UNDERSTANDING APPERCEPTION TODAY by KARL AMERIKS

Kant's theory of the mind has received considerable attention in the last decade, but most of the approaches of recent years can be seen as a resuscitation of strategies initiated by earlier generations of Kantscholars. For example, my own work in excavating the rationalist commitments that may remain intact behind Kant's critical arguments in the Paralogism has some obvious parallels with the "metaphysical" approach to Kant favored in the 1920s by German philosophers such as Heinz Heimsoeth，Max Wundt, and Martin Heidegger. Yet another tradition， going back to Fichte, has been resurrected in our time by the influential work of Dieter Henrich and carried forward in an important new study by Frederick Neuhouser, whose focus on Fichte's concept of the self as "self-positing" resembles notions recently said to be found in Kant by Henry Allison and Robert Pippin. These interpretations contrast with the standard analytic and anti-idealist approach represented most recently by C.T. Powell's new work, which can be seen as a development of suggestions by earlier English-speaking philosophers such as Peter Strawson, Wilfrid Sellars, and Jay Rosenberg. Similar but looser and more distant relations of indebtedness can be found between Patricia Kitcher's work and the Humean tradition, which has bequeathed a set of "naturalistic" problems and approaches, if not answers, to a line of interpreters that extends from Herbart to Robert Paul Wolff.

From a systematic perspective, one can characterize these four major strands in recent work on Kant's theory of mind in the following way. The first two strands are broadly rationalist and continental,and stress, respectively, Kant's alleged Leibnizian or Cartesian concerns. I take myself to represent the first option while the "Fichteans" represent the second.The other pair of strands comes from a broadly empiricist camp.The first strand has a special concern with the conditions of empirical self-knowledge (Strawson and Powell), while the second strand stresses the general empirical implications of Kant's notion of synthesis (Kitcher).

In all these schools of interpretation the manifold ambiguity of Kant's doctrine of the "transcendental unity of apperception" plays a central role. Some aspects of this doctrine are clear enough. Kant speaks of this unity as “transcendental" because he takes it to be a necessary condition of all our experience. He calls it a unity of "apperception”because it is the condition that all items of that experience must be able to be accompanied by the representation "I think." And he speaks of this apperception as involving a "unity" in order to highlight the fact that this "I' has a kind of simplicity or self-sameness that contrasts with the multiplicity of items that are its possible objects. Moreover, it is called an“original synthetic" unity because it is not derivable from any one of the representations by themselves, and because even as a group the representations do not "combine themselves."

This minimal gloss on Kant's terms is common ground for all interpreters, but the intrinsic complexity of the doctrine of apperception still allows for several different points of departure. Because I have discussed the rationalist approaches elsewhere, my main focus here will be on the most impressive recent broadly empiricist reactions to Kant on the self. In focusing on these approaches, I will be abstracting here from many of the noumenal, moral, and metaphysical issues that I believe are central to the ultimate understanding of the Kantian self，in order to concentrate as much as possible on its purely apperceptive aspect, for it is this aspect that is dominant even in these approaches, just as it is in the main line current German interpretations.

Before developing any evaluation of current empiricist approaches to apperception, some further clarification of Kant's doctrine in its own terms is required. In the most fundamental claim of his doctrine of apperception, Kant says there is an "I think" that is "transcendental" precisely because it necessarily can accompany all of one's representations. Much attention has been focused on the phrase "I think"here, but equal attention needs to be given to what it is that this"I think" is supposed to be accompanying. Exactly what is it, that this"I think" thinks of? One might believe that here Kant is speaking directly of objects, psychological objects at least. But for at least two reasons that cannot be exactly right.

The first reason is that to say "I think" is to use a phrase that calls out for completion with a that-clause rather than a mere object term; we say“I think that this is how it feels,etc," rather than merely“I think x”or “I think x,y.” Secondly, although it might seem that Kant means to attach the transcendental “I think” directly to representations as such, his real claim is about one's "own" representations, i.e.,representations that have the quality of not being, as he says, "nothing to me." Although this is often forgotten, Kant's doctrine is not that all representation requires apperception, for he holds that there are whole species of beings (e.g., dogs, and no doubt cats as well) who have representation but not apperception, and there are probably whole layers of human existence (e.g., our “peripheral” or subconscious or infantile representations) that have a similar form. So, Kant's claim is only that representations that are“something to”one, are what must be able to be accompanied by the transcendental “I think." And this means that his doctrine of apperception is precisely not a claim that each human representation as such, that is, just as are presentation, must be susceptible to apperception. And yet it is his view that all“relevant”representations (representations that are not"nothing to" one) must already have some kind of personal quality.

I propose that the simplest way to understand this “personal”quality is to presume that the individual representations in question are already at least implicitly of the form,“I think that so and so . . ." It might at first seem that this would make redundant the “transcendental” “I think”, the “I think” that necessarily can accompany all the representations that are "something to" one. However, a transcendental “I think” is still worth introducing because it has a special collective function，for even if in fact the same “I” is distributively involved in the set:

(E) I think that x, I think that y, I think that z . . .;

nevertheless, (E) is not the same as:

(T)I think that (I think that x, I think that y, I think that z...).

The difference here is not merely that (T) is more complex than any part of (E) or even the whole set (E). There are at least two extra features of (T) that are noteworthy. The first is that it implicitly includes the claim (which may or may not be a correct claim) that all the uses of "I" within it are co-referential; i.e., I believe Kant understands (T) to include the claim that the I which thinks that x, is the same as the I which thinks that y, etc., as well as the thought that this is the very same I which thinks that that I thinks that x, etc. A second, and more controversial, point is that, given the a priori status of transcendental apperception, it seems that Kant understands the possibility of (T) to be a truth condition of the components of (E). More exactly, I take Kant's view to be that, for any of our thoughts contained in such sets as (E), there are corresponding thoughts such as (T), although, given the psychological limits of finite minds, this need not involve the"real possibility"'of one all-inclusive thought(T) which contains all (E)-thoughts.That is, he believes that for any component of an (E)-set, there needs to be at least some possible (T)-thought which contains that component,but there need not be one really possible (T)-thought that includes all such components. At most, what he calls for is roughly what Dieter Henrich has spoken of as just "the coordination of all possible‘l think’ -instances."

It is not immediately clear from a non-Kantian perspective why someone would insist that (E), empirical apperception， requires (T),transcendental apperception. (This terminology can be a little confusing since of course an instance of (T) can also be an example of what Kant would call "empirical apperception," but the point to keep in mind is that the specific capacity for thoughts at least as complex as (T) is what is crucial for introducing the notion of transcendental apperception.) One ground could be the meaning of "I" in these contexts，for，Kant may be reasoning, what could it mean to be an Iwhich is a correlate of any of the "first-level" thoughts "I think that x," etc., if that I could not be experienced as identical with other I's, such as those referred to in (T)? But if the full meaning of (T) is used here, then this amounts to what suddenly appears to be a fairly substantive claim, viz., that a "real subject" of thought could not be such that some of its instantiations could not be recognized by it as such,i.e., as instantiations of one and the same subject. This claim is striking because, if one allows, as it has just been noted that Kant does, that there are representations in beings incapable of thoughts, then it can well seem (with a little sympathy for something like the old idea of a“chain of being") that there also are or could be thoughts in beings incapable of thoughts of thoughts. In other words, such beings could be said to have "empirical”but not“transcendental" apperception (hence the designations“E" and "T").

lt can sound sadly dogmatic to insist such beings are impossible, or to ascribe such an insistence to Kant, but can one avoid such dogmatism and still hold onto Kant's doctrine of apperception? One response would be to concede the remote possibility of such beings and to stress that Kant's doctrine is meant primarily as a doctrine for us; we know that we (at least) do in fact have more than punctual, uncollected, “merely empirical” apperception. Nonetheless, just as we may still have some representations without even empirical apperception, it can also seem that, even if we have some thoughts which are also parts of thoughts like (T), we still could also have some other ("first-level") thoughts which are without any possible thought of those thoughts, let alone a collective "transcendental" thought (a set of thoughts like (T)) which connects it with all other such ("first-level,"i.e.,"(E)-leve1") thoughts. To deny that this could happen is to hold for us what I will call "the Strong Apperception Thesis" (SAT), i.e., the thesis that all one's empirical apperception requires transcendental apperception.

It may seem that very little depends on such an esoteric thesis, and yet one way of looking at much recent work on Kant on the self is to see it as focusing on precisely this point, as exploring whether SAT, or some principle close to it, about a kind of possible self-consciousness for all our consciousness, is the very core of Kant's philosophy.

The evaluation of SAT itself clearly hinges on how one understands the“possibility” that it denies. Understood as restricted to humans, SAT insists on the claim that any “I think that x” episodes that we have must be able to be connected with other such episodes in the transcendental way that was discussed earlier (when analyzing what was called "(T)"). I take it that, if the issue here were one of mere logical possibility, then it would be odd to reject such a claim,and yet it would also be uninteresting to uphold it. Since in some sense any thing can always be made more complex than it is，any particular empirical apperception should always be able to be inflated into an aspect of a higher level apperception. But SAT is meant to imply more than this triviality; it implies that there is something about the very type of "thing" that our empirical apperception is that it requires some real, even if implicit, connection to transcendental apperception. A similar idea about "types" clearly appears to be behind Kant's notion that, if mere animal awareness were "inflated" into apperception, then this should be described not as a mere enrichment but as a leap into a different kind of awareness. This is to say that as "mere animal" awareness it necessarily lacks the "real potential" for apperceptive representation, just as，for Kant, mere human awareness lacks the real potential for intellectual intuition.

It can appear very difficult to determine how one is even to go about beginning to settle the question of whether or not all our empirical apperception does have such a "real potential" relation to transcendental apperception. The whole question may seem to be a matter of the most speculative psychology. Of course, one could assume a strong “realist”attitude to this issue, and hold that there is some hidden truth here, such that the potential is either there or it is not. But this attitude does not do much by itself to help one to decide how to evaluate SAT. A more"Critical"approach would be to ask how a use of SAT might help in the resolution of other philosophical issues. If it turns out that affirming SAT appears to be, as some interpreters suggest, the best way to make sense of various distinctive features of our self-awareness, then this can build a strong indirect case for it,and for affirming the disputed“real potential”. On the other hand, it may rather be true, as I will be arguing, that if we keep in mind all the distinctions (and then some) that were developed along the way in introducing SAT, then there can be a slightly more modest way to understand our self-awareness, such that one can employ a broadly Kantian notion of apperception without going so far as to insist, as he apparently does, on SAT itself, let alone stronger amendments that others have attached to it.

Some philosophers have proposed (or at least attributed the thought to Kant) that because we may get cognitive access to our consciousness only by making a reflective object out of it, this is what makes it "our" consciousness in the first place. I believe this proposal involves a mistake that can be corrected by seeing that it can be true of a state of awareness that, without any objective reflection on the self having taken place, the state is structured by the form“I think that x,”and therefore is already in a personal, even if implicit, sense an instance of "one's" consciousness. In that sense it is an instance of a kind of "self-consciousness" even if it is not expressly consciousness directed to “a self” or explicitly a "consciousness of consciousness.”Instead of working as a point against Kant, all this rather just recalls his notion of “first-level”representations that are specifically "one's own," i.e., not “nothing to me.” One could also draw upon SAT here, and argue that such a first-level state is structured by an "I think" because there can appear to be no point in calling the original state "one's own" if it could not be really connected with other similar states that can be (even if they need not actually be) reflectively, i.e.,"transcendentally" represented. However, even without insisting on SAT -- that is, without assuming that all such "I think" states must be"really" accessible - one could rather simply accept that self-consciousness is marked by what some have called a special phenomenon of "self-familiarity"i.e., that one is immediately and infallibly acquainted with oneself in a way that precedes any appeal to objective criteria. Dieter Sturma has developed an explication of Kant's view along these lines,and thus has spoken of the self in this sense as a "quasi-object." Strawson and Powell have also understood “transcendental self-consciousness” in these terms, as involving the fact that I can't be mistaken about something's being (presently) my experience. Note, however, that the fact that one can't be wrong in this way when one thinks of oneself still does not show that, for any experience one is having, one really can always think of it as such.So, without commitment to SAT, one could invoke the more modest idea of self-familiarity and still give a sense to regarding a simple“I think that x (e.g., "I think that it is warm")18 state as something that is already a kind of self-consciousness (because, at the least,if one did think of it, one couldn't be mistaken about the self having it),even if it is precisely not a reflection upon a distinct "object-self."

One source of resistance to this modest view of apperception may come from the belief promoted, for example, by the interpretation of Dieter Henrich, that separate and prior awareness of the identity of the self as an enduring object (a person) underlies and discloses the conditions sufficient for the objective unity of experience that is asserted in the Transcendental Deduction. But what Kant claims is only the opposite: that any objective assertion about one's own unity must meet the general necessary conditions of any objective unity; for a set of representations to be ascribed to my [one] consciousness, they must meet whatever are the conditions for any one [objective] consciousness. This is not the inverse and absurd claim that for any representations to be ascribed to one consciousness they must be ascribed specifically to my consciousness or my self. And even though the doctrine of apperception also says that any objective unity requires the possibility of the correlative unity of "a" consciousness this unity of consciousness is, as the Paralogism teaches, a formal unity, not necessarily the unity of a particular objective self. Nonetheless,current interpreters often propose Cartesian arguments that attempt to demonstrate the objectivity of our experience as a consequence of knowledge of one's own personal identity, a strategy that oddly presumes such knowledge has a privileged status. Whatever the attraction of such arguments for the project of defeating skepticism,they seem quite contrary to Kant's argument in the Deduction.

If one turns to the American tradition, and in particular to Patricia Kitcher's important new work，Kant's Transcendental Psychology (hereafter =“PK”), one finds an interpretation that appreciates most of these points while trying to use Kant's notion of apperception to explain the peculiarity of self-consciousness. Kitcher proposed that we understand self-consciousness not primarily in reflective terms but rather in terms of a“system”of mental states that process and combine information. While I am very sympathetic to her suspicions of some views that are close to the Strong Apperception Thesis, I am not so persuaded of the exegetical or intrinsic validity of her positive account of apperception.

Kitcher's general concern is with aspects of apperception that do not bear directly on the nature of the self. On her account, apperception is to be understood as a very general cognitive task of synthesizing data for knowledge, a task which can be shown to involve for us various connections which go beyond the mere associative relations of what is derived from sensation. Kant's "transcendental psychology" is then the account of the specific kinds of cognitive capacities that these connections require (e.g., forms of intuition and schematization, cf.PK, 209). Kitcher sees that the emphasis, within Kant's Transcendental Deduction, on the unity of apperception can sometimes make it appear as if Kant is concerned originally with the self as such and is arguing from conditions for knowledge of one's self to structures of the world at large. But, as her analysis of the B Deduction shows, Kant's argument is rather that“synthetic connection and so the categories are crucial for the entire range of our cognitive capacities, however, and not just for our ability to ascribe states to ourselves.Moreover, the categories and synthetic connection are not sufficient for self-ascription”(PK,127; cf. 92-4).

Although the main argument of the Deduction, according to Kitcher thus does not concern the self specifically (but rather the general application of the categories to our empirical domain), the notion of apperception that is central to that argument can still be examined for its implications about the self. In particular, Kitcher proposes that Kant uses the phenomenon of apperception to establish a real unity of self, one that defeats Hume's skepticism, although Kant does not go so far as to make the mistake of claiming to know a "metaphysical" subject, the self as, e.g., a simple immaterial being. Transcendental psychology provides a set of positive descriptions of a“thinking, or better, knowing, self”(PK,22), a self that Kitcher insists is the phenomenal self,even though the method she says one must use in learning how to describe it is philosophical and not straightforwardly empirical. It proceeds by determining what is needed for our epistemic tasks given certain "facts" about our mode of cognition,e.g., that one aspect of it (sensibility) is characterized by receptivity rather than spontaneity. Since these "facts" are so general, they can be said to underlie experience rather than to be found "in" it, and because the determination of what they involve gives us“general specifications for a mind capable of performing various cognitive tasks,”(PK,26) the knowledge that is generated can be called philosophical even if it "has striking affinities with empirical psychology"(PK,26).

One advantage of Kitcher's discussion is that she stresses throughout that Kant's direct concern is not with“experience”in any loose and primitive sense, but rather specifically with"cognitive" experience,i.e., representation that can be“something to" somebody and involves representing objects and making judgments (PK,134—5).Also, unlike those who insist on SAT, Kitcher sees that Kant need not assume that reflexive awareness is an"inseparable component of what it is to perceive, remember, etc. (PK 107). She allows that we do not have to be able to be in direct reflective touch with each one of our own mental states, although she adds - and here she still maybe going too far - they need to be indirectly "connectible" to our actual present consciousness by some string of thoughts (PK 103,113,118f). (At PK 119,134, and 138, she claims that each conscious state must have an epistemic impact on others, and, at PK 136, she recommends determining the "size" of the self by what it has access to at any moment-and thus it can vary over time.) In any case, by stressing that Kant does allow precognitive states (what she calls"non-representative sensations", PK 68,114; cf. also 136-8, on parts of one's life that may fail to leave "traces"), or layers of such states within cognitive consciousness, Kitcher has room for the thought that the self is more than its cognitive states, and that there may be a point in attaching a state to a real subject when it is only a state of obscurely desiring, or moving, or dreaming, i.e., even at a time when no conscious cognition is going on. Nonetheless, even if the self extends beyond its higher mental states, the question remains of what specifically constitutes this mental unity of a real subject. That is,one can ask not only "objective"questions about what the general structures of cognizable objects must be (or what must be the specific structures of bodies, in order for mental states to be ascribed to them in a way that others can accept) but also "subjective"questions about what general structure or capacities we must have in order to be able to know such objects.

The main subjective issue that concerns Kitcher here is the need to prove, against Hume's skepticism, that the mind does have a real unity. She contends that Kant's notion of apperception provides the necessary proof: since a state of apperception is a complex unified representation, one which involves other prior representations, the resultant unity of apperception testifies to an "existential dependence" of our mental states. She concludes that since there has to be such a necessary real relation presumed in any cognitive representation (i.e., any representation with a content about putative objects), then there is a non-question-begging argument that shows the self is itself a real unity, contrary to all that the Humean is allowed to say on his own principles.

Here are some objections to this intriguing argument. One objection is that the Humean already has all that Kitcher claims, since on one version of Hume's own account association is itself a causal process involving real relations between states of mind. One response Kitcher has to this gambit is to translate Hume's text (PK,250, n. 35;cf. Hume's famous phrase, "if the first object had not been, the second never had existed"') into a weaker claim, into a claim that says only that various sequences of mental states have occurred in constant conjunction. But whether or not Hume should be taken to mean only the weaker claim, one can still object that it is not clear exactly what kind of stronger claim can be warranted by the phenomenon of apperception as opposed to association. Certainly association itself, like apperception, could be (and usually is) understood as involving a real dependence of mental states. Thus, it can be argued that I would not be able associatively to image a pink object if I hadn't seen something like it, say a pale red one; just as one could say，as Kitcher would want to, that I could not have the apperception, "I think that the cat is on the mat" without having had “cat” and "mat" thoughts. But then whatever one thinks are the substantive claims about the mind that follow from apperception, it could be contended that just as much follows already from association.

On the other hand, one could also contend, skeptically, that at any moment I don't really have to have prior and separately existing"cat”and“mat”thoughts; I could just perceive or project all these thoughts at the same time (a similar argument,incidentally, could be used to challenge SAT). It is true that Kant speaks of the synthesis of apperception as drawing upon given "prior" thoughts，but a skeptic could contend that the actual existence of such past thoughts is not necessary. After all, that past existence is not even clearly relevant, for suppose one rather had at this moment a thought which was not exactly the same as the prior one, but rather something quite close to it that just happened to“fly into one's head." There is no reason to think that this must impugn the content of the particular apperceptive thought. On Kitcher's own account，what really counts is the content of the states that are combined in apperception (cf.PK 103),so it is not clear that where the content comes from really should matter. Kitcher speaks of the need for a judgment to be "generated" from particular intuitions (PK 110), but there is no proof that this could not be a logical construction rather than a process of temporally separated states;and, after all, even if there is a complex causal background to a present thought, one needs a complex present thought in any case.

One could even allow that a present thought has some relation to a past idea,as in Kant's own example of the analogy of the mind with a series of 'elastic' balls that transmit properties over time without ever forming a real unity. Kant's point here is precisely that there could be“a series of substances of which the first transmits its state together with its consciousness to the second . . . and yet it would not have been the same person in all these states (A364n.; note that this passage does suggest Kant is thinking in terms of sameness of person as requiring at least sameness of substance). So for Kant himself a "real unity" of mind over time does not turn out to be entailed by just any individual apperceptive act as such. And, for any state of mind that is claimed to require a real connection of distinct temporal states, it should be possible to substitute one very complex present state, one that need not be really connected with anything outside it; the content should be all the same. (I bracket“twin earth”problems in order to avoid anachronisms here.) Of course, one could counter that this is to make too much of skeptical hypotheses which neither Kant nor we really take seriously. But unfortunately for Kitcher's argument, this would be a fair response for any context except the present one, where the claim is to be able to present an argument with premises that even a skeptic would have to accept. It is true that if one were facing a skeptic who presumes that all a mind could contain at any moment were absolutely isolated and simple ideas (cf.PK,115), then the “fact” of apperception would be hard to account for -- but that is a very odd kind of skepticism.

In sum, if the mind does require a real unity, apperception alone, as Kitcher understands it, does not seem sufficient to establish this. It is significant that a key passage at A121 that she cites (PK79) stresses that our cognitive representations require something more than association not because otherwise they would be wholly without content, but rather because they would give us mere "accidental heaps." Kant's reply to Hume here would appear to turn on an appeal to not a mere unity but to a necessity. (At PK 102, she cites A107, which makes a similar claim, and she notes the distinction between establishing actual connections and necessary ones, but she does not note the implications of the possibility that Kant's notion of apperception is tied to claims of necessity.) Kitcher knows that Kant makes this extra claim, but she plays it down because she is so suspicious of it. She may be right that the claim has problems, but, without it , Kitcher has a much harder time of trying to show what it is specifically about transcendental apperception that yields special knowledge about the self that the Humean can't have.

Another response Kitcher might take here would be to claim that all along the"real unity" of mind that she finds proven by the phenomenon of apperception is something that she, as a Kantian，precisely does not want to be understood as the identity of an enduring thing, rather than as merely a continuity of a certain kind of causal process. The arguments of the Paralogisms might be understood in this light as meant to liberate us from thinking of the self as a mysterious substratum of noumenal substance. But even if one does not want to get involved with such mysterious thoughts, there are intrinsic and textual difficulties (as C.T.Powell has argued) in letting the self that exercises apperception become, in Kitcher's own provocative words, “no more than contentually interconnected systems of cognitive states”(PK,122).

A natural way to try to save the self as such here would be to bring it in as what Kitcher calls the "combiner," underlying the system of thoughts. However，unlike interpreters such as Prauss and O'Neilland Allison and Pippin, Kitcher resists appealing to a spontaneous "combiner" self as the ground of apperception. This resistance is understandable (if one wants to focus on skepticism), but I find unclear Kitcher's reason for dismissing such an appeal as not even "coherent" (PK,122). She contends "acts or processes of synthesis could not be performed by agents. They are unconscious activities" (PK122; l have not been able to follow her other claim, that making the self the "combiner" would involve an absurd appeal to "faculty identity"). This seems unclear because even if there are various kinds of "unconscious" syntheses, still, apperception, the paradigmatic synthesis, surely can be and is conscious. Of course, it may be that the absolute spontaneity of the combining act of the agent of apperception isn't determinable by consciousness alone. Elsewhere, I have argued that such an appeal is controversial for reasons that Kant himself stressed. But that is another issue. Problems about the scope of our freedom or consciousness should not affect our constant and Kantian belief that the self is more than a mere system.

Another advantage alleged to arise from not appealing to such an underlying self is that the supposedly obscure metaphysical notion of a“perfect bond”(i.e., a substantival one) between one's states(which Leibniz uses) could be replaced by an explanatory appeal to mere "synthetic connections" of states (PK 125). But this appeal simply does not guarantee a true "unity" of self, since then there is nothing to prevent many states from belonging to various selves; "synthetic connections" can be organized in all sorts of ways. Kitcher thus appears to be inviting a Parfitian conclusion that persons should not bethought of in terms of strict identity. However, she surprisingly resists Parfit's contention that there is "no fact of the matter about personal identity," and she counters that "synthetic connection" is a deep fact about mental life; it underlies mental capacities that enable us to be persons(PK133). I see this move as a flight from the original question of what is required for the unity of a particular person, and as a shift to the very different question of what kind of capacities cognizers in general need. This may be, as she says, an important "preliminary" question，but it does not resolve the original issue at all, and for that issue a real enduring self, a substance can still be the natural solution.

Although there is not room here to discuss at equal length C. T.Powell's Kant 's Theory of Self-Consciousness,one cannot help but wonder if it does any better on the problem of proving the unity of our consciousness. The remarkable fact is that although Powell criticizes Kitcher's belief that she has found an argument to defeat Hume, Powell still thinks that here we can get a result from Kant that, although weaker than Kitcher's conclusion, still does defeat Hume. The conclusion is that although apperception does not prove strict identity of a self“throughout”a number of mental states,nor even that there certainly are real connections between distinct states overtime, still we must "consider" these states to be connected. Thus experience must supposedly be thought of "as if" it is in a single consciousness, and to this extent Powell believes Hume is defeated. Moreover, although he concedes that the main argument of the Deduction could be written without reference to apperception (in his sense of "self-apperception"), and could go simply from the conceptualizability of experience to its objectivity (insofar as the former involves corrigibility, hence the possibility of a mistake, hence the idea of something distinct to be mistaken about), still he insists that the discussion of apperception has this value: it shows us, contra Hume, that experience must be "systematic" and "represented as" of one subject. Yet it is still unclear why it is claimed that experience"must" be like this, even if in fact it usually is this way. As we have seen in disputing SAT, we could suppose that there are experiences that really don't connect with others "systematically," or that for some reason can't actually come to our reflection as "of" one subject (even though, once they do enter reflection they do appear this way). Even if it were true that whenever we do represent our experience we regard it as being of "a" subject, the Humean could still contend this is just an understandable fiction. So although Powell goes a step further than Kitcher, backing off from literal "real connection" claims to psychological claims like "we can't but think in terms of a deep unity," the skeptic still seems safe. This need not be a severe problem in itself, but it is a problem for those who claim to have a philosophical refutation of Hume.

l will end by remarking on a common and very controversial exegetical problem. Kitcher argues that since apperception is a phenomenal aspect of the self，it (or its possible underlying substrate) can be nothing more than that, it could not be the sign of an absolutely free and, say, non-spatio-temporal side of the self (PK 140). The problem here is the strong conclusion that we can understand apperception "only" (PK 140) as phenomenal. This is simply to beg the question by ruling out from the beginning Kant's great confidence (which is tempered by a very shrewd critique of specific claims) that investigations of space-time and of morality will show that we can and have to affirm that there is more to the self than what is phenomenal. It seems that the main current interpretations of Kant on the self rush to questionable extremes-on the one hand overly bold assertions that there is a strong unity throughout experience, such as SAT asserts, and that this points to a kind of self-consciousness that is pivotal to the whole Deduction，or on the other hand, overly bold denials of any underlying substantival unity of self, a unity that may at least cohere with, if not point to, the metaphysical nature of the self that Kant did believe was necessary for his whole Critical, and that is to say practical, system. In particular despite all the focus on the "functional" character of reflection and apperception, that is stressed in all current schools of interpretation, Kant himself is clear enough that he does, after all, want to say something about what the "nature" or the stuff of the self is, namely that is definitely is not material, because it, like everything else, cannot be spatio-temporal in its ultimate essence. In his view it does not constitute itself through reflection or synthesis，but rather, like everything else, it first of all simply exists as the non-material Ding an sich that it is, no matter how even we may have to come to know it theoretically.