Most readers of Foucault are eventually struck by a methodological conflict in his work: While Foucault is adamant that systems of knowledge are constituted within social contexts and maintained by power relations, he is equally convinced of the need for theoretical critiques, such as his own, of the history of knowledge. Such critiques are not intended to be positivist history; rather they have the unmistakably normative aim of freeing us of certain contingent and undesirable features of our (scientific) patrimony. In Anglo-American terms, Foucault walks like a social constructivist, but he often talks like a normative philosopher. By conventional measures, the methods of these schools are incompatible, and this is no doubt one reason that Anglo-American readers have tended to damn or to praise Foucault for his relativist leanings.1 Either they overlook the normative aims of his work, or they simply ignore the tension between constructivist means and normative ends.
My purpose in this chapter is to explicate the sources of this seeming inconsistency by analyzing two key notions in Foucault’s early work, by which I mean his writings up to and including The Archaeology of Knowledge . The first of these notions is the notion of archaeology itself, a form of historical investigation of knowledge that is distinguished from the mere history of ideas in part by its unearthing what Foucault calls “historical a prioris.” This concept is the second of the two I will be concerned with. Such historical a prioris, on Foucault’s account, are “unconscious” matrices governing the space of possible statements (énoncés) that occur in the writings of a given historical age. They are conceived simultaneously as conditions on the knowledge of a given scientific culture and as the framework that historians and philosophers use to classify the writings of the age they are studying. Both notions, I shall argue, are derived from Husserlian phenomenology. But both are modified by Foucault in the light of Jean Cavaillès’s critique of Husserl’s theory of science, according to which not only Husserl, but also most logicist philosophers of science and mathematics, are laboring under Kantian “transcendentalist” delusions.
According to Cavaillès, transcendental philosophies of science seek to justify logical and methodological norms by deriving them from foundational acts of consciousness. This approach is easily discerned in Kant: In the first Critique, the normative authority of logic and mathematics is both explained and justified by revealing their origins in those faculties of the mind whose operations are conditions for the possibility of unified and conscious experience. Drawing on his work on the history and philosophy of mathematics, Cavaillès argues that Kant and his successors, above all Husserl, are unable to account for the evolution of modern logic and mathematics on this scheme. Husserl, for instance, extends the transcendental program by focusing on the role of intentionality. Already in his earliest writings on mathematics and logic, he seeks to explain the normative character of these sciences by means of a phenomenological analysis of original intentional acts: Geometry, for instance, concerns “ideal” mathematical objects, such as triangles. These ideal objects can be the objects of intentions only to the extent that such intentions are systematically developed out of more quotidian ones concerning immediate spatio-temporal experiences. Such a process of idealization is not, however, fundamentally different from that involved in the objectification of other objects of experience. For instance, we also constitute a three-dimensional tree standing behind our immediate perceptions by imagining the tree viewed from multiple points of view. In both cases (the tree and the triangle), the objectification of the object goes hand in hand with certain constitutive operations of consciousness, which operations are the conditions on our establishing intentions regarding trees and triangles. A phenomenological investigation lays bare the normative content implicit in our intentional states by analyzing such primitive constitutive acts. Such acts are, as in Kant, the acts of a single, conscious subject.
I will not have occasion to go into Cavaillès’s objections to logicism and the work of the Vienna Circle at any length in this chapter; however, it is worth touching on the parallels he sees between Husserl’s and their work. Logicist theories of mathematics seek to reduce the content of mathematical propositions to logical truths. For both Frege and Wittgenstein, this analysis requires a further definition of these logical truths, and in both cases this definition presumes the existence of prior and structured meanings, for instance Frege’s third realm of senses, or Wittgenstein’s logical space. In order for the reduction to do any work, these structures must be assumed as given, for otherwise the scope of the phrase “logical truth” would not be well-defined, and thus the point of the reduction would be nullified. That is to say, if the principles of logic could not be identified a priori, then we might have to introduce supplementary logical principles as we moved along. And in such a case the claim that mathematics was in some sense reducible to logic would at the very least lose its persuasive force.2 Traditional logicism, on Cavaillès’s reading, founds the authority of logic by invoking the prior existence of meaningful symbols, and to its credit offers more than a merely formal justification of logical laws. Later versions, such as that of Carnap, abandon this presupposition; however, they do so at the cost of making logic part of mathematics—thus of abandoning the initial reductive project.3
Cavaillès is opposed to all these theories of science and mathematics and proposes in their stead what he calls a “philosophy of the concept,” which is a philosophy that tries to understand the sciences by examining the history of concepts and the norms that govern their use. Such an approach is to be distinguished from its transcendental foils in that it does not demand or expect the closure that transcendentalism cannot do without. It does not, in other words, imagine that the ground of normative rules is to be found by revealing hidden cognitive structures, or what Husserl calls “sedimented” cognitive acts. Cavaillès complains that all such theories inevitably lead to the result that there cannot really be anything unexpected in science: The real objects with which science is concerned can never penetrate the fabric of cognition, because the latter must screen the real world precisely in order that it may the source of norms. Transcendentalism, on this view, is constitutionally opposed to realism, and it also cannot admit any real developments in scientific methodology.
It is, I shall be suggesting, just this critique of transcendental philosophy that provides the philosophical background to Foucault’s early work. For example, in his contribution to an issue of the Revue de Métaphysique et de Morale dedicated to George Canguilhem’s work, Foucault gives a brief summary of the state of French philosophy and history of science in the period after the war. In an oft-cited passage,4 he suggests that there was a fundamental division cutting across the series of more easily recognizable cleavages between Marxists and anti-Marxists, Freudians and anti-Freudians, and so forth, namely, the opposition between a philosophy of the subject, of consciousness, and a philosophy of knowledge, “of the concept” as he puts it.5 This division filtered the reception of Husserlian phenomenology in France, with the result that Husserl’s thought was reworked on the subjectivist side by authors such as Sartre and Merleau-Ponty into the existential philosophy, whereas the conceptualists—among whom he includes Bachelard, Canguilhem, Cavaillès, and himself—were more interested in Husserl’s early works on the philosophy of science and mathematics.
As I have just indicated, the term philosophy of the concept is introduced by Cavaillès during his critique of what he regarded as Husserl’s inordinate appeal to the subject in the latter’s final work, The Crisis of European Sciences. The research program that both Cavaillès and Foucault mean to describe with this term is, one might say, empirically rich but philosophically poor. Bachelard’s writings on physics, Cavaillès’s work on mathematics, Foucault’s and Canguilhem’s research on the life and human sciences offer detailed analyses of the emergence of new conceptual frameworks in the sciences; it is work whose interest is largely independent of its theoretical underpinnings. However, their theoretical assumptions are at best vaguely articulated, and their alternative philosophy of the concept is defined largely by negation: It is not a transcendental philosophy of science, in that it does not assign a central epistemological role to the conscious acts of subjects. Cavaillès is hostile to any project that seeks the sources of logic and evidence, method and data, in the minds of scientists. Similarly, when Foucault in the 1960s describes his work as concerned with the unconscious of science, he means to underscore what it is not, namely a phenomenological investigation of the history of the sciences. He is quite right to emphasize the distinction between his work and that of Husserl. But his need to do so stems from the overwhelming similarities between his theory of archaeology and Husserl’s work. Indeed, both of our key concepts from the Archaeology of Knowledge (archaeology and historical a priori) are of Husserlian origin.6 And both are repeatedly explained there in terms of what they are not: The discourses that comprise fields of knowledge and the rules and ontologies attached to them are, for instance, not conditions on the experience of historical actors. Of course, they do have the function of unifying fields of objects by means of rules, and in this sense their role is parallel to Kant’s requirement that the understanding unify the data of conscious experience. But here again, Foucault denies that his theory is a Kantian one, claiming instead that his rules of discourse differ from Kantian rules of thought in being rules governing language. They are not “condition[s] of validity for judgments, but ... condition[s] of reality for statements” (Foucault, 1969, 167). The historical a prioris constituted by these rules are not conditions on what could be thought, but on what could be said in a given science in a given age.
This insistence on the theoretical priority of language to thought is indeed characteristic of the French backlash against phenomenology in the period after the war, a backlash in which Foucault was only one of many participants. As I shall suggest below, the seeds for this reaction were planted by Husserl himself in his late recognition of what Merleau-Ponty called the “problem of language,” that is to say, the problem of explaining how signs convey meaning in the absence of concomitant intentional acts. The division between Foucault and his structuralist allies on the one hand,7 and phenomenologists such as Husserl, Merleau-Ponty and Sartre on the other, consists above all in the former’s insistence that meaning should be analyzed by looking to the implicit structures of sign systems in preference to considering it as a cognitive phenomenon. This French “linguistic turn” is analogous to simultaneous developments in the English-speaking world, and it is characterized by a similar skepticism regarding the theoretical value of meanings or intentions, at least insofar as these are conceived as elements of consciousness.
As we just saw, in Cavaillès’s critique it is not the problem of language that plays the central role, but rather the transcendental approach to the philosophy of science and logic. Nevertheless, the two topics are intimately connected, in that the problem of accounting for language and meaning within the framework of a transcendental philosophy is central to the late work of Husserl that Cavaillès criticizes.8 And it is also, on his view, a central topic of early logicism.9 The difficulty, in a nutshell, is that transcendental theories assume that the meanings of physical signs are mental events (senses, judgments, possible experiences). But it is equally clear that language has a capacity to generate, or at least convey, meanings that are not the products of our own conscious activity. Let me emphasize that this subordination of language to thought is not an accidental feature of such philosophies. It is, on the contrary, essential to their aim of explaining the binding character of normative principles. The reason that propositional logic is valid for linguistic statements, for example, is that it derives from constraints on propositional thought. Transcendental theories argue that in placing conditions on the structure of thoughts, these constraints also determine general connections among kinds of thoughts, for instance those holding among their categorial forms. These formal relations are codified in normative sciences such as logic, which tells us what sorts of transitions between sentences necessarily preserve truth.
This form of argument is therefore committed to the principle that meaning is determined by thought, and that language—the system of signs that refer to, or bear meanings—must conform to this determinacy requirement. Such a requirement is clearly exemplified by Frege’s demand that every well-formed sentence of a logical calculus must have a sense: If this were not the case, then logical (syntactical) operations on other sentences could generate meaningless expressions, so that logic would not be truth-preserving. In other words, although logic is applied to sentences, it derives its normative force from meanings or thoughts. The problem, once again, is that it seems to be a simple fact about actual experience, and actual scientific practice, that we operate with meaning-bearing signs without any clear idea of their senses. We inherit both the vocabulary and grammar of the languages we speak, including those of formal scientific languages. And this fact puts pressure on the transcendental theorist: Since a speaker of these languages may never consciously have fixed their meanings, the theorist must explain where the meanings of such expressions are to be found, and such explanations run the risk of extravagance. Frege, once again, is obliged to postulate an entirely separate “realm” of senses, which are neither signs nor their referents. Husserl seeks refuge from this problem by arguing that the meanings of such signs are sedimented. At some time in the past, there were indeed conscious intentional acts that assigned meanings to these signs. If we want to know what the implicit rules of use governing these signs actually are, we have to inquire into their phenomenological roots. He concludes that it is an essential task of the epistemology of the sciences to identify and reactivate these sedimented meanings and rules.
The anti-phenomenological response to this problem of language is simply to drop the idea that meaning consists in hidden intentional acts and to accept the fact that sign systems acquire their meaning from elsewhere. One thereby shifts the investigation away from the cognitive acts of thinking subjects in favor of, for instance, social structures or speech acts. But it is evident that this solution carries a number of problems with it: Defenders of an intentionalist approach will rightly object that, whatever the legitimacy of social or behaviorist theories of meaning, they will never get at the very phenomenon that was to be analyzed. This objection is in fact programmatic, for both Frege’s and Husserl’s approaches were conceived as foils to naturalism: Frege’s senses and Husserl’s intentions cannot be construed as psychological without rendering the logical relations they determine purely contingent. In such a case, they would fail to be Denknotwendigkeiten, “necessities of thought,” as Frege insists they must be. On my view, the negative definition of the philosophy of the concept that Cavaillès and Foucault adopt avoids the problem of language that emerged within phenomenology by denying the centrality of consciousness to meaning, but it does not adequately address the problems that remain once one has done so. The philosophers of the concept want the normative component of transcendentalist theories without the element of consciousness. This desire leads them to posit unconscious structural conditions on language, and these unconscious conditions are no less problematic than the hidden intentional acts they were intended to replace.
In the following, I will develop Cavaillès’s criticisms of the transcendental philosophy of science by considering in sequence his critique of Husserl’s late work, The Crisis of European Sciences, and then the criticisms of Kant with which he opens his On Logic and the Theory of Science. In conclusion, I will expand on the philosophy of the concept subscribed to by Bachelard, Canguilhem, Cavaillès, and Foucault and will consider above all the philosophical difficulties raised by their theory. My discussion will take the form of a close reading of two passages from the last page of Cavaillès’s book, the first cited by Foucault in his essay on Canguilhem (Foucault, 1994b), the second cited by Canguilhem in an essay on Foucault (Canguilhem, 1967):
Still, the justificatory evidence of transcendental analysis is necessarily unique: while there is consciousness of progress, there is no progress of consciousness. Whereas one of the essential problems of the theory of science is just that progress is not augmentation by juxtaposition, in which the earlier subsists with the new, but a perpetual revision of contents by extension and erasure.10
There is no one consciousness generating its products, nor simply immanent to them, rather it is each time in the immediacy of the idea .... Progress is either material, or between singular essences, and its motor is the need to exceed each of these. It is not a philosophy of consciousness but a philosophy of the concept that can provide a theory of science. The generative necessity is not one of an activity, but of a dialectic.11
Cavaillès’s main objection here is quite simply that transcendental philosophy, at least in its traditional forms, can never explain how science can change. And the reason it cannot do so is that it is committed to the idea of an eternal and unchanging form of reason. Because the latter is, so to speak, already with us (in our minds, in our past, or in our language), the philosopher’s task is inevitably directed toward the past. Both Canguilhem and Foucault endorse this analysis, because they are interested in analyzing the development of scientific rationality. In order to see what is at stake philosophically, I will unpack each of these quotations in turn.
The first passage I have cited is a reductio. Transcendental analysis seeks the normative roots of logic and mathematics in original (and in this sense constitutive) intentional acts. But this means that such analyses cannot make sense of a basic property of science, which is that it surpasses and obliterates its past. Therefore, such analyses are falsely premised: It is not true that there is only one ultimate domain of justification and evidence, namely human consciousness. As we have seen, this concern with the normative origin of scientific logic drives Husserl’s phenomenological project. However, it only becomes an explicitly historical concern in his late work.
Husserl argues in the Crisis that we need a historical analysis of scientific concepts because these concepts are, in his terminology, sedimented, by which he means that they are concepts whose original meanings are concealed from us. As science progresses, it lays down layers of formalized systems, whose elements are increasingly remote from their experiential base. His primary example is that of geometry. Geometry was originally a technique of measurement immediately related to the life-world experiences and needs of the first “proto-geometers.” In an initial, critical step, this knowledge was cast in axiomatic form. The formalization of geometry made it possible to reason concerning mathematical objects and to produce intersubjectively valid, eternal truths concerning these ideal entities. But this procedure severed the link between mathematical concepts and their life-world correlates. One could now think about geometry mechanically, by manipulating signs, so that one could do geometry without thinking about its phenomenal origins.
The next sedimentary layer Husserl considers in the Crisis is that of Galilean physics, which draws on axiomatized geometry without regard to its origins, ultimately positing it as the structure of physical space. It lays down a new layer of formalized science, namely the system of Newtonian mechanics, which also involves concepts with life-world correlates, such as that of force. These are also fixed symbolically, they sediment, and they are then reified as the world of mechanistic physics. Because this world is posited as the causal undercarriage of reality, it gives rise to a “crisis” in the sciences, for it now seems as though the phenomenal life-world, which is the true epistemological foundation of knowledge, is nothing but a supervenient illusion. The crisis can only be resolved, in Husserl’s view, by reversing the process of sedimentation in order to reveal the genuine grounds of the sciences. It is this part of the phenomenological method that Fink characterizes as archaeology, a remark Cavaillès cites as an apt description of Husserl’s project.
Thus Husserl thinks we need a historical epistemology of the sciences because the life-world is buried under a sediment of scientific concepts. By inquiring into the origins of these conceptual layers, we “reactivate” their original meanings, and thereby reveal both the source of scientific norms and the original source of meaning. Calling such an inquiry a true epistemology of the sciences, he anticipates the obvious objection. Why on earth do I need to ask about the conditions under which concepts were first developed in order to give account of their legitimate application? Surely I have the statements of geometry in front of me so that I can work back to their axioms and then inquire into their validity? Isn’t that what it means to conduct an epistemological inquiry? Husserl responds to his own interjections by arguing that all concepts, indeed all cultural products, have an implicit historical dimension. We know that they were produced in the past, and we could not possibly make sense of questions concerning their legitimacy without having this fact in mind. For instance, to ask whether a cultural product such as a hammer is a good hammer is to know that it is a hammer and that it was produced by people to carry out certain tasks. These prior purposes are implicitly invoked whenever I consider the hammer as a tool, instead of as a lump of steel and wood. Similarly, I do not come to learn concepts, indeed I could not learn them, without first accepting them as cultural products with an intentional history, even if I do not know exactly what this history was. In order to apply them in new situations, I must at least imagine some past application.
What distinguishes this account of Husserl’s from his earlier work is its insistence on the historical dimension, which is a late consequence of his desire to overcome Kant’s formalism. Typically enough, Husserl’s theory of scientific norms assigns a central role to intentions and their objects. In order to understand what our logic should look like, we inquire into intentional acts, which inquiry of course addresses their referents. Indeed, an essential aim of phenomenological analyses is to elucidate the simultaneous constitution of intentions and their objects. In doing so, we will come to understand what structures of consciousness are essential to our having intentions of a particular kind, for instance, mathematical ones. We are supposed to get a transcendental logic, but one that is founded on real experience, as opposed to intuitions and categories borrowed from the established sciences and posited as hidden cognitive faculties. Such an investigation therefore eschews aprioristic demonstrations in favor of phenomenological investigations of fields of concrete experience. In his earlier works, Husserl still imagines these constitutive intentional acts as occurring within the confines of a single consciousness. Just as in the related constitution-systems of Carnap, we are to imagine a single consciousness starting from a phenomenal base of givens in order to work its way up to the high-level concepts of scientific theory.
But by the time of the Crisis, Husserl sees this approach as overly simplistic. It is evident that the fields of ideal objects that make up scientific ontologies are not given in immediate experience—indeed they cannot be, because in order to have the degree of intersubjective validity and temporal invariance that we require from the sciences, their theoretical objects have to be distinct from the immediate experiences of given individuals. They are, Husserl emphasizes, first constituted as intentional objects when the appropriate written formalisms are introduced. But these formalisms are the products of a long cultural tradition. The reason that we need a historical investigation into the origin of geometry or into the origins of Galilean physics is that the formalisms that anchor the possibility of referring to ideal scientific objects are something that each individual consciousness has inherited. In order to know the logic that is appropriate to scientific objects, we have at least to imagine a situation in which they were constituted and sedimented, or in Husserlian terminology, the situation in which their ideal form was concretely available to a consciousness. Husserl recognized, in other words, that language could not be accounted for with his usual phenomenological methods, for the simple reason that the meanings of words are not something that the individual consciousness creates.
Because such a historical investigation must be undertaken from the point of view of the present, and because it is concerned with layers of sedimented meanings that “constituted” the objects of past societies and cultures, it cannot be carried out without postulating what Husserl calls a historical a priori (Origin of Geometry, pp. 377–78). Such a historical a priori might, for example, consist in the social structures and the measurement practices that formed the epistemological context in which axiomatic geometry emerged. This a priori is prior in a double, if not an equivocal sense: It is a prior condition on the cognitive actions of the proto-geometers; however, it is also an a priori condition on our conceiving of the emergence of their proto-science. This term makes clear that Husserl’s history of the sciences cannot be understood as an empirical one. In writing a transcendental history of scientific concepts, we are aiming to discern paradigmatic formative events. Like the rational reconstructions of Lakatos, who was also inspired in this regard by Hegel, Husserl’s reconstruction of the history of geometry doesn’t try to get the facts straight. On the contrary, he would claim that there cannot really be any question of getting the historical facts straight until an analysis of his sort has been performed: If we are trying to imagine what went on in Greek society at the time geometry was created, we must obviously ascribe to the Greeks some sort of framework of meaning that we too can understand. There cannot be a question of reconstructing past intentions without assuming a conceptual framework common both to our forebears and to us. History, as we look at it, can be comprehensible only if we assume the existence of such schemes. At the same time, we regard our present scheme as the result of a historical process. These strata of meanings and traditions, which culminate in our own, form what Husserl calls the “internal structure” of history. A phenomenological investigation of the history of scientific concepts thus involves a curious fusion of empirical and a priori methods. It draws on historical data and a priori insights concerning human cognition in order to reconstruct the succession of conceptual schemes that produced the present one.12
So in Husserl’s transcendental analysis, there is not only a search for the historical bases of current scientific concepts; there is also a search for their absolute normative ground. This ground is to be found in the primitive intentional acts of our imagined forebears. Thus Cavaillès’s objection that, according to Husserl’s theory of science, “if we are conscious of progress, still there is no progress of consciousness.” We could not imagine the epistemological context of the emergence of geometry if we could not imagine what Husserl calls the “life-world” of other human beings. This life-world includes what some analytic philosophers today would call the world of subjective facts, the everyday world of colors and shapes, as well as the world of naive physics and psychology, for instance other people with their wishes and wants. Both our supposed Greek ancestors and we must inhabit such a world, whatever the current state of scientific development, for only on this assumption is the conceptual scheme of other human beings living in other cultures in any sense imaginable. It is, one might say, the base level historical a priori that will be brought to light by the archaeology of scientific concepts. But in uncovering conceptual strata and tracing the norms of science back to their phenomenal origins, we are engaged in a conscious regress. And to Cavaillès, this whole analysis is falsified by the true nature of scientific progress. Science does not consist of nested axiomatic systems; indeed, this is not even true of the history of mathematics, as Cavaillès argues in his work on logic and set theory. Earlier systems of science are not preserved as science develops; rather science develops by revising and erasing these systems, which may indeed be completely obliterated in the process. Epistemologically speaking, the conscious acts of our predecessors are irrelevant to us once we have rewritten their earlier system. It follows that the idea of a trans-temporal phenomenal ground of scientific knowledge is otiose, and with it the idea of a transcendental consciousness (or, more modestly, an internal structure of the history of reason) spanning the common human life-world.
Before turning to the second of the two quotations under consideration, let me summarize what we have learned so far. Transcendental philosophies of science seek to ground norms in the origins of conscious experience. They explain their validity by arguing that the latter are implicit in the structure of experience, whether this experience is my own or that of my predecessors. However, this philosophy is backward looking, and it cannot really account for the historical facts: Norms change, and ontologies are rejected and eventually forgotten. In the second passage cited above, Cavaillès suggests his alternative to this regressive philosophy. Instead of deriving it from the philosophy of consciousness, Cavaillès argues, we should conceive the doctrine of the sciences as being one of “concepts and dialectic.” This remark invokes a specifically philosophical objection centering on Kant’s transcendental deduction, and which Cavaillès develops in the opening pages of his book. This critique is, once again, extended to include Husserl’s philosophy, but it is also intended to apply to the philosophy of science of the Vienna Circle. I will concentrate in the following pages on Kant and Husserl, because it is the distinction between “conceptual” and phenomenological theories of historical epistemology that is important to understanding Foucault’s and Canguilhem’s application of Cavaillès’s ideas. I will then conclude with a few critical remarks concerning this application.
Cavaillès’s main objection to Kant’s transcendental arguments concerns the latter’s attempt to justify the necessity of logic by invoking the unifying function of the categories for experience. This is why he insists on the fragmentation of consciousness, that consciousness is “lost in the immediacy of the idea” or that “the term consciousness does not have a univocal application.” For, on Cavaillès’s view, Kant cannot reconcile the generality of logic (Cavaillès sometimes speaks of its “absoluteness”) with its transcendental origins. That is to say, Kant proves the validity of logic by invoking its constitutive role in experience; however, this form of deduction precludes the resultant logic’s having general application. To see why this is so, one needs to recall why Kant’s project requires a transcendental deduction of the categories, which are of course the basis of general (Aristotelian) logic.
If he is to justify the validity of this logic for all possible experiences, Kant cannot merely appeal to the fact that we do think by means of the categories. Even if we assume that they result from pure logical functions of the understanding, this does not explain why they are objectively valid, in other words why the understanding is always right to employ them.
Kant’s solution is that these functions of the understanding are conditions for the unity of apperception. There would be no consciousness if these functions were not actively constituting our unified experience by connecting the data of intuition. In the language of the first Critique, to say that consciousness obtains is to say that there is a unification of a manifold in a series of cognitive acts. Kant’s strategy is therefore to isolate the cognitive acts responsible for the unification of experience and then to extract logic by means of a dual process of abstraction.
Cavaillès, in his opening discussion of the role and the source of rules in science, characterizes Kant’s deduction of these rules as follows: First, the empirical is reduced to the pure content of a mathematical form wholly determined by intuition. Second, the content of thought in general is sloughed off, leaving us with the pure functions of the understanding. Cavaillès objects to both these steps on the grounds that (1) we must be able to make sense of a content independent of its form in order to imagine its being eliminated, but then the form is not a condition on our conceiving the content; or (2) form and content are indeed indissolubly wed, but then the necessity inheres in this single composite, and we are not isolating the necessary condition by eliminating the part we call the matter or content. Thus Kant is wrong to think that he has deduced the validity, and thus the normative character of general logic, by means of his appeal to consciousness. The logic that governs judgments, if it can be shown to be valid, must involve an ontology, as does the logic of Aristotle’s Analytics. And if it does not, then it is entirely empty. “In a philosophy of consciousness,” Cavaillès concludes, “either logic is transcendental, or it does not exist.”13
This attempted refutation of Kant, although highly compressed, does indeed strike at a notoriously weak link in the argument of the Critique. Cavaillès’s point, in a nutshell, is that if Kant is consistent in maintaining that every cognition (Erkenntnis) contains a unifying concept and unified intuitions, then he cannot speak of a prior, non-unified multiplicity that is subsequently unified by the pure functions of the understanding. Conversely, and as Kant himself acknowledged, these pure functions and their associated categories are empty without the contents they unify. Kant’s claim is that the categories, once they have been shown to be implicated at the origin of each cognition, have in consequence also been shown to be valid for all possible subsequent cognitions. This would indeed parallel Aristotle’s (or at the very least the scholastic) justification of the logical role of the categories by means of their metaphysical one; in other words that because they are fundamental modes of being, they are also fundamental modes of propositions. But of course Kant doesn’t want to do that: His categories are not modes of being, but modes of cognition. What they govern are all and only those experiences that can be “anticipated,” to use Kant’s language, namely the extensive and intensive magnitudes of which the multiplicity of phenomenal experience is composed.14
The problem, Cavaillès argues, is that this transcendental justification of logic has been falsified by our repeated encounters with unanticipated scientific objects that have contravened our logical principles. This flaw in Kant’s system has been most recently revealed, in Cavaillès’s opinion, in Hilbert’s and Gödel’s work on the foundations of mathematics. Hilbert’s project of axiomatizing the various branches of mathematics was intended to provide these with logically secure foundations, although Hilbert did not seek to reduce mathematics to logic. He took the axioms to express synthetic, as opposed to analytic truths, and thus far he sided with Kant (and Poincaré) against the logicists.15 But Hilbert, because he wished to establish the completeness and consistency of his systems, thereby put in play the logical rules involved. A system is complete when all truths are deducible within in it. Only in this case are the synthetic truths of mathematics strictly regimented by logical principles. Only then does the business of axiomatizing isolate the synthetic content of the science in question in the axioms, as Hilbert wanted. From this point of view, Gödel’s result touches the heart of Kant’s system, for it shows that there is no “general logic” adequate to mathematical knowledge, that is to the sciences of pure intuition. How much less should we expect there to be one for those concerning empirical intuitions? For Cavaillès, this result is the gravest symptom of a general ailment, namely that the Kantian philosophy of science “completely ignore[s] the contribution of the object to the structure of theory” (Cavaillès, 1960, 14). It does so because it demands that the essential structure of theory be determined by rational principles. These rational principles are in turn principles of the unity of consciousness, and thus the unity of science is nothing more than a projection of this unity onto the natural world. “There is no science in the sense of an autonomous reality that can be characterised as such, but rather rational unification following a fixed type of a diversity organized by the understanding” (14). There is no room for discovery and temporal contingency in this picture of science, just as there is no room allowed by the transcendental deduction for data beyond those that can be constructed or anticipated in pure forms of intuition. Both in the specific mathematical cases that Cavaillès dealt with in his earlier technical work, and in the general case of transcendental theories of science, the foundational character of logic can be preserved only by cutting its connection to the real complexity of scientific objects. Conversely, if logic is to retain its normative role in the sciences, we must concede that the latter cannot be deduced transcendentally.
Husserl, as we have seen, is taken to task on similar grounds: His search for normative origins is regressive, and it cannot make sense of an intervention in the world of science from outside of consciousness (indeed, the methodology of the epoché deliberately shuts out this possibility). Nevertheless, Cavaillès concedes that Husserl’s work is superior to that of both Kant and his logicist descendants to the degree that it seeks the sources of logical and mathematical norms in concrete intentional acts. Indeed, Husserl, in the sections of the Crisis that were supposed to form its main body, and which were published only in the 1960s, characterizes his aims in a manner completely consistent with Cavaillès’s take on the book, in other words with reference to the Transcendental Deduction. In these passages of Part III of the Crisis, Husserl argues that his life-world was implicitly assumed by Kant in the deduction, but that Kant failed to pursue the matter properly. Kant was right, claims Husserl, to begin the “descending” analysis of the A-Deduction in the world of everyday unified experience, but he passed too quickly over this “realm ... of unexamined evidences of being” that are “constant presuppositions of scientific and philosophical thought” (Crisis, § 28). Kant acknowledged the existence of the life-world but failed to see that the deduction should be carried out on this level. In consequence, the logic and mathematics he extracted he had implanted artificially, that is to say Aristotelian logic plus the building blocks of Newtonian physics. Whereas, Husserl believes, the phenomenological method will yield the proper logic precisely because it is grounded in actual, as opposed to idealized, phenomenal experience.
You can imagine how Cavaillès responds to this: Husserl’s method doesn’t really get around the problems with Kant’s. This “pseudo-temporal” origin of concepts is not only a fantasy, but even if it weren’t, it wouldn’t get around the dilemma that faced Kant already: Even if the original constitutive acts of consciousness force a given logic on us, why should, indeed how could, that logic be shown to be binding on experience at a later stage of science? Expanding on his critique of Kant, Cavaillès comments on Husserl that “ if transcendental logic really does found logic, then there is no absolute logic .... If there is an absolute logic, it can only derive its authority from itself, and it is not transcendental” (1960, 66). Applied to the notion of a historical epistemology, this critique yields the result that such an enterprise is “an abdication of thought” because it looks backward to the necessities supposedly implied in foundational acts, instead of considering the open-ended future in which new objects and systems of concepts are developed. The reason we need a logic of the concept, or a dialectic, according to Cavaillès, is that logic is in fact always concerned with a range of possible experiences about which we know as yet nothing. Moreover, the logic we have in hand may always reveal itself inadequate to a domain we know well. This may appear to undermine the normative status of logic. But in fact it does so only if we are in the grip of the philosophy of consciousness. That is to say, only if we think that the validity of logic must be determined once and for all by the structure of cognition.
Despite the merits of Cavaillès’s critique, it is difficult to discern the outlines of the philosophy of the concept that he, along with Bachelard and Canguilhem, intends to offer as an alternative. Within the philosophy of mathematics, Cavaillès favors Tarski’s semantic approach over those of Carnap and Frege, arguing that Tarski did not fall prey to the formalist’s desire to develop mathematics out of a single stock of privileged syntactic rules, but tried instead to discern the logic implicit in extant theories that are “rooted” in actual mathematical objects and practices. 16 In general, Cavaillès and his associates favor a philosophy of science that accords equal weight to the contributions of the objects of theory (which are conceived as independent of both the theories and the theoreticians concerned with them) to those of the theories themselves, and to those of the “critical rectifications” of working scientists. Their approach shares, in other words, many of the virtues and vices of recent analytic work in philosophy of science. On the positive side, the philosophers of the concept are pluralists and, to some extent, realists; they seek a theory of science that is true to the actual practice of science; and they retain an interest in normative questions. However, the theory that results is something of a grab bag, and it is better defined with reference to the formalist and transcendental theories that it rejects than it is through any individual philosophical or epistemological theses.
This is certainly true of Foucault’s use of the term philosophy of the concept, as I argued at the opening of this paper. He uses this term in order to distinguish his own archaeological investigations from their most obvious forebear, namely the phenomenological archaeology called for by Husserl. Foucault’s archaeology, as I will argue in conclusion, aims at uncovering the historical a prioris forming the deep structure of the sciences of various historical periods; however, these structures are more heterogeneous than those imagined by Husserl. But just like Cavaillès, Foucault does not want to throw the baby out with the bath water; he does not want his historical analyses to collapse into mere social history. This normative aspect of the French theorists I discuss in this chapter clearly distinguishes their work from recent English-language history and sociology of science, and it thereby reveals an essential difficulty inherent to any properly philosophical study of the history of the sciences.
Foucault’s archaeological project up until his Archaeology of Knowledge can be characterized as a subversion of Husserl in the light of Cavaillès’s critique. By this I mean that the overall structure of Husserl’s scheme is preserved, while the key epistemological tenet is rejected. In an interview from this period, Foucault characterized his project negatively, as “trying to discover in the history of science something like the unconscious.” The working hypothesis, he continues, “is that the history of science, the history of knowledge, does not simply obey the general law of the progress of reason, it is not human consciousness, it is not human reason that is in some sense the owner of the laws of its history” (Foucault, 1994a, 665–66). What does this scientific unconsciousness consist in? Well, among other things, it is a subliminal matrix that determines the identification and ordering of scientific objects, as is explained in The Order of Things. More generally, it corresponds to what Foucault calls in The Archaeology of Knowledge an historical a priori. Husserl’s historical a priori was, as we saw, the internal structure of meaning that (1) conditioned the field of objects and methods that historical scientists, such as the proto-geometers, were concerned with; and that (2) enables us to understand their intentions and traditions. Foucault’s historical a priori differs from Husserl’s in that it does not describe the framework of past intentional acts. It is instead supposed to be the framework that made possible the formation of past énoncés, or statements, which are to be identified and analyzed without any reference to the conscious intentions of individual speakers.
Thus Foucault’s a priori resembles that of Kant and Husserl in being an ontological ordering that is, for this very reason, a source of normative principles. That is why he calls it the source of the “rationality” of historical periods. But it is so only in his qualified sense, for it is not “human reason” but some other ordering principle that is at work in the sciences of historical ages. And it is not human thought but human speech that is normed. The most obvious examples of such historical a prioris in Foucault’s work are the three epistemes that organize the scientific discourses of the Renaissance, the classical, and the modern ages in The Order of Things. These structures cut across disciplines, they are in large measure invisible to intentional historical actors, and yet they fundamentally determine the things that scientists in these periods do and say, the things they take to be significant, unworthy of notice, or indeed self-evident. In this regard, they resemble Wittgenstein’s “grammatical” rules or Kuhn’s related disciplinary matrices. In all three cases, one supposes that there is a historically contingent a priori at work, whose impact is not conscious to scientists or speakers all while it fundamentally determines their selections of problems and methods, as well as the sorts of ontologies they are willing to countenance. Archaeology, like phenomenology, seeks to understand the meaning of past science by revealing these historical a prioris.
Let me then conclude by elaborating on this critical distinction between Husserl’s and Foucault’s conceptions of the historical epistemology of the sciences. I will do so by citing a criticism that Canguilhem made of Kuhn in the Introduction to his Idéologie et rationalité dans l’histoire des sciences de la vie (1977) and which he took to show the superiority of the French “philosophy of the concept.” Canguilhem concedes that Kuhn’s notions of paradigms and of normal science are useful to the extent that they highlight the historical development of scientific reason. But, Canguilhem objects, they are too Kantian: By invoking the ideas of intentions and of regulating acts, Kuhn’s theory allows these two to become unstuck. More seriously, since Kuhn “accords them an empirical mode of existence, as facts of culture” (Canguilhem, 1977, 23), what seemed to be a critical philosophy turns out to be social psychology. Canguilhem then goes on to contrast Kuhn to Bachelard and Cavaillès, for whom the rationality of mathematical physics inheres in the normative practices of mathematicians. The development of scientific norms must, on the view of Bachelard, Canguilhem, and Cavaillès, be sought within the actual practice of science, for “a science is a discourse that is normed by its critical rectification” (1977, 21). Canguilhem objects, in other words, to the same tactic that Cavaillès criticized in Kant, in which one considers norms (the pure functions of the understanding) in distinction from the thoughts or intuitions they necessarily govern. In so doing, one prepares the ground for the claim that, although empirical intuitions are not merely products of the understanding, they are nonetheless bound by its laws. But in Kant the necessity in question is a properly transcendental one. In Kuhn’s work, Canguilhem objects: “The paradigm is a user’s choice. The normal is what is shared ... by a collective of specialists in an institution” (1977, 23). The norms in question are conceived naturalistically, and this means that there is no distinction to be made, from the point of view of the historical epistemologist himself, between rational and irrational science, except perhaps a pragmatic one.
On Canguilhem’s hostile reading, Kuhn follows one of the two dead ends that present themselves once norms and contents come unstuck: His normal science is a social fact. The other path is the one followed by Husserl: Make the norms a product of a self-validating consciousness that interacts with scientific objects only after they have been processed by that very consciousness. This leads to a static and internalist view of scientific development: The autonomous contribution of objects and experiments collapses, while the norms become irrevocably fixed. Husserl’s historical a priori, because it is ultimately a product of consciousness, will not permit real critical reflection, for such reflection is defined as a regression to life-world origins. And it rules out the possibility that changes in norms be provoked by unexpected inputs on the side of the objects themselves. This is why Foucault, while following Husserl in identifying aprioristic elements in past science, nonetheless feels compelled to situate them on a plane that will prevent such a collapse. It is on this plane, he claims, “that a culture, stripping itself of the empirical orders prescribed to it by its primary codes, stops being guided by them passively ... and liberates itself enough to determine that they are perhaps not the only ones possible, nor even the best ones” (Foucault, 1966, 12). This plane of ordering is neither that of Kuhn’s social conventions, nor is it that of a Husserlian historical a priori, which is a structure of consciousness.
Of course, one cannot help but suspect that Foucault’s solution to Canguilhem’s and Cavaillès’s dilemma comes at a high price, for we would certainly like to know where this plane of unconscious orders should be taken to lie, in other words what the substrate of the scientific unconscious is supposed to be. Furthermore, if this a priori is neither a structure of consciousness nor a sociobiological entity, how does it become actual? That is to say, in what sense does it cause or constrain the things that scientists say? If this a priori acts causally, then it provides an explanation of certain regularities that we observe in past scientific cultures. But then it would be what Canguilhem dismissed as mere social psychology. On the other hand, the demand that it be unconscious means that the historical actors are themselves unaware of it. If it is a system of rules governing their speech and conduct, then these rules are not something that they deliberately instituted and willfully follow. They are at best, like the basic rules governing Wittgensteinian forms of life, rules they follow blindly. And the most a historical epistemologist can say is that the authors of this period act as if they were following rules of this sort. No one can deny him his right to interpret historical events on this pattern. But if we are to speak here of a historical a priori, then we can mean by this at most the a priori of the historian. An essential feature of this notion in Husserl—that the historical a priori was binding on the historical actors—has gone missing. With it has gone the possibility of any properly normative stance toward these past scientific cultures. And that is just the criticism that Canguilhem levels at Kuhn.
My conclusion is that Foucault’s theory of science, at least up until the point that he wrote Archeology of Knowledge, was largely defined through his rejection of Husserl’s model, a rejection that he and others found in their reading of Cavaillès. And though that rejection is well grounded in the critique of transcendental logic and of the philosophy of consciousness that we find there, it retains too much of Husserl’s historical epistemology for it to be coherent. For the notion of an unconscious a priori deliberately preserves features of the transcendental philosophies it was supposed to help supplant. These features—the notion of an a priori, and of the simultaneous constitution of logic, rationality, and a field of objects—are located, as Foucault explicitly maintains, neither on the level of facts, nor on that of consciousness. If they are conditions of knowledge, the knowledge they condition is most likely our own. That this terminology is absent from Foucault’s later work is no doubt a reflection of his own misgivings in this regard.17