LET US NOW CONSIDER the other skeptical impulse which, according to Kant, this time in the Prolegomena, awoke him from his dogmatic metaphysical slumber and gave his philosophy a new direction: “David Hume’s reminder.” The general nature of this reminder is clear enough from the Prolegomena: Hume’s skeptical reflections concerning causation. It is much less clear, though, exactly what it was in Hume’s rather various skeptical reflections on causation that awoke Kant, when it did so, and how it did so. Let me therefore attempt to answer those questions.
Close examination of the Prolegomena—and of a corresponding discussion of Hume in the Critique1—reveals that, although Kant does not distinguish between them very clearly, he in fact has three quite distinct Humean views concerning causation in mind. First, he has in mind Hume’s argument that particular causal connections and laws cannot be known a priori by reason alone but only from experience, since their denial is never self-contradictory or inconceivable in the way that is required for a priori knowledge to be possible.2 Second, he has in mind Hume’s position that the component idea of necessity which is contained in the idea of a cause must be traceable, like all other ideas, to a corresponding antecedent impression, and that the search for such an impression reveals causal necessity to consist, not in a property of causally related items themselves, but instead in what Hume describes as the “customary transition of the imagination from one object to its usual attendant” in the subject’s thought.3 Third, he has in mind Hume’s view that the principle that every event has a cause (henceforth: the causal principle) can only be known “from observation and experience,” since (once again) its denial is not self-contradictory or inconceivable in the way required for a priori knowledge to be possible.4
The first of these Humean views seems to have exerted an influence on Kant by the mid-1760s at the latest, when Kant embraced it wholeheartedly in Dreams of a Spirit Seer (1766).5 Kant notes in the same work and in a letter to Mendelssohn from the same period that this principle concerning the knowledge of particular causal connections and laws carries strong negative consequences for traditional metaphysics’ claims to knowledge of the supersensible, because it rules out as illegitimate any claim to know about the causation of items of experience by supersensible items, or vice versa, or about causal relations between supersensible items themselves.6 The first part of this point (the exclusion of claims to know about the causation of items of experience by supersensible items) was especially important, because it precluded any metaphysics that, like Kant’s own before the period in question,7 presumed to infer from items of experience to supersensible items as their explanatorily required causes. This, then, was evidently Kant’s first significant debt to Hume.8
However, this first Humean view is clearly not what Kant mainly has in mind in the Prolegomena as the influence that awoke him from his dogmatic slumber in metaphysics. This can be seen from three related facts. First, Kant says that the Humean views which roused him were ones which only presented him with a problem to solve but that he “was far from following [Hume] in the conclusions to which he arrived.”9 But this does not characterize Kant’s attitude to the first Humean view, whose point he simply accepted in the 1760s,10 and continued to accept in the critical philosophy;11 and whose negative implication for causal inferences from experience to the supersensible he likewise accepted permanently.12 It does, on the other hand, characterize his attitude to the other two Humean views, those concerned with causal necessity and the causal principle. Second, the answers to Humean problems which Kant goes on to sketch in the Prolegomena are answers, not to the first Humean view, but rather to these other two Humean views—namely, in the preface, the Metaphysical/Transcendental Deduction of the a priori concepts of the understanding (including the concept of cause or causal necessity),13 and, in sections 27–30, the Analogies of Experience (which seek to prove the causal principle, among other principles). Third, a passage from the Critique identifies Hume’s commitment to these other two views as constituting the skepticism that produces the transition from dogmatism in metaphysics to criticism.14 It is therefore on these other two Humean views that we should mainly focus.
In order to see when and how these other two Humean views impinged on Kant’s development, we need to return to the writings of the precritical Kant. Four years after publishing Dreams of a Spirit Seer in 1766 Kant went on to publish the Inaugural Dissertation of 1770. In many respects this work marked a huge advance beyond Dreams of a Spirit Seer towards the critical philosophy. In particular, Kant now for the first time sharply distinguished the faculties of sensibility and intellect (or understanding), developed the phenomena-noumena distinction, and, above all, advanced the theory that space and time, as pure intuitions and pure forms of sensible intuition, are merely ideal—all in ways which would require little modification in the Critique. However, in another respect the Inaugural Dissertation represented a major slide backwards. For Kant was once again indulging in the sweet slumber of supersensuous metaphysics. Thus according to the Inaugural Dissertation the intellect furnishes us with a knowledge of noumena or “things which cannot by their own nature come before the senses of the subject,” including in particular God.15 If Kant’s letter to Garve alludes to his awakening by Pyrrhonian skepticism from the long metaphysical sleep that ended in the mid-1760s, his remarks on Hume’s reminder in the Prolegomena allude mainly to his awakening by Humean skepticism from this briefer metaphysical snooze in the Inaugural Dissertation.
The initial impetus behind this second awakening seems in fact to have come from Kant himself rather than from Hume, however (in crediting Hume he errs on the side of generosity). In a famous letter which he wrote to Herz in 1772 Kant has two second thoughts about a supersensuous metaphysics of the sort that he had adopted in the Inaugural Dissertation.16 First, he has a worry concerning the ability of concepts to refer to supersensible objects in the manner envisaged by the Inaugural Dissertation. His worry is as follows: It is clear enough, he argues, that a concept can refer to an object if the object is the cause of the concept, as in the case of concepts belonging to sensibility (sensibility having been defined in the Inaugural Dissertation as the capacity of the subject to be affected by the presence of an object).17 It is also clear enough that a concept can refer to an object if the concept is the cause of the object—as would notionally be the case for concepts which belonged to a divine archetypal intellect. But since the intellect’s concepts of supersensible noumena as envisaged by the Inaugural Dissertation refer in neither of these two ways, it is unclear how they can refer at all, as the Inaugural Dissertation supposed they did.18 Second, he has an additional worry concerning the possibility of knowing about supersensible noumena in the manner envisaged by the Inaugural Dissertation. His concern this time is the very simple one that (over and above problems about reference) it is unclear how one could have a knowledge of things not attained through experience of them.19
The two remaining Humean views concerning causation correspond closely to these two worries about metaphysics, and seem to have contributed to Kant’s development away from the metaphysics of the Inaugural Dissertation some time shortly after Kant wrote to Herz in 1772. As many commentators have argued, it is likely that the publication in 1772 of Beattie’s An Essay on the Nature and Immutability of Truth in a German translation played an important role here, because this for the first time made Hume’s view concerning the causal principle accessible to Kant.20
The exact role played by the two Humean views in the development of Kant’s thought is by no means easy to determine from the scant evidence available. But it seems in both cases to have had two sides in tension with each other: On the one hand, the general principles which lay behind these Humean views suggested to Kant more refined ways of reformulating the two prima facie objections which he had already himself raised against supersensuous metaphysics in the letter to Herz. On the other hand, and in tension with that, the specific subject matters of Hume’s views—the concept of causal necessity and the causal principle—furnished Kant with especially instructive test cases for the (now refined) prima facie objections, suggesting that there had, in fact, to be something wrong with them, and thus pointing the way towards a metaphysics which might survive their attack. Let me now develop this interpretation in a little more detail.
Consider first Hume’s analysis of the concept of causal necessity. The general principle which drove Hume to give this analysis was the principle that every idea requires a corresponding antecedent impression as its source. Kant had in fact long been familiar with such a position.21 Moreover, he had himself long found it tempting, although never accepting it unequivocally.22 Hume’s principle seems to have reminded Kant of this position, and to have thereby alerted him to the possibility of a more radical version of his own worry in the letter to Herz concerning the concepts of supersensuous metaphysics. Setting aside the merely notional theological case of concepts which cause their objects, Kant’s own worry in the letter had basically been that concepts not derived from sensibility and thereby causally dependent on their objects could not refer. Now Hume with his “No impression, no idea” principle had gone one step further than that: concepts not derived from sensible antecedents could certainly not refer because they could not even exist. It appears that Kant found Hume’s principle sufficiently plausible to take this more radical worry about the concepts of supersensuous metaphysics seriously in addition to his own original worry.23 Accordingly, these two worries would later together constitute the fundamental problematic of the Metaphysical and Transcendental Deductions in the Critique. For example, both are in Kant’s mind at the very start of the Transcendental Deduction, where he first says that the problem involved is whether a priori concepts have “a meaning, an imagined significance” (i.e., the more radical Humean worry),24 and then that it is how they can “relate to objects which they yet do not obtain from any experience” (i.e., Kant’s own original worry).25
To say that Kant found Hume’s “No impression, no idea” principle and the radical worry about the very existence of the concepts of supersensuous metaphysics which flowed from it plausible is not, however, to say that he ultimately accepted them.26 Indeed, I suggest that one reason why he found Hume’s application of the principle to the particular concept of causal necessity so enlightening was precisely that this application afforded a sort of prima facie reduction to absurdity of the principle: Thanks largely to the relative perspicuity of this particular concept’s existence and character, the application of the principle to it could be clearly seen to lead unavoidably to a conclusion that was false—namely, Hume’s conclusion that the concept of causal necessity, if it exists at all, expresses, not a relation between causally related items themselves, but instead merely the mind’s reaction to constant conjunction. As Kant saw things, the concept clearly did exist, but it equally clearly did not express that. Therefore the Humean principle had to be false.27
A second reason why Kant found Hume’s example enlightening, I suggest, is that he drew from it similar instruction concerning his own original, less radical worry about the ability of concepts lacking sensible antecedents to refer. Hume’s examination of the concept of cause, once corrected by Kant so as to excise what he considered to be its manifestly mistaken derivation of the essential component concept of necessity from a subjective impression, had shown it to be a concept containing an essential component concept not derivable from sensible antecedents.28 Yet it must have seemed clear to Kant on reflection that the concepts of cause in general and causal necessity in particular nevertheless do succeed in referring, since, despite their failure to be derivable without remainder from sensation, they are found to be instantiated in experience in a broader, everyday sense of “experience” (after all, we do in some sense see rocks causing windows to break, etc.), and are moreover indispensable to the outlooks of common sense and natural science alike.29 Consequently, Kant’s own original worry that concepts could only refer insofar as they were derived from sensible antecedents now looked as though it had to be mistaken.
To put these two points in another way: As long as one had focused on such hazy, lofty, and dispensable metaphysical terms as, for example, “God,” Hume’s radical worry and Kant’s own original worry had looked quite plausible. Perhaps such terms really did lack meaning or at least reference.30 But once one realized that the same two worries would apply, if at all, just as much to such seemingly perspicuous, experiential, and indispensable concepts as “cause” as well, it was rather the worries that came to look misconceived.
Consider next the other relevant Humean position concerning causation alluded to by the Prolegomena: knowledge of the causal principle must be based on experience, since its denial is not self-contradictory or inconceivable in the way required for a priori knowledge to be possible. Once again, this position seems to have played two roles in Kant’s development in tension with each other, this time in relation to the prima facie objection already raised by Kant himself against supersensuous metaphysics that it is unclear how the intellect could achieve knowledge concerning things of which it has no experience. On the one hand, the general doctrine behind Hume’s position suggested to Kant a more cogent reformulation of his own prima facie objection. On the other hand, the specific subject matter of Hume’s position, the causal principle, provided a particularly instructive test case for the now-refined prima facie objection, suggesting that there had in fact to be something wrong with it.
The general doctrine that lay behind Hume’s position on the causal principle was what has come to be known as Hume’s “fork,” long familiar to Kant in the form that Hume had given it in the Enquiry. As articulated in the Enquiry, the “fork” consisted of the following three claims. First, all known truths are divisible into two kinds: on the one hand, “relations of ideas,” defined by Hume as propositions “intuitively or demonstratively certain,” by which he seems to mean such that their denials either are or imply contradictions; on the other hand, “matters of fact,” defined as all the rest. Second, while relations of ideas “are discoverable by the mere operation of thought,” or in other words a priori, matters of fact are only knowable by means of experience, or a posteriori—since, beyond present sensory experience and memories of past sensory experience, only causal reasoning can furnish us with this kind of knowledge, and it turns out that causal insight itself “arises entirely from experience.”31 Third, and consequently, if it is found that a proposition—Hume adds with feigned casualness, “of divinity or school metaphysics for instance”—is neither certain by virtue of its denial being or implying a contradiction nor known by means of experience, then we may “commit it . . . to the flames: for it can contain nothing but sophistry and illusion.”32 Since it seemed clear to Hume that the causal principle was not known by virtue of its denial being or implying a contradiction, his only option, short of rejecting it as altogether “sophistry and illusion,” was to say that it was known by means of experience.33
How did Hume’s “fork” enable Kant to reformulate in a more cogent fashion his own prima facie objection to supersensuous metaphysics from the letter to Herz, namely the objection that it is unclear how the intellect could achieve knowledge concerning things of which it has no experience? The “fork” shared with this Kantian objection an assumption that a posteriori knowledge of things is, generally speaking, philosophically unproblematic. But the “fork” also made explicit, and evidently alerted Kant to, something that his own objection had overlooked, namely the fact that there is in addition a substantial and well-defined class of a priori cognitions which are unproblematic: those that are true simply in virtue of the law of contradiction (in Hume’s terminology, relations of ideas; in Kant’s, analytic judgments).34 Kant of course readily took this important qualification to heart.35 In this way the “fork” enabled him to formulate a more refined version of his own original objection to supersensuous metaphysics. The objection was now no longer that it was unclear how there could be a priori knowledge generally, but instead that it was unclear how there could be a priori knowledge which did not belong to the unproblematic class founded on the law of contradiction (in Kant’s terminology: a priori knowledge which was not “analytic” but “synthetic”). This objection, already in effect advanced by Hume as the thrust of his “fork,” now reappears as the central puzzle of the Critique, on the settlement of which, according to the work itself, the very fate of metaphysics depends: “How are a priori synthetic judgments possible?”36
In tension with this influence exerted by its underlying general principle, the “fork,” however, Hume’s position concerning the causal principle was also important for Kant because the causal principle, specifically, provided him with an illuminating test case for the now-refined objection to supersensuous metaphysics (in effect, the objection that there is no such thing as synthetic a priori knowledge). Kant agreed with Hume that the causal principle was a clear case of genuine knowledge—for it was supported by confirming instances in experience, and was moreover fundamental to the outlooks of common sense and natural science alike.37 He also agreed with Hume that the principle was obviously synthetic. However, the conclusion to which Hume had been unavoidably driven when he had applied his denial of synthetic a priori knowledge to the principle, namely that the principle was not known a priori but only a posteriori, seemed to Kant to be clearly false—especially because of the strict universality pertaining to the principle.38 Consequently, scrutiny of the example of the causal principle afforded, in Kant’s view, a strong prima facie disproof, or reduction to absurdity, of the denial that there could be synthetic a priori knowledge.39
To put the point in another way: As long as one had focused on such lofty and questionable synthetic a priori claims as, for example, that God had such and such characteristics or performed such and such acts, then the worry that synthetic a priori knowledge was impossible had looked plausible. Perhaps one could not know anything of that sort. But once one realized that this worry, if it applied at all, equally challenged such empirically confirmed and fundamental convictions as our conviction in the causal principle, then it was the worry that came to look misconceived.
In short, the two Humean skeptical views about causation enabled Kant both to refine the objections to supersensuous metaphysics which he had himself articulated in the letter to Herz and to show that, even so refined, there seemed in fact to be something wrong with them, so that the prospects for metaphysics were after all brighter than they had implied.
The latter side of this twofold process is reminiscent of a strategy that is characteristic of the Scottish common sense school: when a philosopher like Hume advances a philosophical principle which entails a deeply counterintuitive consequence, one holds fast to the intuitive position that is under attack and converts his modus ponendo ponens argument against that position into a modus tollendo tollens argument against his principle (i.e., one converts his “If p then q, p therefore q” into “If p then q, -q therefore -p”). Beattie in particular uses this strategy, and I suspect that he influenced Kant here (in other words, I suspect that his influence on Kant went beyond merely providing him with information about Hume’s position on the causal principle, that it also had this more philosophical dimension). This might seem unlikely in view of Kant’s sharp defense of Hume against the Scottish common sense school in the Prolegomena.40 However, in its support, consider, for example, Kant’s remark in his 1799 open letter on Fichte’s Wissenschaftslehre that “the Critique . . . is to be understood exclusively from the point of view of common sense [des gemeinen . . . Verstandes].”41 Kant’s line of thought thus includes what could be called a common sense moment.
However, unlike the common sense school, Kant conceived this as only a prima facie response still in need of reinforcement and elaboration (this, I take it, is the real reason behind his defense of Hume against the common sense school in the Prolegomena). Until Hume’s tempting principle “No impression, no idea” and Kant’s own original tempting principle that (in effect) concepts can only refer insofar as they are derived from sensibility had received a more decisive and thorough refutation than this sort of appeal to particular counterintuitive consequences could constitute, they continued to represent a puzzle about the status of the concept of cause and other a priori concepts (an a priori concept being for Kant by definition one that fails to conform to Hume’s principle, one that is not derivable from sensation, at least not without remainder).42 This puzzle bore both a stick and a carrot for the discipline of metaphysics. The stick was that until the principles were more decisively and thoroughly refuted they still in some, even if diminished, degree threatened the claims of a priori concepts to exist and refer, and hence also the standing of metaphysics. After all, proponents of the principles in question might just bite the bullet in the face of the commonsense response and insist that the counterintuitive consequences flowing from them for such terms as “cause,” and hence for common sense and science, simply must be correct. The carrot was the prospect that a more satisfactory solution to the puzzle might, in addition to dispelling this residual threat, afford new insights into such a priori concepts and into disciplines like metaphysics which depended upon them.
Similarly, until the tempting denial that there could be synthetic a priori knowledge had been given a more decisive and thorough refutation than this sort of appeal to particular counterintuitive consequences could constitute, it continued to represent a puzzle concerning the status of the causal principle and other apparently synthetic a priori principles. And, once again, the puzzle bore both a stick and a carrot. The stick was that, until more cogently answered, it continued to cast at least some doubt on the claims of synthetic a priori principles to constitute knowledge, and hence on disciplines such as metaphysics which relied on them. After all, as in the conceptual case, the philosopher who denied that there could be synthetic a priori knowledge might just bite the bullet and accept the particular counterintuitive consequences involved.43 The carrot was that a full solution to the puzzle might afford new insights into principles of this kind and into disciplines such as metaphysics which relied on them.
In sum, I suggest that Kant’s reflections in or shortly after 1772 on Hume’s treatments of the concept of causal necessity and the causal principle in these ways brought him to a deeper (though not yet final) understanding than he had achieved alone in his letter to Herz of two major puzzles bearing on the possibility of metaphysics: a double puzzle about the existence and reference of a priori concepts, and a puzzle about the possibility of synthetic a priori knowledge. It is mainly this process that Kant has in mind in the Prolegomena when he credits Hume with having roused him from the slumber of dogmatic metaphysics and given his thought a quite new direction.