Beginning in 1788, in the Critique of Practical Reason, and throughout the next decade, Kant defines religion as “the recognition of all duties as divine commands.”1 In one sense, this formula puts “duty” and “divine command” on par—Kant is saying it is legitimate to view certain actions both as duties and as divine commands. But of course Kant goes on consistently and unambiguously to make God the junior partner. In the Critique of Practical Reason he adds that, “everything remains disinterested and grounded only on duty”; five years later, in Religion within the Boundaries of Mere Reason, he writes that, “I must first know that something is duty before I can acknowledge it as divine command.”2
The legacy of Kant’s subordination of God to duty has long been contested. In the arena of cultural criticism, Charles Taylor and Richard Bernstein assess its effects very differently—but both see it as a decisive moment in the development of a secular age.3 Rorty and Geuss, among others, are not so impressed: following Nietzsche, they find that Kant’s ethics still has us bowing down to an external force, only now repackaged as the Moral Law.4 In the religious studies community the tendency has been to see in Kant’s definition of religion an attempt permanently to fix its meaning either by pegging it to an alleged essential property or by in some way artificially restricting its scope. Thus, Talal Asad writes that, “Kant was able to produce a fully essentialized idea of religion which could be counterposed to its phenomenal forms.” And Bruce Lincoln finds that Kant started us down a road toward a “view of religion as delimited, and therefore definable.”5
In this chapter I want to sort out Kant’s views on definition and essentialism; I think they have much to contribute to our contemporary discussion. I have already noted a prima facie tension between Kant’s theory of concepts and the idea of picking out essential properties of objects. When examined in greater detail the reality is somewhat more complex. It will emerge that, on the one hand, Kant’s theory of concepts requires an opposition to essentialism at least as deep as Asad’s and a dimmer view even than Lincoln’s of the prospects for strict definition. At the same time, essentialism does play a role in Kant’s theory of concepts. How all of this can be true at the same time requires some explaining. No doubt when we try to understand such large-scale social transformations as the rise of secularism we must implicate large-scale material factors—historical, economic, and other. But even here we will see that the very small—the theory of concepts—plays a key role. Let us begin with what Kant says about definition.
Here is the central passage from the Critique:
To define properly means just to exhibit originally the exhaustive concept of a thing within its boundaries. Given such a requirement, an empirical concept cannot be defined but only explicated. For since we have in it only some marks of a certain kind of objects of the senses, it is never certain whether by means of the word that designates the same object one does not sometimes think more of these marks but another time fewer of them. Thus in the concept gold one person might think, besides its weight, color, and ductility, its property of not rusting, while another might know nothing about this. One makes use of certain marks only as long as they are sufficient for making distinctions; new observations, however, take some away and add some, and therefore the limits of the concept are never fixed. And in any case, what would be the point of defining such a concept?—since when, e.g., water and its properties are under discussion, one will not stop at what is intended by the word “water” but rather advance to experiments. The word, with the few characteristics which we attach to it, is more properly to be regarded as merely a designation than as a concept of the thing; the so-called definition is nothing more than a determining of the word. . . . Instead of the expression “definition” I would use exposition, which is always cautious, which the critic can accept as valid to a certain degree while yet retaining reservations about its exhaustiveness. (A727–28/B755–56ff.)
Kant begins with the bold-sounding claim that no empirical concept can be “defined properly.”6 But to so define requires a complete enumeration of a concept’s sub-concepts or marks—not something remotely achievable for finite creatures. Why, then, does Kant belabor the point? Apparently he wants to contrast his approach to definition with that found in Leibniz. According to Leibniz a strict or “real” definition requires identifying a thing’s “principle of individuation.” Leibniz writes that, “The most important point . . . is that individuality involves infinity, and only someone who is capable of grasping the infinite could know the principle of individuation of a given thing.”7 In fact, as Kant sees it, the situation is much worse than our inability to grasp the infinite. Kant’s denial of real definition is rooted in the claim that we can never identify even a single essential property of an object. As he says elsewhere in the Critique, we cannot make explicit a single property “by which the object can always be known with certainty” (A241–42). Thus the passage reflects one of Kant’s central humanizing goals: to legitimate the cognition of objects by means of incomplete concepts.
If real definition is foreclosed to those of us with finite, discursive understandings, we are at least able to give “explications” or “expositions”—what he elsewhere terms “nominal” definitions of empirical terms (A241–42 n.). As the quoted passage makes clear, this activity is pragmatic, goal-directed, and open-ended in the sense that it is always subject to refinement through further inquiry (“the limits of the concept are never fixed”). We are to “advance to experiments,” to revise our concepts in the light of experience. Some of this refinement will take the form of conceptual decomposition, in making as explicit as we can the sub-concepts that make up the Umfang or sphere of the concept at issue. If we are giving a nominal definition then our interest is only in decomposition; we are inquiring into which predicate concepts form part of the subject concept. Kant’s idea is that such inquiry will consist in applying the principle of non-contradiction to the subject term; that is, we must ask what can be denied of the subject without contradiction.8 For example, Kant tells us that we cannot deny without contradiction that “All bodies are extended” (A7/B11). Applying the same test to “All bachelors are unmarried men” seems to show that “unmarried” and “man” are contained in “bachelor.” The underlying thought here is that to affirm that “Not all bachelors are unmarried men” is not merely odd or unhelpful or in general not conducive to human flourishing; rather, that what it asserts about the subject concept is, as a matter of fact, self-contradictory.
What is the account of definition on offer here? Judging by Kant’s actual procedure in the quoted passage, and by what he says elsewhere, the idea seems to be that we construct a nominal definition by assembling analytic judgments about the term in question. Thus, Kant apparently counts weight, color, ductility, and rust-resistance as derivable from an analysis of the subject concept “gold.” “Gold is a yellow metal” and “Gold is rust-resistant” would then be analytic judgments, each one forming part of what in various places Kant terms the “logical essence” of the concept.9 This is the point at which an element of essentialism does play a role in Kant’s thinking. Note that it is a fallibilist essentialism: no element on the list of analytic judgments is immune from revision; definition in this sense can only be approached, never attained. In fact, the results of applying the principle of contradiction may require, as Arthur Melnick puts it, “careful consideration and analysis.”10 To take a storied example, consider the judgment, “Every event has a cause.” In line with his rationalist heritage, Kant had formerly taken “cause” to form part of the concept “event.” This is the issue over which, in the Prolegomenon, Kant famously says he had been “slumbering.” As he sees it, it took Hume’s prodding to “awaken” him to the fact that one can deny this judgment without contradiction. It is not a matter of Kant coming to adopt Hume’s definition of “event.” It is a matter of discovering what can and cannot meaningfully be asserted about the subject-matter in question.
We will turn in the next section to Kant’s definition of religion, at which point a central question will be in what sense there is even a subject-matter to be inquired into. In preparation for that discussion we must look more closely at his account of nominal definition. No doubt it rests squarely on the distinction between analytic and synthetic judgment—a distinction which over the last half-century has come in for no little scrutiny. For the moment, I only want to observe that one may defend this distinction and at the same time doubt that “meaning” can be captured with much precision in the context of linguistic interpretation. We will turn to linguistic interpretation in chapter six. But note that, in giving Kantian-style nominal definitions we are not trying to puzzle out what someone means by “gold” or “religion.” We are asking what can and cannot be denied of gold and religion without contradiction. In defining gold, we know to keep out such things as its current price per ounce and the fact that it can still be panned from the Sacramento River. For Kant, these make for synthetic, a posteriori judgments and so have no place in the concept’s logical essence. And neither do we commonly rest our definitions of empirical terms on essentialist intuitions. When we say what we mean by “water” we normally put the fact it is composed of two parts hydrogen and one part oxygen on par with other of its properties—for example, the fact that it is colorless, odorless, potable, and a universal solvent. And, as has often been noted—for example, by Putnam against a former self11—we could not be impugned for applying the same concept to this substance even if its chemical composition were to come in question. For Kant, as we have seen, supporting this practice is the spatial theory of concepts, according to which reference to objects is always by way of concepts which are themselves composed of nested sub-concepts—nested, in principle, without end. Thus, the spatial theory of concepts requires us to “advance to experiments” even as it enforces a modesty as to whether, at any given time, conclusive results are in. This is presumably one reason why, in the quoted passage, Kant imagines a definition commanding some degree of intersubjective agreement—yet something short of consensus. That is, he imagines some critics who “retain reservations,” and others who are ignorant about certain properties (rust-resistance) of the object in question (gold).
Let us pause for a moment in our examination of Kant’s notion of nominal definition to advertise several themes that lie ahead. I have said that Kant’s notion of nominal definition rests squarely on the distinction between analytic and synthetic judgment. And we have already seen that that distinction in turn pivots on the principle of non-contradiction. Now it is notorious that Kant never offers a defense of the principle of non-contradiction. He seems to have taken conformity to the principle as required for meaningful signification of any kind—i.e., that even someone who affirms “both p and not-p” must, on pain of failure to mean anything, deny its negation and so must rely on the principle after all. The prospect of forms of thought or experience “beyond logic” has attracted religious thought of various kinds. One thinks, for example, of Wittgenstein’s famous remark at Tractatus 6.13 that “logic is transcendental”—often interpreted as inviting a form of mysticism. It illustrates one of the core phenomena that interest me in this book; namely, how, while Kant is mounting a frontal assault on religious illusion in the form of dogmatic claims about God and the soul, his very terms of engagement seem at various points to invite religious reflection. We will return more than once to the principle of non-contradiction. A second advertisement: Notice in Kant’s remarks about gold the radicalization of the Putnam-Austin slogan discussed in chapter one. For Kant, too, enough is enough. In drawing on empirical concepts in our everyday communicative contexts perhaps no two of us call into play the same conceptual microstructures (the same nested conceptual hierarchies), even if the differences are subtle. But Kant’s point is not simply that, for finite creatures, vagueness is an unavoidable but unremarkable fact about communication. True, we can still “make distinctions”—that was the point of the Putnam-Austin slogan. The deeper point is that, since all concepts are general, absent general concepts, there is no thought and so no experience—at least not in the fundamental sense of grasping (by means of concepts) that the world is thus and so. That is the deeper error in Korzybski’s position, and it gives some advance notice of the line of thought we will pursue in the next chapter. The question there will be the status, not merely of general concepts, but of the search for systematic, encompassing generalizations—as for example, the attempt to join Buddhism, Christianity and Islam into “world religions.” As it is here, the main point is not simply about extending control over our environment. It isn’t even about admitting that we cannot manage without them. (Geuss points out that Nietzsche combines an anti-essentialism with the recognition that we cannot do without encompassing generalizations.12) The point will be that, without them, I have no environment—no experience that the world is thus and so.
To return now to Kant’s treatment of nominal definition: two final points. Suppose we endorse Kant’s idea that a nominal definition must include only what is essential to the concept being defined. But Kant thinks—in fact, thinks he has proved against Hume—that “cause” is necessarily connected to “event.” This in turn might seem, as it does to C. I. Lewis, to require the inclusion of “cause” in any satisfactory definition of “event.”13 Here the main point is the basis of the connection between subject and predicate. “Cause” cannot be teased out of “event” by application of the principle of contradiction. Rather, as Kant sees it, the nature of its attachment comes by way of the possibility of experience: Kant argues that, unless events are necessarily connected experience is not possible. Thus, the basis of the connection is epistemological, not logical, and so it is not, by Kant’s criteria, part of the subject’s logical essence. Of course we are free to include “is caused” in our working definition of “event,” but this does not threaten Kant’s principled reason for restricting nominal definition to a listing of analytic judgments. Again, one might worry about how much of a concept’s logical essence must be set out before we cross the threshold of nominal definition. Are we to say that a nominal definition is legitimate even if it lists just a small fragment of the concept’s logical essence? But this worry supposes that objects are presented to us not just—as I put it in the previous paragraph—by way of concepts, but solely by way of concepts. This seems to be how Leibniz thinks of divine cognition. God comprehends all that is true of a substance all at the same time; human cognition is then but a faint, finite approximation. By contrast, and as we saw in chapter one, Kant has objects presented to us in intuition. Again, from chapter one: Kant cannot allow concepts to be composed of individuals. That would be to allow existence to be established by concepts alone; hence the appeal of the relation of a point to its including space, and, thus, of the spatial model of concepts. The point is that existence is established through, and objects are presented in, empirical intuition. Thus, to think that reference to an object could be compromised by an incomplete definition or concept is to look in the wrong direction. Concepts are general; any one of them may apply to more than one object. Because intuition is immediate and non-conceptual, its presentational function cannot be compromised as we revise the content of our definitions and concepts.14
A detailed consideration of these issues awaits our discussion of intuition in chapter four, at which point the claim that experience includes a non-conceptual element will take center stage. It connects up with an important strand of Western religious thought, and, historically, has presented another inviting point of contact with Kant’s epistemology. At this point, however, we are almost ready to consider Kant’s definition of religion. In preparation we have reviewed his denial of real definition and his doctrine of nominal definition. Two varieties of definition remain: diagnostic and stipulative. We can be brief. In a series of Reflexionen, apparently from the 1770s, Kant describes a diagnostic definition as one that merely spells out a minimal number of sub-concepts sufficient to separate out a concept from others “for a purpose.” We might say: sufficient but not necessary conditions. The idea seems to be that a diagnostic definition permits identification, but not through properties derivable through the principle of contradiction.15 Thus, we have Euthyphro’s suggestion that the pious is what is “loved by all the gods.” Lastly, as the name suggests, a stipulative definition “is a concept which I have invented.” Kant emphasizes that “this arbitrary concept of mine does not assure me of the existence or even the possibility of its object. I do not even know from it whether it has an object at all” (A729/B757). An apt summary of the skeptical views of “religion” we reviewed in chapter one.
Let us turn now to the definition of religion that Kant favors in published work from 1788 to 1798. We want to observe Kant’s theory of concepts in action. In particular, we want to see whether or to what extent Kant respects the anti-essentialist, pragmatic attitude toward definition that it requires. The most detailed discussion occurs in Religion within the Boundaries of Mere Reason (1793), where Kant writes that “Religion is (subjectively considered) the recognition of all our duties as divine commands.” In a footnote, he adds, “With this definition many an erroneous interpretation of the concept of a religion in general is obviated.”16
Notice that, once Kant lays down this definition, the spatial theory of concepts takes hold right away. The initial definition gives priority neither to “duty” nor to “divine command.” It applies to anyone who takes the sphere of one to coincide with the sphere of the other—including but not limited to conservative Prussian Rosicrucians and liberal pietist Lutherans—but it does not spell out the nature of the inferential links between the two sub-concepts. That spelling-out begins in the very next sentence:
That religion, in which I must first know that something is a divine command in order that I recognize it as my duty, is revealed religion (or a religion that requires a revelation); by contrast, that religion in which I must first know that something is duty before I can acknowledge it as a divine command is natural religion.
In this passage Kant is dividing the extension or sphere of the genus “religion in general” into the species “natural” and “revealed.” Natural religion partly decomposes into such analytic judgments as,
I must recognize something as a duty before I can recognize it as a divine command.
It is possible there is a God.
It is possible there is a future life.
Grace may have its effects.
Whereas revealed religion partly decomposes into,
I am able to recognize something as a divine command before I can recognize it as a duty.
I know there is a God.
I know there is a future life.
I know this particular event is the result of divine grace.17
Once he has natural and revealed religion in place other, related subspecies then come in for similar treatment. “By acts of cult we can achieve justification before God” is part of the concept “superstition” (193); “enthusiasm” contains the judgment that “I can achieve justification through actual contact with God” (193–94). Recall that the spatial theory of concepts requires that every concept be composed of lower ones. Thus, in the context of natural religion, “God” includes the compound concept “moral originator of the world” (Religion, 191 [6:171]). The imprint of Kant’s theory of concepts is evident in all of this.
What shall we say of these definitions: real, nominal, diagnostic, or stipulative? Clearly they are not stipulative—at least not in intent. Kant writes as though he were “explicating” real features of the world. Nor is there anything to suggest that he thinks he is providing real definitions; what he elsewhere says cannot be provided. As I read him, Kant is claiming that none of these judgments can be denied of their matching concept without contradiction—in just the way he thinks the concept “gold” contains weight, color, malleability, and rust-resistance as analytic characteristics, in just the way that “is unmarried” cannot be denied of “bachelor.” “A future life as a tennis player is possible” cannot be analyzed out of “natural religion,” just as “Smith’s winning the lottery is the result of divine grace” cannot be gotten out of “revealed religion”—in just the same way as “declined today on the spot market” cannot be analyzed out of “gold” or “is bald” cannot be found in “bachelor.” As Kant sets it up, the latter connections are synthetic and a posteriori. I take it, moreover, that Kant is invoking the test of non-contradiction when he observes that, “Strictly speaking Judaism is not a religion at all . . . No religion can be conceived without faith in a future life. Judaism as such, taken in its purity, entails absolutely no religious faith” (Religion, 155; Ak. 6: 125–26). Here Kant is saying faith in a future life is to religion as the color yellow is to gold. The point is not about the essential properties of an object (as it would be with real definition), nor is it merely about distinguishing one concept from another (as it would be if the definition were merely diagnostic); it is (in Kant’s estimation) part of the decomposition or exposition of the concept, analytically contained in it in the way that extension is contained in body. Indeed, without the distinction between analytic and synthetic judgment to frame the remark, one cannot see what Kant is doing in denying that “Judaism” forms part of the sphere or Umfang of “religion.”
I take it, then, that Kant means to be giving nominal definitions of the genus “religion” and its sub-concepts, beginning with “natural religion,” and “revealed religion”—making explicit their logical essence by enumerating some of the analytic judgments of which they are composed. But why should we see these definitions as nominal rather than stipulative? The question might be motivated by two kinds of doubt: first, Kant clearly has strategic goals in this matter, among them the exclusion of Judaism and the opposing of “superstition,” “enthusiasm,” and “priest craft” in Prussia and beyond. To some extent surely Kant is shaping the content of his concepts to suit his purposes. Second, there is the thought that his definitions walk too narrow a Kantian line to be anything but stipulative. And it is true that, as Kant sets them up, the relationships of implication and of genus and species between the four elements of “natural religion” are conspicuously consistent both with the cognitive modesty required by the Critique of Pure Reason (they must deny “assertoric knowledge,” assertorisches Wissen, of the existence of God and a future life), and with the motivational integrity of the categorical imperative (they must imply that “there are no particular duties to God in a universal religion”).
How, then, to decide between nominal and stipulative? One thought might be to ask after the degree to which Kant’s audience would have endorsed his definitions—to use the fact that a concept is publicly held as a check against any attempt to simply invent new content for it. After all, Kant is using the German word Religion, and he wants to be understood by his readers. This minimal goal requires him to preserve enough of the ordinary use of the term—that is, enough of the commonly accepted Merkmale—so as to be “accepted as valid.” We have seen that he does not need, and recognizes that he does not need, consensus—that partial ignorance and error are to be expected. (This is fallout from what Brandom calls “the normative pragmatics of judgment,” to which we will turn in chapters five and six.) Taking this route, we must ask, What intersubjective community of human cognizers does Kant have in mind in this instance and what might they reasonably be thought to hold in common?
The answer, while no doubt of historical interest, is no help in this context. For this entire line of thought is opposed by an objection rooted in Kant’s own theory of concepts—what in chapter one appeared as the threat of emptiness. Kant repeatedly tells us that we cannot give sense to the genus-and-species structure of concepts on the strength of conceptual connections alone. It is not enough that Kant is actually engaging with a community of persons with common procedures, texts, etc. They might all be signifying nothing together. Today we might think of the astrological community “inquiring,” together, into nothing. (An example suggested by Kant’s mention of fate and fortune as “usurpatory” concepts.18) At some point our concepts have to be tied to the world, exhibited in concreto, as Kant likes to say: “For every concept we demand, first, the logical form of a concept (of thinking) in general, and then, second, the possibility of giving it an object to which it is to be related. Without this latter it has no sense, and is entirely empty of content” (A239/B298). The issue then is whether Kant’s “natural religion” and its neighbors should take their place, alongside fate and fortune, on his own list of empty, “usurpatory” concepts. This would be the Kantian expression of the contemporary skepticism about the general term “religion” that we encountered in chapter one.
If intersubjective agreement is not enough to give content to our concepts we must, then, turn toward the world, to the difference they make or could make in possible experience. We must locate a subject-matter whose characteristics can be investigated.19 But now we come to an even more serious problem: All three of the concepts in question include the recognition of duty as an essential property. But, as Kant has it, moral judgment incorporates an element of freedom and so is not observable.20 That is, it is never possible to determine, based on evidence available to the senses, whether someone is subordinating to duty what she takes to be a divine command or the reverse—or even whether recognition of duty is in play at all. In the case of “magnetic field” we have seen that Kant is able to turn to the material world (to the arrangement of iron filings) to exhibit the concept in concreto—but such a move is not available here. Here we are confronted not with an abstract term in the empirical arena, but with an exercise of the will whose place in the world in principle cannot be exhibited. This consideration moves Kant to note that, “religion is not a public condition” (Religion, 153; Ak. 6: 124)—a formula that puts him rather at odds with the present-day emphasis on material religion.
The situation, then, is as follows: Kant wants to be giving nominal definitions of the basic genus and species structure of religion. That requires observing how the relevant concepts figure in analytic and synthetic judgments, which in turn requires that they be connected up with the world. But the inclusion of duty as an essential component makes that connection impossible. This is why Kant says we cannot speak of the history of “religion on earth” (Religion, 153; Ak. 6: 124) but only of “types” of faith. Thus, “of all the public religions so far known, the Christian alone is of this [moral] type” (Religion, 95; Ak. 6:52)—by which he means, and can only mean, the (his) concept of Christianity. (Here a tortured reading of the Sermon on the Mount proves very helpful. Kant has no trouble finding in Judaism an ethics tied to fear of punishment and desire for reward while at the same time ignoring these same elements in the Sermon [Religion, 181ff.; Ak. 6: 159ff.].) But this makes for only a verbal advance. Our puzzlement over what gives content to “religion” is simply transferred onto “Christianity.”
To appreciate what resources Kant has available we must keep in mind his claim that his theoretical philosophy legitimates the large contours of the everyday empirical knowledge that we take for granted. This is Kant’s repeated claim that his “transcendental idealism” is equivalent to “empirical realism.” There is a powerful case to be made. In the Critique of Pure Reason Kant argues, for example, against various forms of skepticism, that substances really do endure and that causes really do produce effects (the first and second Analogies of Experience) and that objects really do exist in space outside me (the Refutation of Idealism). We must also keep in mind Kant’s confidence that his moral philosophy merely elaborates the contents of “common rational moral cognition”—that the concept of duty “already dwells in the natural sound understanding.”21 The idea is that, in decomposing the concept “natural religion” Kant is not just asking his critics to “accept as valid” the twin theses that knowledge of God is in fact beyond our reach and that actions are in fact not made right by being commanded by God. Rather, on the epistemological side, he is relying on arguments set out in the first Critique, and, on the moral side, on the basic tenets of his practical philosophy, especially on the experience of freedom. Thus, the project of arriving at a nominal definition of religion and its sub-concepts is not grounded in mere intersubjective agreement but rather in a subject-matter that, as he sees it, supports the claims of cognitive modesty and motivational integrity forming its logical (note: not its real) essence. When taken on its own terms, Kant’s attempt to give nominal definitions of “religion” and its sub-concepts stands or falls with the wider prospects of his epistemology and ethics.
A number of critics have remarked on the progressive thinning-out of the concept of God in Kant’s later works, and have urged us to see it as an artifact of the relentless march of modernity, of Kant’s own growing recognition of the primacy and self-sufficiency of the moral community.22 No doubt these play indispensable roles. But neither should we ignore the part in this drama played by the very small; namely, Kant’s theory of concepts. For we can see that its effects are going to be felt no matter what is happening on the grander stage. It is, as it were, the engine cranking away behind the scenes, exerting a continuous, humanizing pressure. For students of religion its effects may so far be seen along two axes: as vindicating the legitimacy of incomplete concepts at the expense of Leibniz’s transcendent God, and as thinning-out the concept of religion to the point where it picks out nothing on earth.
Kant’s favored definition of religion has of course come in for much criticism. Let us begin with Asad’s lament over, as he puts it, “the gradual evacuation of specificity.” In what follows, Asad first comments on and then quotes a passage from Kant’s Toward Perpetual Peace, published in 1795, two years after the Religion:
By 1795 Kant was able to produce a fully essentialized idea of religion which could be counterposed to its phenomenal forms: “There may certainly be different historical confessions,” he wrote, “although these have nothing to do with religion itself but only with changes in the means used to further religion, and are thus the province of historical research. And there may be just as many religious books (the Zend-Avesta, the Vedas, the Koran, etc.). But there can only be one religion which is valid for all men and at all times. Thus the different confessions can scarcely be more than the vehicles of religion; these are fortuitous, and may vary with differences in time or place.23
Asad concludes: “Far from being a concrete set of practical rules attached to specific processes of power and knowledge, religion has come to be abstracted and universalized.”
Let us put off consideration of the passage from Kant and begin with Asad’s prefatory comment. He is, as we have seen, quite right: Kant does in fact produce an essentialized idea of religion. But this is no criticism—or, if it is, it is a criticism not so much of Kant’s treatment of “religion” but of his entire picture of concepts and of nominal definition. As Kant sees it, all concepts have essential components. It is essential to the idea of a bachelor that he is unmarried but not essential that he is bald (if he is)—essential in the sense that to deny the first, but not the second, is to contradict one’s self. We may quibble with Asad’s “fully” essentialized, for we have seen that Kant explicitly envisages the legitimate use of concepts in partial ignorance of—and while at same time, after “advancing to experiments,” making revisions to—the analytic judgments out of which we had taken them to be composed. At the same time that Kant affirms essentialism about concepts, he denies it of objects—that is, he denies the possibility of real definition on the grounds that to identify essential features of objects is beyond the reach of finite creatures.
Asad next has Kant counterposing his definition of religion “to its phenomenal forms.” Here he is apparently taking Kant to be using “phenomenal” in Kant’s own technical sense—as half of the distinction, roughly, between objects as they are presented to the senses (phenomena) as contrasted to the thought of those same objects apart from all possible human cognition (noumena). As I noted in the introduction, and as Nietzsche is acutely aware, this distinction is another place in Kant’s theory of knowledge that has seemed to invite religious reflection. We will consider it in the conclusion. But in the case at hand I find no evidence that Kant is invoking his transcendental idealism. Rather, he is simply comparing his concept of religion to existing ecclesiastical forms. (In the previous section, we noted that Kant’s definition of religion makes trouble for the very idea of any such comparison: it is not clear how we can ever know that someone has given duty its proper motivational priority over the divine. So perhaps we ought to say that Kant is trying to compare his concept of religion to existing ecclesiastical forms.) But the general notion of comparing ideas to the world is both innocuous and momentous: innocuous because it trades on the simple thought that thinking that something is so does not make it so; momentous because it makes for what Kant calls the “nominal definition of truth”—“the agreement of cognition with its object” (A58/B82), employed by Kant throughout his Critical period. In these passages in the Religion Kant is writing in the spirit of empirical realism.
Turning now to the passage from Perpetual Peace: if not as offering a real definition of religion, how shall we read it? After all, we cannot rule out that Kant fails to respect his own strictures against real definition. The passage occurs in a footnote from which Asad has omitted the first sentence. In the main text Kant has just remarked that nations naturally wish to keep the peace “by ruling the whole world, where possible. But nature wills it otherwise. It makes use of two means to prevent peoples from intermingling and to separate them: differences of language and of religion” [Verschneidenheit der Sprachen under der Religionen]. Then follows the footnote, which I reproduce here from the Gregor translation, and with the first sentence reinstated:
Different religions: an odd expression! just as if one could speak of different morals. There can indeed be historically different creeds, [to be found] not in religion [nicht in die Religion] but in the history of means used to promote it, which is the province of scholarship [ins Feld der Gelehrsamkeit], and just as many religious books (the Zendavesta, the Vedas, the Koran, and so forth), but there can be only one single religion holding for all human beings and in all times [nur eine einzige, fur alle Menschen und in allen Zeiten gültige Religion]. Those can therefore contain nothing more than the vehicle of religion, what is contingent and can differ according to differences of time and place.24
The first thing to note is that Kant has no qualms, in the main text, about the plural, Religionen. He seems to be saying that the world contains different languages and religions in the sense that it contains different trees, or shoes. Why, then, does he find “different religions” an odd expression? The reason, I think, has to do with our main theme: the relationship between general terms and their objects. Objects are particular, and differences may turn up even between those that have so far proved conceptually indistinguishable—thus, Religionen. Concepts are general, each having a logical essence reflected in analytic judgments, and each of which may apply to a range of objects; that is, to a certain kind of thing—thus, Religion. What is odd is the thought of different trees or shoes—not different kinds of trees or shoes, but trees or shoes that cannot be grouped together as trees or shoes. Having recognized a plurality of religions (species, sub-concepts) in the main text he is, in the footnote, trading on the fact that we have thereby committed ourselves to the general concept “religion” (the genus). Thus, when the passage is considered in context it reflects Kant’s logical essentialism but also his anti-essentialism about real properties. In short, it reflects his theory of concepts.
The question of our commitment to the general term is of course very much with us today. Recall Tweed’s definition of religion: “Religions are confluences of organic-cultural flows that intensify joy and confront suffering by drawing on human and suprahuman forces to make homes and cross boundaries.” He comments:
Readers will notice that in my definition I shift from the singular to the plural, marking the boundaries of religions, not religion. That’s not because I want to resist talk about the field’s constitutive term . . . but rather to emphasize that interpreters—even armchair theorists—never encounter religion-in-general. There are only situated observers encountering particular practices performed by particular people in particular contexts.25
Probably all of us would endorse the sentiment motivating Tweed’s avoidance of generality, which is, I take it, the desire to do justice to who and what we encounter. Still, I think Tweed’s avoidance of the genus “religion” is just what Kant finds so odd. The problem is that even judgments in the singular about this or that practice or person or context must, whether we like it or not, be laced with generality. In cognizing that practice or person or context I must apply a general term—that ritual, this priest, that festival—that may well apply to others of its kind. That, as we have seen, is one of the basic elements of Kant’s picture of discursivity. Put in terms of the theory of concepts: we cannot escape generality just by leaving a particular genus behind—for the concepts at the next lower level are themselves composed of further, narrower concepts, as are they, and so on without end. Concepts are general all the way down.
Ironically, the notion of encountering a particular practice or person or context without having to apply general concepts reintroduces Leibniz’s theological model of cognition. It imagines, with Leibniz, one who can avoid generality through a complete determination of the object of cognition—one who can somehow cognize an aspect of the world in all its specificity all at once.26 That would indeed do justice to a particular practice, person or context! The Kantian, humanizing route is to recognize that our concepts are always general and incomplete—but, for all that, legitimate. The point is deeper than the accuracy or respect we owe to what we encounter. It is about what sort of encounter the nature of concepts will allow. George Santayana mocks the notion of religion-in-general from a related angle: “The attempt to speak without speaking any particular language is not more hopeless than the attempt to have a religion that shall be no religion in particular.”27 But the point about generality is not that anyone subscribes to the religion of religion-in-general; rather, that anyone who recognizes members of a species is thereby committed to recognizing, or at least (our topic in chapter three) searching for, the containing genus.
In the last chapter we saw that worries about reification are blocked by the generalizing function of concepts. We must always take “Islam” or “Jainism” to be open to further determination—not because we are methodologically enlightened inquirers, but, here again, because the concepts we employ in the course of inquiry will allow nothing else. The same consideration serves to block the flight from generality. We may, with Asad, lament the gradual evacuation of specificity in the study of religion, but let us not forget that its hoped-for return will be general in form.
Again, Lincoln takes Kant to task for inaugurating a “view of religion as delimited, and therefore definable.”28 For Lincoln, Kant’s definitional efforts represent the culmination of the centuries-long effort to make room, as against the medieval church, for secular learning. My concern here is not so much to disagree—Kant does delimit religion with a view toward giving a definition—as to resist the implication that delimitation can somehow reduce generality to definable proportion. In fact, delimitation and generality are unrelated—that is the point of Kant’s denial of an infima species. No matter where we are on the progression from genus to species to narrower subspecies to even narrower sub-subspecies—we have not moved the generality meter at all. At every stop along the way our concepts are no more or less general than those found at the previous level.
Lincoln expresses reservations about Asad’s remark that “there cannot be a universal definition of religion not only because its constituent elements and relationships are historically specific but because that definition is itself the historical product of discursive processes.”29 Lincoln wonders why historical specificity should block definition and, provided we are talking about nominal definition, it is a fair question. But our Kantian reflections should make us hesitate equally over the second half of Asad’s remark. Here again my aim is not to disagree—partly because it is not quite clear what Asad means by “universal” definition, but mainly because Asad is again quite right: Reflection on the decomposition of a concept is indeed a discursive process—here let us call to mind the image of Kant puzzling over the decomposition of “event,” soon to awaken from his dogmatic slumbers. But it should by now be clear that discursivity is no enemy of definition. On the contrary, the structure of discursivity—the unavoidability of applying general concepts to what strikes us in experience—makes definition possible. Lincoln urges us to understand “efforts at definition as provisional attempts to clarify one’s thought, not to capture the innate essence of things”—an excellent summary of Kant’s doctrine of nominal definition and the theory of concepts that supports it.
We have so far considered objections to Kant’s definition of religion that trade on doubts about his handling of generality and specificity. Other objections depend on applying the definition to actual human communities. Thus, according to Philip Quinn, Kant’s definition
illustrates one sort of problem that arises when a philosopher attempts to define religion. It is generally acknowledged that Theravada Buddhism is a religion. However, its doctrines do not include belief in a personal divinity capable of issuing commands. So it seems that recognition of duties as divine commands is not a necessary condition for being a religion. It thus appears that Kant’s formulation does not specify the correct extension for the concept of religion; it seems to fail to provide conceptually necessary and sufficient conditions for being a religion. If we agree that this is a failure, we may wish to excuse Kant on the grounds that he knew much less about the full extent of religious diversity outside the West than we do.30
I think we have to say that, on the grounds that Quinn marks out, the situation is worse for Kant than Quinn portrays. As we have seen, Kant’s definition does not fit and cannot be made to fit even a single historical tradition or community, whether inside or outside the West. Of course we must add that Kant is well aware of this. He says as much himself. Quinn is criticizing Kant for failing to do something Kant says cannot be done.
But let us put this aside for the moment and join the issues raised by Quinn. Can we fault Kant for not providing necessary and sufficient conditions for being a religion? Not without at the same time rejecting his wider theory of discursivity. For to provide such conditions would require us to locate the object’s real essence and thereby to provide a real definition—a feat Kant holds to be out of reach for finite creatures. (We could perhaps provide sufficient but not necessary conditions. That would yield a merely “diagnostic” definition—but clearly Quinn means to be aiming higher than that.) And what about Quinn’s complaint that Kant fails to “specify the correct extension for the concept of religion”? There is evidently a fundamental tension between this requirement and Kant’s claim that empirical concepts “never remain within secure boundaries.” On this point Kant’s view is friendly toward certain themes from Mark Wilson’s recent work Wandering Significance. Wilson criticizes what he calls “the ur-philosophical desire for semantic fixity”—that is, “a deep attachment to the notion that the contents of our concepts stay largely invariant over time.”31 Semantic fixity makes a natural partner with “classical gluing,” the idea that predicates are somehow glued to properties and that learning a language involves coming to a permanent “grasp” of that connection.
By contrast, Kant invites us to “advance to experiments.” Let us put to one side Quinn’s tacit appeal to essential properties and semantic fixity, and let us see his gesture toward Theravada Buddhism as taking up this invitation; in effect, as asking how Kant’s definition fares when we apply some non-theistic pressure. Even when constructed along narrower lines—designed, say, to put pressure only on the element of duty in “religion is the recognition of all our duties as divine commands”—our experiments could force substantial semantic revision. They could show that Kant is mistaken about the nature of duty—that it is in fact, say, utilitarian or Durkheimian or communitarian; or our inquiry could show that the whole notion of universal obligation is illusory. (We are familiar with arguments for each of these outcomes.32) But why stop there? Clearly, Quinn thinks Kant’s error is much larger than simply mistaking the nature of duty. And, indeed, Kant must be fully open to the pragmatic purport of his own account of nominal definition; the thought might be that we cannot say in advance how secure or insecure the boundaries of that definition will be. Perhaps others will step forward to better, in Wilson’s terms, “map out the informational terrain.”33 And of course in due time the humanities and especially the social sciences did begin to step forward.
The question is how far such experiments—again, a prospect explicitly envisioned and endorsed by Kant—how far can they move the boundaries of Kant’s definition of religion?
Here I find helpful Wilson’s discussion of the surprising flexibility of even such an apparently simple concept as “hardness,” for which, it turns out, there is no single test. Wilson points out that, in the course of everyday life, we tend not to notice that we apply different tests to different substances—squeezing or impressing for plastics and rubbers, scratching and cutting for ceramics, striking and rubbing for metals, and so on. Apparently we do tacitly understand that hardness for brittle substances is not hardness for malleable ones.34 Wilson writes of the concept “hard” being “tugged,” “dragged” and “prolongated” as it is applied in diverse contexts. Moreover, as our everyday tests are taken up into the various manufacturing industries they are often improved, calibrated and assigned to such specialized instruments as:
Brinell or Vickers indenters (vigorous squeezing and then releasing); superficial Rockwell testing (mild squeezing and partial releasing), durometer (squeezing without releasing), sclerometer (scratching), scleroscope (a different instrument that raps its specimen), the Charpy impact test (hitting with a hammer) and so forth.
The upshot, says Wilson, is that “our employment of ‘hardness’ silently distributes itself into a patchwork of sheets, locally distinguished by a certain vein of probing (scratching, tapping, etc.), that sit over various varieties of material stuffs and continue smoothly into one another.”35
Might we view our deployment of “religion” in this way? Might we see, say, Marx, Durkheim, and Freud as undertaking some “quality manufacturing”; as exerting differential pressure on the concept “religion” the likes of which Kant never imagined? In order to answer in the affirmative we have to be able to see Kant’s and the social scientific treatments of “religion” as “continuing smoothly into one another”—and this we cannot do. On this point our hand is forced by the extreme abstractness of Kant’s view of moral compulsion. Strictly speaking, reason—our own reason—compels us to act freely, and this compulsion is not detectable as such in the world, and so it cannot be tugged in the direction of economics or dragged toward the sociological, psychological, or any other material aspect of the world. But if we relax the Kantian strictures a bit, then we can begin to detect the semantic drift occasioned both by the increasing sophistication of the social (and other) sciences and by the complexity of the world. Durkheim, for example, was well aware that Kant defines religion as the recognition of all our duties as divine commands. His treatment of “duty” and “divine” are instructive.
In The Elementary Forms of Religious Life, Durkheim argues that society can command in categorical fashion. Whereas considerations of mere “social utility” mark the realm of magic, “religious prohibitions are categorical imperatives.” By this Durkheim means they threaten not merely physical penalties (“as from not following the advice of his doctor”). Rather, the religious prohibition is rooted in “the respect evoked by the sacred object, and its purpose is to prevent any disrespect.”36 Now in appealing to the notion of duty and in contrasting it with mere utility Durkheim thinks of himself as dragging the Kantian concept out of its a priori fog and into the sociological light. Again, from a Kantian point of view, this cannot be right. No matter how we strengthen the sense of communal authority it cannot command categorically in Kant’s sense—that is, its form cannot be other than hypothetical. Nor does it pretend to be: Durkheim is clear that what gives religious prohibitions their force is the survival of the individuals of which society is composed. Thus we cannot see the Kantian and Durkheimian tests for “is religious” as we see ceramic and rubber being tested for “is hard” by, respectively, a sclerometer and a durometer. But this is precisely where strict adherence to the Kantian line does not serve us well, for by insisting on taking it as having only a priori application we obscure what real semantic movement is taking place. No matter how we assess Kant’s ethics our interest at the moment is in tracing the history of, as Geuss puts it, “the whole concept of absolute obligation”37 as having been tugged by Christianity (notwithstanding Kant’s own analysis of the Sermon on the Mount, as installing the sub-concepts “reward in heaven” and “punishment in hell”), dragged by Kant (as pruning away all inferences to or from material ends), and prolongated by Durkheim (as adding inferences to and from the sub-concept “society”). Whether, with Nietzsche and Geuss, we think humankind would be better off freed from any concept in the vicinity of absolute obligation or whether, with Kant and Durkheim, we think the concept just needs to be purified—either way, the transitions at stake look reasonably smooth.
Durkheim does his best to tell a parallel story about the concept of the divine. Thus, society is “a being that man conceives of as superior to himself,” “on whom he depends,” one that “fosters a sense of dependence,” “categorically demands our cooperation,” “subjects us to restraints, privations, and sacrifices,” “is the object of genuine respect,” “pursues its own ends,” and so on.38 He is often sensitive to the fact that his remanufactured, sociological sense of “is divine” takes us some distance from the God of the Hebrew or Christian Bible as well as from the ethnography of Australian totemism (his official object of study in the Elementary Forms)—and also to the fact that the road from one to the other is anything but smooth. Awareness of the gaps, as I have argued elsewhere, prompts such modest claims as that religion is “practically true” and expresses the facts “correctly enough.” At the same time Durkheim is not immune to spasms of semantic euphoria: thus, “god and society are one and the same.”39 Put in terms of the spatial theory of concepts, practical or symbolic truth results when “society” and “divine” share the sorts of spatial parts (sub-concepts) listed above (and support the same inferences between them) but diverge over others. In sum, I am portraying Durkheim as being largely persuaded by Kant’s “duty” and “divine.” However, his unhappiness at the prospect of locating them in what he perceives as the obfuscatory realm of the a priori makes an opening for what Wilson calls the “prolonging impulse”—for dragging them, in this case, into the sociological arena.
To return, finally, to Quinn: Can Kant’s definition of religion be tugged, dragged and prolongated all the way to Theravada Buddhism? Can we answer Quinn’s objection by appealing to the elasticity of “divinity”? No. But once we put Quinn’s talk of necessary and sufficient conditions to one side we can see that the natural conclusion is not that Kant has failed to “specify the correct extension for the concept of religion.” Rather, we are left with the simple thought that Quinn has changed the subject—that he, and, for the most part, we, now use “religion” in a very different sense than Kant recommends.40 No doubt the change of subject is for the better; for most of our present purposes Kant’s definition is not terribly helpful.41 Today the academic study of religion is explicitly comparative, and takes seriously the material basis of religious experience and practice—for these and other reasons, Kant’s favored definition has for some time made an inviting target. But notice that, whatever our new and improved general concept of religion turns out to be, it will display those formal features familiar to us from Kant’s spatial theory of concepts and reflected in his own favored definition. First, it will be composed of further, more fine-grained sub-concepts. Thus, when Durkheim defines a religion as “a unified system of beliefs and practices relative to sacred things, that is to say, things set apart and forbidden—beliefs and practices which unite into one single moral community, called a church, all those who adhere to them,”42 he is explicitly decomposing the concept “religion” into “beliefs,” “practices,” and “community,” as well as the concept “sacred things” into “set apart” and “forbidden,” each of which sub-concepts admit further specification, etc. Second, this conceptual structure will have a logical essence—a requirement given elegant expression by Lincoln when he remarks that, as he sees it, “four domains—discourse, practice, community, and institution—are necessary parts of anything that can properly be called a ‘religion’.”43 And third, the concept will “never remain within secure boundaries”; rather, we will “advance to experiments,” reserving to ourselves the right to add a fifth domain or, perhaps, to think better of one of the four now on offer.
In all of this, Kant is with Wilson in seeing “conceptual instabilities as one of the unavoidable inconveniences intrinsic to linguistic life.”44 We will shortly turn to what Kant sees as the a priori context of linguistic life, but it is striking that his theory of concepts and of definition should reflect such deeply pragmatic commitments.
Beginning with William James, probably no single claim has been more consistently attacked than that religion can be identified through some one or several essential properties. Thus, James cautions that “the word ‘religion’ cannot stand for any single principle or essence, but is rather a collective name. The theorizing mind tends always to the oversimplification of its materials.” Smith warns that “Students of religion need to abandon the notion of ‘essence.’” Ivan Strenski notes that “if Smith is correct, we in Religious Studies have carried on naively as if the word ‘religion’ stood for a natural kind and thus has not required critical reflection about the historical vicissitudes of the term.” And Geuss, after surveying four types of definitions of religion, points out that they all
share in their more primitive forms a common assumption, namely that there is such a thing as a timeless essence of religion which can be formulated in a definition of religion. Nietzsche, Wittgenstein and Foucault have, each in his own way, subjected this assumption to massive criticism, so it is, to say the least, no longer unproblematic.45
I hope it is clear that Kant belongs in the thick of this crew, but also that his brand of anti-essentialism comes with a series of riders. He is with James in taking “religion” to be a “collective name”—but for reasons having nothing to do with what aspect of the world it picks out; all concepts are collective, that is, general. He is with Smith in thinking that real essences are indeed to be abandoned—but adds that we must take our concepts to have essential components, on pain of married bachelors and Durkheimian religions of solitary individuals. And he supports Strenski’s thought that “religion” cannot stand for a natural kind—but for the reason that there are no natural kinds; that is, no substances whose defining property or properties can somehow be cognized more directly than by way of the application of general concepts. Rather, all cognition is through concepts composed of further sub-concepts without end, constantly refined and reconfigured through further experience and inquiry.
On this point Kant goes further even than Nietzsche, the arch anti-essentialist. In The Genealogy of Morality Nietzsche famously remarks that “only that which has no history can be defined.” Nietzsche holds, and Geuss following him, that such items as triangles, water, mass, and gene admit of strict definition because they are “not part of history.” By contrast, Christianity, punishment, conscience, morality, freedom, democracy are so ensnared in the shifting sands of human, historical circumstance as to make definition impossible.46 From Kant’s point of view, the distinction between items with and items without a history is dogmatic and insufficiently radical. Water and Christianity must both be given what content they have based on the application of general concepts to what strikes us in the course of experience. At most, each is susceptible of nominal definition. History is not to the point.
Asad has been a persistent critic of the tendency toward essentialism in the study of Islam.
The argument here is not against the attempt to generalize about Islam, but against the manner in which that generalization is undertaken. Anyone working on the anthropology of Islam will be aware that there is a considerable diversity in the beliefs and practices of Muslims. The first problem is therefore one of organizing this diversity in terms of an adequate concept. The familiar representation of essential Islam as the fusion of religion with power is not one of these. But neither is the nominalist view that different instances of what are called Islam are essentially unique and sui generis.47
The nominalist contemplates a world in which particulars (this Imam, that text) resist assimilation to general concepts (Islam), and in which general concepts in turn resist assimilation to wider genera (religion). In the next chapter we will ask what is at stake in this kind of ascent to broader and broader generality, unity, and systematicity—what legitimacy does it have and what is its basis? For the moment, we may note that Kant shares Asad’s rejection of both essentialism and nominalism. Kant’s spatial theory of concepts accepts the burden, in Asad’s terms, of “organizing diversity” by making explicit the role in empirical cognition of general, hierarchically nested concepts. The objects we cognize cannot have timeless essences—so love cannot be an essential property of Christianity any more than power can be of Islam, or, for that matter, than H2O can be of water. At the same time, the fact that cognition of particulars is always by way of general concepts guarantees that nominalism will always be late to the scene. We have seen that Kant stands with James in rejecting real essences. Not so as concerns the theorizing mind.