We shall focus on three areas covered in Bermudez's (1998, 1999) book: the project of giving an account of "I"-thoughts which meets various desiderata, the account of the initial problem of circularity in trying to explain "I"-thoughts in terms of the public-language "I," and the account of concept possession. We shall raise problems of clarity as well as of substance that we find in these areas.
1. Bermudez's (1998, 1999) ambitious and fascinating book attempts to answer the interrelated questions of what it is for a creature to be self-conscious and to be able to think "I"-thoughts, and of how we acquire this capacity. The book covers a wealth of topics in cognitive science, each of which would be worth a detailed discussion. We shall focus on three general areas: the project of giving an account of "I"-thoughts that meets four desiderata laid out in the earlier parts of the book, the circularity alleged to be involved in analyzing the capacity to think "I"-thoughts in terms of the capacity to use "I" in the public language, and the account of concept possession.
2. Bermudez holds, and we agree, that a successful philosophical account of what it is to have "I"-thoughts needs to explain a number of special features of such thoughts. Three of these are made explicit in Chapter 1. First, "I"-thoughts are different from other thoughts that refer to the same thing. Bermudez's thought expressed by "I will be the next person to receive a parking ticket" is different from his thought "Bermudez will be the next person to receive a parking ticket," since Bermudez might be amnesiac and not know his own name (pp. 2-4). Let us call this feature of "I"-thoughts "the Perry feature" (after John Perry, who illustrated it in a memorable way in 1979. Second, "I"-thoughts, or at least some "I"-thoughts, have immunity to referential error through misidentification. Suppose that I experience pain and come to think, "I am in pain." To say that my thought is immune to error through misidentification is to say, according to Bermudez's formulation, that the following is possible: I am justified in believing that someone is in pain, but mistaken in believing that it is I who am in pain (pp. 6-7). Third, there is what we might call the self-reflexive feature: if a person is having an "I"-thought, then she must know that she is thinking about herself. "[O]ne cannot think a genuine 'I'-thought if one is ignorant that one is thinking about oneself" (pp. 2-3).
3. A satisfactory account of "I"-thoughts should explain all three of the features above. One can also find a fourth requirement in Bermudez's book. In criticizing the view of Hugh Mellor that "I"-thoughts are necessary for the explanation of action, Bermudez points out that the action of a hungry animal's eating the food in front of him might be explained by reference to a number of different beliefs, that is, to beliefs with different contents. The triggering belief might have the content (which we might express by) "There is food here," "That is food," "Food," or finally "I am facing food," only the latter of which is an "I" content (pp. 28-39). A satisfactory account of "I"-thoughts should be able to distinguish between these various contents. We shall call this requirement "the Mellor requirement."
4. Bermudez's theory of "I"-thoughts is complex. For him, we can understand the full-fledged form of self-consciousness -- in other words our sophisticated "I"-concept -- by seeing it as the end point of a development of successively more sophisticated forms of non-conceptual self-consciousness (p. 269). Bermudez takes us through a number of levels of non-conceptual self-consciousness in Chapters 5-9. First, from the beginning of life the child "exteroceptive perception is a source of nonconceptual first-person contents . . ." (p. 271). Second, somatic proprioception (bodily self-awareness) provides the child with an awareness of itself as a spatially extended physical object responsive to the will. These first two domains provide us with very primitive forms of self-consciousness that are nevertheless "essential" (p. 272) and fundamental in giving an account of full-fledged self-consciousness. The last two forms of self-consciousness depend on the first two. The third form is one in which the child is able to distinguish between its experiences and what they are experiences of (self vs. world). The fourth involves a concept, or better proto-concept, of the self as a bearer of psychological properties. All of the chapters contain a wealth of philosophical and psychological insights.
5. About each form of self-consciousness one would like to have clear and explicit answers to the following questions. What is the content of an "I"-thought of this form? In what sense is it an "I"-thought: which of the special features of full-fledged "I"-thoughts does it have? What constitutes "I"-thoughts of this form? How are the special features of "I"-thoughts of this form to be explained? And finally, what reasons are there to believe that young children have "I"-thoughts of this form? Our criticism of Bermudez's book in this section, which is on the whole a friendly one, is that often the answers to a number of these questions are not clear and explicit. To illustrate our worry here, we shall make a few brief remarks about the first and fourth forms of self-consciousness.
6. The first form of self-consciousness, which Bermudez calls "the self of ecological optics," is discussed in Chapter 5. Drawing on the work of J. J. Gibson, Bermudez argues that visual perception contains information not only about the environment but also about the self, and so involves a primitive, indeed the most primitive, form of self-consciousness. Information about the self is carried by visual perception in a number of ways. Perception of the boundedness of the visual field, perception of body parts such as the nose, as well as the pattern of flow in the optic array all provide information about the self. For example, the optical flow yields information about both active and passive movement of the self (p.112). The key idea here is that "the pick-up of self-specifying information in exteroceptive perception is a source of nonconceptual first-person contents from the beginning of life" (p. 271).
7. According to Bermudez, even a very young child has this first form of self-consciousness. It has a rudimentary concept, or proto-concept, of itself. It has what we can call "I-eco" thoughts (for the "I" of ecological optics). Which of the special features of full-fledged "I"-thoughts do "I-eco" thoughts have? Obviously, they lack the self-reflexive feature. Do they have the first two features, the Perry feature and the immunity feature? Supposedly so, but one would like this to be spelled out explicitly. What makes an "I-eco" thought an "I"-thought at all?
8. What, for Bermudez, constitutes having an "I-eco" thought? Here again, one would like this to be more explicitly and carefully spelled out. It seems that Bermudez wants to analyze "I-eco" thoughts in terms of three things: (1) being a state which carries, or perhaps is sensitive to, information about the self [Section 5.1], (2) being a state with representational content (i.e. a state which is appropriately characterized as an intermediary between sensory input and behavioral output) [p. 115], and finally (3) being a state which has an intimate connection to action (pp. 117-118). Suppose that an infant is facing food and starts crawling towards it. What is it for its perceptual state(s) here to have first-person content? According to Bermudez, the state will have to carry "self-specifying information," information that the infant is facing something that is food. But what -- and here we are thinking of the Mellor requirement -- is the difference between carrying the self-specifying information and merely carrying the information that we might express by "There is food"? We need to know more about how the notion of information works when the subject of the vehicle of information is also the source of the information. Again, as is well known, informational content in its standard sense is DE RE content (Dretske, 1981, pp. 66-67), that is, it is referentially transparent in genuine singular term position (Peacocke, 1983, pp. 6-7). This means, however, that even if we grant that the infant's perceptual state carries self-specifying information, we will have to deal with the Perry feature. What is the difference between the child's "thoughts" (which we might express by) "This thing is facing food" and "I am facing food," where both of the thoughts refer to the child? Bermudez (following Perry) gives the answer that "[p]erceptual contents are first person contents in virtue of their immediate connections with behavior" (p.118). But doesn't this answer run again into Bermudez's objection to Mellor? "I"-thoughts are not necessary for action: thoughts with different contents may trigger action. So the thought, "This thing (or body) is facing food," may well be triggering the child's action.
9. We would like to distinguish between constitutional and epistemological issues. The question, "What constitutes such and such a state?" is different from the question, "What reasons are there for holding that some creature is in that state?" Bermudez, at least sometimes, seems to blur this distinction. If one is clear about what constitutes "I-eco" thoughts, then it is perfectly all right to appeal to inference to the best explanation in discussing whether young children have such thoughts, as Bermudez does in Section 5.3. It is another thing to think that saying that "I-eco" thoughts are states which are "crucial to the intentional explanation" (p. 122) of certain types of behavior takes us very far along the way of giving an account of what constitutes such thoughts. Accusing Bermudez of blurring the distinction is perhaps unfair in the case of his discussion of "I"-eco thoughts, but it may be fairer in other cases. States with representational content, according to Bermudez, are states that are needed to explain "connections between sensory input and behavioral output [which] are not invariant" (p. 115). To say this, however, is not to say that much about what constitutes representational states. It is true that earlier on Bermudez says that the account -- and here this means constitutional account -- that he wants to give of intentional mental states is a broadly functionalist one (p. 37, fn. 3), but it is hard to find a really explicit statement of such a theory.
10. We turn now to the fourth form of self-consciousness. By the time the child is ready to acquire language, and in particular the use of the linguistic "I," it has reached the fourth form of nonconceptual self-consciousness, in which it is able to have thoughts about others' mental states as well as its own. This final nonconceptual form incorporates the other three forms. It is somewhat unclear, however, what all this involves. Does the child at this stage have a concept, or rather proto-concept, of itself that is self-reflexive? Remember that the feature of self-reflexivity was this: if a person is having an "I"-thought then she must know that she is thinking about herself. If we have been following Bermudez, it seems on the one hand that we should say that the child at this last prelinguistic stage does not have a self-reflective concept, or proto-concept. This is because on Bermudez's view self-reflexivity is tied up with having mastery of the linguistic "I." On the other hand, it might seem that the child does have a self-reflexive "concept." According to Bermudez, rather complicated first-person prelinguistic Gricean intentions and thoughts are needed in an account of mastery of the linguistic "I." To have such Gricean thoughts, the child needs to have the resources to have thoughts about its own thoughts, and perhaps thoughts that its own thoughts are about certain objects. This means that it could well have the capacity to have the thought, still of course prelinguistic, that we would express by "The object of my present thought is me." And if all this is so, could it not be that the child also has a "concept" of "I" that is in fact self-reflexive? Admittedly, we are substituting speculation for detailed argument here. At any rate the answer to this important question is not made explicit in the book. Furthermore, if there is any truth to the idea that by this last prelinguistic stage the child already has a self-reflexive concept, then it becomes puzzling what additional capacities it acquires when it has mastered the linguistic "I."
11. A central motive for Bermudez in giving a prelinguistic account of "I"-thoughts is to overcome what he takes to be an important circularity problem. "I"-thoughts are typically expressed in English by "I"-sentences, and so it is tempting to suppose that one could give a philosophical account of the capacity to think "I"-thoughts in terms of the capacity to use "I" in the public language. According to Bermudez, however, the temptation should be resisted, since such an account will be circular. A satisfactory account of someone's mastery of the public-language "I" will presuppose, in a sense, that they are already capable of having "I"-thoughts. Bermudez's circularity claim here is plausible, we concur, but there is again some lack of clarity in the way that he unpacks the claim.
12. What kind of circularity is Bermudez worried about here? Bermudez assumes, and we will follow him here, that the capacity to understand a word in a public language is to be articulated in terms of a creature's having mastery of a rule governing that word (see Section 1.4). To understand "I," then, is to have mastery of the appropriate rule governing "I." Circularity will arise if mastery of the "I"-rule presupposes that the child already has the capacity to think "I"-thoughts. An analogy may be helpful here. Assume that "red" is a primitive term in English (it cannot be defined in terms of other words in English), and consider the rule "'Red' refers to (the property) RED." Although we could not of course teach a child the meaning of "red" by telling it the rule, we could get it to achieve mastery of the rule through training. Nevertheless, one might think that such training would be psychologically impossible unless the child already possessed something like the concept of red: this of course would have to be something pre-linguistic. Suppose we wanted to give an account of what it is to possess this pre-linguistic "concept" of red. It would be circular to say that to possess such a concept it suffices that one understand "red" in English. Of course giving a non-circular account of what it is to possess the pre-linguistic ability does not bother many, since it seems relatively clear what such an account would appeal to, e.g. the ability to discriminate red objects from non-red objects.
13. Bermudez holds that a similar kind of circularity holds in the case of trying to explain the capacity to have "I"-thoughts in terms of the linguistic capacity to use the first person, "I." Acquiring mastery of the rule for "I" presupposes that a creature already has a pre-linguistic "I" "concept." Such a pre-linguistic "concept" also seems much more difficult to explain than the pre-linguistic "concept" of red. As we saw in the first section, any account of "I"-thoughts will have to deal with a number of very puzzling features of such thoughts, such as the Perry feature and the immunity feature.
14. We have two points to make about Bermudez's exposition of the circularity. First, one wants to be clear about what rules of language are supposed to look like. In particular, do all linguistic rules need to refer to the psychological states, for example the intentions, of the speaker? If we think of rules of language as rules that a hearer must know in order to understand utterances of another, then do rules have to have the form, e.g. "When a speaker uses 'red,' in doing so he intends to refer to RED," or will the following form suffice: "Any token of 'red' refers to RED"? If the latter, then what is wrong with the following linguistic rule for "I": "Any token of 'I' refers to the speaker"? Bermudez asks how we are to construe such a rule (p. 15). We could, somewhat cavalierly, try to construe it in the following way: Jane's token "I" refers to Jane, John's token "I" refers to John, etc. What would be wrong with that? If there is nothing wrong with such a rule, then we have avoided the circularity that Bermudez is worried about.
15. Bermudez obviously thinks that such a rule will not do. We find, however, that his reasons are somewhat difficult to extract from the text. One of his reasons seems to be that a full understanding of the linguistic "I" requires that a hearer understand the meaning of sentences in which "I" occurs inside "that"-clauses in propositional attitude ascriptions. It might be replied, again perhaps too quickly, that a full understanding of "red," as well as many other words, including proper names, also require understanding of such words inside "that"-clauses; nevertheless, the simple rule above for "red" seems to suffice.
16. Bermudez does give a further reason why the simple token rule above will not do. The reason is that the rule will have to appeal to the producer, rather than merely the speaker, of the token, and that when we try to unpack "producer of 'I'" we shall have to bring in the producer's intention that an "I" token refer to himself. Moreover, the previous "himself" has to be construed as Castaneda's indirect reflexive "himself*," and this will introduce circularity into the account. According to Bermudez, we shall have to bring in the producer in order to account for examples such as that of the secretary receiving dictation from his boss and writing "I look forward to hearing from you." We agree that this reason is a good one. Our only criticism here is that it is spelled out not in the initial discussion of circularity (in Section 1.4) but only towards the end of the book (on p. 276).
17. Quite apart from concerns about criteria for first-person content, it is possible to be troubled by Bermudez's account of why "conceptual abilities are constitutively linked with . . . linguistic abilities in such a way that conceptual abilities cannot be possessed by nonlinguistic creatures" (p. 42). In defense of this Priority Principle [PP], Bermudez offers what appear to be two arguments, each of the form, "Capacity C is necessary for concept possession; only language users can have C; therefore, only language users can possess concepts." In each case, acceptance of the first premise invites trouble: individuals to whom Bermudez explicitly attributes certain concepts turn out not to have them; concepts we would ordinarily attribute to each other seem to be beyond anyone's grasp and, worst of all, possession of any concept whatsoever seems to involve just the sort of circularity Bermudez has been at pains to exorcise.
18. The first of Bermudez's links between linguistic and conceptual abilities ties concept possession to the ability to make genuine inferential transitions and this ability, in turn, to language use. The argument is explicit.
(1) "Mastery of a concept is tied up with grasp of its inferential role: [i.e., with] its contribution to the inferential powers of propositions in which it features" (p. 61).
(2) "Drawing inferences requires grasping general rules of inference" (p. 70).
(3) Grasping general rules of inference requires being able to provide justifications for at least some inferential transitions (p. 71).
(4) "[P]roviding justifications is a paradigmatically linguistic activity . . . a matter of identifying and articulating the reasons for a given classification, inference, or judgment" (p. 71).
(5) Prelinguistic creatures are incapable of engaging in these forms of justification.
(6) Therefore, prelinguistic creatures are incapable of concept possession.
"Mere sensitivity to the truth of . . . transitions . . . is not enough for [concept] possession," Bermudez tells us. "Rational sensitivity is required, and rational sensitivity comes only with language mastery" (p. 71).
19. A second argument for PP can also be discerned in Bermudez's denial that infants possess a primitive object concept. Psychologists err, Bermudez thinks, in interpreting infants' astonishment at an anomalous display as evidence that they have engaged in a concept-based inference. Not only are babies incapable, for reasons just canvassed, of genuinely inferring from "a is an object; objects are solid" to "nothing will pass through a"; they are unable even to formulate a premise of the form "a is an object." According to Bermudez, "Identifying something as an object is something that one does for REASONS. An individual identifies something as an object [only] if he knows what the criteria for objecthood are and can recognize that they are by and large satisfied" (p. 69, emphasis in original). Put more generally, possession of any concept, Fx, requires "a (theoretical) understanding of the reasons for thinking that something exemplifies the concept of an [F]" (p. 73). This carries us beyond the previous requirement of simply understanding what, in general, counts as justification by requiring in addition a grasp of the particular properties that are, in each case, "constitutive" of F-hood. Bermudez does not make an explicit connection between this and linguistic abilities. One way to bridge the gap, however, is to say that the requisite grasp of properties essential to F-hood is available only to competent language users because only they are in a position to justify application of a concept Fx to an instance on the basis of the definition (or, perhaps, criteria) associated with a natural language term that expresses it.
20. If this is correct, then Bermudez sees a constitutive link between concept possession and linguistic abilities because he believes (1) that concept possession requires both a capacity for justification and a grasp of (at least some adequate subset of) properties associated definitionally (or criterially) with category membership and (2) that each of these, requires mastery of a natural language. Whether (2) can be defended is an interesting question that we shall not pursue here; (1) clearly comes at a high price.
21. First, insistence on (1) appears to fly in the teeth of Bermudez's own wish to attribute the concepts human being, living animal and internal organ to 4 year olds on the basis of their projective predilections.
[Susan] Carey told . . . [4 year olds] that people had a greenish internal organ called a "spleen." She then showed them a toy mechanical monkey and a live earthworm and asked them which was more likely to have a spleen. Although a toy mechanical monkey obviously looks much more like a human being . . . the children decided that the worm was more likely to have a spleen . . . . It seems natural to say that they made use of inference patterns linking the concepts human being, living animal, and internal organs (p.70).
Although Bermudez does not tell us exactly how these inference patterns might go, it seems questionable, to say the least, whether Carey's young children really meet his criteria. Take, for example, the concept animal. Summarizing her observations, Carey says, "[C]hildren under 7 have not mapped the word 'animal' onto the biological concept that encompasses all animals, for they exclude people from its extension" (1985, p. 140); "they will vehemently deny that people are animals" (p. 138). Moreover, while it is true that young children are more likely to project various biological properties from humans onto other animals than onto a toy monkey, they are unlikely to use information about biological properties of animals other than humans as a basis for such projections. "[W]hy," Carey asks, "will the 4-year-old not project spleens to other animals from dogs or bees at all?" (1985, p. 138, stress added). Bermudez's strategy for connecting concept possession with linguistic capacities would seem to deny Carey's 4-year-olds the very concepts for which he wishes to give them credit: their judgments and projective preferences suggest that they have neither a grasp of the definitions of the corresponding English language terms or of features essential to animalhood.
22. Moreover, if we take Bermudez seriously with regard to the need to justify concept application by invoking definitional properties, attribution of mundane concepts to adults becomes equally problematic. There is strong opinion in the philosophy of language that many, perhaps most, lexicalized concepts lack necessary and sufficient conditions for their application and, notoriously, that they are absent in the case of natural kind terms like 'animal.' If this is correct, Bermudez's explication of the Priority Principle would appear to have as an unwelcome consequence that nobody has a concept of living animal.
23. Even if we set aside the issue of grasping essential properties and focus attention, instead, on Bermudez's requirement for justification, trouble looms. Perhaps there are rational grounds for concept application that do not involve grasping a definition and perhaps Carey's 4-year-olds might be capable of invoking them, if not here, at least elsewhere. Indeed, Carey mentions that "children as young as 4 produce justifications [for their judgments] that show they appreciate the force of [various] . . . arguments" (p. 87). The problem then is explaining how, on Bermudez's view of concept possession, the children could have come by their notion of what counts as a justified inference.
24. As far as we can see, Bermudez makes possession of a concept of justification a prerequisite for possession of any concept whatsoever, including, presumably, possession of the concept of justification. This creates circularity of just the sort that led Bermudez to write this excellent book. The project of showing how we could arrive at full-fledged self-consciousness on the basis of our initial capacities and the contribution of the world is a truly impressive undertaking, for which we should all be grateful. In trying to accomplish this, Bermudez has been willing to set aside at least one mainstay of traditional analytic philosophy, namely, the view that all thought contents must be conceptual. Perhaps there is something to be said for unburdening himself a bit further before undertaking the next leg of the journey.
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
Carey, S. 1985. Conceptual change in Children. Cambridge, MA: MIT Press.
Dretske, F. 1981. Knowledge and the Flow of Information. Cambridge, MA: MIT Press.
Peacocke, C. 1983. Sense and Content. Oxford: The Clarendon Press.
Perry, J. 1979. "The Problem of the Essential Indexical." Nous 13: 3-21.