2. Against the Heuristic Reading

The ideas of reason (the soul, the world, and God) are individually examined in the three main chapters of the Dialectic (the Paralogisms, the Antinomies, and the Ideal). Their conclusions dovetail: none of the ideas has a legitimate constitutive use, i.e., a use that issues in cognition of the supersensible object it represents. But this does not preclude these concepts from having a positive role in our cognitive activity. In the Appendix, which concludes the Dialectic, Kant says that they have a legitimate regulative use, which contributes to the systematic unity of our cognition of nature (A644/B672).

Variants of the heuristic reading share the basic view that the regulative function of the ideas is to guide scientific investigation toward a certain goal. The sense in which the ideas are necessary is this: without their guidance, we could not possibly expand empirical cognition of nature in such a way as to achieve systematic unification of it (systematic unification is often understood in terms of integration of scientific theories into a single unified theory). Different versions of the heuristic reading vary in the way they conceive of the character of this guidance and the goal toward which it directs our inquiry. We will turn to representative versions of this reading below.

The heuristic reading has two implications which will become important as points of contrast with my own reading. First, the use of the ideas is limited in scope. They are necessary only for a specialized use of our cognitive faculty, namely, scientific investigation and theorization. They are optional in relation to our cognitive activity in general. For most of us who do not directly engage in scientific research, the ideas have no significance (from the theoretical standpoint). Second, the ideas do not play a role in enabling cognition of any kind, be it experience (i.e., empirical cognition grounded in understanding and sensibility) or another kind of cognition distinct from it. Scientific cognition, on the heuristic reading, differs from common cognition only in extent and in degree of systematicity. In these two respects, the cognitive role attributed to the ideas sharply contrasts with that of the categories or the pure concepts of the understanding. The categories are involved in every ordinary exercise of the faculty of cognition, and they are conditions of the possibility of cognition of a certain kind, namely, experience.

The heuristic reading is intuitively attractive. However, one should note that it is not unequivocally confirmed by textual evidence. More importantly, it does not satisfactorily address some of the central interpretive problems. First, there is what we might call the Relevance Problem: how could the concepts of unconditioned—and, hence, noumenal—objects have any relevance for empirical scientific inquiry? While it is not implausible that progressive systematic unification would require some conception of a goal or endpoint, it is far from obvious that such a conception needs to involve any objects that are unconditioned or lie beyond possible experience. Second, it is unclear how the ideas can give rise to normative principles that can steer the course of scientific research in any specific direction. Call this the Normativity Problem. The ideas are (on the fact of it) descriptive rather than prescriptive. They represent objects, ones in which we are to regard “all the connection of things in the world of sense as if they had their ground” (A681/B709). But what difference would it make if the scientist working in the laboratory were to be committed (or not committed) to such an attitude towards the object of her investigation? What course of action is enjoined by such an as-if attitude? Finally, even if we can make sense of the idea that scientific inquiry necessarily presupposes some concept of the unconditioned as its guidance, it remains to be explained why Kant attributes this role to all and only the three specific ideas (soul, world-whole, and God). Call this the Specificity Problem.

Let’s take a closer look at the Relevance Problem. This problem can be sharpened in terms of two different gaps that seem to bar the ideas from having any role in scientific investigation. One is a gap between the conditioned objects of such investigation and the unconditioned objects of the ideas. Scientific inquiry is exclusively concerned with appearances and thus conditioned objects. The unconditioned condition of appearances is not an object of scientific inquiry. Even a complete unified physical theory, for Kant, would only teach us about appearances and not their unconditioned conditions (their noumenal grounds). The other is a gap between the unconditioned cognition that scientific inquiry seeks and the unconditioned object represented by the ideas. Granted that a complete system of cognition must involve an unconditioned cognition (an epistemic first principle), the difficulty here is that an unconditioned cognition need not be about an unconditioned object or directly map onto one. An unconditioned cognition can be understood as a maximally general cognition from which less general cognitions can be derived.Footnote ⁴ But it is not obvious that a cognition cannot be maximally general unless it has something unconditioned as its object.

Consider Zuckert’s (Reference Zuckert, Breitenbach and Massimi2017) interpretation. The ideas function as “placeholders” for yet-to-be-discovered objects of scientific investigation. The reason why such placeholders must refer to unconditioned objects is that they are unknowable, and our concepts of them are “almost empty.” In this way, the ideas do not encroach upon empirical investigation. They do not pretend to say anything substantive about the object we may discover yet represent “something that is (projected as) there (a purported object), beyond the current state of knowledge, and either the source of, part of, or identical to a rationally unified world” (98). The ideas thus “give only very general, vague guidance concerning how to investigate nature” (100). For Zuckert, the gap between objects of the ideas and objects of scientific investigation is not a problem but rather what enables the ideas to fulfill their function.

There are two problems with this proposal. First, indeterminateness does not seem to require the representation of something unconditioned. It seems possible to represent a projected object of inquiry with the requisite openness without representing it as unconditioned. Why could the scientist not think of the next object she will discover both indeterminately and as conditioned (and so on indefinitely)? Given the second gap mentioned earlier, the scientist can still (it seems) rationally assume that nature is “rationally unified” in the sense that cognitions of its kinds and laws will ultimately form a system based on one highest principle (e.g., the fundamental law of nature) without simultaneously assuming the existence of any unconditioned object, in nature or outside of it. Second, to the extent that this reading addresses the Relevance Problem, it does so at the cost of exacerbating the Normativity and the Specificity Problem. If the relevance of the idea of the unconditioned is explicated in terms of indeterminateness, the worry is that its direction might become so indeterminate or generic (“Keep going! Do not give up!”) that it makes no procedural difference to scientific investigation. It could still be useful (for instance, psychologically) but not indispensable. Emphasis on indeterminateness also sits uneasily with the fact that Kant attributes the regulative function to ideas representing specific objects rather than a generic idea of something unconditioned. While Zuckert maintains that the specific ideas are nonarbitrary because they “hypostasize” different demands of reason,Footnote ⁵ her view leaves it ambiguous whether the ideas themselves are individually necessary or whether they are Kant’s expository devices for making salient different aspects of a single demand of reason.

Massimi (Reference Massimi2017) and Spagnesi (Reference Spagnesi2023) have defended readings that aim specifically to tackle the Normativity Problem. Both argue that the role of the ideas is purely prescriptive; they serve as standards internal to rules the conformity to which entails no commitment (illusory, rational, or hypothetical) to the existence of any unconditioned objects. For Massimi, the rules in question are rules for the correct use of the understanding, which secure “the ability to communicate [judgments] so that inter-conversational agreement can be reached” (79–80). But she does not explain why such agreement requires concepts of the unconditioned. Spagnesi characterizes these rules likewise as rules for the use of the understanding but sees them as operative at the level of empirical synthesis (rather than at the intersubjective level). They instruct the understanding on how to “initiate and continue the regress to the conditions of what is conditioned in appearance” (12). For example, the idea of the soul as a simple substance is a constituent of the following rule: “If inner appearances are given as composite, then initiate and continue the regress in the series of their conditions in accordance with the idea of simple substance” (11). We begin with something given in experience (inner appearances) that is conditioned (composite); the rule then instructs us to “look for an empirical condition that stands in the same conditioning relation in which a simple substance would stand to composite appearances” (11). This can only be partially and progressively realized, for nothing unconditioned (e.g., simple or uncaused) can be given in experience. The rule effectively enjoins us to keep looking for a further condition and so on indefinitely.

This reading does not satisfactorily explain why the ideas of the unconditioned are necessary. If to continue the regress “in accordance with the idea of simple substance” just means to continue to look for a condition that is simpler but still composite, why does the rule need to refer to something (absolutely) simple? More generally, any rule prescribing an indefinitely progressive search for ever more basic conditions, it seems, can be stated in a purely relative way, e.g., “Whenever something composite is given, find something simpler out of which it is composed.” The idea of something simple or compositionally unconditioned seems dispensable. Further, Kant holds that the regress in empirical synthesis, which we are enjoined to initiate and continue, can never be completed; we can never encounter any condition among appearances that is not itself conditioned. A version of the first gap reemerges between the indefinitely progressive synthesis of conditions and its completion. On Spagnesi’s reading, the ideas serve as standards of completeness of regress. Their role is to instruct us when to stop. But if stopping is not a possibility, why does it make sense to include such instruction in the rule at all?

Finally, let us consider recent responses to the Specificity Problem. In response to this problem, commentators typically turn on the assumption that the regulative roles of the ideas are specific to a domain of objects, corresponding to one of the special sciences. Here is how Frederick Rauscher formulates the basic idea:

On this view, the ideas function as concepts of the highest genus under which all of the objects of a given special science fall. Kraus (Reference Kraus2020, 235ff.) has further developed this view with respect to the idea of the soul. The idea of the soul “serves—according to the principle of homogeneity—as the highest genus concept that first defines the domain of psychological phenomena.” It delineates the domain of “inner nature” and guides the empirical investigation and classification of mental states and powers toward a comprehensive system of psychological predicates.

Despite its intuitive appeal, this reading raises several problems. First, the ideas do not seem fit for the role of genus concepts; they are not concepts of classes of things but of individual entities. Further, the ideas are concepts of unconditioned conditions. The primary relation between the object of an idea and an object of experience must be one of real conditioning, i.e., real explanatory relation. But the genus-species relation between concepts does not map onto the real conditioning relation between things. To ascend from a lower genus (gold) to a higher one (metal) is to ascend from a less general concept to a more general concept under which it is contained but not necessarily from a concept of something conditioned to that of its real condition (e.g., from a composite to its simpler constituent). The reading is thus faced with a version of the second gap (cognition versus object).

These considerations also put into question Rauscher’s and Kraus’s fundamental assumption, namely, that the discussion of the principle of homogeneity and other “logical” principles in the first part of the Appendix directly bears on the topic of the regulative use of the ideas, discussed in the second part. In particular, it raises a question about the place of these principles in relation to Kant’s distinction between the logical and the real use of reason. Let me explain. The logical use of reason concerns logical conditioning (or containment) relations among concepts and is governed by the “Logical Maxim,” which enjoins us “to find the unconditioned for conditioned cognitions of the understanding, with which its unity will be completed” (A307/B364). Reason in its logical use thus seeks to unify empirical concepts into a single conceptual hierarchy. In contrast, the real use “deals with objects” (A306/B363); it concerns real conditioning relations among objects. The real use is governed by the “Supreme Principle of Reason,” which asserts the existence of an unconditioned condition for every conditioned thing (more on this below). It is relatively obvious that the ideas of reason pertain to the real use, for they are concepts of the unconditioned conditions assumed by the Supreme Principle and their positive use concerns our cognition of real conditioning relations among objects. How about the logical principles?

My suggestion is that the logical principles belong to the logical use of reason and thus do not immediately bear on how we should understand the regulative use of the ideas (though there may be ways in which the two uses of reason can be viewed as complementary).Footnote ⁷ For the role of the logical principles is to guide the formation of a genus-species hierarchy of natural kind concepts according to their relative degrees of generality, i.e., their logical conditioning relations.Footnote ⁸ Note that this reading of the logical principles is compatible with the two following claims. First, that natural kinds are individuated by their essences, and thus a genus-species hierarchy implicates at least one kind of real conditioning relation—one between an essence and the properties it grounds, e.g., Gold’s essence and its ductility.Footnote ⁹ Be that as it may, this conditioning relation is not what reason tracks in subsuming a lower concept under a higher one: the containment relation between <metal> and <gold> does not map onto any real conditioning relation between distinct entities. Rather, it is the containment relation among concepts that is the topic of the use of reason under the guidance of the logical principles. Second, that the logical principles give rise to “transcendental” presuppositions (A660/B688ff.) that nature is unified in such a way that the search for ever higher genera and ever lower species is possible. These are indeed claims about the object (nature), but it does not follow that the purview of the logical principles extends to the real use. The reason is that these presuppositions do not obviously commit one to the existence of objects that are unconditioned in the real sense. The presupposed unity of nature is (prima facie, at least) compatible with the non-existence of the unconditioned in the real sense. If so, the logical principles do not after all express the same demand that drives reason in its real use (in other words, they are not specifications of the Supreme Principle). These considerations are admittedly inconclusive, but they indicate that the burden is on the proponents of the standard view to explain the connection between the logical principles and the Supreme Principle before they can rely on the former to explicate the regulative use of the ideas. A full case for the view I have suggested, however, will have to remain a task for future work.

Perhaps we could read Kraus’s view more charitably as saying that the highest genus concept of empirical psychology is not the idea of the soul itself but the derivative concept of soul-states.Footnote ¹⁰ But this leads to another problem: if psychology is an empirical investigation of inner appearances, why must it regard mental states as states of a noumenal substance, which in principle cannot be an object of experience? Empirical mental states are accidents of a phenomenal substance.Footnote ¹¹ Again, we stumble upon the first gap (conditioned versus unconditioned). I do not wish to deny that, for Kant, the regulative use of the idea of the soul involves regarding inner appearances “as if” they were states of a noumenal soul (Kant says this at A672/B700 and A682–3/B710–1). What is at issue is how this claim is to be interpreted. I disagree with Kraus that it is best interpreted as the claim that the idea of a noumenal soul (or of its states) functions as the highest genus concept of empirical psychology. Textual counterevidence to Kraus’s reading comes from Kant’s discussion of the highest genus concepts of natural science in the Metaphysical Foundations. He says that the highest genus concept of the science of extended nature (physics) is the empirical concept of matter. Its counterpart, the highest genus concept of the science of thinking nature (psychology), is the empirical concept of a thinking being (4:470). This is obviously not identical to the a priori concept of a noumenal thinking substance, just as the concept of matter is not identical to the idea of the world-whole.Footnote ¹²

We have seen that previous approaches to the regulative use of the ideas, despite their varied emphases and orientations, encounter versions of the same basic difficulty: they are unable to close either or both of the two aforementioned gaps (conditioned versus unconditioned; cognition versus object). I want to argue that this is due to a shared feature, namely, their failure to recognize that the ideas in their regulative use do not simply guide the extension of cognition of the understanding but enable cognition that is distinct in kind from it. Once we recognize that the ideas are conditions of the possibility of a different kind of cognition, we will be in a better position to address the basic difficulty and other difficulties examined in this section.

3. Comprehension as Cognition of Reason

The heuristic reading explicates the role of the ideas in terms of a kind of cognition that is possible independently of employing them: cognition arising from the cooperation of understanding and sensibility (experience or empirical cognition in the narrow sense).Footnote ¹³ Scientific cognition is not distinct in kind from “common” empirical cognition. We can acquire a great deal of empirical cognition without using the ideas. To follow their guidance, on this reading, is not a matter of doing something different (cognitively speaking) but doing the same thing better. The use of the ideas makes it possible for our cognition to attain “greater extension alongside greater unity.” But what is thereby further extended is the same kind of cognition and the same kind of unity. On such a view, reason appears to be merely auxiliary to the understanding—a “merely subordinate” rather than a self-standing faculty (A305/B362). Kant’s characterization of reason as a cognitive faculty distinct from the understanding and supreme seems exaggerated and misleading.

I will argue that although the ideas are not conditions of the possibility of experience, they are conditions of the possibility of cognition of a different kind.Footnote ¹⁴ Recall Kant’s remark that “all our cognition starts from the senses, goes from there to the understanding, and ends with reason” (A298/B355). My suggestion is that in passing from the understanding to reason, we pass from one kind of cognition to another, more demanding kind of cognition. Thus, Kant’s characterization of reason as the supreme faculty of cognition can be vindicated.

Consider Kant’s claim that understanding and reason are responsible for two different kinds of unity: “unity of reason” (Vernunftseinheit) and “unity of understanding” (Verstandeseinheit):

Kant explicitly says that reason grounds “unity a priori through concepts” that is not of the same kind as the unity of understanding or “the unity of a possible experience” (A307/B363). The unity of reason does not apply directly to objects of experience but rather to the concepts of the understanding. Such concepts are unified under higher-order representations, which Kant calls “principles.”

That the unity of reason is different in kind from any unity that can be achieved by the understanding implies that it is not grounded in an act of synthesis guided by the categories. If so, the unity of reason must be grounded in a different type of synthesis guided by some other a priori concepts. This is precisely how Kant characterizes the role of the ideas in this passage:

The role of the ideas with respect to the unity of reason is understood in analogy with the role of the categories with respect to the unity of understanding. I take this to mean that the ideas are rules of synthesis that ground the unity of reason. This synthesis differs from the synthesis performed by the understanding in two important respects. First, the synthesis of reason aims at absolute totality, which cannot be thought except through an idea of the unconditioned. Whereas the understanding progressively and incompletely synthesizes the manifold of given representations into the experience of objects, reason, in some sense, anticipates the completion of its synthesis through the ideas. The ideas “express” the unity of synthesis of such a kind that depends on an idea of its completion. Second, the synthesis of reason is characterized as a “synthesis of conditions.” It specifically concerns the connection among conditions and thus with conditioning relations: “all pure concepts have to do generally with the synthetic unity of representations, but concepts of pure reason (transcendental ideas) have to do with the unconditioned synthetic unity of all conditions in general” (A334/B391, emphasis added).

Appealing to the distinction in kind between the unity of understanding and of reason only goes some way toward establishing my thesis that reason enables a distinct kind of cognition. For the proponents of the heuristic reading could concede that reason is concerned with unity that is distinct in kind from the unity of understanding, while denying that reason enables a distinct kind of cognition. They could maintain that the “unity of reason” refers to an ideal that our cognition can approximate but never realize (to realize it would require us to cognize the unconditioned conditions of appearance). And as long as such unity is not realized, we have not yet attained any cognition that is different in kind from the cognition of the understanding.

To reply to this objection, we may begin by noting that Kant does identify a distinct kind of cognition that has its source in reason and is possible for us. He gives it the name ‘comprehension’.

Kant makes three claims here. First, comprehension is different from mere cognition of the understanding (Verstehen), which he refers to as “empirical cognition” and “experience” in the first sentence. Second, comprehension involves the ideas of reason, which parallels the way the categories are involved in experience. The suggestion is that the ideas are conditions of the possibility of comprehension. Third, experience is necessary but not sufficient for comprehension. This indicates that comprehension is distinct in kind from experience. Comprehension differs from experience not only in virtue of having a common structural feature (the same kind of unity) to a greater degree but also in virtue of possessing an additional structural feature—a differentia—which experience lacks.

The fact that comprehension seldom appears in the first Critique could cause suspicion of overemphasis. Two points can be made to allay this worry. First, as we will see shortly, there is reason to think that “comprehension” is synonymous with two other terms that Kant uses more frequently: “rational cognition” (Vernunfterkenntnis) and “cognition from principles” (Erkenntnis aus Prinzipien, cognitio ex principiis). Kant introduces reason as the faculty of cognition from principles, which “is something entirely different from mere cognition of the understanding” (A302/B358). He uses “cognition from principles” alternately with “rational cognition,” which refers to a priori cognition through reason (cf. A836/B864). Second, “comprehension” repeatedly appears as the term for a distinct kind of cognition in the Jäsche Logik and various transcripts of Kant’s logic lectures. It will be useful to consider some of these passages in detail.

In the Jäsche Logik, comprehension appears as the highest among what Kant calls “degrees” (Grade) of human cognition, which are distinguished from one another “in regard to the objective content of our cognition in general” (9:65).Footnote ¹⁶ These form a ladder of progressive stages of human cognitive achievement. With minor deviations, the same Stufenleiter is found in several logic lecture transcripts. In all of them, comprehension is ranked highest. Despite the terminology, the difference between any two degrees is one in kind. Consider, for instance, the first degree, (merely) “to represent something,” and the second degree, “to represent something with consciousness.” The difference between representation with and without consciousness is obviously one in kind and not in degree. I take this to carry over to the difference between understanding and comprehension as well. Understanding comes in the middle of the ladder (fifth in the Jäsche). To understand (verstehen, intelligere) something is “to cognize something through the understanding by means of concepts, or to conceive” (9:65). In contrast, to comprehend (begreifen, comprehendere) something is “to cognize something through reason or a priori.” We can comprehend only what we have previously understood, but not all objects of our understanding are comprehended: “one can conceive much, although one cannot comprehend it.” The use of reason in this regard is more limited than that of the understanding: “The field of understanding [Verstehen] or of the understanding [Verstand] is thus in general much greater than the field of comprehension [Begreifen] or of reason [Vernunft]” (9:65).

Three claims can be made on the basis of these passages from the logic lectures. First, the range of cognitive achievements possible for a human cognizer extends beyond experience (or “understanding” in these passages) to higher or more demanding kinds of cognition. Second, this does not entail that we have cognition of objects other than objects of possible experience. Comprehension presupposes understanding, and we can understand (i.e., subsume under concepts) only objects that can be given in intuition. Although comprehension differs in kind from understanding, it is not directed at a different kind of objects. Third, the extension of our cognition beyond experience (with respect to the kind of cognition but not objects) is made possible by the use of reason.

To comprehend something, Kant says, is to cognize it through reason or a priori. But what does it mean to cognize something a priori? There are two possible interpretations. One construes “a priori” in the familiar Kantian sense of “nonempirical” or “independent of experience.” The other construes it in the more basic, Aristotelian sense: “a priori cognition” refers to “cognition from explanatory grounds”. This notion of a priori cognition is a descendant of Aristotle’s epistêmê. In the Posterior Analytics, epistêmê is defined as what we possess when “we know of the explanation [aitia] because of which the object holds that it is its explanation, and also that is not possible for it to be otherwise.”Footnote ¹⁷ The Aristotelian notion of a priori cognition was still current in the eighteenth centuryFootnote ¹⁸ and, as we will see shortly, in Kant’s works.

The first interpretation appears consistent with the example given in the Jäsche Logik: “Nothing can be comprehended more than what the mathematician demonstrates, e.g., that all lines in the circle are proportional” (9:65). Here one might be inclined to regard the sense in which mathematical cognition is a priori to be that it is nonempirical. But the Wiener Logik passage suggests that comprehension can be gained through observation and experiment, so empirical cognition, too, can count as comprehension if it meets certain criteria. Kant gives two examples: comprehending the lunar eclipse through the moon’s path and comprehending the combustibility of gun powder through saltpeter’s driving force (24:846). Neither of these phenomena can be cognized independently of empirical observation, which shows that comprehension is not limited to non-empirical cognition.

Textual evidence speaks more strongly in favor of interpreting “a priori” in the Aristotelian sense. The Logik Blomberg states that to comprehend something is “to have insight into something sufficiently,” and to have insight “into what gold is I must investigate one of its marks in particular and abstract from it its ground. E.g., why it does not rust, why is it ductile, heavier than others” (24:135, italics added). To comprehend gold thus involves cognizing the explanatory grounds of its various properties. The previous examples from the Wiener Logik can be read along the same line: “I can comprehend the eclipse of the moon, because I know the moon’s path” (24:846); the moon’s path is the explanatory ground of the eclipse. I can comprehend gun powder’s combustibility through saltpeter’s driving force, for this is the explanatory ground of combustibility.

Further, Kant links rational cognition or cognition from principles with cognition from grounds, as attested in the Metaphysik L₂: “rational cognitions, which are from principles <ex principiis>[…] arise from this, that one cognizes their grounds and obtains them a priori” (28:531). In the Lectures on Pedagogy, Kant characterizes reason, one of the “higher powers of understanding,” in terms of cognition from grounds: “By means of reason one grasps grounds” (9:476).Footnote ¹⁹

Kant’s attribution of cognition from grounds or a priori to reason betrays his adherence to the Leibnizian conception of reason as the faculty for discerning the explanatory connection between ground and consequence. Consider this passage from Leibniz’s New Essays:

An “a priori” ground is one that enables us to know not only that p is true but also why p is true (it is the “cause” of the truth itself). And reason, as is for Kant, is the faculty through which we cognize such explanatory connection. This explains why cognizing something “through reason” is equivalent to cognizing it “a priori” (cf. 9:65). In the first Critique, Kant writes that “a necessary problem of pure reason” concerns “the complete comprehensibility of what is given in appearance,” for which “we need its grounds but not its consequences” (A411/B438).

Textual evidence we have considered suggests that comprehension is after all ubiquitous in Kant’s writings though under three different designations: (i) “comprehension” (Begreifen), (ii) “rational cognition” (Vernunfterkenntnis), and (iii) “cognition from principles” (Erkenntnis aus Prinzipien, cognitio ex principiis). These terms refer to the same thing, namely, a kind cognition of an object that depends on a grasp of its explanatory grounds.

We may conclude that when Kant characterizes comprehension as “a priori” he means to identify it with a priori cognition in the Aristotelian rather than the more familiar Kantian sense. In fact, he considers the latter as derivative of the former. In the lengthy discussion of apriority at the opening of the Metaphysik Mrongovius, the basic sense of “a priori” is Aristotelian: “If I begin from the consequences, then I cognize something a posteriori; if I begin from the grounds, then I cognize a priori” (29:748).Footnote ²¹ Suppose God is the ground of the world; if I cognize the world from God, then I cognize it a priori. Cognition “taken from experience [is] eminently a posteriori” because experience only presents us with “last consequences” of inner determinations of things. The Kantian sense of “a priori” thus turns out to be derivative of the Aristotelian sense: “from now on when we call cognitions a posteriori, then we are always understanding these to be from experience, because experience contains the last consequence of our cognition, for which we seek grounds by means of reason” (29:748).

To cognize something a priori in the Aristotelian sense is to cognize it as a consequence of an explanatory ground. In the pre-Critical Nova dilucidatio, Kant calls this ratio cur (literally “reason why”), which is either ratio essendi (ground of being) or ratio fiendi (ground of becoming).Footnote ²² To cognize something a priori is to cognize it in such a way that yields explanatory understanding of its features, i.e., understanding why it has those features. To comprehend an object, a, one must not only cognize that a is F but also understand why a is F, by grasping the explanatory ground of a’s F-ness and the explanatory relation in which it stands to a’s F-ness. Comprehension thus can be defined as the kind of cognition that essentially involves an explanatory understanding of an object in virtue of grasping its explanatory ground.

Let us return to the objection we set out to respond to. Recall that the proponents of the heuristic reading can accept that

1. (a) the unity of reason is distinct in kind from the unity of understanding,

2. (b) the regulative use of the ideas is aimed at the realization of the unity of reason,

   while denying that

3. (c) the regulative use of the idea enables cognition that is distinct in kind from mere cognition of the understanding.

For they can insist that the unity of reason refers to a goal or a “task” (Aufgabe) that we employ the ideas to represent but which we cannot possibly accomplish. And insofar as we have not fully realized the unity of reason, we have not yet achieved anything that differs in kind from cognition of the understanding. We can now reply to this objection by appealing to another distinction in kind that Kant makes, between comprehension and understanding, both of which are possible for us. The unity of reason, then, does not characterize a kind of cognition that we merely approximate but in principle cannot achieve. On the contrary, it characterizes a kind of cognition that is achievable for us, although our possession of it is always partial insofar as we cannot cognize the complete explanatory ground or the unconditioned condition of any object.

But is comprehension possible for us? Kant does not dispute the commonsense assumption that we are in possession of some explanatory understanding of objects, even though we lack knowledge of their complete explanation. The Jäsche Logik states that “all our comprehension is only relative, i.e., sufficient for a certain purpose; we do not comprehend anything without qualification [schlechthin]” (9:65). While the field of comprehension is narrower than the field of understanding, it is by no means empty. Kant only denies the possibility of comprehension without qualification or absolute comprehension. This is an expression of epistemic humility. To comprehend a’s F-ness absolutely or without qualification would mean to cognize the complete explanatory ground or the unconditioned condition of a’s F-ness. But the impossibility of absolute or complete comprehension does not rule out the possibility of partial comprehension, which proceeds from a grasp of a partial ground or explanation.Footnote ²³ I will argue that even this modest sort of comprehension depends on the ideas of reason for its possibility.

4. Ideas as Pure Concepts of Comprehension

I have presented a case for my thesis that reason, through its pure concepts, enables a distinct kind of cognition, i.e., comprehension or cognition from principles. This is best characterized as the kind of cognition that essentially involves a type of explanatory understanding that depends on a grasp of real grounds. The exercise of the understanding on the sensibly given can yield an empirical judgment that things are a certain way, but not an understanding why they are that way. This requires the exercise of reason and the application of its pure concepts—“concepts of reason serve for comprehension, just as concepts of the understanding serve for understanding” (A310–11/B367). My aim in this section is to fill in the details of this reading by showing why it makes sense to think that the possibility of comprehension depends on these concepts, even though our comprehension always remains partial.

The ideas are concepts of unconditioned conditions in the real sense. Real (as opposed to logical) conditioning relations are explanatory relations that hold between two things insofar as one metaphysically depends on and is explained by the other.Footnote ²⁴ Something is conditioned if and only if it metaphysically depends in some respect on another thing (its condition) and is explained by it. Unconditioned conditions are metaphysical explanations that admit no further explanation and, hence, are explanatorily basic. The ideas thus are concepts of complete explanations (here and in what follows I use “explanation” exclusively in the ontic sense, i.e., explanations are things rather than cognitions). It is also important to note that an object can be unconditioned in one respect and conditioned some other respects. For instance, absolute simples are compositionally unconditioned, but they may be causally conditioned insofar as they have a cause (e.g., God). One obvious connection between the ideas of reason and comprehension is this: cognition of unconditioned conditions such as represented by the ideas would allow the cognizer to have complete comprehension of whatever it is they condition.

We comprehend an object when we cognize one of its conditions and the conditioning relation between them. But, as we have seen, Kant denies that we can have complete comprehension. For we cannot cognize any unconditioned object. For every conditioned object, every condition of it we can cognize is always in turn conditioned. Here we encounter a (by now familiar) gap between the unconditioned objects (of the ideas) and the conditioned objects (of possible comprehension). If our comprehension exclusively involves conditioned conditions, why should we think that concepts of unconditioned conditions have any role to play in relation to it?

I will argue that this gap can be closed. For partial comprehension, based on a grasp of partial explanation, presupposes the possession of a conception of complete explanation. To see that Kant endorses this condition on partial comprehension, consider the Supreme Principle of Reason: “When the conditioned is given, then so is the whole series of conditions subordinated one to the other, which is itself unconditioned, also given (i.e., contained in the object and its connection)” (A307–8/B364). The principle says that if something conditioned exists, a complete explanation of it also exists, which comprehends all partial explanations within it. Note that the Supreme Principle does two things. First, it affirms the existence of an unconditioned condition for any given conditioned object. Second, it relates the unconditioned condition (a complete explanation) to conditioned conditions (partial explanations) as whole to parts.

How is the Supreme Principle related to the exercise of reason that issues in comprehension? Kant says that the Supreme Principle governs the real use of reason. In my view, the real use of reason is the exercise of reason that underlies comprehension. The real use “deals with objects” (A306/B363), i.e., it concerns objects and their explanatory connections, whereas the logical use concerns cognitions and their inferential connections. Kant maintains that there would be no real but only logical use of reason unless we “assume” (annehmen) the Supreme Principle (A307/B364). This suggests that the Supreme Principle is constitutive of comprehension.Footnote ²⁵ For we have no real use of reason without making use of the Supreme Principle, and without the real use, no comprehension.

Commentators have argued that synthetic a priori principles, such as the Categorical Imperative and the Principles of Pure Understanding, are both constitutive and normative of the relevant kind of rational activity or the exercise of the relevant rational faculty.Footnote ²⁶ These principles are constitutive in the sense that they express the essential (albeit normative) structure of the activity in question. They set the standards according to which an instance of activity of the relevant kind is liable to being assessed, and being liable to being assessed according to those standards is a condition on counting as an instance of that kind of activity. For to engage in a rational activity of a certain type (or to exercise a rational faculty) is to follow the principles governing it or be guided by them in one’s activity. It is in virtue of being constitutive and normative in this sense that synthetic a priori principles are constitutive in another sense, namely, that of being conditions of the possibility of the kind of rational activity they govern.

I suggest that the relationship between the Supreme Principle and comprehension be understood along the same line. The Supreme Principle is the synthetic a priori principle governing the activity of comprehension.Footnote ²⁷ A cognition counts as an instance of comprehension only if it is liable to being assessed according to the standard set by the Supreme Principle. What is this standard? Recall that the Supreme Principle asserts not only that any conditioned object has a complete explanation but also that each of its partial explanations is comprehended within that complete explanation. The assumption of this principle can be understood as imposing a constraint on what counts as a (partial) explanation whose cognition contributes to (partial) comprehension. Roughly, for any object, x, we seek to comprehend, a different object, y, counts as a (partial) explanation of x only if there exists a complete explanation, z, of x such that y is a part of z. If such a constraint is involved in our guiding conception of comprehension, it follows that partial comprehension requires the possession of some conception of complete explanation.Footnote ²⁸ For without such a conception, we could not be guided by the Supreme Principle in exercising our capacity for comprehension. In this way, we can close the gap between the conditioned objects of our comprehension and the unconditioned objects of the ideas.

Kant does not offer an independent argument for the claim that partial comprehension requires possession of a conception of complete explanation. I have presented it as simply following from his assertion that the Supreme Principle is the principle governing the real use of reason.Footnote ²⁹ But the claim itself is not implausible considered as a claim about our commonsense conception of explanation (or reason why something is the case). Consider what it means to comprehend something in Kant’s sense. Comprehending a involves answering the question why a is F. An answer to this question must identify a b which explains (is a real condition of) a’s being F. But this gives an adequate answer to the initial question only if b itself does not in turn raise the same why-question, that is, only if b is a complete explanation or an unconditioned condition relative to that why-question.Footnote ³⁰ The point is that a grasp of what it is to answer the question why x is F depends on a grasp of what it is to answer it adequately. For y is explanatory of x’s being F only if grasping y constitutes a step toward an adequate answer to that question. If y does not bring us closer to this, the initial why-question is merely postponed but not answered to any extent. Thus, y is explanatory of x’s being F only if y is part of a complete explanation of x’s being F.

There is a children’s song in Thai that goes something like this: Why does it rain? It rains because the frog croaks. Why does the frog croak? Because its stomach hurts. Why (each time the melody repeats) does its stomach hurt? Because the rice is uncooked. Why? Because the firewood does not burn. Why? Because it is wet. And why? Because it rains! One way to understand the humor here is that none of the because’s, it turns out, is a genuine because. For none of them brings us any closer to an adequate answer to the question why it rains (but instead back to the question itself). So understood, the song presupposes that we (the listener) understand the very idea of explanation (or reason-why) in terms of complete explanation. My suggestion is that it is plausible to view the Supreme Principle, the a priori principle governing comprehension, as an explicit and philosophically regimented articulation of what is implicit in our pre-philosophical conception of an explanation (or reason-why).

According to our analysis, it is part of our guiding conception of comprehension that its possibility depends on the existence of an unconditioned condition.Footnote ³¹ But this does not entail that comprehension requires cognition of an unconditioned condition. Comprehending x involves taking x and its condition, y, to be conditioned by something else, z, which is unconditioned, but it does not involve cognizing z itself.Footnote ³² Relatedly, objects of possible comprehension (both the comprehended object and its condition) are exclusively sensible objects or objects of experience. Although comprehending such objects requires the ideas of the unconditioned, the relevant use of the ideas does not yield cognition of their unconditioned objects. Further, the claim that the Supreme Principle—and the ideas of reason, which arise from it (cf. A308/B364)—are constitutive with respect to comprehension is perfectly compatible with Kant’s claim that they are merely regulative with respect to experience. The regulative-constitutive distinction is applied relative to a kind of cognition (cf. A179–80/B222–23; A664/B692). A principle may be constitutive of one kind of cognition while being merely regulative of another (though very often Kant simply writes “regulative” as a shorthand for “regulative with respect to experience”).

By explaining why comprehension presupposes concepts of the unconditioned, we have addressed the Relevance Problem. What about the Normativity Problem? On the transcendental reading, the regulative use of the ideas does not generate normative principles that we can adopt in explicit deliberation, within the context of scientific investigation or otherwise. The ideas do not prescribe something like a procedure that we can follow. Nonetheless, there is a sense in which the ideas guide our cognitive activity, namely, by figuring in the principles (various specifications of the Supreme Principle) that express constitutive norms of the exercise of reason as the capacity for comprehension. They are involved in the conception of explanation that is implicated in our conception of comprehension and sets a condition on what counts as an instance of it.

But in what sense is such a conception normative for us at all if we cannot decide whether a given cognition meets the condition imposed by it? We cannot know whether any given object has an unconditioned condition. To that extent, we do not know whether any given object qualifies as an object of comprehension—whether it is indeed comprehensible.

Although we cannot use the ideas to “sort” our cognitions into those that qualify as comprehension and those that do not, they are still normative in the following way. The ideas determine how we ought to think about objects and our cognitions of them. They comprise a background conception of the world against which, and in conformity with which, our cognitive engagement with it unfolds. This is analogous to one way we can think of the categories as normative. The categories determine how we ought to think about objects (e.g., of alterations that they are governed by causal laws). The categories are “the form of thinking of an object in general” (A51/B75) or “the concepts of an object in general” (B128); they determine how we must think about an object for it to be an object of our thought at all. The ideas of reason add another layer to our conception of objects. The ideas form part of the concept of an object (not in general or überhaupt but) of comprehension. An object of comprehension is one that we can make sense in terms of the real explanatory nexus in which it is embedded. The concept of an object of comprehension involves the notion that any such object must be embedded within an explanatory whole, a system of interconnected explanations descending from a complete explanation.Footnote ³³

Further, the implicit conception of comprehension, which involves the ideas, functions as the guiding conception of our broader cognitive activity, to which belongs the exercise of the understanding. In this way, we can explain the sense in which the ideas are not only constitutive of comprehension but (precisely because of that) also regulative of experience. What this means is that they give direction to the use of the understanding that constitutes experience:

The unity toward which the understanding is directed is obviously the unity of reason, characteristic of comprehension. But what does it mean for reason to direct the understanding in this way?

Consider how experience is related to comprehension. First, comprehension presupposes experience. Experience constitutes “one part” of it (A310–11/B367). The logic lecture passages previously considered make clear that comprehension is not only more demanding than understanding (Verstehen) but also presupposes understanding of the same object. Kant also notes in different places (e.g., A306–7/B363) that the use of reason does not directly relate to objects but takes experience of objects, delivered by the understanding, as its input.

Second, comprehension is the end or telos of our entire faculty of theoretical cognition. This is suggested by the place of comprehension in the progression of our cognition as well as the status of reason as the supreme faculty of cognition. Reason completes the work of our cognitive faculty “on the matter of intuition” by bringing it “under the highest unity of thinking” (A298/B355). In the second Critique, Kant remarks that “reason, as the faculty of principles, determines the interest of all the powers of the mind but itself determines its own. The interest of its speculative use consists in the cognition of the object up to the highest a priori principles” (5:119). The faculties of the mind form a teleological system in which all their ends or interests are set by and subsumed under reason’s own.Footnote ³⁵ The implication is that comprehension (cognition from principles) constitutes reason’s self-given end (with respect to its theoretical use)Footnote ³⁶ and at the same time an overarching end toward which all the lower cognitive faculties are oriented.

If our faculty of cognition as a whole is teleologically oriented toward comprehension and comprehension cannot be acquired without experience of appropriate objects, it follows that we ought to focus the use of the understanding on acquiring experience that maximizes our comprehension. It is in this sense that reason “regulates” the understanding, namely, by determining what counts as a worthwhile use of it. From the viewpoint of the understanding, any experience is presumably as good as any other. Reason demands that we privilege experiences that are explanatorily relevant for any aspect of reality that calls for explanation, though it leaves it open how we might go about doing this (developing a new scientific instrument, modifying one’s theoretical framework,Footnote ³⁷ engaging in interdisciplinary collaboration, etc.). The ideas are involved in this demand to the extent that the notion of explanation that figures in it implicates a conception of complete explanation.

On the transcendental reading, the ideas guide the use of the understanding at a more abstract level than on the heuristic reading. They do not posit any particular objects as goals of scientific investigation nor make any concrete recommendation concerning its procedure. Rather, the ideas are constituents in the implicit conception of our cognitive activity that guides it at the broadest level (broadest because it picks out the overarching end of the faculty of cognition). Empirical natural science is one (but not the only) disciplined use of the understanding that responds to the demand reason makes on us. As an explanatory enterprise, science deepens our comprehension. But, if my reading is correct, the significance of reason and its ideas extends beyond science, narrowly construed. They govern, “in a fundamental and unnoticed way” (A329/B387), all cognitive activities that contribute, directly or indirectly, to the explanatory understanding of reality. These include pure natural science, mathematics,Footnote ³⁸ philosophy, and, perhaps, more mundane and less disciplined attempts to make sense of ourselves and the world around us.

5. Ideas as Concepts of Real Conditions

It remains for us to address the Specificity Problem. I have argued that there is reason to think that comprehension requires possession of some conception of complete explanation. But rather than a generic concept of an unconditioned condition, Kant thinks that comprehension requires three specific concepts of unconditioned conditions. Why?

We may begin by considering what distinguishes one idea of reason from the others. First, recall that there are different kinds of real conditioning relations. An object can be unconditioned in respect of one kind of conditioning but not of others. All three ideas are concepts of real conditions that are themselves unconditioned, but they differ from one another with respect to the kind of conditioning relative to which their object is unconditioned. Second, while Kant recognizes numerous kinds of real conditioning relations, he thinks that they fall under three basic kinds, forming a system that can be completely derived in the same manner as the system of the categories (cf. A338/B396). We can say that the ideas are concepts of unconditioned conditions with respect to the three basic kinds of conditioning relations. Third, the ideas have their source in “the nature of reason itself” (A327/B384). They are “generated in an entirely necessary way by reason according to its original laws” (A338/B396). If the ideas are formed a priori, we must also possess a priori concepts of the basic kinds of conditioning relations. We can then derive the ideas from the Supreme Principle by specifying it in terms of each of the basic kinds of conditioning relations. In this way, the structure of our comprehension is determined a priori by reason, analogously to the way the structure of our empirical thinking is determined a priori by the understanding.

Taking these three points together, we can give the following answer to the Specificity Problem: comprehension requires all three ideas because there are three basic forms or varieties of comprehension, determined a priori by reason. Since there are three basic kinds of conditioning relations, there are three basic respects in which objects can be conditioned and thus three basic ways in which they can be comprehended. The ideas correspond to the forms of comprehension, just as the categories correspond to the forms of empirical judgment.

More concretely: comprehending x involves grasping not only its condition, y, but also the relation between x and y as a conditioning relation of a specific kind. But this yields comprehension only if there is an unconditioned condition of which y is a part. So, comprehending x further involves conceptualizing y as a part of an unconditioned condition relative to that kind of conditioning relation. This explains why comprehension requires a different concept of complete explanation for each basic kind of conditioning relation.

A difficult question, which I will be able to address only in a preliminary way, is how we can make sense of the idea that there are three basic forms of comprehension. One answer, which is suggested by the standard answer to the Specificity Problem, is that each form of comprehension is distinctive to a special empirical science. In Section 2, I already raised some issues with the view that construes the specific role of each idea in terms of a special domain of science. Another problem that arises presently is that the basic kinds of conditioning relations do not neatly map onto the purported special sciences. Consider, for instance, what we might call the cosmological conditioning relations, discussed in the Antinomies. Some of these relations, such as composition, pertain exclusively to outer objects, which may seem to justify associating the cosmological idea with physics. But other relations, such as causation, are not exclusive to physics as opposed to psychology or natural science as a whole. So, it is unclear how each special science is supposed to be understood so as to map onto each basic kind of conditioning relations.

It is worth noting that the putative relevance of the special sciences is insufficiently supported by textual evidence. Kant never affirms it but comes closest to doing so in a passage from the Appendix in which the roles of the ideas are introduced in relation to “psychology,” “cosmology,” and “theology” (A672/B700). But occurring in their usual triadic sequence, these terms unambiguously refer to the metaphysica specialis (cf. A334/B391) and not any disciplines of empirical science. While one could insist that “psychology” means empirical psychology, “cosmology,” and “theology” do not lend themselves to analogous interpretations. Kant reserves “cosmology” for rational cosmology and instead uses “physics” or “physiology” to refer to the empirical science of corporeal nature (e.g., A846/B874). As far as I’m aware, he never uses “theology” to refer to any empirical discipline.

Giving up the assumption that the labor is divided among the ideas according to domains of science or objects of inquiry opens the way for an alternative reading in which one and the same object can be comprehended in multiple ways, each of which involves the use of a different idea. If an object can be conditioned in more than one respect, it follows that it can be comprehended in more than one respect. There are as many ways to comprehend a given object as there are conditioning relations in which it stands to other objects. The tripartite division of the system of ideas reflects the basic respects in which objects of experience in general are conditioned and thus can be comprehended.

Support for this reading comes from Kant’s explication of the derivation of the ideas in terms of the elements of cognition in general: the subject, appearance, and the object.Footnote ³⁹ The system of the ideas is derived from “the universal of all relation that our representations can have,” namely, “1) the relation to the subject, 2) the relation to objects, and indeed either as appearances or as objects of thinking in general” (A333–34/B390–1, translation modified). These elements give rise to three types of “unconditioned synthetic unity of all conditions in general,” each of which is “contained” in an idea. These correspond to what appears later as the three classes of “the universal conditions of thinking” (A396) or “conditions of all representations in general” (A405/B432).Footnote ⁴⁰

These passages are schematic and obscure. But they suggest the following way to construe the idea that there are three basic respects in which objects are conditioned and can be comprehended. First, they are conditioned by the subject insofar as they are transcendentally ideal, and thus, in some sense, subject-dependent. Second, objects can be considered as part of a series of appearances and thus as conditioned by another appearance. Third, objects as appearances are conditioned by things in themselves. Kant seems to think that the third kind of conditioning relations figures in the explanation of real possibility in terms of actuality (cf. A571/B599ff.). For this reason, he describes the relevant class of conditions as the “condition of the possibility of everything that can be thought” (A334/B391) or “conditions of the possibility of objects in general” (A406/B433).

With this account of the relationship between the ideas and the real conditioning relations in hand, we can explain how the characterization of the theoretical use of the ideas in terms of comprehension relates to the characterization of it in terms of systematicity. Consider this representative passage:

First, I submit that the systematic unity in this passage is equivalent to the unity of comprehension. This is confirmed by the fact that Kant immediately glosses it as “unity of reason,” which, as we have seen, designates the unity characteristic of comprehension. He also says later that “the unity of reason is the unity of a system” (A680/B708). Second, “cognition” consequently refers not to just any kind of cognition but specifically to comprehension. It follows that “interconnection based on one principle” must consist of epistemic relations among cognitions that reflect or map onto real conditioning relations among their objects. Third, systematic unity in this sense requires an idea “of the form of a whole of cognition, which precedes the determinate cognition of the parts.” I read this as equivalent to an idea of complete comprehension, which depends on cognition of an unconditioned condition. This last point can thus be seen as expressing, in terms of the form and connection of cognition, the claim that comprehension requires an idea of an unconditioned condition.

I have assumed that at issue in the present context is the notion of systematicity that specifically pertains to comprehension, i.e., to cognition whose interconnection tracks real conditioning relations.Footnote ⁴¹ This entails that the “one principle” on which the whole of cognition is based would have as its object an unconditioned condition that grounds a corresponding whole of objects (cognitions of which are parts of that whole of cognition). From the fact that there are different kinds of conditioning relations, two consequences follow. First, an idea of a whole of cognition can be specified in terms of each kind of conditioning relation. This would yield different ideas of systems of cognition, each of which tracks a different kind of conditioning relations. Second, although each system flows from one “principle,” i.e., a cognition of an object that is unconditioned in a certain respect, this object may be conditioned in another respect by an object outside of its system. From this arises the prospect of integrating different systems of cognition into a larger system via real conditioning relations linking their unconditioned objects with members of another system.