Kant's treatment of a priori concepts has occupied the attention of almost anyone who has tried to read him. Although the attention paid to pure concepts is understandable, I believe that an important difficulty in his account of empirical concepts has been widely overlooked.
Kant's discussion of empirical concepts in the Critique of Pure Reason is notable mainly for its brevity; apparently he did not think it necessary to provide an account of empirical concepts at all, but rather treats it as unproblematically true that we obtain them by abstraction from experience. In the Aesthetic he distinguishes a priori from a posteriori concepts and states about the latter, without elaboration, that empirical intuition is that "upon which these are grounded" (A47/B64). In the Deduction, similarly, he seems to assume that the relation between our empirical concepts and their "ground" or "proof" in experience is perfectly straightforward. The suspicion that he accepts some sort of abstractionist account of concept formation is strengthened by the many passages in the Logic where he holds that, for example, empirical concepts "spring from" experience, "from which they have been extracted as to their content" (1974:97; Ak. 9:92).
However, it is a mystery how Kant could consistently hold that "the empirical concept springs from the senses through the comparison of objects of experience" (ibid.). In the terms of the Critique of Pure Reason, where concepts are characterized as rules for synthesis (e.g., at A105), the problem is this: empirical concepts could only be abstracted from experience of a synthesized sensible manifold. Such synthesis requires the possession of rules of synthesis (concepts) and thus it seems that there would be experience to abstract from only if the concept has already been used to synthesize the manifold. Our mystery is how Kant could hold an abstractionist view of empirical concept formation if a major claim of his critical philosophy is precisely that it is only through experience of a synthesized manifold that "experience is brought into existence" (A86/B118).
This line of thought suggests that Kant cannot hold a simple abstractionist view where the acquisition of a concept depends on experience of a manifold synthesized by that very concept. Consider, however, a more sophisticated abstractionism. Because they stipulate necessary rules for synthesis in experience as a whole, in a sense the categories provide rules for the construction of empirical concepts: empirical concepts must conform to the forms of synthesis specified by the categories. Could not our most basic empirical concepts be derived from whatever experience synthesis according to the categories provides, these basic concepts, in turn, providing the basis for the acquisition of other empirical concepts? In much the way that the concept 'igloo' can be acquired on the basic of the more basic concepts 'house' and 'ice', could not the categories provide the material from which we abstract certain basic empirical concepts, which in turn make possible experience from which we can abstract other concepts?
The question, in other words, is whether or not some empirical concepts can be abstracted from a manifold which has been synthesized solely in accordance with the categories. If so, then perhaps the most basic empirical concepts simply "fall out" from the transcendental synthesis applied to the empirical manifold. However, two difficulties suggest that this modified abstractionist view would not escape our problem. Consider whether it is plausible to suppose that any unities whatsoever can be achieved through the application of the categories alone. Although much is unclear about the problem Kant took himself to be addressing in the Schematism, this much seems clear: the problem of unschematized categories is meant to show that effective application of the categorial rules of synthesis presupposes our using certain empirical concepts. But if whatever unities required by the categories are achieved only through the employment of empirical concepts, then the modified abstractionism could not get off the ground, viz. because synthesis in accord with the categories prior to the acquisition and application of empirical concepts could not provide any basis for abstraction.
Even if, pace the problem of unschematized categories, it were plausible to suppose that we simply are "programmed" to acquire some basic empirical concepts on the basis of our experience of a merely transcendentally synthesized manifold, it hardly seems plausible to hold that we are thus "automatically" provided all empirical concepts. But how, then, would we account for the acquisition of all the rest of the empirical concepts, which allegedly arise on the basis of experience conceptualized in accord with the basic ones? To claim that our understanding is programmed to allow the rest to fall out of this experience seems no better than holding that all fall out from the transcendentally-synthesized manifold. At the very least, both positions fail to do justice to Kant's insistence that empirical concepts arise on the basis of experience; they also threaten to collapse the distinction between pure and empirical concepts. Yet the modified abstractionist cannot admit that non-basic concepts are acquired in some other way, because then his account simply has nothing to say about how this is possible. On reflection, the modified abstractionist view seems to rest on a rather basic misconstrual of Kant's philosophy: the notion of concepts "falling out from" more basic unities is wholly inappropriate. Surely Kant did not maintain that the categories simply can be "deduced" from the unity of apperception and that empirical concepts are likewise "derivable" from the categories. Rather, his is the much weaker claim, which leaves room both for a sharp distinction between pure and empirical concepts and for the spontaneity of the understanding, that the categories must conform to the unity of apperception and empirical concepts must conform to the categories.
Insofar as any empirical concepts are acquired in response to differing "empirical matter" in the manifold, and are not simply programmed to fall out of the necessary forms of synthesis provided by the categories, the same old question arises: how is it possible to account for the capacity of the understanding to synthesize in accordance with a novel concept, if not through the application of rules of synthesis which we could apply only if we already possessed that very concept?
The central question is at what stage in the synthesis of the manifold a capacity for appropriate differential response comes into play. Where in Kant's picture of mental activity might such a capacity be located? An obvious candidate is the faculty of the imagination.
It will be useful to approach Kant's discussion of imaginative synthesis through his doctrine of the blindness of intuition. Consider his famous dictum that "thoughts without content are empty, intuitions without concepts are blind" (A51/B75). In exactly what sense are unconceptualized intuitions "blind"? Is Kant making the strong claim that we are given in intuition-i.e., given prior to conceptualization-no experience of objects' features? The worry, of course, is that if sensibility presents us with "bare particulars," if we aren't given in intuition information about particulars' features or relations, then it seems impossible that from this the faculty of understanding could construct empirical concepts.
Unpacking Kant's claim about the blindness of intuitions requires coming to grips with a rather daunting number of difficult Kantian notions, among them sensation, perception, matter, and form. In rather crude terms his general picture is this: concepts and intuitions are the two sorts of representations we synthesize to form experience. Intuitions are connected with the sensibility, the faculty of receptivity or capacity of receiving representations. Through intuitions, "objects are given to us by means of sensibility" (B33/A19). These empirical intuitions rest on "affectations:" objects affect our minds in certain ways to which we are receptive; when we receive representations, sensations are produced in us. The reason why without intuitions concepts are empty is that "in no other way can an object be given to us...[and consequently] all thought must...relate ultimately to intuitions" (ibid.). Concepts, on the other hand, are the products of our faculty of understanding or spontaneity. In the Critique of Pure Reason at least, spontaneity is-along with receptivity-one of the "two fundamental sources of the mind...[from which] our knowledge springs" (A50/B74); it is "the power of knowing an object through these [given] representations" (ibid.), "spontaneity in the production of concepts" (ibid.). Intuitions without concepts are blind because, as the representations occurring prior to synthesis and judgment, they are in an important sense undetermined. This is because (excepting whatever structure is provided by space and time, the forms of sensibility) receptivity provides simply the matter of intuitions but not the form through which we are able to experience empirically real objects in space and time.
How might Kant respond to our worry about our ability to notice similarities? It bears repeating that the crucial issue is in what sense, exactly, are intuitions without concepts are undetermined? At one extreme, Kant's claim might be that intuition provides us with no awareness of any determinate features of objects. If this is what he means, then Kant must also be committed to holding that we cannot experience (and thus cannot notice) resemblances without conceptualizing our intuitions. Although by itself this is not tantamount to the circularity objection I sketched above, it nonetheless seems perilously close to it because it is difficult to understand how Kant could (1) maintain that we must conceptualize to have any experience of features and nevertheless (2) deny that (for example) we need to conceptualize with the concept of red to be able to notice objects' redness. At the other extreme, intuition might "give" us experience of particulars and their features. Although this position avoids the circularity objection, it seems to make vacuous Kant's doctrine of the blindness of intuitions and, indeed, seems to be in serious tension with his account of the role of judgment in experience.
An intermediate reading might be this: Kant's claim that intuitions are blind amounts to the relatively weak claim that, although intuitions provide us with experience of properties (e.g., of color), intuition alone is insufficient to provide us with experience of objects, i.e., with experience of empirically real objects in space and time. The mere possession of simultaneous and successive sensations, images and feelings, does not, for Kant, qualify as having an objective experience. In Kant's technical sense, an objective experience is one that is or is not veridical because it presents one with putative objects ostensibly arranged in events and states of affairs. This suggests that intuitions could be blind for Kant in the sense that, although intuition provides "subjective" experience of sensations, feelings, etc., through it we are not given any experience of objects.
This interpretation goes some distance towards understanding how Kant might answer the circularity charge. In effect, this reading provides room for an account on which we are capable of noticing at least some resemblances independently of our conceptualizing according to the corresponding empirical concepts. However, although it might provide the basis for a non-circular account of the acquisition of basic sensory concepts (e.g., redness, hardness, etc.), it seems that many other empirical concepts (e.g., that of a tree) must have their source in experience of objects. Moreover, it is clear that mere subjective experience is insufficient for the acquisition of even simple sensory concepts. Even if Kant's account of our subjective experience does not appeal to our use of empirical concepts, it is still a separate question whether the same is true about his account of our ability to notice resemblances among our sensory experiences. Even if intuition does in some sense provide us with experience of features, it is still unclear whether it could also provide us with the means of noticing similarities between the features of successive and simultaneous sensations?
As we have seen, a crucial question is exactly what is provided by "matter" and not by "form." At A20/B34, Kant characterizes form as "that which so determines the manifold of appearance that it allows of being ordered in certain relations." A central doctrine of the Aesthetic is that "form must lie ready for the sensations a priori in the mind" (ibid.); this claim is crucial to his thesis that space and time, as forms of intuition, are transcendentally ideal. Matter, on the other hand, is associated with sensation and empirical intuition. Kant characterizes it as "that in the appearance which corresponds to sensation" (ibid.).
The strongest of the three interpretations of the blindness of intuitions requires two further claims, viz., that intuitions are no more contentful than sensations and that sensations themselves provide us with no experiences of features. Sometimes Kant does write as if the matter of sensation is wholly undetermined. Consider this series of definitions Kant provides at the beginning of the Aesthetic:
The effect of an object upon the faculty of representation so far as we are affected by it is sensation. That intuition which is in relation to the object through sensation, is entitled empirical. The [i.e., prior to synthesis and judgment] undetermined object of an empirical intuition is entitled appearance. That in the appearance which corresponds to sensation I term its matter. (A20/B34)
Note, however, that even in this passage, and despite Kant's talk of an "undetermined object of empirical intuition," the textual evidence is mixed. Kant also speaks of the intuition being "in relation to an object through sensation," which suggests that intuitions do carry with them some sort of experience of objects, or at least of the features of objects.
On balance, I think that Kant's doctrine of the blindness of intuitions should be given a relatively weak interpretation, according to which unconceptualized intuitions provide us with subjective experience of features. This is to provide an answer to the first of the two questions raised above. Yet our other questions are still unanswered; it is still unclear whether Kant thinks that intuition provides us with the resources to notice resemblances among different sensations we experience, or even whether we have to notice resemblances at all in order to acquire sensory concepts. It is therefore still unclear whether Kant can provide a non-circular abstractionist account of empirical concept acquisition.
Now, finally, we can turn to the suggestion that the imagination might provide the key to this problem. In his discussion of the threefold synthesis of the Subjective Deduction, Kant appears sensitive to our worry about circularity. In the Deduction in B, for example, he writes of the categories being involved in a "transcendental synthesis of the imagination....[which is] an action of the understanding on the sensibility" (B152). Unfortunately, it is not clear whether he believes that another action of the understanding required for recognition is conceptualization in accordance with empirical concepts. He characterizes these acts of synthesis as actions of "the reproductive imagination, whose synthesis is entirely subject to empirical laws" (ibid.). The synthesis of reproduction, which provides for our being "conscious that what we think is the same as what we thought a moment before" (A103), seems most relevant to our problem. He writes:
If we were not conscious that what we think is the same as what we though a moment before, all reproduction in the series of representations would be useless. For it would in its present state be a new representation which would not in any way belong to the act whereby it was to be gradually generated. The manifold of the representation would never, therefore, form a whole since it would lack that unity which only consciousness can impart to it....[S]uch consciousness...must always be present; without it concepts, and therewith knowledge of objects, are altogether impossible. (A103-4)
Here Kant comes tantalizingly close to addressing explicitly the worry about circularity. Although his main concern in this section is with pure and not empirical concepts (viz., here he seems to anticipate claims about the categories and the transcendental unity of apperception), nevertheless one point he makes seems to be precisely that recognition of similarities in successive experiences cannot be accounted for wholly in terms of our use of empirical concepts. Such an account would be circular, he suggests, because this recognition itself helps make it possible for us to apply concepts to intuitions. Alas, there are few hints in the Critique of Pure Reason of how Kant thinks it is possible to avoid such circularity.
The Logic promises to provide some of the details of how Kant thinks we acquire empirical concepts. Unfortunately, it quickly becomes clear that the account he gives here only attempts to explain cases where empirical concepts are acquired on the basis of other, previously acquired concepts. Consider this example:
In order to make our presentations into concepts, one must be able to compare, reflect, and abstract....For example, I see a fir, a willow, and a linden. In firstly comparing these objects, I notice that they are different from one another in respect of trunk, branches, leaves, and the like; further, however, I reflect only on what they have in common, the trunk, the branches, the leaves themselves, and abstract from their size, shape, and so forth; thus I gain the concept of a tree. (1974:100; Ak. 9:94-5)
It seems to be no accident that in this example it is taken for granted that we have available such concepts as trunk, branches, leaves, etc. How, if we don't already have those concepts, can we notice that the three trees "are different from each other in respect of" those features? Clearly, we cannot, at least not if we take seriously Kant's claims that intuitions are singular representations and concepts represent common features of our sensory experience. In other words, Kant can only explain how we acquire concepts by appealing to a process of noticing similarities and differences that itself requires the employment of other concepts. Obviously, such an account cannot provide a full theory of concept acquisition; what it cannot do is show how it is possible for us to acquire concepts in the first place.
What we require is a fuller account which isn't just logical in Kant's sense of explaining how we acquire one concept by using others. Thus far, however, our examination of the Critique of Pure Reason and Logic has shed almost no light on how Kant might answer the Kantian question of how empirical concepts are possible for us. If sensibility only gives us singular representations, then it seems that we can only recognize similarities through comparison and reflection by bringing intuitions under concepts.
I shall end by pointing out a Kantian resource which promises to provide the basis for a Kantian account of empirical concept formation, viz. his conception of reflective judgment introduced in the first Introduction to the Critique of Judgment. Reflective judgments may offer precisely what we require: they are cognitions in which the unification of the manifold is not determined by a concept the understanding already possesses, but rather which is guided by a regulative principle to generate a new empirical concept. Reflective judgment promises a new way of understanding how judgment, imagination, and the understanding might work together to construct empirical concepts.
Absent some such account, there is a deep problem about how Kant could account for the possibility of empirical concept formation. To the extent that we can discern a worked-out account in Kant's texts which we have considered, it seems either circular, because is seems committed to the position that we can't can notice similarities without already possessing the corresponding concepts, or woefully incomplete, because it offers no hint of how such a capacity might be possible.