The eponymous paradox of Bermudez (1998, 1999) suggests that self-consciousness lies beyond the bounds of psychological explanation. Bermudez contends that the only way to avert this unlovely result is by invoking the notion of nonconceptual thought. This contention is rooted in adherence to the Priority Principle, which dictates that conceptual capacities are essentially off-limits to pre- and nonlinguistic creatures. I argue that the latter claim can and should be resisted, and that the paradox of self-consciousness can be equally well (if not better) disarmed by denying the dependence of first-person conceptual thought on first-person language.
P1. The only way to explain self-consciousness is by explaining the capacity to think first-person thoughts.
P2. The only way to explain the capacity to think a certain type of thought is by explaining the capacity for the canonical linguistic expression of thoughts of that type.
P3. First-person thoughts are canonically expressed by means of the first-person pronoun. [NOTE 1]
P4. The capacity to think first-person thoughts is an essential ingredient of competence with the first-person pronoun.
P5. If P1-P4 are true, then self-consciousness cannot be explained in a noncircular way.
P6. Noncircularity is a necessary condition on genuine explanation.
C. Self-consciousness cannot be explained.
The argument is plainly valid, and each of the six premises is plausible on its face. Hence the paradox. [NOTE 2]
2. The pivotal step in the paradoxical reasoning is, of course, the second one. It consists in supposing that the only explanatory access we have to thoughts of a given type is via the capacity for the canonical linguistic expression of thoughts of that type. As Bermudez notes, this claim has some appeal, but it hardly seems irresistible. An obvious way to contest it would be to argue that, at least with respect to certain types of thought, the capacity to think thoughts of those types is dissociable from the capacity to express those thoughts linguistically. For if such dissociability could be made plausible, it would seem odd indeed to insist that the latter capacity must be explanatorily prior to the former. What possible grounds could there be for this insistence?
3. As Bermudez observes (pp. 41-42), however, skepticism about this strategy for resolving the paradox can be induced on the basis of the following considerations:
Conceptual Requirement Principle [CRP]. The range of thoughts attributable to a subject is exhaustively determined by the range of concepts in that subjects possession.
Priority Principle[PP]. Conceptual abilities are constitutively linked with linguistic abilities in such a way that conceptual abilities cannot be possessed by pre- or nonlinguistic subjects.
These two principles jointly entail the impossibility of thought without language. So if P2 is to be given up on the grounds suggested, then at least one of these will have to be given up as well. Trouble is, its not clear which -- if either -- we can do without.
4. On this score, Bermudez makes two points: first, that the paradox of self-consciousness can be resolved by rejecting the Conceptual Requirement Principle (CRP); second, that this is the only method of resolution available, given the options at hand (p. 272). Apropos of the first point, he argues that there exists a type of first-person thought which is nonconceptual in character, hence available to subjects who lack mastery of the first-person pronoun. The second point rests upon an argument which purports to show that the Priority Principle (PP) is non-negotiable -- thereby rendering the denial of that principle, as opposed to the denial of CRP, a closed option.
5. For present purposes I shall remain agnostic on the first point: that is, I will leave open the possibility of a nonconceptualist resolution of the paradox. But as far as the purported necessity of such a resolution is concerned, my approach will be less ecumenical. In particular, I will argue that Bermudez's effort to justify PP does not succeed, and that there are independent reasons to reject the principle. I'll close by suggesting that, other things being equal, a strategy which preserves CRP is preferable to a strategy of the kind defended by Bermudez, which gives it up.
6. Bermudez defends the Priority Principle by appealing to a broad constraint on the nature of psychological explanation. His strategy is to argue on a priori epistemological grounds that concept-involving modes of explanation do not legitimately apply to the behavior of creatures without language, insofar as languageless creatures lack the minimum epistemic qualifications for concept possession and use (pp. 69-71). Here is how the argument goes:
P1. Explanations of behavior which advert to a subject's conceptual abilities are acceptable only when the subject can employ the relevant concepts in rational judgment.
P2. The capacity for rational judgment requires both (a) the capacity to form warranted beliefs and (b) the capacity to identify warrants for those beliefs.
P3. The capacity to identify warrants for one's beliefs requires linguistic competence.
C. Concept-involving modes of explanation do not apply to pre- or nonlinguistic creatures.
Though the reasoning above is valid, its soundness can be questioned at various points. I'll indicate three of them.
7. First, it is far from clear that one cannot be warranted in one's beliefs unless one can supply some of those beliefs with a warrant. On a reliabilist view of justification, for example, whether or not a belief is justified may depend solely on its causal ancestry, viz. whether it was produced by a reliable, or truth-tracking, psychological process. [NOTE 3] There is no reason to suppose that the believer must be able to identify that ancestry in general. Why insist on this ability in so much as a single case? At this point Bermudez simply appeals to epistemological intuition (p. 71). But intuitions of this sort are notoriously unstable, and the particular intuition in question is weak at best. What's needed is further argument, and none is on offer.
8. Second, it is important to distinguish between the ability to justify a belief -- that is, to have beliefs with justificatory content -- and the ability to communicate those beliefs linguistically. Failing to observe this distinction simply begs the question as to whether thought is dissociable from language. Yet this is just what Bermudez does when he writes that "[p]roviding justifications is a matter of identifying and articulating the reasons for a given classification, inference, or judgement. It is because prelinguistic creatures are in principle incapable of providing such justifications that the priority principle is true" (p. 71). The core claim here is that "identifying and articulating" the justificatory basis of a judgment is an essentially linguistic activity. But terms like 'identify' and 'articulate' are potentially equivocal: they admit what I shall call, following Sperber and Wilson (1995), both cognitive and communicative readings. Cognitively identifying or articulating a justification is something done in thought, the communicative counterpart of which is something done in language. Of course, it might be argued that cognitive justification is parasitic on communicative justification; but once again all we get from Bermudez is assertion.
9. Third, a methodological objection. It seems clear that the question as to whether thought is possible without language is largely an empirical one. So we should expect it to be settled largely on a posteriori grounds, not from the armchair. A priori reasoning may play a role here, but probably no more than a peripheral one. That Bermudez lays such an emphasis on intuitions about the nature of justification (a decidedly elusive topic) is somewhat surprising, given his repeated insistence on the naturalistic slant of his inquiry. [NOTE 4] After all, the very idea of naturalism in philosophy militates against this way of proceeding.
10. Now on to grounds for skepticism about priority. Evidence for the dissociability of conceptual capacities from linguistic ones comes from studies of the ability to recognize oneself in a mirror. In the 'mark test' paradigm first described by Gallup (1970), a subject whose forehead has been surreptitiously marked with dye is situated in front of a mirror and her actions observed. If the subject reaches up to examine the mark or engages in other self-exploratory behaviors using the reflected image as a guide, it seems plausible to suppose that self-recognition has occurred. For it is natural to explain these behaviors by saying that the subject has identified herself, in a first-person manner, as the individual reflected in the mirror. It is also natural to suppose that this sort of identificatory capacity has a conceptual basis -- in particular, that it implicates a subject's self concept (however rudimentary).
11. The relevant empirical facts are these. First, developmentally normal humans recognize themselves in mirrors as early as 15 months of age, but they do not begin to use personal pronouns before 24 months (Lewis and Brooks-Gunn (1979)). Second, and more decisively, at least two non-language-using animal species, viz. chimpanzees and orangutans, recognize themselves in mirrors (Povinelli and Prince (1998)). In the face of these facts, advocates of priority need to provide a nonconceptualist account of mirror self-recognition -- and how such an account would go is anyone's guess. [NOTE 5] Certainly the dominant interpretation of the data is a conceptualist one.
12. This sort of worry ought to be especially pressing for a pro-priority theorist like Bermudez, given certain of his other theoretical commitments. One reason for this emerges from his discussion of visual agnosias, and in particular from his remarks about the taxonomic divide between apperceptive and associative agnosias. Bermudez glosses this in terms of a distinction between deficits of object perception (apperceptive) and deficits of object recognition (associative), noting that the latter "appear to be more cognitive than perceptual" (p. 80). He then goes on to compare the visual world of the pre-verbal infant with that of an "idealized global associative agnosic -- a patient whose representations of the world are devoid of semantic features," adding that the components of the visual system which are damaged in associative agnosics are "best viewed as generating representations of the world at the conceptual level" (p. 81). Finally, in a tacit allusion to the Priority Principle, he notes that associative agnosias are correlated with lesions in language-specialized areas of the brain.
13. These remarks make it clear that, in Bermudez's view, object recognition is a concept-involving capacity. But the capacity to recognize oneself -- in a mirror or otherwise -- is surely just a special case of that more general capacity. [NOTE 6] So if Bermudez is right to insist on the essentially conceptual character of object recognition, then an adequate nonconceptualist explanation of mirror self-recognition is probably not in the cards.
14. As noted above, Bermudez insists that the only plausible way to resolve the paradox of self-consciousness is by rejecting CRP -- in particular, by positing a type of first-person thought which is not composed of concepts. The main reason for this insistence lies with his commitment to PP, which precludes the possibility of a subject's possessing a self concept without being able to employ the first-person pronoun. In reply I've argued that such a commitment is dubious, and that the paradox can also be resolved in a manner consistent with the idea that thoughts invariably exhibit conceptual structure. As a result, Bermudez's claim that jettisoning CRP is the only way to escape the paradox cannot be sustained.
15. Now I want to suggest something stronger, namely, that given the choice between an escape strategy which gives up CRP (like Bermudez's) and one which retains this principle, we do better to opt for the latter. Here's why. Both strategies suppose that some thoughts are built up from concepts; the question is whether all thoughts are like this. Pro-CRP strategists say yes, whereas anti-CRP strategists say no. It is important to note the ontological implications of rejecting this principle. For in saying that some thoughts are nonconceptual, one commits oneself not just to the idea that there are more thoughts than previously supposed, but that there are more kinds or types of thought. This becomes clear from the fact that conceptual and nonconceptual thoughts have fundamentally different properties. For example, it's almost universally agreed that conceptual thought is both productive (since there is no finite upper bound on the number of distinct thoughts a thinker can potentially host) and systematic (since the capacity to host certain thoughts entails the capacity to host other thoughts with a similar structure). By contrast, nonconceptual thought need not have either of these properties (pp. 92-93). This is good prima facie evidence of a strong taxonomic or typological distinction between conceptual thought, on the one hand, and nonconceptual thought, on the other.
16. Once such a distinction is admitted, however, the anti-CRP position can be seen to suffer on methodological grounds. For the principle of parsimony (a.k.a. Ockham's Razor) dictates that types of entity are not to be multiplied beyond explanatory necessity. Ceteris paribus, then, the theory with the leaner ontology is to be preferred. Pending evidence of psychological phenomena which cannot be explained save by adverting to nonconceptual thoughts -- or other compelling reasons to give up CRP (e.g., a sturdier argument for PP) -- the principle should be retained. In the meantime we have no good reason to believe in nonconceptual first-person thought, and good reasons not to.
Thanks to Murat Aydede, Sara Bernal, Shaun Nichols, and Sven Rosenkranz for advice, comments, and encouragement.
 Or a functionally analogous linguistic device. So-called pro-drop languages, such as Spanish and Italian, allow for subject pronouns to be unexpressed. In such languages the verbal inflection tends to be rich; even so, personal pronouns exist and are used, both nominatively and otherwise.
 The term 'paradox' does not really fit here. A paradox is an argument whose soundness cannot be denied without doing real violence to intuition, and the argument under consideration does not meet that condition. A good example of an argument which does is the sorites:
P1. Someone with 0 hairs is bald. P2. If someone with n hairs is bald, then so is someone with n+1 hairs. C. Everyone is bald.
Measured by the soritical standard, Bermudez's use of the term 'paradox' in the present setting seems like false advertising ('reductio' would be more apt). I adopt it here mostly for the sake of expository convenience.
 See, for example, Goldman (1979).
 For programmatic remarks to this effect, see Bermudez (1998), p. xii and pp. 47-48.
 It may be no accident that Bermudez never addresses, or even mentions, mirror self-recognition -- despite ubiquitous discussion of the topic in the psychological literature. Even a cursory glance at this literature reveals how remarkable an omission this is, especially in philosophical work as avowedly naturalistic as Bermudez's. Almost every developmentalist writing about self-consciousness, it seems, has something to say about mirror self-recognition; see e.g. Butterworth (1990), Kagan (1998), Meltzoff (1990), and Neisser (1993), among many others. The situation in primatology is the same; see references in Povinelli and Prince (1998).
 Note that prosopagnosia, or impaired face recognition, is standardly classified as an associative agnosia. See Pallis (1955) for a case study of a prosopagnosic who regularly mistook his reflection in a mirror for that of a stranger.
Bermudez, J. L. (1998). The Paradox of Self-Consciousness. MIT Press.
Bermudez, J. L. (1999) Precis of "The Paradox of Self-Consciousness". PSYCOLOQUY 10(035) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1999.volume.10/ psyc.99.10.035.self-consciousness.1.bermudez http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?10.035
Butterworth, G. (1990). Self-perception in infancy. In Cicchetti, D. and Beeghly, M., editors, The Self in Transition: Infancy to Childhood. University of Chicago Press.
Gallup, G. G. (1970). Chimpanzees: Self-recognition. Science, 167.
Goldman, A. I. (1979). What is justified belief? In Pappas, G., editor, Justification and Knowledge. Reidel.
Kagan, J. (1998). Is there a self in infancy? In Ferrari, M. and Sternberg, R. J., editors, Self-Awareness: Its Nature and Development. Guilford Press.
Lewis, M. and Brooks-Gunn, J. (1979). Social Cognition and the Acquisition of Self. Plenum Press.
Meltzoff, A. N. (1990). Foundations for developing a concept of self. In Cicchetti, D. and Beeghly, M., editors, The Self in Transition: Infancy to Childhood. University of Chicago Press.
Neisser, U. (1993). The self perceived. In Neisser, U., editor, The Perceived Self: Ecological and Interpersonal Sources of Self-Knowledge. Cambridge University Press.
Pallis, C. A. (1955). Impaired identification of faces and places with agnosia for colors. Journal of Neurology, Neurosurgery, and Psychiatry, 18.
Povinelli, D. J. and Prince, C. G. (1998). When self met other. In Ferrari, M. and Sternberg, R. J., editors, Self-Awareness: Its Nature and Development. Guilford Press.
Sperber, D. and Wilson, D. (1995). Relevance: Communication and Cognition. 2nd edition. Blackwell.