriminate the relevant particular in the perceptual scene. Mutatis mutandis, something similar ought to be the case when indexicals are concerned. Still, being concepts and not more basic representations, demonstrative and indexical concepts can be used or deployed in other contexts, once a subject is capable of making the appropriate adjustments. Frege, I believe, was alluding to that idea in his famous, yet puzzling remark: Frege-2: If someone wants to say the same today as he expressed yesterday using the word 'today' , he must replace this word with 'yesterday' . Although the thought is the same, its verbal expression must be different so that the sense, which would otherwise be affected by the differing times of utterance, is re-adjusted. The case is the same with words like 'here ' and 'there' (1956, 296).
Evans' The Varieties of Reference (1982) and the literature that stemmed from it can be seen as developing that point and as offering, primarily, a theory of particulars' individuation and of the dynamics of indexical and demonstrative thought.
An identification, in contrast, goes beyond the mere individuation of a particular, and establishes an identity between the object presented -that is 4 I use small caps for concepts. 5 See Coliva and Sacchi (2001, chapter 1) for a detailed treatment of russell Principle and its history. It is not my aim to provide an account of concepts canonically expressed by proper names. From a cognitive point of view, they do not work as stopping points unless we are able to anchor them by means of demonstratives or indexicals such as that (person), or he/she/you, etc.
individuated -thus 1 and the object presented -that is, individuated -thus 2 . When things go right, it is the same object thought about in two different ways and therefore the identity is true and informative, even if necessary. That is the case with (i) Hesperus = Phosphorus (ii) Mary Ann Evans = George Eliot When things go wrong, the objects are different; the identity is false and potentially misleading. Error through misidentification, properly understood, can only occur when such an identity judgment goes wrong. Immunity to error through misidentification, in contrast, depends either (A) on the absence of any identification component and is actually due to the fact that the judgment in question depends on the exercise of a merely individuative (that is to say, discriminative ability), which allows us to single out one specific particular among many; or else, (B) on the fact that the identity is a priori and necessarily true.
I think the former is the case when (logically) immune to error through misidentification I-judgements are at stake. I also think there are cases of the second kind, however. They do not arise in the domain of contingent truths, though. Still, the identification between '3' and '√9' fulfills the requirements of immunity to error through misidentification as thought of as depending on the a priori truth of a necessary identity. However, 'I = the thinker of this thought (or of this token of I)' , which is an "identity" some theorists might want to appeal to in order to account for logical immunity to error through misidentification of at least some I-judgements, is not like '3 = √9' . In particular, the latter is informative, while the former is not. While the former correctly individuates the first-person concept, it does not really establish an identity between an object thought of thus 1 and thus 2 . The arithmetical analog of 'I = the thinker of this thought (or of this token of I)' , therefore, is not '3 = √9' , but '3 is the successor of 2' . While the latter is such that anyone who possesses the concept 3 (whether it is itself ultimately analytic or synthetic a priori is something we should not care about in this context) cannot fail to assent to it, it is not enough merely to have the concept 3 in order to see that '3 = √9' . (Of course, also '3 = √9' may turn out to be analytic, for instance if you endorse neo-logicism. Still, it would not figure among the axioms, let alone the definitions of the theory, but only among its theorems. Hence, also for someone who is persuaded that all arithmetical truths are analytic, there is clearly a difference between definitions and axiomatic truths, and those truths that deductively follow from the axioms and/or the definitions of the theory). All this suggests that '3' works like an indexical, at least for us who are accustomed to a decimal system, for it constitutes the stopping point of our inquiry when, for instance, we need to find out the number which satisfies the function 'being the positive square root of 9' . 6 What this shows, once more, is that immunity to error through misidentification, if it exists, cannot be accounted for by appealing to the semantic or functional role of the first-person concept, or to its (Kaplanian) role or its equivalent in thought. The problem is not one of detail. It is not whether it is correct to think of that role as 'the thinker of this thought' or 'the thinker of this token of I' , or whatever have you. The problem is much more basic and fundamental. Whatever (implicitly) defines a concept -in this case the first-person concept -cannot play double duty and provide us with an independent identification -on the right hand-side -of the object presented on the left hand-side. It is not by chance, therefore, that in mathematics we do not use the same symbol for definition and for identity. Similarly, in philosophy of language and mind we should be mindful of the fact that those definiens are, indeed, such and not different ways of presenting an already individuated object -that is, oneself. To be clear to the point of running the risk of sounding trivial, but in the context of the reflection on immunity to error through misidentification repeating the obvious may be important, we should keep in mind that (2) I = df the thinker of this thought (or the thinker of this token of I) is altogether different from (3) I = AC While (2) individuates a concept and tells us, philosophers, where we should start looking for the distinctive pre-conceptual abilities which eventually allow us to think about ourselves in a first-personal way, (3) states an identity, based on an identification, between, if I am not mistaken about my name and my biography, myself thought of as myself and myself thought of as the bearer of that name (if not also through some further descriptive information).
But now, if (2) is not in any decently precise sense an identification and is, once appropriately understood, a definition of the first-person concept, or at the very least a characterization of its conceptual role, it is obvious that it cannot be appealed to in order to explain those cases in which an I-judgment occurs and yet it is (logically) immune to error through misidentification, 6 Obviously, I cannot take up the task of defending this idea here; nor can I develop any further the comparison between immunity to error through misidentification in the domain of arithmetic and in the domain of demonstrative and indexical judgements. I believe, however, this would be an exciting area of research. by saying that in those case the relevant identification is between oneself and the thinker of this thought (or the thinker of this token of I). For there is simply is no such identification. Nor there can possibly be. Eminent theorists, therefore, have conflated identification and definition and have thought of appealing to the latter to explain the former, and in particular immunity to error through misidentification. 7 This, as we have seen, cannot be right. The right account of (logical) immunity to error through misidentification in the empirical domain, therefore, is one that builds on the idea that the relevant self-ascriptions are so immune because they are not based on any identification component at all. The conceptual role of the first-person concept, in contrast, if correctly specified, should account for the fact that that concept enjoys the real guarantee. To repeat, it is such that, just by employing it, one cannot fail to know which person one is referring to by means of it.
Moreover, in order to distinguish logical from merely de facto immunity to error through misidentification, the explanation had better be fine-grained enough to allow for levels at which the relevant identifications may nevertheless be present. After all, we all have the intuition that even if, de facto, 'My legs are bent' when based on proprioceptive feelings is immune to error through misidentification, this is only a contingent truth due to the fact that in this world we receive proprioceptive information just from our bodies. Still, that does not rule out the possibility that in a (metaphysically and epistemically) different possible world things may be otherwise. We therefore need a theory that makes room for that possibility, while acknowledging that this is not what actually happens in this world (luckily). In Coliva (2006Coliva ( , 2012 I have provided the details. Here I am filling out the more general picture which motivates that strategy and does so in a very general way. Hence, also those who are unconvinced about the details of my previous proposals should eventually see why something along those lines had better be right. To repeat a point already made but which is worth insisting upon: it is one thing to allow for cases in which 'My legs are bent' , even when based on proprioceptive information, may involve a mistaken identification of the person whose legs are so bent. This will presumably be due to the presence, in the grounds (generally speaking 8 ) of a mistaken identification component of the form 7 This mistake can be found in earlier work by Peacocke (2008) and it seems to me to motivate Higginbotham's (2010) strategy. 8 In Coliva (2006Coliva ( , 2012 I have argued that we should distinguish between grounds and background presuppositions of a judgment like that. Then, depending on one's overall theoretical preferences, the relevant identification component can figure in the latter and make the judgment affected by error through misidentification relative to its background presuppositions; or else, once given a subject enough information about her current state, that identification would be moved to the grounds she could offer if she were asked to make sense of her (mistaken, as it happens) belief 'My legs are bent' .
(4) I = the person whose legs are responsible for the proprioceptive information/feelings I am now having It is an altogether different thing to hold that, given that mistaken identification, the person who thinks 'My legs are bent' is actually thinking about someone else. The point I have been making about the distinction between the real guarantee and immunity to error through misidentification should make us aware of the fact that this is not really a possibility. Even when there is such a mistaken identification component as part of one's grounds (generally speaking) for one's judgment, one is indeed thinking of oneself and ascribing to oneself a property one does not have. (Actually, one's judgment might be true; yet, if it were, it would be ill grounded for it would be based on someone else's legs being bent).

Some comparisons
Let me briefly compare this account with some claims advanced in the literature on immunity to error through misidentification and the first person.
First of all, let us very briefly compare my distinction between immunity to error through misidentification and the real guarantee with Pryor's (1999) suggestion that the interesting kind of immunity is what he dubs 'wh-immunity' . I have already said why, in my view (Coliva 2006), wh-error through misidentification is not a case of error through misidentification at all. The idea is that either it is simply a case of erroneous predication, or else it is a case of the familiar kind of error through misidentification, which involves a mistaken identification component, one containing a rigid definite description as one of the two concepts flagging on opposite sides of the identity. To see why, let us go back to Pryor's example concerning a smelly skunk. Either, after smelling a skunky smell, the subject wrongly ascribes to a visually presented animal the property of being a skunk, when in fact it isn't one; or else, she wrongly believes that that animal is a skunk because of a mistaken identification component as part of her (perhaps implicit) grounds for her judgment. Namely, (5) That animal = the animal which is (actually) responsible for the skunky odor I am smelling Nevertheless, here I should briefly like to say why immunity to wh-misidentification is no immunity to error through misidentification. The main reason is that since, in my view, wh-error through misidentification differs from error through misidentification only if it is actually taken to be a case of mistaken predication, the corresponding kind of immunity should amount to the impossibility of making a predication mistake. Clearly, however, this is not what immunity to error through misidentification (logical or de facto) amounts to. That is to say, even when I self-ascribe an occurrent mental state to myself, based on introspective awareness, I may make a mistake of predication, even though, in my view, I cannot take someone else's mental state to be mine. For, if I am aware of that mental state on that basis -that is, introspection 9 -it is mine, even if it can (metaphysically) originate in someone else's brain activity. For instance, if I have a terrible toothache, it hurts me; I try to get rid of it one way or the other, etc. Never mind if it originates in someone else's brain activity. That is to say, I may not be its author, or its causal origin, but it is still mine. Indeed, it affects me, it moves me to act in certain ways, and there is something it is like to have it, just like when it originates in the normal way from my brain. 10 However, even in the normal run of cases, I may go wrong in identifying the kind of mental state I am having. After all, it may be a gums' ache, rather than a toothache.
Still, maybe what Pryor is really interested in, and calls wh-immunity to error through misidentification, is in fact what I call 'the real guarantee' . If so, however, there is no possibility of failure and hence of wh-error through misidentification, even when one takes someone else's hair to be blowing in the wind, or even when one takes someone else's legs to be bent on the basis of one's proprioceptive feelings. That is to say, even in those cases I would know whom I thinking about as having her hair blown by the wind, or as having bent legs. Hence, wh-error and wh-immunity to error through misidentification cannot be used, as Pryor does, to reconcile Evans and Shoemaker by saying that even if self-ascriptions based on proprioception do not involve identification components, and are therefore such as to lead to immunity to error through misidentification of the usual (non-wh) kind, they can they still be affected by wh-error through misidentification. For, 9 Here we can remain neutral on what introspection comes down to. Surely, the etymology of the word does us a disservice, though, for it suggests that there is something to be seen within ourselves. This kind of Cartesian conception of introspection has, however, largely fallen into disrepute nowadays. Virtually all theorists agree that, to the extent that there is an inner epistemology, it is not based on having one's mental state in view. They differ on the positive account of introspection they offer, though. I have extensively dealt with this topic in Coliva (2016). 10 Still, I agree that proprioception does not suffice for making the other person's legs mine. It seems to me that the criteria of bodily identity and mental identity vary considerably. Hence, I can perfectly well have proprioceptive feelings, which are entirely mine, but do not suffice to make their causal origin -that person's legs -mine. In contrast, even if my introspectively available mental states originated in someone else's brain activity, they would be mine, despite the fact I would not be at their causal origin. If we call the person who is at the causal origin of a mental state 'the author' of that mental state and the person who is introspectively aware of it 'the owner' , it may be that a given token mental state is authored by one person and owned by two different people.
if wh-immunity to error through misidentification is identical to the real guarantee, then all I-judgements enjoy it. Recently, Cappelen and Dever (2013, pp. 133-135) have claimed that indexicality is neither necessary nor sufficient for immunity to error through misidentification. Now, given all I have been saying about the difference between the real guarantee and immunity to error through misidentification, I agree that indexicality is not sufficient for immunity to error through misidentification. We have seen plenty of cases in which there is an I-judgment, hence an idexical one, possessed of the real guarantee, and yet such as to be used in judgment on the grounds of a mistaken identification component. What more is needed in order to have immunity is precisely that the epistemic structure sustaining the judgment be free of any identification component. 11 Yet, it seems to me that indexicality is necessary for immunity to error through misidentification. Two are the main considerations marshaled against that view by Cappelen and Dever. First, John Campbell's 'John the Ripper case' .
Cappelen-Dever 2: If Holmes believes that Jack the Ripper is the killer, he may be wrong, for there may be no killer. But he cannot be wrong because he is thinking about the wrong person: he is thinking about the killer, if there is one, or about no one, if there is not. Descriptive names are not indexical, so here we have immunity without indexicality. (p. 134) Yet, descriptive names are either devices for general reference (so Holmes is thinking of whoever satisfies the property of being the killer of the relevant victims), in which case the issue of their being immune to error through misidentification does not even arise, for that minimally requires a singular thought. Or else, they are devices for singular reference. That, however, involves a demonstrative, de re thought about that person, who -rightly or wrongly -is taken to have committed those murders, even if one is not yet able perceptually to single it out among many. In the latter case, then, immunity to error through misidentification would occur, but it would be 11 They object to the sufficiency of identification-freedom for immunity to error through misidentification by arguing that there may be inductively justified I-judgments, free of any identification component and yet liable or indeed affected by error through misidentification (cf. Cappelen and Dever 2013, 136). now, to the extent to which I understand this under-explained possibility, it seems to me that for induction to justify a generalization about oneself, it must trade on the identity of the subject who, at different times, has instantiated a given property. Thus, there is the assumption, in the background presuppositions of one's judgment, that the person one takes to have instantiated those properties in the past is indeed oneself and not someone else. If that background presupposition is false -that is, if one takes oneself to have instantiated F at t 1 , when in fact, at t 1 , someone else instantiated it and one wrongly took that person to be identical to oneself -then the inductive generalization may be false, or at least ill grounded, and that would be due to a mistaken identification component in its background presuppositions. due to the occurrence of a demonstrative thought (albeit not of a perceptual kind). 12 Hence, the case of 'John the Ripper' would not show that devices for singular reference other than indexicals and demonstratives can be used in judgements which are immune to error through misidentification.
The second consideration Cappelen and Dever put forward hinges on the idea that immunity to error through misidentification can occur with 'identity-neutral' contents. Maybe proprioception just gives one certain identity-neutral experiences, which however invariably correlate with one's being in the relevant state. In response, it should be noted that the 'identity-committed' character of proprioceptive contents is the one feature virtually all theorists who have been discussing these issues have largely agreed on. Indeed, it seems to be the hallmark of proprioceptive (vs. allo-ceptive) feelings (or information). Hence, if we take that away we would not have proprio-ception. We would merely have perception of what happens to be one's body but is not felt as one's own, even if it is perceived through the operation of some inner-sense mechanism, rather than by means of sight, hearing, touch or smell that are the senses through which we usually perceive other people's bodies.
This point could be sidestepped by saying that it would be little more than terminological. I think this is easier said than done, for, if proprioception were identity-neutral, then subjects could sensibly wonder whose legs they are feeling 'from the inside' , whereas they usually do not sensibly wonder that. Moreover, even if it is conceivable they could wonder about that, they would be under the impression of feeling their legs. In fact, as I have maintained in Coliva (2006Coliva ( , 2012, in that case the identification component (4) would not license the passage from an identity-neutral to an identity-committed content, but from the ascription of the property of seeming or feeling one's legs as bent to the ascription of the property that one's legs are bent. Still, for the sake of argument, let us allow for the possibility that proprioception is in fact identity-neutral and let us consider the following example by Cappelen and Dever (2013, 135): Cappelen-Dever 3: Call Gareth's two legs 'LL' and 'RL' . (One) way of characterizing the situation is by saying that Gareth knows that LL is on top of RL, and cannot be wrong about which two legs he is thinking about.
The idea they wish to put forward by means of that example is that it is a 'theoretical choice to describe proprioception as delivering self beliefs, rather than LL, RL, etc. beliefs' (2013, 135). As I have tried to show, it is not 12 The fact that one would not be able to single him out perceptually may cast doubt on the fact that descriptive proper names actually work as devices for singular reference. That would just leave us with the first option discussed in the main text, which poses no problem for the necessity claim. so much a theoretical choice as almost an analytic claim, held by a lot of theorists who have been working on proprioception, memory, and, more generally, on those sources of information (or of information retention) that work from the inside of a subject. Let us grant, however, the theoreticity of the choice for the sake of argument. Would that show that immunity can be had without indexical or demonstrative contents? Obviously not. For surely proprioceptive contents do not come with names of one's bodily parts. So, the kind of proprioceptive contents Gareth has are that leg 1 (for LL) e that leg 2 (for RL). 13 And now, since it is in the nature of the case Cappelen and Dever are conceiving that proprioception, though not containing I-contents, necessarily gives one correct information about one's body (otherwise they would not even seem to have a case against the fact that immunity involves indexicality), all we are left with is the idea that in such a case there is a singular, individuative judgment, concerning that leg 1 and containing a complex predicative component in which it figures a demonstrative concept -that is, being on top of that leg 2 . 14 The conclusion, therefore, is the following conditional one. If one is prepared to allow that proprioception may not deliver first-person contents, immunity to error through misidentification is nevertheless tight to demonstrative contents. In particular, it arises whenever there is no identification but merely an individuation of a particular (let it be a physical object, a place or a moment in time), through a demonstrative content, which then gives rise to a judgment containing a demonstrative concept.
Thus, indexicals and demonstratives are necessary for immunity to error through misidentification, even though they are not sufficient for it, as there can be indexical or demonstrative judgements which are not the result simply of an individuation but of a (possibly mistaken) identification, occurring in the grounds (generally speaking) of the judgment itself.

Conclusions
In this paper, we have seen that what makes indexical and demonstrative contents special is a distinctive cognitive function they alone perform. Namely, the function of being stopping points of inquiry. We have focused on I-contents and have claimed that such a function depends on the fact that they enjoy what elsewhere I have dubbed 'the real guarantee' . Namely, the 13 I use the underscored notation to indicate proprioceptive contents. 14 That seems to me the most charitable reading of their example. For if the judgment were based on identifying that leg 1 with LL and that leg 2 with rL, there would be room for error through misidentification. guarantee of knowing, just by having I-thoughts, which person one is thinking about. We have then seen that that guarantee, which is enjoyed by all I-judgements, should not be conflated with immunity to error through misidentification, which only some I-judgements partake. Error and immunity to error through misidentification are epistemic phenomena, that depend on the grounds on which a given I-judgment is based, and, in particular on whether they contain or lack any identification component. After comparing the real guarantee with Pryor's notion of wh-immunity to error through misidentification, we have seen that while Cappelen and Dever are right in holding that indexicality, in particular I-indexicality, does not suffice for immunity to error through misidentification, it is necessary for it, contrary to what they maintain. Or, at the very least, in order to have immunity we need contents which, like indexical ones, can play the role of stopping points of inquiry and those contents, besides first-personal (and other indexical ones), are only demonstrative contents. When used in merely individuative judgements, then, all indexical and demonstrative concepts guarantee immunity to error through misidentification, precisely because the relevant judgements do not depend on any identification component. It remains to be fully investigated what makes the possession of those concepts possible in the first place. The intriguing suggestion, to be left for future development, is that they depend on more primitive representational contents, which single out their referents in such a way that one cannot fail to know which person, object, place or time one is thinking about. In other words, what remains to be investigated is how these more primitive representations can secure the real guarantee possessed by those concepts. It is the brief of this paper that such a guarantee is what distinguishes indexical and demonstrative concepts from all other kinds of singular concepts and that it is a much deeper phenomenon than immunity to error through misidentification, which deserves to be scrutinized by any serious work on these concepts, and in particular on the first person.