Circularity or reduction: What is the real issue?
Comments on José Bermúdez The Paradox of Self-Consciousness

by Andrew Brook e-mail

Dept. of Philosophy
Carleton University
Ottawa, Ontario, Canada K1S 5B6

This book has many virtues. Despite its somewhat daunting title, it is very readable, in part because most of it is beautifully written. It makes use of a mass of empirical information from cognitive and developmental psychology, something still too rare in philosophy of mind. It attempts to build a completely naturalistic picture of how full-blown adult self-consciousness could be the product of simpler forms of awareness of self, e.g., the ability to discriminate between what is in fact oneself and what is not. It does so in the service of a solution to a properly philosophical problem, namely, a possible deep tension buried in the view of self-consciousness currently most popular among philosophers. And finally, it sustains a single complex argument from beginning to end
Along the way, the book discusses some fundamental issues in the foundations of psychology and the philosophy of mind such as the nature of concepts and the possibility of thought without language.

The Project

Readers thoroughly familiar with Bermúdez’s book may be able to skip the next few pages. The last half dozen paragraphs of this introduction to Bermúdez’s project are essential to understand the critique that follows and should not be skipped.

Bermúdez has summarized his project as follows (1999):

      This book addresses two fundamental questions in the philosophy and psychology of self-consciousness: (1) Can we provide a noncircular account of full-fledged self-conscious thought and language in terms of more fundamental capacities? (2) Can we explain how full-fledged self-conscious thought and language can arise in the normal course of human development? I argue that a paradox (the paradox of self-consciousness) arises from the apparent strict interdependence between self-conscious thought and linguistic self-reference. Responding to the paradox, I draw on recent work in empirical psychology and philosophy to cut the tie between self-conscious thought and linguistic self-reference. The book studies primitive forms of nonconceptual self-consciousness manifested in visual perception, somatic proprioception, spatial reasoning and interpersonal psychological interactions.

Perhaps the tension is not exactly a paradox, Bermudez’s claim notwithstanding, because there is not, strictly speaking, any element of contradiction or self-refutation in it, but it is something pretty serious. Let’s see how it develops and what it looks like more specifically.
First Bermúdez lays out what he calls a deflationary theory of self-consciousness, at the centre of which are what have come to be called 'I'-thoughts. 'I'-thoughts involve a distinctive kind of self-reference. Modifying Bermudez’s own example, consider:

    I think: (1) The person AB is about to be attacked by a poisonous spider

    I think: (2) I am about to be attacked by a poisonous spider

Even though (1) and (2) are both thoughts about the same person, namely me, these are quite different thoughts. Roughly, I could truly think (1) of AB without knowing that AB is me, and, on the other hand, I could truly think (2) of myself, knowing that it is myself to whom I am referring, even if I do not know or have ceased to know that I am AB. As Shoemaker (1968) has put it, I can refer to myself, refer to myself as myself knowing that it is myself (the subject of my experiences) to whom I am referring, without identifying myself under concepts; I can refer to myself without knowing identifying facts about myself. (Shoemaker calls this reference to self without identification. It is different from, though related to, his better known notion of immunity to error through misidentification with respect to the first person. We will return to the distinction.) It is thoughts like (2) that are called ‘I’-thoughts. What makes them special is that one cannot use ‘I’ and other first person pronouns in this way "without knowing that one is thinking about oneself" (1998, 3). Following Castañeda, when reference is made to an ‘I’-thought in the third person, indirect discourse, Bermúdez marks the third person pronoun with an ‘*’ (as in ‘AB said that he* was about to be attacked by a poisonous spider’).< 1 >
The core claim of Bermudez’s deflationary account is that the ability to have ‘I’-thoughts is so central to self-consciousness of the kind that adult human beings have that elucidation of this ability is all that is required to elucidate what is distinctive about self-consciousness (1998, 295). This is actually a stronger claim than he needs. To get his problem going, all he needs is that the ability to think ‘I’-thoughts is necessary for the kind of self-consciousness that adult human beings have. It is natural to think that full-blown self-consciousness distinctively involves other things, too, but Bermúdez can remain entirely neutral on this further issue. So let us leave the deflationary theory and focus on ‘I’-thoughts themselves.
If thinking ‘I’-thoughts is necessary for self-consciousness of a certain kind, then part of an account of self-consciousness of that kind will be an account of our capacity to think ‘I’-thoughts. Here is how Bermudez’s tension arises. On a certain widely accepted assumption about the relationship of thought and language (Bermúdez calls it the Thought-Language Principle), the way to gain an understanding of the capacity to think ‘I’-thoughts is via an analysis of the first-person pronouns used in them (plus perhaps other things but that does not matter). However, mastery of first-person pronouns (‘I’, ‘me’, ‘myself’, ‘my’) requires, it would seem, the capacity to think ‘I’-thoughts (1999, 24). We have a nice tight little circle.
In fact, according to Bermúdez, we have two nice tight little circles. The first is an explanatory circle, in which to explain either side of the circle, we need the other side. The second is a capacity circle, in which we need either capacity to account for our ability to acquire the other. The second circle is independently serious because it makes our acquisition of these reciprocal capacities a complete mystery. (This violates what Bermúdez calls the Acquisition Constraint: If a capacity exists, there must be an explanation of how a normal human being could acquire that capacity in the course of normal human development [1998, 19].)
Note that no circle involving use of pronouns and ability to think thoughts of a certain kind arises for thoughts like (1). No use of pronouns is involved. That is why Bermúdez distinguishes thoughts like (2) from thoughts like (1).
Bermudez’s response to the circle is to go after the Thought-Language Principle and specifically the idea that the way to understand the capacity to think ‘I’-thoughts is via an analysis of the first-person pronouns used in them. He wants to show that we can gain a full understanding of our capacity to think ‘I’-thoughts without appealing to our ability to use first-person pronouns by appealing instead to the capacities behind simpler kinds of self-consciousness, capacities that do not use first-person pronouns or, perhaps, language at all, and building up a picture of the capacity to think ‘I’-thoughts out of these simpler capacities.
There is a relationship between the Thought-Language Principle and the distinction between thoughts like (1) and thoughts like (2). Not only does the circle not arise for thoughts like (1); it would also be more plausible to hold that our capacity to refer to beings who may in fact be ourselves can be understood as an extension of simpler, nonlinguistic capacities than that our capacity to think ‘I’-thoughts can be. It seem far less plausible to maintain that we can explain the central capacity in thoughts like (2), namely, the ability to refer to ourselves knowing that it is ourselves to whom we are referring, in terms of anything nonlinguistic than thoughts like (1). In short, thoughts like (2) are the hard case for Bermudez’s kind of account.
To find the simpler kinds of self-consciousness out of which to build full-blown capacity to think ‘I’-thoughts, Bermúdez examines four domains: perceptual experience and Gibson’s account of it (Chapter 5), somatic proprioception and awareness of one’s own body (Chapter 6), point of view and spatial reasoning at and over time (Chapters 7 and 8), and awareness of one’s own and others’ psychological states (Chapter 9). In each of these, Bermúdez claims to find one or more forms of consciousness of self. The former two are synchronic and involve relatively simple kinds of self-consciousness. The latter are diachronic and the self-consciousness is more complex. What is distinctive about all four of them, according to Bermúdez, is that they are nonconceptual (i.e., the creature involved need not have the concepts required to specify the contents of the act of awareness involved [1998, 49]). If so, then all four are nonlinguistic and therefore could not be dependent on any capacity to use anything linguistic such as first-person pronouns.
In the final chapter (Chapter 10), Bermúdez puts these various forms of simple consciousness of self to work to try to break the two circles. What he wants is an account of ‘I’- thoughts that does not require that the thinker of those thought have mastered first-person pronouns. We would then be free to use the capacity to think ‘I’-thoughts in an account of what it is to have mastery of first-person pronouns.

Bermudez’s strategy in Chapter 10 is straightforward. First he gives an account of the requirements for using first-person pronouns where the user knows that she is referring to herself. This account is a variation on the well-known Gricean communication-intention approach and consists of a complex pattern of ‘I’-thoughts. Then he argues that we can give an account of this pattern of ‘I’-thoughts without invoking the capacity to use first-person pronouns. The pattern of ‘I’-thoughts is this:

      An utterer u utters ‘I’ to refer to himself* [see above for this use of ‘*’] if and only if u utters ‘I’ in full comprehension of the token-reflexive rule that tokens of ‘I’ refer to their producer and with the tripartite intention
      a. that some audience a should have their attention drawn to him*,
      b. that a should be aware of his* intention that a’s attention should be drawn to him*,
      c. that the awareness mentioned in (b) should be part of the explanation for a’s attention being drawn to him. [1998, 283]

There does not seem to be any risk that to know the token-reflexive rule, a user of ‘I’ must have the capacity to use first-person pronouns, so Bermúdez focuses on (a) - (c). At least (a) and (b) involve ‘I’-thoughts (the ‘*’ means that the utterer must know that the person to whom ‘him’ refers is himself). And the question is, can we give an account in which they can be satisfied without the utterer having to use first person pronouns?
I will focus on Bermúdez's way of dealing with clause (a) but let me say a quick word about how he approaches (b) and (c). About (b), oversimplifying a bit, Bermúdez argues that prelinguistic infants interacting with their mothers have the ability to form the appropriate kind of intention and to know that they have such an intention (At any rate, attribution of such complex intentions and knowledge is the best explanation of the behaviour we observe). No capacity to use first-person pronouns is required. Similarly with respect to the kind of ‘bringing-it-about’ at the heart of (c). In fact, it is not clear that (c) involves an ‘I’-thought at all. If so, the problem does not even arise for it and Bermúdez's account has to handle only (a) and (b).
There are questions that could be asked about these two clauses and the accounts of them. For example, in (b), must u be aware of his intention to draw a's attention to him or is it enough that he have the intention, as in clause (a)? I will let them pass because I want to focus on clause (a) itself, the requirement that the speaker intend that some audience a should have their attention drawn to him* (i.e., to the person he knows to be himself) by his use of 'I'. Bermúdez argues that it can be satisfied without the utterer having the capacity to use first person pronouns.
Indeed, says Bermúdez, clause (a) could be satisfied without attributing a capacity to use first-person pronouns in at least two ways.

    One way ... would be through an intention to draw an audience’s attention to the material self as revealed in proprioceptive self-consciousness. Another way would be through an intention to draw an audience’s attention to the material self as a spatial element moving within, acting upon, and being acted upon by, the spatial environment [1998, 284]

Proprioceptive self-consciousness and awareness of "the material self as a spatial element moving within, acting upon, and being acted upon by, the spatial environment" are two of the forms of nonconceptual self-awareness that Bermúdez has delineated in earlier chapters.

Does the Project Work?

Unfortunately, it is not clear that either will do the job. I say ‘unfortunately’ because it would be wonderful to have a successful marriage of full-blown adult self-consciousness and simpler forms of self-consciousness of the sort that Bermúdez proposes.

        First concern: I am not sure exactly how Bermúdez wants to frame the problem. I have a fair idea what an ‘I’-thought is but I am not sure what he means by ‘mastery of first person pronouns’. He could mean either mastery of the appropriate words in some natural language or mastery of the ability to make the appropriate kind of reference to and/or judgment about oneself. The first would make the project silly – I doubt that anyone has ever held that the way to understand ‘I’-thoughts is via analysis of our knowledge of a few English words – but the second runs a serious risk of undermining it much more profoundly. Here is why. One cannot have an ‘I’-thought "without knowing that one is thinking about oneself" (1998, 3). But to refer to oneself in such a way that one knows that one is thinking about oneself is to make precisely the kind of reference, the kind of judgment, that mastery of first person pronouns affords. That is to say, part of having an ‘I’-thought simply is doing what mastery of the first person pronouns allows us to do. If so, there are not two separate elements here, the thought and the mastery of the kind of reference and/or judgment that first person pronouns affords. From this it would follow, in turn, that there is no circle. A thing cannot be in a circle with itself.
So let us rejig the project slightly and turn it into a project of investigating whether clause (a) could be satisfied, more generally whether 'I'-thoughts could be thought, using just proprioceptive awareness and awareness of oneself as a spatial element moving in, acting upon and being acted upon by the spatial environment. In short, can full-blown, 'I'-thought self-consciousness be reduced to, accounted for in terms of these simpler forms of self-consciousness, whether or not the former involves the capacities that mastery of first person pronouns would give us?

         Second concern: Right away we notice a slippage in Bermúdez's terminology, specifically in his term 'first person contents'. Sometimes it seems to mean merely some information about the being who is in fact oneself. Other times it clearly means the kind of reference to, judgment about oneself that yields 'I'-thoughts - reference or judgment in which one knows that the person to whom one is referring is oneself. (This is one of the few spots of unclarity in an otherwise lucid book, but it is a bad one.) What we need to get out of proprioceptive self-consciousness and consciousness of oneself as a spatial element is the latter, not just the former. The trouble is, these two kinds of simple self-consciousness may provide only the former, not the latter. This leads to a third concern.)

        Third concern: Can we get 'I'-thoughts out of proprioceptive self-consciousness and/or conscious of oneself as a spatial element at all? It is far from clear that we can. Clause (a) has it that a speaker of 'I' or cognates wants an audience's attention drawn to himself*. Notice first that this is not the same thing as wanting to draw attention to the body that is (in fact) his. The speaker may not know which body is delivering proprioceptive sensations to him, which body is the one travelling through the spatiotemporal path carved by the base point of his experiences, which body is responding to his desires and emotions. (If the latter seems far-fetched, think of Bermúdez's own favourite example, infants. Some body is in fact responding to their desires and emotions, by sucking, crying, etc., but it is very unlikely that the infant knows which one - or indeed anything about bodies at all.)
In general, the kind of awareness of self provided by ‘I’-thoughts is very, very different from any kind of awareness of one’s body or even, perhaps, awareness of oneself as an object in the world of objects in general. (We will return to the latter thought.) One difference can be brought out by an example.
Suppose that I have a proprioceptive sense that my left leg is folded under my right. Now recall Shoemaker's notion of immunity to error. I could be wrong that it is my leg that is giving me these sensations. Another leg appropriately hooked up to my proprioceptive nervous system could probably give me exactly the same sensations. But I could not be wrong that the sensation is my sensation. Not at any rate if I am aware of it by having it - if I am aware of the sensation in other ways, e.g., by inferring it from behaviour, by reading about it, whatever, then I could be in error in identifying it as mine but not if I am aware of it by having it. That really does confer immunity to error through misidentification with respect to the first person.
What exactly it is about the ‘I’-thoughts kind of consciousness of self that confers immunity is a matter of some debate. (I discuss the issue and most of the other issues in the remainder of this review in 1994, 89-90). Whatever it is, the kind of consciousness of self involved has some other striking features.

1.This kind of consciousness of self confers immunity, the point we have just seen.

2. When one is conscious of oneself in the way involved in ‘I’-thoughts, one is conscious of oneself as the single common subject of a great many different experiences, as Kant put it (1781/7, A350). A number of experiences are tied together so that they are all part of a single field or state or act of consciousness.

3. The reference to oneself by which one achieves awareness of oneself as subject seems to involve no "identification", an idea well discussed by Shoemaker 1968 and Perry 1979 and introduced earlier. Generalizing the notion a bit, reference to self may not proceed by way of attribution of properties to oneself at all. That is, it may be nonascriptive. One argument for this view is that one is or could be aware of oneself as the subject of every one of one’s conscious experiences and in roughly the same way, i.e., as their subject. If so, consciousness of self is not what Bennett calls ‘experience-dividing’ – statements expressing it have "no direct implications of the form ‘I shall experience C rather than D’ " (Bennett 1974; Brook 1994, 80). If so, no experience-dividing attributions could be necessary for such consciousness.< 2 >

4. Certain indexicals may be essential, i.e., ineliminable. (This is another idea that goes back to the 1960s; Perry 1979 is one of the fullest statements of it.) One argument goes as follows. To know that I wrote a certain book a few years ago, it is not enough to know that someone over six feet tall wrote that book, or that someone who teaches philosophy at a particular university wrote that book, or ... or ... or ... , for I could know all these things without knowing that it was me who has these properties (and I could know that it was me who wrote that book and not know that any of these things are properties of me). Nor would it help to add details of a more identifying kind – the person whose office number is 123 in building ABC, the person whose office phone number is ... . If I don’t know that that office is my office, that that phone number is my phone number, I could know all these things and still not know that it was me who wrote the book. And vice-versa – through bizarre selective amnesia, I could cease to know all such things about myself and yet continue to know that it was me who wrote the book. As Shoemaker puts it,

      ... no matter how detailed a token-reflexive-free description of a person is, ... it cannot possibly entail that I am that person [1968, 560].< 3 >

(There is a connection between this point and the idea that the consciousness of self in question is not by way of identifying or even any attributions. If some indexicals are essential, the consciousness of self involved must at least outstrip all attributions.)

5. In the kind of consciousness of self in question, I may not appear to myself as an object, i.e. an item contained within a world of items and in various relationships to other items in that world, at all. If, when I appear to myself as subject, I appear nonascriptively, i.e., as not characterised as anything, when I am aware of myself as subject, I not only do not appear to myself to be a physical object, I do not appear to myself to be an object among other objects at all. However I appear, my type, boundaries, etc., are entirely unspecified. Kant held this view (1781/7, A382, A402, B429). It is also at least related to Wittgenstein’s (1933-4) distinction between consciousness of the self as subject and as object, a distinction that Bermúdez discusses (1998, 5).

Certainly the view that one is aware of oneself as an object of any sort faces some real puzzles. For example, how can an entity appearing as subject of all its experience also appear as subject to be an object of only a few of them (a puzzle that exercised Kant [1781/7, B155])? Equally, virtually any of one’s conscious experiences can be the basis for awareness of oneself as subject, but one could appear as an object, including a physical object, of only some of them.
Now, conscious of self of the kind delineated in (1) - (5) is the kind involved in clause (a). Some states and events give me the capacity to be aware of them and of myself as their subject, the thing that has them, namely, representational states and events. I can become aware of all others only via further states or events that represent them. When I intend to draw someone’s attention to myself in the way that clause (a) is about, I intend to draw their attention to the thing that is having such representations, not (except incidentally) to anything being represented in them. This is the kind of consciousness of oneself that has features (1) - (5). So let us ask: How does Bermudez’s attempt to understand consciousness of self of the kind given by ‘I’-thoughts in terms of proprioceptive consciousness of self and consciousness of being a spatial element fare in the light of (1) - (5)?
The answer, I think, is: Not well. His account simply builds no bridge at all between proprioceptive consciousness of self and consciousness of being a spatial element, in both cases awareness of oneself as an object of some kind but at most awareness of the thing that is in fact oneself, and awareness of oneself as subject.

So do we remain caught in Bermúdez's explanatory and/or capacity circles? This is not so clear. First, as we have seen, there may be no such circles. That would be the case if ability to refer to oneself in the way provided by the first person pronouns is simply part of thinking an 'I'-thought. Second, even if one or both circles is real, there would still be no problem if we can explain them both and account for our acquisition of both capacities in terms of some common third factor. In fact, this brings us close to Bermúdez's underlying project, to provide an account of full-blown, 'I'-thought self-consciousness in terms of simpler capacities. We could give up all notion of a circle connecting 'I'-thoughts and the ability make first person judgments without this underlying project losing any of its viability or importance. And it is vitally important. Since 'I'-thought self-consciousness did grow out of simpler capacities phylogenetically and does so again in every child ontogenetically, it would be real progress to have a good story, or even a story, of how this happens or could happen. I am not sure that Bermúdez has such a story yet but he has shown us the direction we have to go in if we want to find one.


<1> This point about ‘I’, etc., that the user knows that it is she herself to whom reference is being made, and the corresponding distinction within uses of third person pronouns was first noted by Castañeda (1966). He says that when third person demonstratives (‘he*’, etc.) are used to express ‘I’-thoughts, they are more properly thought of as quasi-indicators than real indicators, for reasons that we do not need to go into here.

<2> Notice that the notion of self-reference without identification is very different from the notion of immunity to error though misidentification with respect to the first person, even though Shoemaker introduced them both in the same paper. Even if there were no such immunity, it could still be the case that we can or must refer to ourselves without identifying ourselves. For some reason, the notion of immunity has received much more attention than the notion of reference to self without identification, especially in the UK.

<3> Though most often associated with Castañeda and Perry, a version of it can be found even earlier in Nagel (1965).


Bennett, J. (1974), Kant’s Dialectic Cambridge: Cambridge University Press.

Bermúdez, J. (1998), The Paradox of Self-Consciousness, Bradford/MIT Press.

Bermúdez, J. (1999), Precis of (1998). Psycholoquy 10(35). <>

Brook, A. (1994), Kant and the Mind. New York: Cambridge University Press

Castañeda, Hector-N. (1966), ‘He’: A study in the logic of self-consciousness. Ratio 8, pp.130-57

Kant, I. (1781/87), Critique of Pure Reason. P. Guyer and A. Wood, trans. and eds. Cambridge: Cambridge University Press (cited as Axxx for the first edition of 1781and Bxxx for the second of 1787)

Nagel, T. (1965), Physicalism. Philosophical Review 74, 339-56

Perry, J.(1979), The problem of the essential indexical. Noûs 13, pp.3-21

Shoemaker, Sydney. (1968), Self-reference and self-awareness. Journal of Philosophy 65, pp. 555-67

Wittgenstein, L. (1933-4), Blue and Brown Books. Rush Rhees, ed. Oxford: Blackwell’s, 1958

top | back to symposium index | Bermúdez's reply