Introduction. Nathaniel Goldberg Kant on Demarcation and Discovery

Transcription

1 Nathaniel Goldberg Kant on Demarcation and Discovery Abstract: Kant makes two claims in the Critique of Pure Reason that anticipate concerns of twentieth-century philosophy of science. The first, that the understanding and sensibility are constitutive of knowledge,while reason is responsible for transcendental illusion, amounts to his solution to Karl Popper s problem of demarcating science from pseudoscience. The second, that besides these constitutive roles of the understanding and sensibility, reason is itself needed to discover new empirical knowledge, anticipates Hans Reichenbach s distinction between the contexts of justification and discovery. Unlike Reichenbach, however, who thinks that there can be a logic only of justification, Kant provides what amounts to a logic of discovery. Though Kant s broader concerns are not Popper s orreichenbach s, using theirs as framing devices reveals two otherwise unnoticed things about the Critique of Pure Reason. First, besides its general epistemological and metaphysical aims, the Critique lays groundwork for the twentieth century s specialized field of the philosophy of science. Second, Kant s solution to the demarcation problem contradicts his logic of discovery, so in this instance the Critique is too ambitious. Introduction Immanuel Kant makes two claims in the Critique of Pure Reason that anticipate central concerns of twentieth-century philosophy of science. The first, in the Transcendental Analytic and Transcendental Dialectic, is that the understanding and sensibility are constitutive of knowledge, while reason is responsible for transcendental illusion. This anticipates Karl Popper s solution to and diagnosis of the problem of demarcating science from pseudoscience. The demarcation problem is the challenge of identifying a criterion by which to distinguish science from those nonscientific disciplines also purporting to make claims about the world. Popper himself traces recognition of the problem directly to Kant.¹ The problem is introduced at Popper (2002, 4). Bennett claims that Kant has his own demarcation problem (1974, 263): demarcating the understanding from reason. As Iexplain in 1, Kant nevertheless uses the word demarcations (CPR A296/B 352) similarly to how Popper does. Other philosophers anticipate the problem of demarcating science from pseudoscience, including Locke and before him Plato when each attempts to distinguish knowledge from opinion. Nonetheless, as Popper notes, his attempt traces directly to Kant s.

2 44 Nathaniel Goldberg Kant ssecond claim, in the Appendix to the Transcendental Dialectic, is that reason is essential to the discovery of new empirical knowledge.² This anticipates Hans Reichenbach s distinction between the context of justification, where statements are rationallyevaluated, and the context of discovery,wherenew concepts and statements potentially leading to new knowledge are devised.³ The aim of a logic of justification is to identify those epistemological principles by which scientific theories are appraised. The aim of a logic of discovery is to identify those principles by which such theories are created and developed. Though in other writing Reichenbach (1965) strives to reformulate Kant s views in light of twentieth-century science,⁴ he does not trace the distinction between contexts or the search for alogic of either to Kant. Noone does, though one should. Kant s broader concerns are not Popper s or Reichenbach s, nor are theirs his. Here I evaluate Kant s two claims in their own right. I nevertheless use Popper s and Reichenbach s concerns as framing devices for two reasons. First,doing so throws into relief the multifaceted and enduring nature ofthe first Critique. Besides its general epistemological and metaphysical aims, we see that it lays the groundwork for what would in the twentieth century become the specialized field of the philosophy of science. Michael Friedman, himself an expositor of Kant s philosophy of science and who also contributes to contemporary philosophy of science, explains: I am very much in sympathy with the idea that an internal engagement with the philosophical tradition can greatly illuminate our current intellectual predicament (1996, 434). Popper and Reichenbach are two twentieth-century German philosophers who know Kant s work well. Understanding Kant s views in light of theirs is itself an internal engagement that can greatly illuminate not only our current intellectual predicament but also Kant s own. Nor is doing so anachronistic. As becomes evident, Kant has something like Popper sand Reichenbach sprojects in mind. And Popperexplicitly does, and Reichenbach can, look to Kant for inspiration. Further, showing how Kant anticipates these later thinkers further establishes his enduring relevance. Popper and Reichenbach are giants of the twentieth century. To the extent that their views trace to Kant, his continued importance is greater than we might otherwise recognize. Ifocus on the first part of the Appendix, On the Regulative Use of the Ideas of Pure Reason, though Kant is also concerned with the context of discovery in the second part,onthe Final Aim of the Natural Dialectic of Pure Reason. Kant is concerned with it as well inthe Critique of the Power Judgment in the Introduction ( IV VI) and second half, the Critique of the Power of Teleological Judgment ( 61 62, 67 68, 72 74). I limit myself to the Critique of Pure Reason. The distinction isintroduced at Reichenbach (2006, 314). See Goldberg (2015, 68 69, ).

3 Kant on Demarcation and Discovery 45 Second, using Popper s and Reichenbach s concerns as framing devices also reveals Kant s ambition. Instead of aiming merely to demarcate science from pseudoscience, as Popper does, Kant aims to demarcate knowledge (and so scientific knowledge) from illusion (and so illusion taken as science). Conversely, while Reichenbach argues that the contexts of justification and discovery differ in kind, Kant is committed to their being the same in kind. Both Reichenbach and Kant would maintain that, because (as Reichenbach might say) the context of justification is governed by a priori principles, it is epistemological. Yet, while Reichenbach is committed to the view that because the context of discovery is governed by empirically derived suggestions it is merely psychological, Kant is committed to the view that it too is governed by a priori principles. For Kant, therefore, the context of discovery is itself epistemological. Hence, while Reichenbach thinks that there can be alogic onlyofjustification, Kant is committed to there being logics of justification and discovery. As I read him, Kant proposes a logic of discovery himself. Unfortunately for Kant, however, his anticipations of Popper and Reichenbach are too ambitious. Kant s solution to and diagnosis of the demarcation problem contradict his logic of discovery. In 1 I present Kant s position on demarcation. In 2 I present his position on discovery. In 3 I analyze Kant s attempt at reconciling those positions. In 4 I show that Kant s attempt fails. In 5 I conclude that Kant should reject his logic of discovery. 1 Kant on Demarcation A natural place to look for Kant s solution to the demarcation problem is the Transcendental Dialectic. There Kant is concerned with principles that actually incite us to tear down all those boundary posts and to lay claim to awholly new territory that recognizes no demarcations anywhere (CPR A296/B 352, my emphasis). As he explains, thosedemarcations are between what is within the limits of possible experience [ ]and beyond these boundaries. What is within possible experience can be known. What is beyond, or transcends,these boundaries cannot be. AccordingtoKant,transcendental illusion results from human reason s natural and unavoidable (CPR A298/B 354) propensity to apply the transcendental categories of the understanding, conditions of the possibility of experience, beyond possible experience.⁵ Reason, as the faculty of systematicity, always goes Kant distinguishes transcendental illusion from empirical illusion, which concerns objects of

4 46 Nathaniel Goldberg to the absolute totality in the synthesis of conditions,and never ends except with the absolutely unconditioned, i.e., what is unconditioned in every relation (CPR A326/B 382). What is unconditioned in every relation is not an object of possible experience. The transcendental dialectic, Kant explains, therefore content[s] itself with uncovering the illusion in transcendental judgments while at the same time protecting us from being deceived by it (CPR A 297/B 354).⁶ Transcendental judgmentsthemselvesgobeyond possible experience,and Kant endeavors to uncover the illusion in three different kinds: transcendental judgments concerning the soul (the Paralogism of Pure Reason), the world (the Antinomy of Pure Reason), and God (the Ideal of Pure Reason).⁷ Nonetheless in the concluding paragraph of the Appendix to the Transcendental Dialectic (CPR A /B ) Kant maintains that the Transcendental Analytic itself has shown that we can have no knowledge beyond possible experience.though Kant dedicates the Transcendental Dialectic to uncovering illusion and protecting us from deception, apparently the Transcendental Analytic has already solved the demarcation problem. What does Kant say in the Transcendental Analytic? Having maintained in the Transcendental Aesthetic that space and time are humanly imposed conditions of the possibility of apriori and empirical intuition, arising from sensibility, in the Transcendental Analytic Kant first claims to identify the a priori concepts, or categories, of human understanding (the Metaphysical Deduction). Then he contends that these categories are also conditions of the possibility of experience, objects of experience, and empirical knowledge (the Transcendental Deduction).⁸ Next he argues that the categories can be schematized according to the a priori form of time and so applied to empirical intuition (the Schematism). Finally he allegedly derives from the schematized categories synthetic apriori possible experience, and from logical illusion, which concerns invalid logical reasoning (CPR A /B ). I focus on the transcendental. Ifollow Guyer and Wood s use of transcendental judgment to name those judgments that for Kant go beyond possible experience. Regardless for philosophical and exegetical reasons transcendent judgment better translates Kant s text. In the body ofthe Transcendental Dialectic Kant discusses the soul, world, and God as three ideas of reason leading to illusion, and in the second part of its Appendix how these three are regulative. In the first part of the Appendix, which is my focus (see note 2), Kant discusses the idea of a focus imaginarius and ideas (and principles) of specification, aggregation, and affinity. Kant neither explains how these latter ideas relate to the former ones nor notes the switch. See Grier (2001, 297). For Kant, ifthe categories are conditions of the possibility ofexperience, then they are also conditions of the possibility of objects of experience (CPR A 158/B 197) and of empirical knowledge (CPR B 147).

5 Kant on Demarcation and Discovery 47 judgments, or principles, directly responsible for all possible experience, objects of experience, and empirical knowledge (the System of Principles). To see how this is meant to solve the demarcation problem, note that Kant can now make good on his claim in the Introduction that all knowledge divides into three kinds. Analytic a priori judgments, and so knowledge, which are independent of sensibility, clarify concepts rather than go beyond them to possible (and sometimes actual) experience of objects in space and time. Synthetic apriori judgments, and so knowledge, are conditions of the possibility of experience, objects of experience, and empirical knowledge. And synthetic a posteriori judgments, or knowledge, which are empirical judgments, are made possible by synthetic a priori judgments. Hence knowledge is demarcated from illusion because it concerns either concepts independent of possible experience, conditions of the possibility of experience, or experience itself, respectively. Claims about anything that exists, and so concerning neither mere concepts nor conditions of the possibility of experience, which could not be experienced, would not be knowledge. Knowledge (and science) stops and illusion (and pseudoscience) starts at the border of anything allegedly existing that cannot be experience. This is consonant with what Kant says about the supreme (or universal and completely sufficient) principle of all analytic and synthetic judgments: Now the proposition that no predicate pertains to a thing that contradicts it is called the principle of contradiction [ ]. Hence we must allow the principle of contradiction to count as the universal and completely sufficient principle of all analytic cognition. (CPR A151/B ) The test for analytic cognition, or knowledge, is that it not be self-contradictory. That test is purely conceptual, not concerning experience. Likewise: The supreme principle of all synthetic judgments is, therefore: Every object stands under the necessary conditions of the synthetic unity of the manifold of intuition in a possible experience. (CPR A158/B 197) All synthetic knowledge,both a priori and a posteriori, must conform to possible experience. Synthetic apriori knowledge is acondition of the possibilityofexperience, while synthetic a posteriori knowledge just is experience: [The categories] serve only for the possibility of empirical cognition. Such cognition, however, is called experience (CPR B 147). All knowledge, according to Kant, either is merely conceptual and so independent of experience or conforms to possible experience. Anything that is neither is illusion. How does the Transcendental Dialectic relate? It diagnoses why transcendental illusion arises at all. While we sense through intuition and understand

6 48 Nathaniel Goldberg through concepts, which are jointly constitutive of knowledge and our focus is on empirical knowledge we reason through ideas. Moreover, while reason uses its ideas to systematize knowledge, reason s natural and unavoidable (CPR A298/B 354) propensity for systematization never ends except with the absolutely unconditioned (CPR A326/B 382). Reason applies the understanding s concepts beyond space and time and so possible experience of objects. Transcendental illusion results. At the start of the Appendix to the Transcendental Dialectic Kant explains: The outcome of all dialectical attempts of pure reason not only confirms what we have already proved in the Transcendental Analytic, namely that all the inferences that would carry us out beyond the field of possible experience are deceptive and groundless, but it also simultaneouslyteaches us this particular lesson: that human reason has anatural propensity to overlap all these boundaries, and that transcendental ideas are just as natural to it as the categories are to the understanding, although with this difference, that just as the categories lead to truth, i.e., to the agreement of our concepts with their objects, the ideas effect a mere, but irresistible, illusion, deception by which one can hardly resist even through the most acute criticism. (CPR A 642/B 670) Hence Kant solves the demarcation problem by identifying sensibility and the understanding as the only faculties constitutive of knowledge and recognizing that neither by itself goes beyond possible experience. Kant diagnoses the demarcation problem by identifying reason as that faculty that does go beyond possible experience via its ideas. 2 Kant on Discovery And yet Kant alsoneedstoexplain how scientific discovery is possible. Since the understanding has both a priori and empirical concepts, constitutive of not merely possible but also actual experience, unless the understanding has access to new empirical concepts it cannot itself go beyond actual experience. If concepts of empirical objects are constitutive of empirical knowledge, and [s]uch cognition [or knowledge], however, is called experience (CPR B 147), then without new empirical concepts to constitute new empirical knowledge, there can be no new experience. Empirical discovery becomes impossible.⁹ Neiman explains: [I]t is not despite, but because of, understanding s role inshaping experience that it is incapable of looking beyond it (1994, 68).

7 Kant on Demarcation and Discovery 49 So Kant must find a way to provide the understanding with new empirical concepts. Having disparaged reason in the body of the Transcendental Dialectic, in its Appendix Kant maintains that reason comes to the understanding s aid. It is necessary for what I take to be Kant s logic of discovery.¹⁰ Kant begins by noting that ideas of reason, though not constitutive of knowledge, have an excellent and indispensably necessary regulative use, namely that of directing the understanding to a certain goal respecting which the lines of direction of all its rules convergeatone point,which, although it is onlyanidea (focus imaginarius) i.e., apoint from which the concepts of the understandingdonot really proceed, since it lies entirelyoutside the bounds of possible experience nonetheless still serves to obtain for these concepts the greatest unity alongside the greatest extension. (CPR A644/B 672) While the understanding and sensibility are constitutive of knowledge, reason is regulative of it. Reason directs (or guides or orders) the understanding by bringing it into greater unity and extension. It does so by positing the idea of a focus imaginarius. As an artist might posit a point of perspective beyond her canvas towardwhich she paints objects as converging,reason posits this focus imaginarius as a point of perspective beyond possible experience towardwhich the understanding constructs rules, or concepts, as converging.¹¹ Because the focus imaginarius is an idea, the understanding never reaches it. As onlyanidea [ ] itlies entirely outside the bounds of possible experience. As imaginary, it is an illusion. [Y]et, Kant continues, this illusion (which can be prevented from deceiving) is nevertheless indispensably necessary if besides the objects before our eyes we want to see those that lie far in the back, i.e., when, in our case, the understanding wants to go beyond every given experience (beyond this part of the whole of possible experience), and hence wants to take the measure of its greatest possible and uttermost extension. (CPR A /B ) Reflecting on the focus imaginarius presumably allows reason to derive what he goes on to present as three particular a priori principles for empirical investigation. The principle of specification maintains that every genus is divisible into species. The principle of aggregation maintains that every species is subsumable under a genus. And the principle of affinity maintains that there can be a contin- Nor might reason be necessary merelyfor alogic of discovery.without reason ssystematicity, Neiman (1994, 59) argues, scientific explanation, and, McFarland (1970, 20) argues, science itself, is impossible. According to Allison (2004,425), Kant has in mind passages innewton s Opticks concerning optical (and so empirical) illusions caused by mirrors. See Grier (2001, 37 38).

8 50 Nathaniel Goldberg uous transition between species of a genus.¹² What good are these principles? Kant explains: Reason never relates directly to an object, but solely to the understanding (CPR A 643/B 671). In particular reason applies its principles to concepts of actual empirical objects, which the understanding already possesses, to produce concepts of possible objects, whose actuality the understanding can then explore.¹³ Consequently the understanding investigates with sensibility whether these concepts of possible objects turn out to be concepts of actual objects objects actually existing in space and time. Scientists thereby can engage in empirical discovery. Moreover, because such discovery would be directed by a priori principles (themselves derived from an a priori idea), it would be epistemological not psychological. Kant has provided a logic of discovery.¹⁴ Kant explains that reason s appealing to a priori principles is how the chemistry of his day proceeded: That there are absorbent earths of different species (chalky earths and muriatic earths) needed for its discovery aforegoing rule [or principle] of reason that made it atask for the understanding to seek for varieties, by presupposing nature to be so abundant that it presumes them. (CPR A 657/B 685) Once the understandinghas the empirical concept of an absorbent earth, the discovery of different species of such earths, chalky and muriatic, required that the understanding look for different species. That required that reason presuppose that objects of experience, and nature, can be further speciated, itself requiring reason s a priori principle of specification. Once the understanding has the empirical concept of an absorbent earth, reason would via the principle of specification direct the understanding to investigate with sensibility whether there are different species of absorbent earths. If the investigation succeeds, then the concept of one or more species of absorbent earths turns out to be aconcept of an Kant discusses these at CPR (A /B ). Kant names them as such at A /B At A /B he calls the first two laws or principles of species (or specification ) and genera. At A /B he calls the three specification, homogeneity, and continuity of forms. At A 662/B 690 he calls them manifoldness, unity, and affinity. As Iexplain in 3 when Iconsider the relevant passage, Kant goes on to argue that reason is therefore analogous to the understanding insofar as each takes a lower faculty for its object (CPR A664/B 692). Rauscher explains: Kant denies that reason can create concepts of objects, but allows that reason can create concepts or ideas that can unite (or otherwise relate) the empirical concepts of the understanding (2010, 295). This is aprocess of concept constitution. See Goldberg (2015, 9), where Ishow how constitution makes possible acquisition.

9 Kant on Demarcation and Discovery 51 actual object of experience. Empirical knowledge results. If the investigation fails, then the concept remains one of a merely possible object. Yet herein lies a problem. If discovery leads to new knowledge, then that is because reason used its principles to constitute with the understanding a concept of a possible object of experience. The understanding then used that concept to constitute with sensibility an actual object of experience along with actual experience and actual empirical knowledge. Reason, the understanding, and sensibility would all be constitutive of knowledge. But that violates Kant s solution to the demarcation problem. In the Transcendental Analytic Kant solves the problem by limiting the knowledge-constituting faculties to the understanding and sensibility. If in the Appendix to the Transcendental Dialectic Kant grants reason constitutive powers for his logic of discovery, then he cannot prevent reason from having such powers generally. But then reason could lead to illusion as diagnosed in the bodyofthe Transcendental Dialectic. Adopting Popper s and Reichenbach s language, we can say that Kant s solution to and diagnosis of the demarcation problem contradict his logic of discovery.¹⁵ 3 Kant s Attempted Reconciliation As Iread him, Kant attempts at CPR (A /B ) in the Appendix to the Transcendental Dialectic to reconcile this contradiction. The passage begins: What is strange about these principles [of specification, aggregation, and affinity], and what alone concerns us, is this: that they seem to be transcendental, and even though they contain mere ideas to be followed in the empirical use of reason, which reason can follow only asymptotically, as it were, i. e., merely by approximation, without ever reaching them, yet these principles, as synthetic propositions apriori, nevertheless have objective but indeterminate validity, and serve as a rule of possible experience, and can be used with good success, as heuristic principles, in actually elaborating it. (CPR A663/B 691) The principles of specification, aggregation, and affinity seem transcendental, i. e., to concern conditions of the possibility of experience, objects of experience, and empirical knowledge.they therefore seem to serveasrules of possible experience in particular. They are even synthetic apriori. Yet the principles contain mere ideas of reason. And no idea is constitutive of possible experience, objects O Shea pinpoints the problem: The obvious danger[ ]isthat to assert that reason s(sensetranscending) ideal of systematic unity is straightforwardly objective suggests precisely the natural tendency toward transcendental illusion that Kant has just spent hundreds of pages [in the Transcendental Dialectic] warning us against (1997, 238).

10 52 Nathaniel Goldberg of experience, or empirical knowledge. Kant sees his way around this when he claims that these principles, as synthetic propositions apriori, nevertheless have objective but indeterminate validity. As synthetic apriori, they are in fact transcendental and so concern conditions of the possibility of empirical knowledge. Nonetheless their objective validity is indeterminate. Based on what Kant says elsewhere, to have objective validity these principles must yield conditions of the possibility of knowledge of objects (CPR A89 90/B 122), i.e., empirical knowledge. Based on what he says here, for their objective validity to be indeterminate they must be constitutive not of such knowledge but of the ways in which conceptsofthe understandingare constitutive of such knowledge.¹⁶ Thus, Kant explains later in our target passage, CPR (A /B ), since every principle that establishes for the understanding a thoroughgoing unity of its use a priori is also valid, albeit only indirectly, for the object of experience, the principles of pure reason will also have objective reality in regard to this object. (CPR A 665/B 693) The principles of specification, aggregation, and affinity have objective but indeterminate validity because they apply directly to the understanding s use of concepts and so indirectly to objects of experience (and ultimately empirical knowledge). What guarantees that we are justified in applying these principles to experience? Kant is clear: one cannot bring about a transcendental deduction of them, which, as has been proved above, is always impossible in regard to ideas (CPR A /B ). And in the Transcendental Analytic such a deduction was needed to establish the categories applicability to experience. Moreover below the target passage Kant admits: The ideas of reason, of course, do not permit any deduction of the same kind as the categories; but if they are to have the least objective validity, even if it is only an indeterminate one [ ], then a deduction of them must definitely be possible, granted that it must diverge quite far from the deduction one can carry out in the case of the categories. (CPR A /B ) Allison calls the passage perplexing (2004, 435). Horstmann (1989, 166), repeating Kemp Smith s (1995, 547) charge, calls it extremely self-contradictory. England (1968, ), Walsh (1997, ), and Guyer (1990, 17 43) make similar claims. While Kant s passage is perplexing and self-contradictory, his ingenuity mitigates the extremeness. (See Grier (2001, ch. 8), and Pickering (2011), for an argument against its being self-contradictory.)

11 Kant on Demarcation and Discovery 53 A deduction must be possible lest scientists in particular be unjustified in applying these principles, even indirectly, to experience. Without such adeduction the principles could not be part of a logic of discovery. In the text from CPR (A /B ) that follows Kant provides what I take to be a deduction of his three principles of reason. Nonetheless the deduction diverges quite far from that of the categories. As I read Kant,the deduction proceeds in three steps. For reasons that become clear below, I call those steps Analogous Faculties, Analogous Schemata,and Master Argument, respectively. Step one, Analogous Faculties, begins it: The understanding constitutes an object for reason, just as sensibility does for the understanding (CPR A 664/ B692). Reason is therefore analogous to the understanding insofar as each takes a lower faculty for its object. Kant is building on his position in the Transcendental Dialectic: If the understanding may be a faculty of unity of appearances [and so objects] via rules, then reason is the faculty of the unity of the rules of understanding under principles. (CPR A 302/B 359) Now,when Kant says that sensibilityconstitutes an object for the understanding, he cannot mean that sensibility is itself an object of experience. Sensibility is that faculty which receivessensations that once intuited and categorized become experience. For it to be an object of experience is either incoherent or infinitely regressive. Kant must instead mean that the understanding takes sensibility for its object. Sensibility becomes the target of influence on the part of the understanding. The understanding constitutes unities, or objects, of intuition, via rules (or concepts). So sensibility constitutes an object for the understanding because the understanding via its rules (or principles) determines objects of experience. There is support in the Transcendental Analytic for this reading: Sensibility gives us forms (of intuition), but the understanding gives us rules. It is always poring through appearances with the aim of finding some sort of rule in them. (CPR A 126) Sensibility constitutes an object for the understanding because the understanding pores through intuition to determine rules (or concepts) to which it conforms. In so doing it unifies objects within it. Kant continues: Thus as exaggerated and contradictory as it may sound to say that the understanding is itself the source of laws of nature [ ], such an assertion is nevertheless correct and appropriate to the object, namely experience. (CPR A 127)

12 54 Nathaniel Goldberg The understanding not only takes experience for its object but also provides experience with laws. Subtract the contribution of the understanding to experience and sensibility remains. The understanding, before providing sensibility with laws, must take sensibility itself for its object, lest it have no object with which to provide laws at all. And, by providing sensibility with laws, the understanding constitutes objects of experience. What would it mean for the understanding to constitute an object for reason? The understanding becomes the target of influence on the part of reason. Though not in the same way, in a nevertheless analogous way reason would determine unities of the understanding. The understanding constitutes an object for reason because reason via its principles somehowdetermines rules of the understanding. Now, if the analogy holds, then principles of reason determine these rules of the understanding by constituting them. Though Kant (CPR A642/B 770, B644/B 672) says that principles of reason merely regulate these rules, he later (CPR A /B ) maintains that any such regulative use presupposes a transcendental use. Such a use, concerning conditions of the possibility of experience and its objects, would be constitutive. Further, in the passageofthe Appendix directlyconcerning us, Kant needsprinciples of reason to have atranscendental, and so constitutive,use. Merely regulating existing rules (or concepts) would do nothing to determine new rules, i.e., rules for possible objects of knowledge not already known. Kant would not have provided discovery with a logic. So reason apparently has some constitutive function analogous to the constitutive function of the understanding. Three passages earlier in the Appendix support this. First, Kant contends that without reason we would have no coherent use of the understanding (CPR A651/B 679). He analogouslycontends in the Transcendental Analytic that without the understanding we would have no coherent use of sensibility. Since the understanding stands in a constitutive relation to objects of sensibility (which would be objects of experience), reason would presumably stand in a constitutive relation to rules of the understanding (which would be concepts of objects of experience). Second, using the principle of specification as an example, Kant argues that sameness of kind is necessarily presupposed in the manifold of possible experience [ ], because without it no empirical concepts and hence no experience would be possible (CPR A654/ B682, my emphasis). Since principles of reason are necessarily presupposed by possible experience, these principles allow us to investigate possible objects. This would presumably be so in virtue of principles of reason constituting rules for possible objects. And third, Kant writes that reason via its principles prepares the field for the understanding (CPR A657/B 685). He explains in the previous sentence:

13 Kant on Demarcation and Discovery 55 For we have an understanding only under the presupposition of varieties in nature [the principle of specification], just as we have one only under the condition that nature s objects have in themselvesasameness of kind [the principle of aggregation], because it is just the manifoldness of what can be grasped together under a concept that constitutes use of this concept and the business of the understanding. (CPR A657/B 685) Principlesofreason allow us to have an understanding,because these principles constitute use of the understanding s concepts (or rules). Kant analogously writes in the Transcendental Analytic that the understanding via the categories prepares the field for sensibility. Rules of the understanding allow us to have sensibility, because those rules constituteobjectsofexperience.thus, Analogous Faculties concludes, reason is analogous to the understanding insofar as each takes a lower faculty for its object. All that is step one in Kant s attempted reconciliation. Kant recognizes that he cannot yet conclude that principles of reason determine rules of the understanding and so indirectly determine objects of experience. In the next sentence (CPR A /B ) of the target passage of the Appendix Kant recalls his claim in the Transcendental Analytic that schemata are required to mediate applying the understanding s concepts to empirical intuition (CPR A / B ). Kant must now locate aschema to mediate applying aprinciple of reason to concepts of the understanding. In Analogous Schemata, step two in his attempted resolution, he does so. Kant knows that reason does not relate directly to sensibility. Unlike the schemata employed by the understanding, reason cannot use as a schema the a priori form of time or any other product of sensibility. Instead, in Analogous Schemata, Kant argues that reason uses the schema of the idea of the maximum of division and unification of the understanding s cognition in one principle (CPR A665/B 693). With this schema, principles of reason can be applied to the understanding,because [ ]all restrictingconditions,which give indeterminate manifolds, are omitted (ibid.). Reason watches the two maxima between which any of the understanding s cognitions, or knowledge, which might result from applying these rules to intuition, must be confined. Reason knows that in determining rules of the understanding it must avoid ordering both too few and too many possible cognitions under one principle. Being so restricted these rules of the understanding lose their indetermination. Thus, Analogous Schemataconcludes,the idea of amaximum is analogous to images and time relations insofar as each is aschema. Step three that Kant takes in the passage atcpr (A /B ) remains. Because this step relies on the first two, Analogous Faculties and Analogous Schemata, and because its conclusion is that principles of reason have ob-

14 56 Nathaniel Goldberg jective but indeterminate validity, step three iskant s Master Argument. Itoccurs in a single, especially contentful sentence: Now since every principle that establishes for the understanding a thoroughgoing unity of its use apriori is also valid, albeit only indirectly, for the object of experience, the principles of pure reason will also have objective reality in regard tothis object,yet not so as to determine something in it, but only to indicate the procedure in accordance with which the empirical and determinate use of the understanding in experience can be brought into thoroughgoing agreement with itself, by bringing it as far as possible into thoroughgoing agreement with the principle of thoroughgoing unity; and from that it is derived. (CPR A /B ) Kant is claiming five things. First, principles of reason determine a thoroughgoing unity of the understanding s use (Analogous Faculties). Second, they do so via the schemaofamaximum (Analogous Schemata). Third,principles of reason do so apriori. (Because the idea of amaximum is merely anidea, and its application is not to objects of experience but to the understanding itself, this claim is analytic.) Hence principles of reason, which are apriori, can determine rules of the understanding. Fourth, the understanding determines objects of experience (Transcendental Analytic). Thus, Master Argument concludes, principles of reason indirectly determine objects of experience. And fifth, these principles do so by indicating the procedure in accordance with which the understanding can be unified (Analogous Faculties). Hence reason, in virtue of the a priori application of its principles via their schema to rules of the understanding, determines the unity of the understanding, and so indirectly determines the understanding s determination of objects. Principles of reason are therefore analogous to rules or concepts of the understanding insofar as each have objective validity,though the validity of these principles is indirect. Principles of reason determine unities not of sensibility but of the understanding. Such principles constitute not objects of experience but concepts of the understanding, themselves concepts of objects of possible experience. Principles of reason regulate the sorts of objects that the understanding can constitute. And, because any subsequent discovery of new experience, objects of experience, and empirical knowledge would proceed from a priori principles of reason, Kant has provided a logic of discovery. 4 The Failure of Kant s Attempted Reconciliation When Popper proposes his solution to the demarcation problem, he does not simultaneously propose a logic of discovery. While Reichenbach might endorse

15 Kant on Demarcation and Discovery 57 Popper ssolution to the demarcation problem, he thinks that alogic of discovery is impossible. So neither faces Kant s challenge of reconciling the solution to and diagnosis of one with a proposal for the other. For Kant himself to have reconciled them, the three steps that Iread him as making in the passage atcpr (A /B ) must succeed. Unfortunately for Kant, the first, Analogous Faculties, and third, Master Argument, fail. Analogous Faculties concludes that reason is analogous to the understanding insofar as each takes a lower faculty for its object.though both reason and the understanding are involved in the unity of a lower faculty, however, only the understanding can take a lower faculty for its object in a requisitely analogous way. Since sensibility constitutes an object for the understanding, the understanding via its rules determines unities, or objects, of empirical intuition. These unities are objective in virtue of being determined by the understanding, which Kant establishes in the Transcendental Analytic as a faculty of objectivity. For the analogy to succeed, a consequence of the understanding s constituting an object for reason must be that reason determines unities, or rules, of the understanding. These rules must be objective in virtue of being determined by reason. Since Kant s point is that a merely analogous relation between rules of the understanding and objects of experience exists between principles of reason and rules of the understanding, it is insufficient to disqualify Analogous Faculties because reason is not a faculty of objectivity in the sense of constituting objects of experience. But it is sufficient to disqualify Analogous Faculties on related grounds. Whereas unities of intuition get their objectivity from the understanding,unitiesofthe understanding do not get theirobjectivity from reason. In the Transcendental Analytic Kant identifies the understanding as the source of its own objectivity: The understanding is thus not merely a faculty for making rules [ ]; it is itself the legislation for nature (CPR A126). That is how Kant can contend that the understanding and sensibility are jointly constitutive of knowledge, solving the demarcation problem. The understanding orders empirical intuition constitutively. For the analogy to hold, reason must order the understanding constitutively. But,because rules of the understanding are constituted by the understanding itself, if reason bears any relation to these rules then it is merely regulative. So Kant cannot recognize in reason powers of determination analogous to those of the understanding.reason is not analogous to the understanding insofar as each takes a lower faculty for its object in the requisite way. Analogous Faculties fails. Because Master Argument relies on Analogous Faculties, Master Argument and so Kant s attempted reconciliation also fails. Nonetheless it is instructive to consider Kant s remaining two steps. The second, Analogous Schemata, con-

16 58 Nathaniel Goldberg cludes that the idea of amaximum is analogous to the apriori form of time insofar as each is a schema. Does reason s knowing to navigate between ordering too few and too many possible cognitions under one principle mediate the determination of rules? Returning to Kant s example, suppose that the concept of an absorbent earth is a rule or concept of the understanding.we want to order this rule under the principles of specification, aggregation, and affinity. Kant seems right that todosodeterminately wemust keep in mind how few and how many possible cognitions can be ordered under any of these principles. Otherwise we could not tell whether an absorbent earth has species, is amember of agenus, or is one memberofafamilyofdissimilar objects atthe same level of specification. We therefore could not determine rules for possible related objects. The idea of a maximum therefore seems to allow principles of reason to determine rules of the understanding.we have no reason to think that it does not. What should we think about Master Argument,which concludes that principles of reason have objective but indeterminate validity? Step three fails because Kant attributes to principles of reason both too much and too little influence. We can see this in two ways. One focuses on how determination, constitution, and regulation relate. The other focuses on whether reason constitutes for the understanding possible objects of experience, as I have been saying, or objects of possible experience, instead. Consider determination, constitution, and regulation. Kant attributes to principles of reason too much influence because he argues that such principles determine rules of the understanding, thereby indirectly determining objects. Such principles allegedly can be transcendental without constituting objects. They would merely regulate them. Hence Kant s argument that principles of reason have objective but indeterminatevalidityreduces to his taking indirect determination to entail regulation instead of constitution. Yet it does not. Kant knows that direct determination is intransitive.ifprinciples of reason directlydetermine rules of the understanding, and rules of the understanding directly determine objects of experience, then principles of reason do not directly determine these objects. As Kant observes, they indirectly determine them. Now Kant apparently assumes that constitution is also intransitive, and that constitution and regulation are analogous to direct determination and indirect determination, respectively. Kant argues that, if principles of reason constitute rules of the understanding, and rules of the understanding constitute objects of experience, then principles of reason do not constitute these objects. They merely regulate them. Unfortunately for Kant, constitution is not intransitive. Grain constitutes flour, and flour constitutes bread. It would be wrong to deducethat grain does not constitute bread. Yet the latter parallels Kant s conclusion concerning principles of reason and objects of experience. The issue is not whether principles of reason

17 Kant on Demarcation and Discovery 59 indirectly determine objects of experience but whether indirectly determining them entails that principles of reason merely regulate them. It does not. Master Argument fails. Conversely Kant also attributes to principles of reason too little influence because he argues that the understanding grounds the objectivity of its own rules. As we heard, this entails that the understanding is not an object for reason in the requisite way: Analogous Faculties fails. Consequently principles of reason would not determine rules of the understanding. Because sensibility is an object for the understanding, anything determining rules of the understanding would indirectly determine objects of experience. Kant s argument depends on this. Yet, because principles of reason do not determine rules of the understanding, principles of reason do not indirectly determine objects. So principles of reason would not have objectivity of any kind. Without Analogous Faculties,Kant smaster Argument fails straightaway. The other way to see that Kant attributes to principles of reason both too much and too little influence is to focus on what reason allegedly constitutes for the understanding. Above I argued that reason constitutes concepts of possible objects of experience. The understanding can then investigate with sensibility whether these concepts of possible objects turn out to be concepts of actual objects. If so then empirical discovery results. Nonetheless, ifreason does ultimately constitute concepts of actual objects of experience, then Kant s solution to and diagnosis of the demarcation problem are contradicted. In the Transcendental Analytic Kant identifies the understanding and sensibility as the only faculties constitutive of such, lest, as Kant explains in the Transcendental Dialectic, illusion results. So, for Kant, reason cannot constitute concepts of possible objects that turn out to be conceptsofactual objects. Nor can reason constituteconcepts of merely possible objects that cannot turn out to be concepts of actual objects. That precludes reason s ever leading to the discovery of anything that actually exists. Kant would have no logic of discovery at all. Perhaps Kant can instead maintain that reason constitutes possible concepts of objects. These possible concepts would themselves have to be merely possible. Were they to turn out to be actual concepts of objects, then reason would be constitutive of actual objects of experience. Kant would be in the same situation as above. So suppose that reason does constitute merely possible concepts of objects. Kant s idea now would be that, once reason constitutes these possible concepts for the understanding, the understanding would investigate with sensibility whether these merely possible concepts correlate with actual concepts. If so then experience, objects of experience, and empirical knowledge result. Because the understanding and sensibility would rely on actual concepts, while reason

18 60 Nathaniel Goldberg would provide merely possible ones, Kant s solution to and diagnosis of the demarcation problem seem safe. Unfortunatelyfor Kant,merelypossible concepts of objects of experience are insufficient for the task at hand. The understanding functions by relying on actual concepts. To constitute objects of possible experience, experience, and empirical knowledge,the understanding requires actual concepts not merely possible ones. A fortiori to function in discovery the understanding requires actual concepts too. But then the understanding must get actual concepts of objects from somewhere beside reason. Where could that be? The understanding cannot generate its own new concepts if the concepts are empirical, which they must be to function in a logic of (empirical) discovery. The understanding can generate only its own a priori concepts. Nor can the understanding appeal to experience for new empirical concepts. If concepts of empirical objects are constitutive of empirical knowledge, and [s]uch cognition [or knowledge], however, is called experience, then without already possessing new empirical concepts to constitute new empirical knowledge, there can be no new experience. Nor can the understanding appeal to sensibility to generate a new concept, empirical or otherwise. Sensibility is afaculty not of conceptsbut of intuitions.accordingtokant s resources inthe Critique of Pure Reason, however, ifnew empirical concepts cannot be generated by reason, the understanding, or sensibility, then they cannot be generated at all.¹⁷ Hence, even if reason does constitute merely possible concepts of objects, because the understanding would still need but be unable to possess actual concepts to correlate with them, reason s merely possible concepts would by themselves be insufficient for discovery. I can summarize this second way of seeing that Kant attributes to principles of reason both too much and too little influence as follows. If reason constitutes concepts of possible objects,then reason has too much influence over the constitution of empirical knowledge. Kant contradicts his solution to and diagnosis of the demarcation problem. If reason constitutes possible concepts of objects, then it has too little influence over the constitution of empirical knowledge. The understanding would still need actual concepts to discover new empirical knowledge and, since those concepts can come from neither reason, the understanding, nor sensibility, and there are no other sources in the Critique of Pure Reason to which Kant can turn, he cannot provide discovery with a logic.¹⁸ Not until the Critique of the Power ofjudgment does Kant distinguish imagination from judgment as faculties.there Kant appeals to (reflective) judgment to provide a different logic of discovery. Kant is presumably aware that this logic of discovery fails since heoffers adifferent one in the Critique of the Power of Judgment. See note 17.