1. The Charge of ‘Regularities Which Regulate Themselves’

In order to make intelligible the specific type of rules that archaeology of knowledge studies, Dreyfus and Rabinow consider and reject several approaches. Since the rules are historically changing, Dreyfus and Rabinow reject a view of them as social laws based on physical laws that operate in the brain. They reject the model of self-conscious rule-following, in turn, because the rules must be unrecognized as such by the subjects whose ways of thinking they shape. It seems more promising, at first, to understand the rules as descriptive regularities of a discursive practice because one could then meet the requirement that the rules be unconscious. However, the conception of rules as descriptive regularities cannot be reconciled with Foucault's other central commitment, namely, that the rules archaeology studies were actually operative with specific effects in particular historical circumstances. As Dreyfus and Rabinow see it, Foucault must choose between attributing historical efficacy to the rules, on the one hand, and holding that the rules are not recognized as such by the thinking subjects, on the other. Since Foucault rejects neither of these two commitments, Dreyfus and Rabinow conclude that he commits himself to an incoherent view that conflates the descriptive and normative registers by attributing causal efficacy to the very descriptions of regularities the archaeologist arrives at through a retrospective analysis of discursive practices.

Dreyfus and Rabinow draw this conclusion at the end of a discussion of the explanatory power of Foucault's archaeological analyses. As they rightly emphasize, archaeology of knowledge is not merely a descriptive enterprise in the history of thought. Foucault's key idea, which is expressed through his use of the notion of a historical a priori, is to identify historically particular systems of thought, ‘discursive formations’, on the basis of the rules of the discursive practices archaeology uncovers (Foucault Reference Foucault1969a: 45–54). By thus circumscribing different systems of thought, each governed by a distinct set of rules that were unknown to the thinking subjects, Foucault seeks to account for systematic patterns in the history of thought without reliance on individual psychology or some metahistorical notion of rationality (Foucault Reference Foucault1968: 693–94; Reference Foucault1969a: 82–83; Reference Foucault1970). Obviously, this explanatory connection requires that the rules of discursive practices archaeology studies were in fact historically efficacious.

By insisting that Foucault must choose between the requirements of implicitness and efficaciousness, Dreyfus and Rabinow suggest that archaeology's explanatory ambitions rest on an incoherent conception of the rules of discursive practices it purports to study. Their diagnosis is that ‘in his account of the causal powers of discursive formations, Foucault illegitimately hypothesized the observed formal regularities which describe discursive formations into conditions of these formations’ existence’ (Dreyfus and Rabinow Reference Dreyfus and Rabinow1982: 83). In order to save archaeology, Dreyfus and Rabinow recommend that Foucault relinquish the idea that his archaeological analyses possess any explanatory power. In their view, Foucault's ‘unclearness concerning the question of causal efficacy surely shows that the archaeologist should never have raised this problem in the first place’ (ibid.). In fact, however, a choice between implicitness and efficaciousness is mandatory only if the alternative that combines them has been excluded. And that independent alternative is provided, as we shall see, by the pragmatist conception of rules.

2. Implicitness and Efficaciousness

For Foucault the task of finding a conception of rules that can combine the requirements of implicitness and efficaciousness is motivated by the contrast he draws between two kinds knowledge—connaissance and savoir. Foucault uses the word ‘connaissance’ to designate specific bodies of empirical knowledge he studies—psychiatry, clinical medicine, and criminology, among others—understood as sets of truth claims. The distinctive focus of Foucault's archaeological analyses, however, is marked by the word ‘savoir’, which he employs to designate a system of rules that escapes the consciousness of thinking subjects and nonetheless defines a particular space of possibilities for them in a given historical context. In 1966, shortly after the publication of The Order of Things, Foucault sums up this distinctive conception of knowledge as follows:

Foucault treats such a system of implicit knowledge as a historically dynamic constellation of constitutive conditions for particular discursive possibilities. The unconscious rules that make up a system of savoir constitute and constrain a set of discursive possibilities that are actualized as truth claims in specific bodies of empirical knowledge, connaissances. For example, Foucault explains, in History of Madness ‘it was that knowledge [savoir] that I wanted to examine, as condition of possibility of knowledges [connaissances], institutions, and practices’ (ibid.) that identify mental illness as an object of theoretical investigation and practical intervention.

Foucault contrasts this epistemological view with the assumption of unrestrained epistemic sovereignty of the knowing subject that, according to him, dominates modern philosophy and is epitomized in phenomenology (Foucault Reference Foucault1966d; Reference Foucault1969b: 817; Reference Foucault1970: 881; Reference Foucault1971a: 1033). In contrast to that ‘humanist’ tradition, which takes the unconstrained freedom of a human subject as its metaphysical foundation and methodological starting point, Foucault assigns primacy, on both counts, to an unconscious ‘system’:

Though this subordination of the freedom of a thinking subject to an unconscious system of rules created an ‘antihumanist’ allegiance between Foucault's archaeology and its contemporaneous structuralist human sciences, Foucault's thoroughly historical outlook made his work always distinct from any strictly speaking structuralist methods and aspirations. In 1967, for example, Foucault underscores this decisive difference in approach:

Archaeology of knowledge studies the ‘secret function’ of statements [énoncés] to constitute a system of savoir that unbeknownst to the speaking subjects defines a particular set of discursive possibilities for them in a given discursive practice. Foucault's key idea is that the rules of savoir are created, sustained, and sometimes transformed through the very activity of making statements without representing the rules as such (Foucault Reference Foucault1969a: 136–37, 192). In short, the rules are simultaneously both implicit and efficacious. Focusing on the history of the human sciences, Foucault illustrates that general requirement by saying that the rules function as ‘a positive unconscious of knowledge [at] a level that escapes the scientist's consciousness and nevertheless partakes of the scientific discourse instead of contesting its validity and seeking to decrease its scientific nature’ (Foucault Reference Foucault1970: 877, emphasis in the original).

Such a system of constitutive rules is not merely a theoretical construction for Foucault, but it is crucial for his archaeological project that ‘the development of this knowledge [savoir] and its transformations. . . put in play complex relations of causality’ (Foucault Reference Foucault1969c: 872) in the history of thought. In other words, it is crucial that a system of savoir functions as a historical a priori. ‘It is this a priori,’ Foucault explains in The Order of Things, ‘that, in a given historical period, carves out in experience a field of possible knowledge, defines the mode of being of the objects that appear in it, endows everyday perception with theoretical powers, and defines the conditions under which a discourse that is recognized as true can be held about things’ (Foucault Reference Foucault1966a: 171). In The Archaeology of Knowledge, then, Foucault notes explicitly that his use of the notion of a historical a priori serves to mark an explanatory connection between the rules of a discursive practice and a particular system of thought: ‘The reason for using this a little barbarous term [historical a priori] is that this a priori must account for statements in their dispersion’ (Foucault Reference Foucault1969a: 167, emphasis added). The explanatory connection is underwritten by the constitutive dependence between a particular set of discursive possibilities and the rules of a given discursive practice. For Foucault a historical a priori ‘is defined as the group of rules that characterize a discursive practice’, and he indicates that constitutive dependence by underscoring that ‘these rules are not imposed from outside onto the elements they connect [mettent en relation],’ but the rules ‘partake in the very thing they connect [sont engagées dans cela même qu'elles relient]’ (Foucault Reference Foucault1969a: 168). As we shall see, grasping how Foucault combines this Kantian view of objects of knowledge as conceptually constituted with a decidedly pragmatist conception of the constitutive rules that function as a system of savoir will be the key to a proper understanding of his archaeological project.

3. Foucault's Pragmatist Turn

To grasp the pragmatism of Foucault's considered view it is instructive to consider how his understanding of the rules of savoir changed in the course of writing The Archaeology of Knowledge, his most ambitious presentation of archaeology. In the book, published in 1969, Foucault presents his considered view that these rules are not articulated as statements of rules but instead enacted implicitly in a discursive practice. But in the unpublished manuscript of The Archaeology of Knowledge Foucault defines savoir as a historically changing group of statements that play a normative function in a discursive practice. In the manuscript, Foucault argues that ‘[t]he grid that constitutes, for a given period [. . . ] the system of savoir, may be called the grid of determining assertions. This set of assertions cannot be said to be true or false within the scientific discourse they make possible’ (Foucault Reference Foucault1967d: 558). In other words, in the manuscript Foucault understands savoir in terms of statements, indeed as a special set of assertions whose role is to define what types of statements can be formulated as intelligible empirical claims to be verified or falsified. Foucault articulates this view very clearly: ‘This network of assertions is what I call savoir’ (Foucault Reference Foucault1967d: 556). Thus both specific bodies of empirical knowledge, connaissances, and the system of rules that constitutes the particular discursive possibilities they realize, savoir, are understood as statements. ‘The savoir [is] the network of assertions that give rise to scientific statements in their possibility; it is the space of their emergence’ (Foucault Reference Foucault1967d: 563).

Ultimately, however, that conception of savoir as an explicitly articulated historical a priori defeats the purpose of an archaeology of knowledge to unearth a system of rules that goes unrecognized by those whose discursive possibilities it shapes. Statements of rules cannot make ‘a positive unconscious of knowledge’. It is therefore not surprising that in the published version of The Archaeology of Knowledge Foucault underscores that the rules of discursive practices archaeology studies are not to be understood as statements. Foucault states that a discursive practice is governed by ‘a group of anonymous, historical rules’ (Foucault Reference Foucault1969a: 153–4), and he is very clear about rejecting the view he had endorsed in the manuscript: ‘These rules are never given in a formulation, they traverse formulations and constitute for them a space of coexistence; therefore one cannot find a single statement that would articulate them as such’ (Foucault Reference Foucault1969a: 192). Given the very idea of savoir as a positive unconscious of knowledge, this considered view is indeed what Foucault ought to maintain.

This shift from an explicit to an implicit conception of the rules of savoir signals a pragmatist turn in Foucault's philosophy. The central role of the concept of practice in Foucault's thought was noted early on by some of his most astute interpreters (Veyne [1978] Reference Veyne and Davidson1997), but only recently has the topic received the wider attention it deserves (Koopman Reference Koopman2011). Foucault himself came to express this pragmatist commitment as ‘a third principle of method: address “practices” as the field of analysis, conduct the study by privileging what “was done”’ (Foucault Reference Foucault1984d: 1453–54). What is done in a discursive practice—the activity of making statements, in contrast to their representational content—assumes a fundamental role in Foucault's understanding of thought when he begins to conceptualize savoir as a group of rules that are implicit in a discursive practice. Implicit rules cannot be attributed to subjects as propositional knowledge—that is, as knowledge that something is (or ought to be) thus and so. Instead, conformity to the implicit rules must be understood as a practical ability of the participants of a discursive practice, namely, as knowledge-how that guides their use of concepts without being represented as a set of rules. Though Foucault does not use the distinction ‘knowing that/knowing how’ (Ryle Reference Ryle1949), his own contrast between connaissance and savoir needs to be drawn in terms of propositionally articulated knowledge on the one hand and practical abilities on the other. Only thus can savoir consist of rules that are both implicit and efficacious, in contrast to bodies of knowledge, connaissances, that consist of truth claims.

Correspondingly, two different conceptions of thought take shape, depending on whether one gives primacy to knowing that or knowing how as the fundamental, self-standing type of knowledge. Let me designate these two alternatives as ‘intellectualist’ and ‘pragmatist’ conceptions of thought, respectively. According to the pragmatist conception, the propositional content of what is represented in thought depends on the activity of reasoning, and in general, as Robert Brandom puts it, ‘believing that things are thus-and-so is to be understood in terms of practical abilities to do something’ (Brandom Reference Brandom2011: 9, emphasis in the original). This means that subjects grasp the fundamental normative standards that govern the use of concepts not in a propositionally articulated form but as an ability to participate in a discursive practice. Thus understood, the very intelligibility of propositionally articulated thought rests on behavioral dispositions whose norms the subject does not know in a propositional form. In contrast, the intellectualist conception of thought takes propositionally articulated knowledge as the primitive, self-standing type of knowledge and views the activity of doing something as derivative. Therefore, according to the intellectualist view, propositional contents form a system of truth-apt representations whose objective purport is intelligible to a subject independently of the subject's mastery of any practical abilities. In other words, the intellectualist and pragmatist conceptions of thought disagree about the type of knowledge in virtue of which a subject is able to use concepts. The intellectualist strategy is committed to accounting for that ability in terms of propositionally articulated representational contents it attributes to a subject, whereas the pragmatist approach insists that understanding the objective purport of such representations presupposes knowledge-how. To be sure, Foucault never presented an adequate philosophical argument for this pragmatist view. Nevertheless, such an argument exists, and I want to show that it can be rehearsed, independently of Foucault's specific concerns and commitments, to justify the pragmatist conception of rules.

4. The Regress of Rules

Kant did not only formulate the view of reasoning as acting on the basis of representations of rules (Kant [1785] Reference Kant and Gregor1996: 4:412, 427), but he was also the first to register the crucial limitation that view encounters due to a regress of rules. In the Critique of Pure Reason, Kant notes the threat of an interminable regress and explains what it entails as follows:

The conclusion of the regress argument—reiterated by Wittgenstein and elaborated by Sellars—is that reasoning is fundamentally governed by normative standards that are not grasped in a representational form but instead are mastered as a practical ability (Wittgenstein Reference Wittgenstein1953: §§ 198–202; Sellars Reference Sellars1954). Whereas Kant views this fundamental ability as a natural talent [Naturgabe], which readily varies from one individual's psychology to another (Kant [1793] Reference Kant and Gregor1996: 8:275–76), Wittgenstein and Sellars, for example, understand the ability as a product of linguistic training in a social practice and thus locate the normative bedrock of reasoning in a shared culture of ‘customs (uses, institutions)’ (Wittgenstein Reference Wittgenstein1953: § 199, emphasis in the original). In any event, the regress of rules undermines a general conception of rules as representations and, specifically, its intellectualist version according to which all rules are represented as statements of rules. The upshot is that following representations of rules requires an ability to conform reliably to standards of correctness that are implicit, that is, not represented. Some have hypothesized that this implicit normative bedrock is a biological feature of the human brain (Fodor Reference Fodor1975: 55—78). But, assuming that concepts are acquired of a piece with language acquisition, as I do, the regress in fact shows that the bedrock of reasoning consists of normative standards that are implicit in a discursive practice.Footnote ²

Thus the regress of rules gives rise to a metanormative question, which Brandom formulates as follows: ‘how to understand proprieties of practice, without appealing to rules, interpretations, justifications, or other explicit claims that something is appropriate’ (Brandom Reference Brandom1994: 25)? The regress undermines the view that all normative standards exist as statements of rules, which Brandom calls ‘regulism’. Since the regress arises from within the perspective of a rule-following subject, it is tempting to try to avoid it by opting for an alternative approach that is independent of the perspective of any agent. Following this strategy, the claims one makes about the rules of a discursive practice would be understood as descriptions of regularities one identifies by observing a practice, and the claims would specifically involve no reference to the particular perspective of any of the participants in the practice. No regress of rules would ensue because such descriptions say nothing about subjects as following rules intentionally. In keeping with Brandom's terminology, let me call this alternative conception of rules ‘regularism’. The problem with the regularist strategy, however, is that by replacing an account of proprieties with descriptions of regularities it loses the very idea that there are normative forces operative in a discursive practice. The strategy therefore has no resources to make sense of the fundamental fact of our discursive lives that we are susceptible to error when applying concepts—that is, of the fact that concepts have criteria of application.

The metanormative challenge, then, is ‘to make sense of a notion of norms implicit in practice that will not lose either the notion of implicitness, as regulism does, or the notion of norms, as simple regularism does’ (Brandom Reference Brandom1994: 29). Now, this twofold general constraint for understanding the normative structure of a discursive practice corresponds to the two criteria of adequacy I have identified for Foucault's conception of the rules of discursive practices on the basis of the specific concerns of his archaeological project. Therefore, it is all the more striking to see Dreyfus and Rabinow insist that Foucault must choose between regulism and regularism:

That is indeed what Dreyfus and Rabinow suppose when they proceed to conclude that Foucault is committed to a conceptual confusion of ‘regularities which regulate themselves’, a confusion that conflates the descriptive and normative registers by attributing normative force and causal efficacy to regularist descriptions that map regularities of discourse. But why do Dreyfus and Rabinow overlook the pragmatist conception of rules that not only provides an independent alternative and thus enables one to avoid a mandatory choice between regulism and regularism, but also, as I have shown, constitutes a conceptual centerpiece in Foucault's archaeological project? This appears all the more perplexing given that Dreyfus himself is a long-standing champion of the pragmatist conception of thought, which he traces back to Heidegger's view in Being in Time that Zuhandenheit (and Umgang) have an ontological priority over Vorhandenheit (and Erkenntnis) (Dreyfus Reference Dreyfus1972: 173; Reference Dreyfus1991: 60–87; Heidegger [1927] Reference Heidegger1962: §§ 15–16). I believe that the most charitable explanation for this omission is that Dreyfus and Rabinow fail to recognize the pragmatist approach in Foucault's conception of rules, because they interpret archaeology, mistakenly, as we shall see, as a sort of structuralism. As they see it, Foucault himself rejects the pragmatist alternative by making a ‘structuralist move’. Therefore, rather than calling into question the pragmatist conception of rules as such, Dreyfus and Rabinow are in fact only arguing against Foucault's entitlement to it jointly with the ‘structuralist move’ they attribute to archaeology. Curiously enough, Dreyfus and Rabinow thus end up criticizing Foucault for abandoning the pragmatist approach that he, in fact, endorses. However, as I hope to show next, Foucault's entitlement to the pragmatist conception of rules is not threatened by this line of criticism, which stems from a failure to grasp the specificity of his archaeological project.

5. The Charge of a ‘Structuralist Move’

According to Dreyfus and Rabinow, Foucault's ‘structuralist move’ makes archaeology of knowledge diametrically opposed to the pragmatist approach, which they favorably attribute to the early Heidegger, the later Wittgenstein, and others. Dreyfus and Rabinow believe that Foucault, as well as Heidegger and Wittgenstein, are all ‘interested in the practical background that makes objectivity possible’ (Dreyfus and Rabinow Reference Dreyfus and Rabinow1982: 62). The crucial issue, according to Dreyfus and Rabinow, is whether these background practices themselves are discursive or not: Hermeneutic thinkers such as Heidegger and Kuhn would agree with Foucault that subjects are surely not the source of discourse. All would agree that the source is ‘an anonymous field of practices’. But those doing hermeneutics would insist that this field is not purely discursive. . . . Changing nondiscursive skills sustain the changing styles of statements, the modalities of enunciation, and the kinds of subjects which are possible. (Dreyfus and Rabinow Reference Dreyfus and Rabinow1982: 69, emphasis added). In contrast, Dreyfus and Rabinow argue, ‘Foucault. . . makes a structuralist move which sharply distinguishes his account of the background practices from that of Wittgenstein and Heidegger. Although he is clearly aware that nondiscursive practices play a role in “forming” objects he insists that the crucial role is played by what he calls discursive relations’ (Dreyfus and Rabinow Reference Dreyfus and Rabinow1982: 62, emphasis in the original).

As Dreyfus and Rabinow put it, correctly, Foucault ‘claims that discursive relations have a certain effect on all other relations’ (Dreyfus and Rabinow Reference Dreyfus and Rabinow1982: 64). Thus, for them, Foucault's ‘structuralist move’ is ‘the extreme and interesting (if ultimately implausible) claim that discourse unifies the whole system of practices, and that it is only in terms of this discursive unity that the various social, political, economic, technological, and pedagogical factors come together and function in a coherent way’ (Dreyfus and Rabinow Reference Dreyfus and Rabinow1982: 65). To Dreyfus and Rabinow this primacy of ‘discursive relations’ means that ‘Foucault is not satisfied to accept social practices as a level of explanation’ (Dreyfus and Rabinow Reference Dreyfus and Rabinow1982: 82). As a result, they conclude that Foucault embraces a diametrically opposite view of the background practices than the view of ‘the existential-pragmatic philosophers,’ exemplified by Heidegger and Wittgenstein:

This line of criticism, I hope to show now, is motivated by philosophical concerns that are alien to Foucault's archaeological project and irrelevant to a judicious assessment of its merits and shortcomings. Specifically, Dreyfus and Rabinow understand the distinction between discursive and nondiscursive practices differently than Foucault does, and this divergence reflects a decisive discrepancy between their respective philosophical outlooks. To show that, let me first explain why Foucault's notion of discursive relations needs to be understood from within his generally Kantian epistemological outlook and then why it is plausible, within Foucault's philosophical outlook, to maintain that all nondiscursive practices depend on discursive practices.

6. Foucault's Kantian Pragmatism

It is important to realize that Foucault's discussion of discursive relations belongs to a section of The Archaeology of Knowledge entitled ‘The Formation of Objects’ where he consistently distinguishes between objects of discourse [objets] and prediscursive things [choses], explicitly excluding things thus understood from the scope of archaeology of knowledge (Foucault Reference Foucault1969a: 64–67). Foucault's generally Kantian epistemological outlook is clearly pronounced: ‘in short, one wants to get rid of ‘things’ altogether, to de-present them. . . . To substitute for the enigmatic treasure of ‘things’ that precede discourse, the rule-governed formation of objects that take shape only in it. To define these objects without referring to the ground of things, but by relating them to the group of rules that allow them to be formed as objects of a discourse and thus constitute their conditions of historical emergence’ (Foucault Reference Foucault1969a: 64–65, emphasis in the original). In contrast to things, thus understood, Foucault underscores that ‘the object does not wait in a limbo for the order that will set it free and enable it to be embodied in a visible and sayable [bavarde] objectivity; it does not pre-exist in itself, kept by some obstacle at the edges of light. It exists under the positive conditions of a complex bundle of relations’ (Foucault Reference Foucault1969a: 61).

It is these relations, which play a constitutive role with respect to objects of discourse, that Foucault calls ‘discursive relations’. In contrast to relations between things (not objects), on the one hand, and relations between linguistic-cum-semantic entities (not statements), on the other, Foucault underscores the specificity of the discursive relations by saying that they belong to discourse as a practice: ‘These relations characterize not the language [la langue] the discourse uses, not the circumstances where it unfolds, but the discourse itself as a practice’ (Foucault Reference Foucault1969a: 63). I believe that commentators have generally failed to grasp the full significance of this claim, because they have not realized that Foucault's notion of discursive relations is a pragmatic category, specifically that these relations are generated by the activity of making statements understood as ‘connecting [une mise en relations] that characterizes the discursive practice itself’ (ibid.). In contrast to relations between things, on the one hand, and relations between linguistic abstractions in a discourse already pronounced, on the other, discursive relations are ‘at the limit of discourse, as it were: they provide it with the objects it can talk about’ (ibid.). Foucault, who has already stated that ‘an object. . . does not pre-exist in itself’ (Foucault Reference Foucault1969a: 61), is quick to clarify that ‘rather (for this picture of offering presupposes that objects are formed on one side and discourse on another), they [the discursive relations] determine the bundle of relations that discourse must bring about [effectuer] to be able to talk about such and such objects’ (Foucault Reference Foucault1969a: 63). Thus understood, the discursive relations are constitutive of the objects of knowledge in a given discursive practice, and the configuration of these constitutive relations is governed by the given rules of savoir, understood as ‘a group of rules that are immanent in a practice and define it in its specificity’ (ibid., emphasis in the original).

This view of discursive relations is an expression of Foucault's decidedly pragmatist elaboration of the Kantian thought that objects of knowledge are actively constituted by conceptual determinations. According to Foucault, the constitutive relations themselves are created and organized by doings in a discursive practice, namely, by the activity of making statements. Given this generally Kantian epistemological outlook, it should come as no surprise that Foucault is not concerned with prediscursive things but objects of discourse. As Marc Djaballah emphasizes in his study of Kantian aspects in Foucault's thought, ‘the objects of discourses have the basic structure of sensible objects in Kant's theoretical philosophy. They are not less than the objects of which Kant deduces the conditions of possibility, but more’ (Djaballah Reference Djaballah2008: 239). Whereas Kant inquires into the necessary conditions for any object of empirical judgment, Foucault's focus lies in the additional sufficient conditions for particular types of objects to become thinkable. Instead of asking how the pure concepts of the faculty of understanding determine the transcendental object X (Kant [1781/87] Reference Kant1998: A109–10), Foucault studies the historical articulation of further conceptual determinations that specify particular types of objects for thought, and he takes these further determinations to be constituted through the activity of making statement in a discursive practice.

7. Discursive and Nondiscursive

But why does Foucault maintain that the organization of all social practices depends on a given configuration of discursive relations understood in this way? Here it is crucial to recognize two versions of the distinction between ‘discursive practices’ and ‘nondiscursive practices’—a broad and a narrow sense of that distinction. For Foucault, all constellations of social practices are discursive in the broad sense that they involve the use of concepts. A system of thought, as Foucault understands it, is a network [reseau] that correlates practices of making statements and practices of doing (other) things as two dimensions of a historically particular form of experience (Foucault Reference Foucault1969c: 874; Reference Foucault1971b: 1075; Reference Foucault1977: 299; Reference Foucault1980: 845–46; Reference Foucault1984c: 1397–1400). According to this view, ‘thought is understood as the very form of action,’ (Foucault Reference Foucault1984c: 1399), and therefore social reality is always discursive in the broad sense: ‘discourse must not be understood as the set of things that are said, nor as the manner of saying them. It is just as much in what is not said, or what is marked by gestures, attitudes, ways of being, patterns of behavior, spatial arrangements. Discourse is the set of constrained and constraining meanings that pass through social relations’ (Foucault Reference Foucault1976: 123). Thus, when Foucault draws a distinction between discursive practices and ‘nondiscursive practices’, it is a narrow distinction within this already essentially concept-involving outlook. In the narrow sense, then, discursive practices consist of the activity of making statements, whereas nondiscursive practices consist of other actions that nonetheless involve an application of concepts. In other words, this narrow distinction marks theoretical and practical uses of reason as two kinds of practice within a system of thought.

The narrow sense of the ‘discursive/nondiscursive’ distinction escapes Dreyfus and Rabinow because they believe, overlooking some decisive differences, that Foucault as well as the early Heidegger and the later Wittgenstein are all simply ‘interested in the practical background that makes objectivity possible’ (Dreyfus and Rabinow Reference Dreyfus and Rabinow1982: 62). Dreyfus and Rabinow believe that Foucault is specifically concerned with the preconditions for scientific knowledge about human beings. They urge that ‘like Kant who woke up from his dogmatic slumber and deduced the categories which were to put physics on a sure footing, Foucault wishes to wake us from our “anthropological sleep” in order to open our eyes to a successful study of human beings’ (Dreyfus and Rabinow Reference Dreyfus and Rabinow1982: 44). And they hold that ‘The Archaeology of Knowledge presents this new method in detail and sketches the theory of discourse on which it is based’ (ibid.). However, as Gary Gutting has already argued compellingly, The Archaeology of Knowledge is a methodological treatise for conducting analyses of a specific kind in the history of thought, and neither these analyses nor Foucault's presentation of their distinctive methodology involve a concern with the necessary preconditions of knowledge or of human sciences in particular (Gutting Reference Gutting1989: 261–72). On the contrary, Foucault emphasizes that his methodological choices are informed by philosophical commitments that are geared toward the goal of ‘making the history of thought overcome its transcendental subjection’ (Foucault Reference Foucault1969a: 264), namely, its conceptual and methodological dependence on the notion of transcendental conditions of experience (Foucault Reference Foucault1969a: 21–27, 264–65). And Foucault gladly admits that ‘for the time being, and without being able to see an end to it, my discourse, far from determining the place from which it speaks, evades the ground where it could find support. It is a discourse on discourses, but it does not intend to find in them a hidden law, a covered origin it would only need to set free; nor does it intend to establish on its own and starting from itself the general theory whose concrete examples they would be’ (Foucault Reference Foucault1969a: 267–68).

Here Dreyfus's own preoccupation with a transcendental inquiry into the preconditions of human experience occludes the specificity of the concerns that motivate Foucault's archaeological project. If one assumes that finding an ontological foundation for essentially concept-involving experience in some prediscursive activities is the philosophical problem that Foucault, among others, should be addressing, then archaeology of knowledge indeed seems to fail due to its lacking ontological foundation (Han Reference Han1998). Most recently, Dreyfus has defended these ontological concerns in his debate with John McDowell regarding the extent to which human experience is conceptually structured. Dreyfus is dissatisfied with ‘conceptualists’ like McDowell—and Foucault—who overlook the topic of a prediscursive foundation of experience because doing that, so Dreyfus argues, makes the conceptualist views unavoidably incomplete. In contrast, Dreyfus insists that an adequate account of human experience must be based on a prediscursive foundation of skillful coping:

However, neither Foucault nor McDowell is striving to formulate a philosophical theory of human experience, and without that ambition Dreyfus's point, however valid it may be, loses its force. As McDowell states laconically in his response to Dreyfus, ‘[n]o doubt we acquire embodied coping skills before we acquire concepts, in the demanding sense that connects with rationality’ (McDowell Reference McDowell2007: 345). Nevertheless, McDowell argues, the experience of concept-using subjects is thoroughly discursive because the embodied coping skills become animated by rationality once we become full-fledged concept-users: ‘I do not have to ignore embodied coping; I have to hold that, in mature human beings, embodied coping is permeated by mindedness’ (McDowell Reference McDowell2007: 339), namely, by the use of concepts in thought and action.

As I have already indicated, Foucault understands social reality similarly as a constellation of practices that essentially involve the use of concepts. In a Reference Foucault1981 interview, Foucault makes this point as follows: ‘One must overcome the sacralization of the social as the only authority on what is real [seule instance du réel] and stop considering as thin air this essential thing in human life and human relations, namely, thought. Thought, it exists, well beyond and below the systems and edifices of discourse. It is something that is often hidden, but always animates everyday behavior. There is always a little bit of thought even in the most foolish of institutions, there is always thought even in silent habits’ (Foucault Reference Foucault1981: 999). For Foucault, then, as he once put it succinctly, ‘there is thought everywhere’ (Foucault Reference Foucault1982: 1170). And, to borrow Foucault's own words, one might say of Dreyfus and Rabinow that their ‘mistake consists of forgetting that people think, and that their behaviors, their attitudes, and their practices are animated by thought [habités par une pensée]’ (Foucault Reference Foucault1984e: 1473).

8. Archaeology of Knowledge as a Diagnostic Project

Once Foucault's archaeological project has been severed from concerns of transcendental philosophy, the question remains as to how archaeology's goal and motivation should be understood instead. Before concluding, let me briefly address this important question. While it is well known that Foucault defined the historical present as the focus of his philosophical attention in a series of discussions, from 1978 to 1984, of Kant's essay ‘What Is Enlightenment?’ virtually no one has noted that Foucault explicitly endorses a diagnostic conception of philosophy already at the peak of his reflections on archaeology of knowledge in the second half of the 1960s. This earlier series of remarks reveals that for Foucault a diagnostic conception of philosophy was initially a bequest from Nietzsche and that archaeology of knowledge seeks to take up that diagnostic task. In 1966, when Foucault replies to a question about philosopher's role in contemporary society, he invokes Nietzsche's diagnostic conception of philosophy: ‘But, speaking of Nietzsche, we can return to your question [what is the role of a philosopher in society]: for him, a philosopher was a diagnostician of the state of thinking. Actually, one can envisage two kinds of philosophers, one who opens up new paths for thought, like Heidegger, and one who plays somewhat of the role of an archaeologist, who studies the space in which thought unfolds, as well as the conditions of this thinking, its mode of constitution’ (Foucault Reference Foucault1966e: 581). If Foucault adopts the role of an archaeologist, he does it for the sake of pursuing this diagnostic task, and as he explains in 1967, it is this diagnostic orientation that confers a philosophical character onto his otherwise merely historical investigations:

The two passages I have quoted belong to a longer series of generally neglected remarks in which Foucault repeatedly endorses a diagnostic conception of philosophy and presents archaeology of knowledge as a diagnostic project (Foucault Reference Foucault1966a: 10; Reference Foucault1967a: 609; Reference Foucault1967c: 641, 648; Reference Foucault1968: 693; Reference Foucault1973: 1302). While a detailed discussion of these remarks is a task for another occasion, it should be clear already that Foucault is not just making a rhetorical move when he retorts to the charges of an imaginary Sartrean opponent in the epilogue of The Archaeology of Knowledge that archaeology ‘constantly differentiates, it is diagnostic’ (Foucault Reference Foucault1969a: 268, emphasis in the original).

Furthermore, acknowledging archaeology's diagnostic character opens up a perspective for a more unified understanding of Foucault's philosophical work than has been available before. In particular, one needs to understand archaeology as a diagnostic project to make sense of Foucault's central but neglected remarks, in 1984, regarding the role that archaeology continues to play in his work. In the essay ‘What is Enlightenment?’ Foucault underscores his distinctive conception of critique, in contrast to the goals of Kant's critical philosophy, by saying that ‘critique is archaeological in its method’ (Foucault Reference Foucault1984b: 1393). Foucault's critique is ‘archaeological—and not transcendental’ (ibid.) precisely because it aims to diagnose the historical present, specifically the current form of thought, as opposed to investigating the necessary conditions of human experience. In another text dating from the same year, Foucault explains that the archaeological dimension of his work uncovers historically particular forms of thought, whereas its genealogical dimension reveals the contingency of these forms by tracing their historical formation through ‘practices and their modifications’ (Foucault Reference Foucault1984a: 17–18). It seems clear to me that, put together, all the remarks on archaeology as a diagnostic project, ranging from 1966 to 1984, constitute an essential strand of continuity throughout Foucault's philosophical career that no longer can be overlooked if one wants to understand, as I do, how his ideas on relations of power and practices of the self elaborate aspects of the distinctive epistemological view that archaeology of knowledge embodies instead of abandoning archaeology as an ill-conceived project.