In Bermudez's impressive and ambitious work, an incongruity arises between genuinely first-person nonconceptual contents derived from somatic proprioception (construed in his preferred fashion) and the requirement that those contents be immune to error through misidentification, a core condition on genuine self-consciousness which he accepts. The tension between these two commitments places pressure upon the Priority Principle, seemingly forcing a choice between abandoning it or the condition that nonconceptual content be immune to error through misidentification.
2. The account is not without its difficulties, however. Chapter 6, perhaps one of the most important chapters in the book, harbors a tension worth exploring. In this chapter, Bermudez attempts to show how somatic proprioception funds nonconceptual first-person contents that meet the two core conditions on genuine self-conscious thought as he (and many others) sees them. First, self- conscious thoughts cannot be about oneself nonaccidentally -- that is, they cannot be about oneself without one recognizing that to be the case. As a second condition, such thoughts must have immediate implications for action (pp. 147-148) [NOTE 1]. Here, I will be concerned with the requirement that self-conscious thoughts be nonaccidental -- cashed out as immunity to error through misidentification relative to the first-person pronoun -- and the way in which it seems to stand at odds with nonconceptual contents derived from somatic proprioception. In the end, we will see that Bermudez faces a dilemma: he must abandon either the strong version of the Priority Principle or the condition that autonomous nonconceptual first-person contents must be immune to error through misidentification.
3. Early on in the first chapter, we are introduced to a distinction between types of first-person content derived initially from Wittgenstein's The Blue Book (1958, pp. 66-68). Wittgenstein distinguishes between what he calls 'I' used as subject and 'I' used as object. The latter, he claims, permits the possibility of misidentifying the referent of the first-person pronoun, whereas the former does not [NOTE 2]. When uttering 'I am in pain' -- the canonical instance of 'I' used as subject -- Wittgenstein offers that the identification of the speaker is not in question. That is, I cannot ascribe a felt pain to someone that by accident turns out to be myself. In a genuine self- conscious ascription of a property, it is no accident that I recognize that I am the subject of the ascription, for it could not be otherwise. To use Wittgenstein's memorable phrase: "The man who cries out with pain, or says that he has pain, doesn't choose the mouth which says it" (Wittgenstein 1958, p. 68).
4. Sydney Shoemaker has done much work to elucidate and to extend this condition, labeling it with the now standard terminology "immunity to error through misidentification relative to the first-person pronoun" (Shoemaker 1968, p. 7). For Shoemaker, as for Wittgenstein, a certain class of judgments permit error in the predicate position but do not leave the identity of the subject of the predication in question, for knowing in a particular way that a property is instantiated simply obviates any need for identifying its source. Bermudez rightly points out, as Evans did before him, that these contents are immune to error through misidentification in virtue of the "evidence base from which they are derived, or the information on which they are based" (p. 6) [NOTE 3]. Fundamentally, Bermudez -- like nearly all other participants in this dialectic -- accepts that contents cannot be considered genuinely self- conscious unless they possess this type of immunity [NOTE 4].
5. We revisit immunity to error through misidentification in chapter 6 where Bermudez attempts to show that autonomous nonconceptual first- person contents based upon somatic proprioception are indeed immune to this sort of error. By this time we have been introduced to an account of representational content that includes as a criterion the possibility of misrepresentation, for with the potential for representation comes the potential for misrepresentation (p. 94). However, a conflict arises for Bermudez when the two criteria are to be concurrently satisfied. For somatic proprioception to be a source of genuine self-conscious content, it must serve as an evidence base that does not leave its subject in doubt, even for creatures lacking any conceptual resources whatsoever. Yet to qualify as representational contents -- that is, to be considered contents at all -- thoughts funded by proprioception must allow for the possibility of misrepresentation. Misidentification is but a special case of misrepresentation, and hence endorsing immunity to error through misidentification apparently precludes misrepresentation, which apparently serves to disqualify proprioceptive states from being representational [NOTE 5].
6. Unlike those who discuss immunity to misidentification as it relates to judgments, it is not at all clear that Bermudez has the philosophical machinery to relieve this tension. Evans, for example, does not fall into a similar predicament, for his 'I'-thoughts possess a conceptual structure that localizes -- as Shoemaker's condition in its long form indicates -- the immunity to error through misidentification relative to the first person pronoun. Misrepresentation can still occur with regard to the predicate position and the ascription of bodily properties, and hence immunity to misidentification and misrepresentation can co-exist in the same thought or judgment. Non-language-using creatures, in contrast, do not have the first- person pronoun at their disposal. Without conceptually structured thoughts, it seems that these types of subjects cannot possess contents that are immune to error through misidentification, for they have nothing that that immunity could be relative to.
7. Or do they? Bermudez argues that inference to the best explanation warrants ascribing "protobeliefs", or belief analogs, to non-language-using creatures requiring intentional explanations to account for their behavior (pp. 117-119). As Bermudez presents them, perceptual protobeliefs [NOTE 6] are nearly as rich as their conceptual correlates: they can embody "nonextensional modes of presentation" in terms of Gibsonian affordances (p. 121), and they are somewhat compositional, though they do not allow for "global recombinability", failing to meet Evans's Generality Constraint (Bermudez 1998, p. 93; Evans 1982, pp. 100-105). So structured, perceptual protobeliefs support primitive inference and the limited generation of further new nonconceptual contents from a set of others. Accordingly, perceptual protobeliefs so construed (including contents based on somatic proprioception) seem capable of supporting something like a discrete subject component, analogous to an 'I'-idea, that localizes immunity to error through misidentification, as well as a predicative component that could misrepresent a state of the world or body.
8. One certainly becomes puzzled at this point. If nonconceptual contents based upon somatic proprioception can support both a component immune to misidentification and a component preserving the possibility of misrepresentation, then what are we to make of the motivation for maintaining a clear conceptual/nonconceptual distinction with regard to contents? Indeed, it seems that inference to the best explanation warrants thinking of the constituents of protobeliefs as "protoconcepts". Much like concepts [NOTE 7], protoconcepts could be defined in terms of their inferential role, where a protoconcept's inferential role can be cashed out in terms of the protopropositions or protobeliefs in which it features. As the analogy deepens between concepts and protoconcepts, we seem to have less and less reason to conclude that creatures lacking language likewise lack conceptual abilities of any sort, however limited or nascent. After all, the set of protopropositions may be quite limited for non-language using creatures, but they nevertheless succeed in satisfying two subtle and sophisticated philosophical criteria. Perhaps that success itself provides compelling evidence of concept possession.
9. Bermudez would no doubt resist this approach since it seems to run afoul of the Priority Principle: The Priority Principle: Conceptual abilities are constitutively linked with linguistic abilities in such a way that conceptual abilities cannot be possessed by nonlinguistic creatures (p. 42). Priority was initially important because it "allows us to make a very clear distinction between conceptual and nonconceptual modes of content- bearing representation" (p. 43), and hence provides us with a means of evading the eponymous paradox. Yet, given that protobeliefs are in some measure compositional and fund limited inference -- indeed are constituted by protoconcepts -- it is no longer clear how we can maintain a very clear distinction between conceptual and nonconceptual contents. Perhaps inference to the best explanation is just not fine-grained enough to support such a sharp distinction.
10. Bermudez would no doubt further rejoin that the protoconcept/ concept analogy runs fairly shallow, for even if non-language-using creatures possessed a range of protoconcepts defined in terms of protoconceptual roles, they do not have an explicit grasp of these roles. Such creatures are merely sensitive to the truth of inferential transitions (p. 70). He writes: Certainly, it is possible to be justified (or warranted) in making a certain inferential transition without being able to provide a justification (or warrant) for that inferential transition. It is a familiar epistemological point, after all, that there is a difference between being justified in holding a belief and justifying that belief. What does not seem to be true is that one can be justified in making an inferential transition even if one is not capable of providing any justifications at all for any inferential transitions. But providing justifications is a paradigmatically linguistic activity. Providing justifications is a matter of identifying and articulating the reasons for a given classification, inference, or judgment. It is because prelinguistic creatures are in principle incapable of providing such justifications that the priority thesis is true. Mere sensitivity to the truth of inferential transitions involving a given concept is not enough for possession of that concept. Rational sensitivity is required, and rational sensitivity comes only with language mastery (p. 71). For Bermudez, then, concept possession is a fairly advanced skill based upon an ability to identify and to provide reasons for beliefs, and limited inferential ability -- even an ability to make inferences that one is justified in making -- does not indicate concept possession.
11. This, however, seems a bit too stringent. Being able to give reasons as reasons is a function of possessing the concepts of justification, belief, reason, among others. Imposing the further requirement that one recognize that one is in fact giving reasons may disqualify attributing conceptual abilities where we normally would be comfortable doing so. To take an example ready to hand (p. 70), the children in Susan Carey's experiments who concluded that a worm was more likely to have a spleen than a toy mechanical monkey are probably not in position to identify their reasons as reasons and to answer a call to justify their inferences. Still, even Bermudez wants to credit these 4- year olds with the possession of the concepts human being, living animal, internal organs, and the inferential relations between them [NOTE 8].
12. It seems that maintaining that proprioceptive nonconceptual contents be immune to error through misidentification entails that the Priority Principle must at least be weakened if not abandoned altogether. Perhaps we can spare the Priority Principle in its strongest form by abandoning the requirement that nonconceptual proprioceptive contents be immune to error through misidentification. That is, we accept that protobeliefs are only minimally structured, and no robust subject concept or protoconcept is available to bear the weight of an immunity claim. It's not clear to me that we sacrifice much explanatory power in making this move, since we can still hold firmly to the second core condition for genuine self-conscious thought -- namely, that nonconceptual proprioceptive contents must have immediate implications for action, which in fact they do (Bermudez 1998, 148). Moreover, preserving this latter condition at the expense of immunity does not cast us back into the eponymous paradox, for armed with this remaining condition we still have a means of determining the class of nonconceptual contents that are a form of genuine primitive self-consciousness. Although this is the path I would advise, most theorists will not be anxious to forgo any commitment to immunity to misidentification. In any event, this is certainly a position in logical space that deserves attention [NOTE 9].
13. Whether one elects to weaken the Priority Principle or to relinquish the condition that nonconceptual contents must be immune to error through misidentification, much work needs to be done to further elucidate the core conditions on the possession of self-conscious thought as well as the relationship between conceptual and nonconceptual content. These are not insignificant projects, to say the least. Nevertheless, problems of the sort I have raised here are perennial in this field and do not diminish the quality of Bermudez's work. His attempts to make tractable this most intractable of puzzles has furthered our understanding of how it is that our thoughts come to be about ourselves. This rich book will no doubt become required reading for all those captivated by the mysterious nature of self-consciousness, as well as for those interested in integrating experimental work with more traditional philosophical enterprises.
 This second condition is also known as the action component (Evans 1982, Chapter 7), and has been discussed most prominently by Castaeda 1966 and Perry 1979. I will have little to say about it here.
 Indeed, Wittgenstein claims that 'I' in cases of its use as subject is not a referring expression at all. This position is endorsed and artfully defended by Anscombe 1975.
 Cf. Evans 1982, Chapter 7. Bermudez also argues, persuasively I think, that Shoemaker's elucidation of immunity to error through misidentification should be stated in terms of justification as opposed to knowledge. For if one can still be mistaken about the instantiation of a predicate -- even if one cannot be mistaken about the first-person identification in that case -- that belief cannot be considered knowledge.
 This condition on self-conscious thought or belief is almost universally accepted in one form or another (that is, I have yet to find a doubter besides myself). See, to list but a few of those not mentioned thus far, Anscombe 1975, Strawson 1994, Cassam 1997, Prior 1999.
 This argument is a species of one Shoemaker has presented fairly persuasively and at length (Shoemaker 1986, 1994) that perception and introspection are at odds because the former permits misidentification while the latter does not.
 Bermudez also briefly discusses instrumental protobeliefs (p. 118), but our discussion can safely ignore them.
 I draw this analogy tentatively, fully aware of the current controversies surrounding a general account of concepts, or the lack thereof.
 In my experience, most college students have quite a bit to learn about providing justifications for their inferences, let alone 4-year olds.
 Although this is not the proper point to launch into an extensive discussion of immunity to error through misidentification, a survey of the literature shows that most of its adherents merely assume its truth to be patently obvious. Even Shoemaker, one of the most careful philosophers, more or less adopts Wittgenstein's distinction tout court. Shoemaker does hazard an account based on a certain class of predicates -- namely P* predicates -- that "can be known to be instantiated in such a way that knowing it to be instantiated in that way is equivalent to knowing it to be instantiated in oneself" (Shoemaker 1968, p. 16). Still, we are given no means of determining these predicates or the way in which we must know them to be instantiated. We are in need of much stronger arguments to promote this condition to a core requirement for a given thought to be self-conscious. This is all said in an effort to show that although the advice in the text to abandon immunity to misidentification may seem radical, it is in fact much less so.
Anscombe, G. E. M. (1975). "The First Person," in S. Guttenplan ed., (1975). Mind and Language. Oxford. Clarendon Press.
Bermudez, J. L. (1998). The Paradox of Self-Consciousness. MIT Press.
Bermudez, J. L. (1999). Precis of The Paradox of Self-Consciousness. Psycoloquy 10(35). ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1999.volume.10/psyc.9 9.10.035.self-consciousness.1.bermudez http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?10.035
Cassam, Q. (1997). Self and World. Oxford. Clarendon Press.
Castaeda, H.-N. (1966). "'He': A Study in the Logic of Self- Consciousness," Ratio 8, 130-157.
Evans, G. (1982). The Varieties of Reference. Oxford. Clarendon Press.
Perry, J. (1979). "The Problem of the Essential Indexical," Nous 13, 3-21.
Prior, J. (1999). "Immunity to Error Through Misidentification," Philosophical Topics 26, 1999.
Shoemaker, S. (1968). "Self-Reference and Self-Awareness," The Journal of Philosophy 65, 555-567.
Shoemaker, S. (1986). "Introspection and the Self," reprinted in Shoemaker 1996, 3-24.
Shoemaker, S. (1994). "Self-Knowledge and 'Inner Sense'," Philosophy and Phenomenological Research 54, 249-314.
Shoemaker, S. (1996). The First-Person Perspective and Other Essays. Cambridge University Press.
Strawson, P. F. (1994). "The First Person -- and Others," in Q. Cassam ed., (1994). Self-Knowledge. Oxford University Press, 210-215.
Wittgenstein, L. (1958). The Blue and Brown Books. Oxford. Basil Blackwell.