Michel Foucault (1969)


The Archæology of Knowledge

Chapter 1
The Unities of Discourse

Source: The Archaeology of Knowledge (1969), publ. Routledge, 1972. The First 3 Chapters of main body of work are reproduced here.

The use of concepts of discontinuity, rupture, threshold, limit, series, and transformation present all historical analysis not only with questions of procedure, but with theoretical problems. It is these problems that will be studied here (the questions of procedure will be examined in later empirical studies - if the opportunity, the desire, and the courage to undertake them do not desert me). These theoretical problems too will be examined only in a particular field: in those disciplines - so unsure of their frontiers, and so vague in content - that we call the history of ideas, or of thought, or of science, or of knowledge.

But there is a negative work to be carried out first: we must rid ourselves of a whole mass of notions, each of which, in its own way, diversifies the theme of continuity. They may not have a very rigorous conceptual structure, but they have a very precise function. Take the notion of tradition: it is intended to give a special temporal status to a group of phenomena that are both successive and identical (or at least similar); it makes it possible to rethink the dispersion of history in the form of the same; it allows a reduction of the difference proper to every beginning, in order to pursue without discontinuity the endless search for the origin; tradition enables us to isolate the new against a background of permanence, and to transfer its merit to originality, to genius, to the decisions proper to individuals. Then there is the notion of influence, which provides a support - of too magical a kind to be very amenable to analysis - for the facts of transmission and communication; which refers to an apparently causal process (but with neither rigorous delimitation nor theoretical definition) the phenomena of resemblance or repetition; which links, at a distance and through time - as if through the mediation of a medium of propagation such defined unities as individuals, œuvres, notions, or theories. There are the notions of development and evolution: they make it possible to group a succession of dispersed events, to link them to one and the same organising principle, to subject them to the exemplary power of life (with its adaptations, its capacity for innovation, the incessant correlation of its different elements, its systems of assimilation and exchange), to discover, already at work in each beginning, a principle of coherence and the outline of a future unity, to master time through a perpetually reversible relation between an origin and a term that are never given, but are always at work. There is the notion of 'spirit', which enables us to establish between the simultaneous or successive phenomena of a given period a community of meanings, symbolic links, an interplay of resemblance and reflexion, or which allows the sovereignty of collective consciousness to emerge as the principle of unity and explanation. We must question those ready-made syntheses, those groupings that we normally accept before any examination, those links whose validity is recognised from the outset; we must oust those forms and obscure forces by which we usually link the discourse of one man with that of another; they must be driven out from the darkness in which they reign. And instead of according them unqualified, spontaneous value, we must accept, in the name of methodological rigour, that, in the first instance, they concern only a population of dispersed events.

We must also question those divisions or groupings with which we have become so familiar. Can one accept, as such, the distinction between the major types of discourse, or that between such forms or genres as science, literature, philosophy, religion, history, fiction, etc., and which tend to create certain great historical individualities? We are not even sure of ourselves when we use these distinctions in our own world of discourse, let alone when we are analysing groups of statements which, when first formulated, were distributed, divided, and characterised in a quite different way: after all, 'literature' and 'politics' are recent categories, which can be applied to medieval culture, or even classical culture, only by a retrospective hypothesis, and by an interplay of formal analogies or semantic resemblances; but neither literature, nor politics, nor philosophy and the sciences articulated the field of discourse, in the seventeenth or eighteenth century, as they did in the nineteenth century. In any case, these divisions - whether our own, or those contemporary with the discourse under examination - are always themselves reflexive categories, principles of classification, normative rules, institutionalised types: they, in turn, are facts of discourse that deserve to be analysed beside others; of course, they also have complex relations with each other, but they are not intrinsic, autochthonous, and universally recognisable characteristics.

But the unities that must be suspended above all are those that emerge in the most immediate way: those of the book and the œuvre. At first sight, it would seem that one could not abandon these unities without extreme artificiality. Are they not given in the most definite way? There is the material individualisation of the book, which occupies a determined space which has an economic value, and which itself indicates, by a number of signs, the limits of its beginning and its end; and there is the establishment of an oeuvre, which we recognise and delimit by attributing a certain number of texts to an author. And yet as soon as one looks at the matter a little more closely the difficulties begin. The material unity of the book? Is this the same in the case of an anthology of poems, a collection of posthumous fragments, Desargues' Traité des Coniques, or a volume of Michelet's Histoire de France? Is it the same in the case of Mallarmé's Un Coup de dés, the trial of Gilles de Rais, Butor's San Marco, or a Catholic missal? In other words, is not the material unity of the volume a weak, accessory unity in relation to the discursive unity of which it is the support? But is this discursive unity itself homogeneous and uniformly applicable? A novel by Stendhal and a novel by Dostoyevsky do not have the same relation of individuality as that between two novels belonging to Balzac's cycle La Comédie humaine; and the relation between Balzac's novels is not the same as that existing between Joyce's Ulysses and the Odyssey. The frontiers of a book are never clear-cut: beyond the title, the first lines, and the last full stop, beyond its internal configuration and its autonomous form, it is caught up in a system of references to other books, other texts, other sentences: it is a node within a network. And this network of references is not the same in the case of a mathematical treatise, a textual commentary, a historical account, and an episode in a novel cycle; the unity of the book, even in the sense of a group of relations, cannot be regarded as identical in each case. The book is not simply the object that one holds in one's hands; and it cannot remain within the little parallelepiped that contains it: its unity is variable and relative. As soon as one questions that unity, it lows its self-evidence; it indicates itself, constructs itself, only on the basis Of a complex field of discourse.

The problems raised by the œuvre are even more difficult. Yet, at first sight, what could be more simple? A collection of texts that can be designated by the sign of a proper name. But this designation (even leaving to one side problems of attribution) is not a homogeneous function: does the name of an author designate in the same way a text that he has published under his name, a text that he has presented under a pseudonym, another found after his death in the form of an unfinished draft, and another that is merely a collection of jottings, a notebook? The establishment of a complete oeuvre presupposes a number of choices that are difficult to justify or even to formulate: is it enough to add to the texts published by the author those that he intended for publication but which remained unfinished by the fact of his death? Should one also include all his sketches and first drafts, with all their corrections and crossings out? Should one add sketches that he himself abandoned? And what status should be given to letters, notes, reported conversations, transcriptions of what he said made by those present at the time, in short, to that vast mass of verbal traces left by an individual at his death, and which speak in an endless confusion so many different languages (langages)? In any case, the name 'Mallarmé' does not refer in the same way to his themes (translation exercises from French into English), his translations of Edgar Allan Poe, his poems, and his replies to questionnaires; similarly, the same relation does not exist between the name Nietzsche on the one hand and the youthful autobiographies, the scholastic dissertations, the philological articles, Zarathustra, Ecco Homo, the letters, the last postcards signed 'Dionysos' or 'Kaiser Nietzsche', and the innumerable notebooks with their jumble of laundry bills and sketches for aphorisms. In fact, if one speaks, so undiscriminately and unreflectingly of an author's œuvre, it is because one imagines it to be defined by a certain expressive function. One is admitting that there must be a level (as deep as it is necessary to imagine it) at which the oeuvre emerges, in all its fragments, even the smallest, most inessential ones, as the expression of the thought, the experience, the imagination, or the unconscious of the author, or, indeed, of the historical determinations that operated upon him. But it is at once apparent that such a unity, far from being given immediately is the result of an operation; that this operation is interpretative (since it deciphers, in the text, the transcription of something that it both conceals and manifests); and that the operation that determines the opus, in its unity, and consequently the œuvre itself, will not be the same in the case of the author of the Théâtre et son Double (Artaud) and the author of the Tractatus (Wittgenstein), and therefore when one speaks of an œuvre in each case one is using the word in a different sense. The œuvre can be regarded neither as an immediate unity, nor as a certain unity, nor as a homogeneous unity.

One last precaution must be taken to disconnect the unquestioned continuities by which we organise, in advance, the discourse that we are to analyse: we must renounce two linked, but opposite themes. The first involves a wish that it should never be possible to assign, in the order of discourse, the irruption of a real event; that beyond any apparent beginning, there is always a secret origin - so secret and so fundamental that it can never be quite grasped in itself. Thus one is led inevitably, through the naïvety of chronologies, towards an ever-receding point that is never itself present in any history; this point is merely its own void; and from that point all beginnings can never be more than recommencements or occultation (in one and the same gesture, this and that). To this theme is connected another according to which all manifest discourse is secretly based on an 'already-said'; and that this 'already said' is not merely a phrase that has already been spoken, or a text that has already been written, but a 'never-said', an incorporeal discourse, a voice as silent as a breath, a writing that is merely the hollow of its own mark. It is supposed therefore that everything that is formulated in discourse was already articulated in that semi-silence that precedes it, which continues to run obstinately beneath it, but which it covers and silences. The manifest discourse, therefore, is really no more than the repressive presence of what it does not say; and this 'not-said' is a hollow that undermines from within all that is said. The first theme sees the historical analysis of discourse as the quest for and the repetition of an origin that eludes all historical determination; the second sees it as the interpretation of 'hearing' of an 'already-said' that is at the same time a 'not-said'. We must renounce all those themes whose function is to ensure the infinite continuity of discourse and its secret presence to itself in the interplay of a constantly recurring absence. We must be ready to receive every moment of discourse in its sudden irruption; in that punctuality in which it appears, and in that temporal dispersion that enables it to be repeated, known, forgotten, transformed, utterly erased, and hidden, far from all view, in the dust of books. Discourse must not be referred to the distant presence of the origin, but treated as and when it occurs.

These pre-existing forms of continuity, all these syntheses that are accepted without question, must remain in suspense. They must not be rejected definitively of course, but the tranquillity with which they are accepted must be disturbed; we must show that they do not come about of themselves, but are always the result of a construction the rules of which must be known, and the justifications of which must be scrutinised: we must define in what conditions and in view of which analyses certain of them are legitimate; and we must indicate which of them can never be accepted in any circumstances. It may be, for example, that the notions of 'influence' or 'evolution' belong to a criticism that puts them - for the foreseeable future - out of use. But need we dispense for ever with the 'œuvre', the 'book', or even such unities as 'science' or 'literature'? Should we regard them as illusions, illegitimate constructions, or ill-acquired results? Should we never make use of them, even as a temporary support, and never provide them with a definition? What we must do, in fact, is to tear away from them their virtual self-evidence, and to free the problems that they pose; to recognise that they are not the tranquil locus on the basis of which other questions (concerning their structure, coherence, systematicity, transformations) may be posed, but that they themselves pose a whole cluster of questions (What are they? How can they be defined or limited? What distinct types of laws can they obey? What articulation are they capable of? What sub-groups can they give rise to? What specific phenomena do they reveal in the field of discourse?). We must recognise that they may not, in the last resort, be what they seem at first sight. In short, that they require a theory, and that this theory cannot be constructed unless the field of the facts of discourse on the basis of which those facts are built up appears in its non-synthetic purity.

And I, in turn, will do no more than this: of course, I shall take as my starting-point whatever unities are already given (such as psychopathology, medicine, or political economy); but I shall not place myself inside these dubious unities in order to study their internal configuration or their secret contradictions. I shall make use of them just long enough to ask myself what unities they form; by what right they can claim a field that specifies them in space and a continuity that individualises them in time; according to what laws they are formed; against the background of which discursive events they stand out; and whether they are not, in their accepted and quasi-institutional individuality, ultimately the surface effect of more firmly grounded unities. I shall accept the groupings that history suggests only to subject them at once to interrogation; to break them up and then to see whether they can be legitimately reformed; or whether other groupings should be made; to replace them in a more general space which, while dissipating their apparent familiarity, makes it possible to construct a theory of them.

Once these immediate forms of continuity are suspended, an entire field is set free. A vast field, but one that can be defined nonetheless: this field is made up of the totality of all effective statements (whether spoken or written), in their dispersion as events and in the occurrence that is proper to them. Before approaching, with any degree of certainty, a science, or novels, or political speeches, or the œuvre of an author, or even a single book, the material with which one is dealing is, in its raw, neutral state, a population of events in the space of discourse in general. One is led therefore to the project of a pure description of discursive events as the horizon for the search for the unities that form within it. This description is easily distinguishable from an analysis of the language. Of course, a linguistic system can be established (unless it is constructed artificially) only by using a corpus of statements, or a collection of discursive facts; but we must then define, on the basis of this grouping, which has value as a sample, rules that may make it possible to construct other statements than these: even if it has long since disappeared, even if it is no longer spoken, and can be reconstructed only on the basis of rare fragments, a language (langue) is still a system for possible statements, a finite body of rules that authorises an infinite number of performances. The field of discursive events, on the other hand, is a grouping that is always finite and limited at any moment to the linguistic sequences that have been formulated; they may be innumerable, they may, in sheer size, exceed the capacities of recording, memory, or reading: nevertheless they form a finite grouping. The question posed by language analysis of some discursive fact or other is always: according to what rules has a particular statement been made, and consequently according to what rules could other similar statements be made? The description of the events of discourse poses a quite different question: how is it that one particular statement appeared rather than another?

It is also clear that this description of discourses is in opposition to the history of thought. There too a system of thought can be reconstituted only on the basis of a definite discursive totality. But this totality is treated in such a way that one tries to rediscover beyond the statements themselves the intention of the speaking subject, his conscious activity, what he meant, or, again, the unconscious activity that took place, despite himself, in what he said or in the almost imperceptible fracture of his actual words; in any case, we must reconstitute another discourse, rediscover the silent murmuring, the inexhaustible speech that animates from within the voice that one hears, re-establish the tiny, invisible text that runs between and sometimes collides with them. The analysis of thought is always allegorical in relation to the discourse that it employs. Its question is unfailingly: what was being said in what was said? The analysis of the discursive field is orientated in a quite different way; we must grasp the statement in the exact specificity of its occurrence; determine its conditions of existence, fix at least its limits, establish its correlations with other statements that may be connected with it, and show what other forms of statement it excludes. We do not seek below what is manifest the half silent murmur of another discourse; we must show why it could not be other than it was, in what respect it is exclusive of any other, how it assumes, in the midst of others and in relation to them, a place that no other could occupy. The question proper to such an analysis might be formulated in this way: what is this specific existence that emerges from what is said and nowhere else?

We must ask ourselves what purpose is ultimately served by this suspension of all the accepted unities, if, in the end, we return to the unities that we pretended to question at the outset. In fact, the systematic erasure of all given unities enables us first of all to restore to the statement the specificity of its occurrence, and to show that discontinuity is one of those great accidents that create cracks not only in the geology of history, but also in the simple fact of the statement; it emerges in its historical irruption; what we try to examine is the incision that it makes, that irreducible - and very often tiny - emergence. However banal it may be, however unimportant its consequences may appear to be, however quickly it may be forgotten after its appearance, however little heard or however badly deciphered we may suppose it to be, a statement is always an event that neither the language (langue) nor the meaning can quite exhaust. It is certainly a strange event: first, because on the one hand it is linked to the gesture of writing or to the articulation of speech, and also on the other hand it opens up to itself a residual existence in the field of a memory, or in the materiality of manuscripts, books, or any other form of recording; secondly, because, like every event, it is unique, yet subject to repetition, transformation, and reactivation; thirdly, because it is linked not only to the situations that provoke it, and to the consequences that it gives rise to, but at the same time, and in accordance with a quite different modality, to the statements that precede and follow it.

But if we isolate, in relation to the language and to thought, the occurrence of the statement/event, it is not in order to spread over everything a dust of facts. it is in order to be sure that this occurrence is not linked with synthesising operations of a purely psychological kind (the intention of the author,, the form of his mind, the rigour of his thought, the themes that obsess him, the project that traverses his existence and gives it meaning) and to be able to grasp other forms of regularity, other types of relations. Relations between statements (even if the author is unaware of them; even if the statements do not have the same author; even if the authors were unaware of each other's existence); relations between groups of statements thus established (even if these groups do not concern the same, or even adjacent, fields; even if they do not possess the same formal level; even if they are not the locus of assignable exchanges); relations between statements and groups of statements and events of a quite different kind (technical, economic, social, political). To reveal in all its purity the space in which discursive events are deployed is not to undertake to re-establish it in an isolation that nothing could overcome; it is not to close it upon itself; it is to leave oneself free to describe the interplay of relations within it and outside it.

The third purpose of such a description of the facts of discourse is that by freeing them of all the groupings that purport to be natural, immediate, universal unities, one is able to describe other unities, but this time by means of a group of controlled decisions., Providing one defines the conditions clearly, it might be legitimate to constitute, on the basis of correctly described relations, discursive groups that are not arbitrary, and yet remain invisible. Of course, these relations would never be formulated for themselves in the statements in question (unlike, for example, those explicit relations that are posed and spoken in discourse itself, as in the form of the novel, or a series of mathematical theorems). But in no way would they constitute a sort of secret discourse, animating the manifest discourse from within; it is not therefore an interpretation of the facts of the statement that might reveal them, but the analysis of their coexistence, their succession, their mutual functioning, their reciprocal determination, and their independent or correlative transformation.

However, it is not possible to describe all the relations that may emerge in this way without some guide-lines. A provisional division must be adopted as an initial approximation: an initial region that analysis will subsequently demolish and, if necessary, reorganise. But how is such a region to be circumscribed? on the one hand, we must choose, empirically, a field in which the relations are likely to be numerous, dense, and relatively easy to describe: and in what other region do discursive events appear to be more closely linked to one another, to occur in accordance with more easily decipherable relations, than in the region usually known as science? But, on the other hand, what better way of grasping in a statement, not the moment of its formal structure and laws of construction, but that of its existence and the rules that govern its appearance, if not by dealing with relatively uniformalised groups of discourses, in which the statements do not seem necessarily to be built on the rules of pure syntax? How can we be sure of avoiding such divisions as the œuvre, or such categories as 'influence', unless, from the very outset, we adopt sufficiently broad fields and scales that are chronologically vast enough? Lastly, how can we be sure that we will not find ourselves in the grip of all those over-hasty unities or syntheses concerning the speaking subject, or the author of the text, in short, all anthropological categories? Unless, perhaps, we consider all the statements out of which these categories are constituted - all the statements that have chosen the subject of discourse (their own subject) as their 'object' and have undertaken to deploy it as their field of knowledge?

This explains the de facto privilege that I have accorded to those discourses that, to put it very schematically, define the 'sciences of man'. But it is only a provisional privilege. Two facts must be constantly borne in mind: that the analysis of discursive events is in no way limited to such a field; and that the division of this field itself cannot be regarded either as definitive or as absolutely valid; it is no more than an initial approximation that must allow relations to appear that may erase the limits of this initial outline.

Discursive Formations

I have undertaken, then, to describe the relations between statements. I have been careful to accept as valid none of the unities that would normally present themselves to anyone embarking on such a task. I have decided to ignore no form of discontinuity, break, threshold, or limit. I have decided to describe statements in the field of discourse and the relations of which they are capable. As I see it, two series of problems arise at the outset: the first, which I shall leave to one side for the time being and shall return to later, concerns the indiscriminate use that I have made of the terms statement, event, and discourse; the second concerns the relations that may legitimately be described between the statements that have been left in their provisional, visible grouping.

There are statements, for example, that are quite obviously concerned and have been from a date that is easy enough to determine - with political economy, or biology, or psychopathology; there are others that equally obviously belong to those age-old continuities known as grammar or medicine. But what are these unities? How can we say that the analysis of headaches carried out by Willis or Charcot belong to the same order of discourse? That Petty's inventions are in continuity with Neumann's econometry? That the analysis of judgement by the Port-Royal grammarians belongs to the same domain as the discovery of vowel gradations in the Indo-European languages? What, in fact, are medicine, grammar, or political economy? Are they merely a retrospective regrouping by which the contemporary sciences deceive themselves as to their own past? Are they forms that have become established once and for all and have gone on developing through time? Do they conceal other unities? And what sort of links can validly be recognised between all these statements that form, in such a familiar and insistent way, such an enigmatic mass?

First hypothesis - and the one that, at first sight, struck me as being the most likely and the most easily proved: statements different in form, and dispersed in time, form a group if they refer to one and the same object. Thus, statements belonging to psychopathology all seem to refer to an object that emerges in various ways in individual or social experience and which may be called madness. But I soon realised that the unity of the object 'madness' does not enable one to individualise a group of statements, and to establish between them a relation that is both constant and describable. There are two reasons for this. It would certainly be a mistake to try to discover what could have been said of madness at a particular time by interrogating the being of madness itself, its secret content, its silent, self-enclosed truth; mental illness was constituted by all that was said in all the statements that named it, divided it up, described it, explained it, traced its developments, indicated its various correlations, judged it, and possibly gave it speech by articulating, in its name, discourses that were to be taken as its own. Moreover, this group of statements is far from referring to a single object, formed once and for all, and to preserving it indefinitely as its horizon of inexhaustible ideality; the object presented as their correlative by medical statements of the seventeenth or eighteenth century is not identical with the object that emerges in legal sentences or police action; similarly, all the objects of psychopathological discourses were modified from Pinel or Esquirol to Bleuler: it is not the same illnesses that are at issue in each of these cases; we are not dealing with the same madmen.

One might, perhaps one should, conclude from this multiplicity of objects that it is not possible to accept, as a valid unity forming a group of statements, a 'discourse, concerning madness'. Perhaps one should confine one's attention to those groups of statements that have one and the same object: the discourses on melancholia, or neurosis, for example. But one would soon realise that each of these discourses in turn constituted its object and worked it to the point of transforming it altogether. So that the problem arises of knowing whether the unity of a discourse is based not so much on the permanence and uniqueness of an object as on the space in which various objects emerge and are continuously transformed. Would not the typical relation that would enable us to individualise a group of statements concerning madness then be: the rule of simultaneous or successive emergence of the various objects that are named, described, analysed, appreciated, or judged in that relation? The unity of discourses on madness would not be based upon the existence of the object 'madness', or the constitution of a single horizon of objectivity; it would be the interplay of the rules that make possible the appearance of objects during a given period of time: objects that are shaped by measures of discrimination and repression, objects that are differentiated in daily practice, in law, in religious casuistry, in medical diagnosis, objects that are manifested in pathological descriptions, objects that are circumscribed by medical codes, practices, treatment, and care. Moreover, the unity of the discourses on madness would be the interplay of the rules that define the transformations of these d' rent objects, their non-identity through time, the break produced in them, the internal discontinuity that suspends their permanence. Paradoxically, to define a group of statements in terms of its individuality would be to define the dispersion of these objects, to grasp all the interstices that separate them, to measure the distances that reign between them - in other words, to formulate their law of division.

Second hypothesis to define a group of relations between statements: their form and type of connection. It seemed to me, for example, that from the nineteenth century medical science was characterised not so much by its objects or concepts as by a certain style, a certain constant manner of statement. For the first time, medicine no longer consisted of a group of traditions, observations, and heterogeneous practices, but of a corpus of knowledge that presupposed the same way of looking at things, the same division of the perceptual field, the same analysis of the pathological fact in accordance with the visible space of the body, the same system of transcribing what one perceived in what one said (same vocabulary, same play of metaphor); in short, it seemed to me that medicine was organised as a series of descriptive statements. But, there again, I had to abandon this hypothesis at the outset and recognise that clinical discourse was just as much a group of hypotheses about life and death, of ethical choices, of therapeutic decisions, of institutional regulations, of teaching models, as a group of descriptions; that the descriptions could not, in any case, be abstracted from the hypotheses, and that the descriptive statement was only one of the formulations present in medical discourse. I also had to recognise that this description has constantly been displaced: either because, from Bichat to cell pathology, the scales and guide-lines have been displaced; or because from visual inspection, auscultation and palpation to the use of the microscope and biological tests, the information system has been modified; or, again, because, from simple anatomo-clinical correlation to the delicate analysis of physio-pathological processes, the lexicon of signs and their decipherment has been entirely reconstituted; or, finally, because the doctor has gradually ceased to be himself the locus of the registering and interpretation of information, and because, beside him, outside him, there have appeared masses of documentation, instruments of correlation, and techniques of analysis, which, of course, he makes use of, but which modify his position as an observing subject in relation to the patient.

All these alterations, which may now lead to the threshold of a new medicine, gradually appeared in medical discourse throughout the nineteenth century. If one wished to define this discourse by a codified and normative system of statement, one would have to recognise that this medicine disintegrated as soon as it appeared and that it really found its formulation only in Bichat and Laennec. If there is a unity, its principle is not therefore a determined form of statements; is it not rather the group of rules, which, simultaneously or in turn, have made possible purely perceptual descriptions, together with observations mediated through instruments, the procedures used in laboratory experiments, statistical calculations, epidemiological or demographic observations, institutional regulations, and therapeutic practice? What one must characterise and individualise is the coexistence of these dispersed and heterogeneous statements; the system that governs their division, the degree to which they depend upon one another, the way in which they interlock or exclude one another, the transformation that they undergo, and the play of their location, arrangement, and replacement.

Another direction of research, another hypothesis: might it not be possible to establish groups of statements, by determining the system of permanent and coherent concepts involved? For example, does not the Classical analysis of language and grammatical facts (from Lancelot to the end of the eighteenth century) rest on a definite number of concepts whose content and usage had been established once and for all: the concept of judgement defined as the general, normative form of any sentence, the concepts of subject and predicate regrouped under the more general category of noun, the concept of verb used as the equivalent of that of logical copula, the concept of word defined as the sign of a representation, etc.? In this way, one might reconstitute the conceptual architecture of Classical grammar. But there too one would soon come up against limitations: no sooner would one have succeeded in describing with such elements the analyses carried out by the Port-Royal authors than one would no doubt be forced to acknowledge the appearance of new concepts; some of these may be derived from the first, but the others are heterogeneous and a few even incompatible with them. The notion of natural or inverted syntactical order, that of complement (introduced in the eighteenth century by Beauzée), may still no doubt be integrated into the conceptual system of the Port-Royal grammar. But neither the idea of an originally expressive value of sounds, nor that of a primitive body of knowledge enveloped in words and conveyed in some obscure way by them, nor that of regularity in the mutation of consonants, nor the notion of the verb as a mere name capable of designating an action or operation, is compatible with the group of concepts used by Lancelot or Duclos. Must we admit therefore that grammar only appears to form a coherent figure; and that this group of statements, analyses, descriptions, principles and consequences, deductions that has been perpetrated under this name for over a century is no more than a false unity? But perhaps one might discover a discursive unity if one sought it not in the coherence of concepts, but in their simultaneous or successive emergence, in the distance that separates them and even in their incompatibility. One would no longer seek an architecture of concepts sufficiently general and abstract to embrace all others and to introduce them into the same deductive structure; one would try to analyse the interplay of their appearances and dispersion.

Lastly, a fourth hypothesis to regroup the statements, describe their interconnection and account for the unitary forms under which they are presented: the identity and persistence of themes. In 'sciences' like economics or biology, which are so controversial in character, so open to philosophical or ethical options, so exposed in certain cases to political manipulation, it is legitimate in the first instance to suppose that a certain thematic is capable of linking, and animating a group of discourses, like an organism with its own needs, its own internal force, and its own capacity for survival. Could one not, for example, constitute as a unity everything that has constituted the evolutionist theme from Buffon to Darwin? A theme that in the first instance was more philosophical, closer to cosmology than to biology; a theme that directed research from afar rather than named, regrouped, and explained results; a theme that always presupposed more than one was aware Of, but which, on the basis of this fundamental choice, forcibly transformed into discursive knowledge what had been outlined as a hypothesis or as a necessity. Could one not speak of the Physiocratic theme in the same way? An idea that postulated, beyond all demonstration and prior to all analysis, the natural character of the three ground rents; which consequently presupposed the economic and political primacy of agrarian property; which excluded all analysis of the mechanisms of industrial production; which implied, on the other hand, the description of the circulation of money within a state, of its distribution between different social categories, and of the channels by which it flowed back into production; which finally led Ricardo to consider those cases in which this triple rent did not appear, the conditions in which it could form, and consequently to denounce the arbitrariness of the Physiocratic theme?

But on the basis of such an attempt, one is led to make two inverse and complementary observations. In one case, the same thematic is articulated on the basis of two sets of concepts, two types of analysis, two perfectly different fields of objects: in its most general formulation, the evolutionist idea is perhaps the same in the work of Benoit de Maillet, Borden or Diderot, and in that of Darwin; but, in fact, what makes it possible and coherent is not at all the same thing in either case. In the eighteenth century, the evolutionist idea is defined on the basis of a kinship of species forming a continuum laid down at the outset (interrupted only by natural catastrophes) or gradually built up by the passing of time. In the nineteenth century the evolutionist theme concerns not so much the constitution of a continuous table of species, as the description of discontinuous groups and the analysis of the modes of interaction between an organism whose elements are interdependent and an environment that provides its real conditions of life. A single theme, but based on two types of discourse. In the case of Physiocracy, on the other hands Quesnay's choice rests exactly on the same system of concepts as the opposite opinion held by those that might be called utilitarists. At this period the analysis of wealth involved a relatively limited set of concepts that was accepted by all (coinage was given the same definition; prices were given the same explanation; and labour costs were calculated in the same way). But, on the basis of this single set of concepts, there were two ways of explaining the formation of value, according to whether it was analysed on the basis of exchange, or on that of remuneration for the day's work. These two possibilities contained within economic theory, and in the rules of its set of concepts, resulted, on the basis of the same elements, in two different options.

It would probably be wrong therefore to seek in the existence of these themes the principles of the individualisation of a discourse. Should they not be sought rather in the dispersion of the points of choice that the discourse leaves free? In the different possibilities that it opens of reanimating already existing themes, of arousing opposed strategies, of giving way to irreconcilable interests, of making it possible, with a particular set of concepts, to play different games? Rather than seeking the permanence of themes, images, and opinions through time, rather than retracing the dialectic of their conflicts in order to individualise groups of statements, could one not rather mark out the dispersion of the points of choice, and define prior to any option, to any thematic preference, a field of strategic possibilities?

I am presented therefore with four attempts, four failures - and four successive hypotheses. They must now be put to the test. Concerning those large groups of statements with which we are so familiar - and which we call medicine, economics, or grammar - I have asked myself on what their unity could be based. On a full, tightly packed, continuous, geographically well-defined field of objects? What appeared to me were rather series full of gaps, intertwined with one another, interplays of differences, distances, substitutions, transformations. On a definite, normative type of statement? I found formulations of levels that were much too different and functions that were much too heterogeneous to be linked together and arranged in a single figure, and to simulate, from one period to another, beyond individual œuvres, a sort of great uninterrupted text. On a well-defined alphabet of notions? One is confronted with concepts that differ in structure and in the rules governing their use, which ignore or exclude one another, and which cannot enter the unity of a logical architecture. On the permanence of a thematic? What one finds are rather various strategic possibilities that permit the activation of incompatible themes, or, again, the establishment of the same theme in different groups of statement. Hence the idea of describing these dispersions themselves; of discovering whether, between these elements, which are certainly not organised as a progressively deductive structure, nor as an enormous book that is being gradually and continuously written, nor as the œuvre of a collective subject, one cannot discern a regularity: an order in their successive appearance, correlations in their simultaneity, assignable positions in a common space, a reciprocal functioning, linked and hierarchised transformations. Such an analysis would not try to isolate small islands of coherence in order to describe their internal structure; it would not try to suspect and to reveal latent conflicts; it would study forms of division. Or again: instead of reconstituting chains of inference (as one often does in the history of the sciences or of philosophy), instead of drawing up tables of differences (as the linguists do), it would describe systems of dispersion.

Whenever one can describe, between a number of statements, such a system of dispersion, whenever, between objects, types of statement, concepts, or thematic choices, one can define a regularity (an order, correlations, positions and functionings, transformations), we will say, for the sake of convenience, that we are dealing with a discursive formation - thus avoiding words that are already overladen with conditions and consequences, and in any case inadequate to the task of designating such a dispersion, such as 'science' 'ideology', 'theory', or 'domain of objectivity'. The conditions to which the elements of this division (objects, mode of statement, concepts, thematic choices) are subjected we shall call the rules of formation. The rules of formation are conditions of existence (but also of coexistence, maintenance, modification, and disappearance) in a given discursive division.

This, then, is the field to be covered; these the notions that we must put to the test and the analyses that we must carry out. I am well aware that the risks are considerable. For an initial probe, I made use of certain fairly loose, but familiar, groups of statement: I have no proof that I shall find them again at the end of the analysis, nor that I shall discover the principle of their delimitation and individualisation; I am not sure that the discursive formations that I shall isolate will define medicine in its overall unity, or economics and grammar in the overall curve of their historical destination; they may even introduce unexpected boundaries and divisions. Similarly, I have no proof that such a description will be able to take account of the scientificity (or non-scientificity) of the discursive groups that I have taken as an attack point and which presented themselves at the outset with a certain pretension to scientific rationality; I have no proof that my analysis will not be situated at a quite different level, constituting a description that is irreducible to epistemology or to the history of the sciences. Moreover, at the end of such an enterprise, one may not recover those unities that, out of methodological rigour, one initially held in suspense: one may be compelled to dissociate certain œuvres, ignore influences and traditions, abandon definitively the question of origin, allow the commanding presence of authors to fade into the background; and thus everything that was thought to be proper to the history of ideas may disappear from view. The danger, in short, is that instead of providing a basis for what already exists, instead of going over with bold strokes lines that have already been sketched, instead of finding reassurance in this return and final confirmation, instead of completing the blessed circle that announces, after innumerable stratagems and as many nights, that all is saved, one is forced to advance beyond familiar territory, far from the certainties to which one is accustomed, towards an as yet uncharted land and unforeseeable conclusion. Is there not a danger that everything that has so far protected the historian in his daily journey and accompanied him until nightfall (the destiny of rationality and the teleology of the sciences, the long, continuous labour of thought from period to period, the awakening and the progress of consciousness, its perpetual resumption of itself, the uncompleted, but uninterrupted movement of totalisations, the return to an ever-open source, and finally the historico-transcendental thematic) may disappear, leaving for analysis a blank, indifferent space, lacking in both interiority and promise?

The Formation of Objects

We must now list the various directions that lie open to us, and see whether this notion of 'rules of formation' - of which little more than a rough sketch has so far been provided - can be given real content. Let us look first at the formation of objects. And in order to facilitate our analysis, let us take as an example the discourse of psychopathology from the nineteenth century onwards - a chronological break that is easy enough to accept in a first approach to the subject. There are enough signs to indicate it, but let us take just two of these: the establishment at the beginning of the century of a new mode of exclusion and confinement of the madman in a psychiatric hospital; and the possibility of tracing certain present-day notions back to Esquirol, Heinroth, or Pinel (paranoia can be traced back to monomania, the intelligence quotient to the initial notion of imbecility, general paralysis to chronic encephalitis, character neurosis to nondelirious madness); whereas if we try to trace the development of psychopathology beyond the nineteenth century, we soon lose our way, the path becomes confused, and the projection of Du Laurens or even Van Swicten on the pathology of Kraepelin or Bleuler provides no more than chance coincidences. The objects with which psychopathology has dealt since this break in time are very numerous, mostly very new, but also very precarious, subject to change and, in some cases, to rapid disappearance: in addition to motor disturbances, hallucinations, and speech disorders (which were already regarded as manifestations of madness, although they were recognised, delimited, described, and analysed in a different way), objects appeared that belonged to hitherto unused registers: minor behavioural disorders, sexual aberrations and disturbances, the phenomena of suggestion and hypnosis, lesions of the central nervous system, deficiencies of intellectual or motor adaptation, criminality. And on the basis of each of these registers a variety of objects were named, circumscribed, analysed, then rectified, re-defined, challenged, erased. Is it possible to lay down the rule to which their appearance was subject? Is it possible to discover according to which non-deductive system these objects could be juxtaposed and placed in succession to form the fragmented field - showing at certain points great gaps, at others a plethora of information - of psychopathology? What has ruled their existence as objects of discourse?

(a) First we must map the first surfaces of their emergence: show where these individual differences, which, according to the degrees of rationalisation, conceptual codes, and types of theory, will be accorded the status of disease, alienation, anomaly, dementia, neurosis or psychosis, degeneration, etc., may emerge, and then be designated and analysed. These surfaces of emergence are not the same for different societies, at different periods, and in different forms of discourse. In the case of nineteenth-century psychopathology, they were probably constituted by the family, the immediate social group, the work situation, the religious community (which are all normative, which are all susceptible to deviation, which all have a margin of tolerance and a threshold beyond which exclusion is demanded, which all have a mode of designation and a mode of rejecting madness, which all transfer to medicine if not the responsibility for treatment and cure, at least the burden of explanation); although organised according to a specific mode, these surfaces of emergence were not new in the nineteenth century. On the other hand, it was no doubt at this period that new surfaces of appearance began to function: art with its own normativity, sexuality (its deviations in relation to customary prohibitions become for the first time an object of observation, description, and analysis for psychiatric discourse), penality (whereas in previous periods madness was carefully distinguished from criminal conduct and was regarded as an excuse, criminality itself becomes - and subsequent to the celebrated 'homicidal monomanias' - a form of deviance more or less related to madness). In these fields of initial differentiation, in the distances, the discontinuities, and the thresholds that appear within it, psychiatric discourse finds a way of limiting its domain, of defining what it is talking about, of giving it the status of an object - and therefore of making it manifest, nameable, and describable.

(b) We must also describe the authorities of delimitation: in the nineteenth century, medicine (as an institution possessing its own rules, as a group of individuals constituting the medical profession, as a body of knowledge and practice, as an authority recognised by public opinion, the law, and government) became the major authority in society that delimited, designated, named, and established madness as an object; but it was not alone in this: the law and penal law in particular (with the definitions of excuse, non-responsibility, extenuating circumstances, and with the application of such notions as the crime passional, heredity, danger to society), the religious authority (in so far as it set itself up as the authority that divided the mystical from the pathological, the spiritual from the corporeal, the supernatural from the abnormal, and in so far as it practised the direction of conscience with a view to understanding individuals rather than carrying out a casuistical classification of actions and circumstances), literary and art criticism (which in the nineteenth century treated the work less and less as an object of taste that had to be judged, and more and more as a language that had to be interpreted and in which the author's tricks of expression had to be recognised).

(c) Lastly, we must analyse the grids of specification: these are the systems according to which the different 'kinds of madness' are divided, contrasted, related, regrouped, classified, derived from one another as objects of psychiatric discourse (in the nineteenth century, these grids of differentiation were: the soul, as a group of hierarchised, related, and more or less interpenetrable faculties; the body, as a three-dimensional volume of organs linked together by networks of dependence and communication; the life and history of individuals, as a linear succession of phases, a tangle of traces, a group of potential reactivations, cyclical repetitions; the interplays of neuropsychological correlations as systems of reciprocal projections, and as a field of circular causality).

Such a description is still in itself inadequate. And for two reasons. These planes of emergence, authorities of delimitation, or forms of specification do not provide objects, fully formed and armed, that the discourse of psychopathology has then merely to list, classify, name, select, and cover with a network of words and sentences: it is not the families - with their norms, their prohibitions, their sensitivity thresholds - that decide who is mad, and present the 'patients' to the psychiatrists for analysis and judgement; it is not the legal system itself that hands over certain criminals to psychiatry, that sees paranoia beyond a particular murder, or a neurosis behind a sexual offence. It would be quite wrong to see discourse as a place where previously established objects are laid one after another like words on a page. But the above enumeration is inadequate for a second reason. it has located, one after another, several planes of differentiation in which the objects of discourse may appear. But what relations exist between them? Why this enumeration rather than another? What defined and closed group does one imagine one is circumscribing in this way? And how can one speak of a 'system of formation' if one knows only a series of different, heterogeneous determinations, lacking attributable links and relations?

In fact, these two series of questions refer back to the same point. In order to locate that point, let us re-examine the previous example. In the sphere with which psychopathology dealt in the nineteenth century, one sees the very early appearance (as early as Esquirol) of a whole series of objects belonging to the category of delinquency: homicide (and suicide), crimes passionels, sexual offences, certain forms of theft, vagrancy - and then, through them, heredity, the neurogenic environment, aggressive or self-punishing behaviour, perversions, criminal impulses, suggestibility, etc. It would be inadequate to say that one was dealing here with the consequences of a discovery: of the sudden discovery by a psychiatrist of a resemblance between, criminal and pathological behaviour, a discovery of the presence in certain delinquents of the classical signs of alienation, or mental derangement. Such facts lie beyond the grasp of contemporary research: indeed, the problem is how to decide what made them possible, and how these 'discoveries' could lead to others that took them up, rectified them, modified them, or even disproved them. Similarly, it would be irrelevant to attribute the appearance of these new objects to the norms of nineteenth-century bourgeois society, to a reinforced police and penal framework, to the establishment of a new code of criminal justice, to the introduction and use of extenuating circumstances, to the increase in crime. No doubt, all these processes were at work; but they could not of themselves form objects for psychiatric discourse; to pursue the description at this level one would fall short of what one was seeking.

If, in a particular period in the history of our society, the delinquent was psychologised and pathologised, if criminal behaviour could give rise to a whole series of objects of knowledge, this was because a group of particular relations was adopted for use in psychiatric discourse. The relation between planes of specification like penal categories and degrees of diminished responsibility, and planes of psychological characterisation (faculties, aptitudes, degrees of development or involution, different ways of reacting to the environment, character types, whether acquired, innate, or hereditary). The relation between the authority of medical decision and the authority of judicial decision (a really complex relation since medical decision recognises absolutely the authority of the judiciary to define crime, to determine the circumstances in which it is committed, and the punishment that it deserves; but reserves the right to analyse its origin and to determine the degree of responsibility involved). The relation between the filter formed by judicial interrogation, police information, investigation, and the whole machinery of judicial information, and the filter formed by the medical questionnaire, clinical examinations, the search for antecedents, and biographical accounts. The relation between the family, sexual and penal norms of the behaviour of individuals, and the table of pathological symptoms and diseases of which they are the signs. The relation between therapeutic confinement in hospital (with its own thresholds, its criteria of cure, its way of distinguishing the normal from the pathological) and punitive confinement in prison (with its system of punishment and pedagogy, its criteria of good conduct, improvement, and freedom). These are the relations that, operating in psychiatric discourse, have made possible the formation of a whole group of various objects.

Let us generalise: in the nineteenth century, psychiatric discourse is characterised not by privileged objects, but by the way in which it forms objects that are in fact highly dispersed. This formation is made possible by a group of relations established between authorities of emergence, delimitation, and specification. One might say, then, that a discursive formation is defined (as far as its objects are concerned, at least) if one can establish such a group; if one can show how any particular object of discourse finds in it its place and law of emergence; if one can show that it may give birth simultaneously or successively to mutually exclusive objects, without having to modify itself.

Hence a certain number of remarks and consequences.

1. The conditions necessary for the appearance of an object of discourse, the historical conditions required if one is to 'say anything' about it, and if several people are to say different things about it, the conditions necessary if it is to exist in relation to other objects, if it is to establish with them relations of resemblance, proximity, distance, difference, transformation - as we can see, these conditions are many and imposing. Which means that one cannot speak of anything at any time; it is not easy to say something new; it is not enough for us to open our eyes, to pay attention, or to be aware, for new objects suddenly to light up and emerge out of the ground. But this difficulty is not only a negative one; it must not be attached to some obstacle whose power appears to be, exclusively, to blind, to hinder, to prevent discovery, to conceal the purity of the evidence or the dumb obstinacy of the things themselves; the object does not await in limbo the order that will free it and enable it to become embodied in a visible and prolix objectivity; it does not pre-exist itself, held back by some obstacle at the first edges of light. It exists under the positive conditions of a complex group of relations.

2. These relations are established between institutions, economic and social processes, behavioural patterns, systems of norms, techniques, types of classification, modes of characterisation; and these relations are not present in the object; it is not they that are deployed when the object is being analysed; they do not indicate the web, the immanent rationality, that ideal nervure that reappears totally or in part when one conceives of the object in the truth of its concept. They do not define its internal constitution, but what enables it to appear, to juxtapose itself with other objects, to situate itself in relation to them, to define its difference, its irreducibility, and even perhaps its heterogeneity, in short, to be placed in a field of exteriority.

3. These relations must be distinguished first from what we might call primary relations, and which, independently of all discourse or all object of discourse, may be described between institutions, techniques, social forms, etc. After all, we know very well that relations existed between the bourgeois family and the functioning of judicial authorities and categories in the nineteenth century that can be analysed in their own right. They cannot always be superposed upon the relations that go to form objects: the relations of dependence that may be assigned to this primary level are not necessarily expressed in the formation of relations that makes discursive objects possible. But we must also distinguish the secondary relations that are formulated in discourse itself. what, for example, the psychiatrists of the nineteenth century could say about the relations between the family and criminality does not reproduce, as we know, the interplay of real dependencies; but neither does it reproduce the interplay of relations that make possible and sustain the objects of psychiatric discourse. Thus a space unfolds articulated with possible discourses: a system of real or primary relations, a system of reflexive or secondary relations, and a system of relations that might properly be called discursive. The problem is to reveal the specificity of these discursive relations, and their interplay with the other two kinds.

4. Discursive relations are not, as we can see, internal to discourse: they do not connect concepts or words with one another; they do not establish a deductive or rhetorical structure between propositions or sentences. Yet they are not relations exterior to discourse, relations that might limit it, or impose certain forms upon it, or force it, in certain circumstances, to state certain things. They are, in a sense, at the limit of discourse: they offer it objects of which it can speak, or rather (for this image of offering presupposes that objects are formed independently of discourse), they determine the group of relations that discourse must establish in order to speak of this or that object, in order to deal with them, name them, analyse them, classify them, explain them, etc. These relations characterise not the language (langue) used by discourse, nor the circumstances in which it is deployed, but discourse itself as a practice.

We can now complete the analysis and see to what extent it fulfils, and to what extent it modifies, the initial project.

Taking those group figures which, in an insistent but confused way, presented themselves as psychology, economics, grammar, medicine, we asked on what kind of unity they could be based: were they simply a reconstruction after the event, based on particular works, successive theories, notions and themes some of which had been abandoned, others maintained by tradition, and again others fated to fall into oblivion only to be revived at a later date? Were they simply a series of linked enterprises?

We sought the unity of discourse in the objects themselves, in their distribution, in the interplay of their differences, in their proximity or distance - in short, in what is given to the speaking subject; and, in the end, we are sent back to a setting-up of relations that characterises discursive practice itself; and what we discover is neither a configuration, nor a form, but a group of rules that are immanent in a practice, and define it in its specificity. We also used, as a point of reference, a unity like psychopathology: if we had wanted to provide it with a date of birth and precise limits, it would no doubt have been necessary to discover when the word was first used, to what kind of analysis it could be applied, and how it achieved its separation from neurology on the one hand and psychology on the other. What has emerged is a unity of another type, which does not appear to have the same dates, or the same surface, or the same articulations, but which may take account of a group of objects for which the term psychopathology was merely a reflexive, secondary, classificatory rubric. Psychopathology finally emerged as a discipline in a constant state of renewal, subject to constant discoveries, criticisms, and corrected errors; the system of formation that we have defined remains stable. But let there be no misunderstanding: it is not the objects that remain constant, nor the domain that they form; it is not even their point of emergence or their mode of characterisation; but the relation between the surfaces on which they appear, on which they can be delimited, on which they can be analysed and specified.

In the descriptions for which I have attempted to provide a theory, there can be no question of interpreting discourse with a view to writing a history of the referent. In the example chosen, we are not trying to find out who was mad at a particular period, or in what his madness consisted, or whether his disturbances were identical with those known to us today. We are not asking ourselves whether witches were unrecognised and persecuted madmen and madwomen, or whether, at a different period, a mystical or aesthetic experience was not unduly medicalised. We are not trying to reconstitute what madness itself might be, in the form in which it first presented itself to some primitive, fundamental, deaf, scarcely articulated' experience, and in the form in which it was later organised (translated, deformed, travestied, perhaps even repressed) by discourses, and the oblique, often twisted play of their operations. Such a history of the referent is no doubt possible; and I have no wish at the outset to exclude any effort to uncover and free these 'prediscursive' experiences from the tyranny of the text. But what we are concerned with here is not to neutralise discourse, to make it the sign of something else, and to pierce through its density in order to reach what remains silently anterior to it, but on the contrary to maintain it in its consistency, to make it emerge in its own complexity. What, in short, we wish to do is to dispense with 'things'. To 'depresentify' them. To conjure up their rich, heavy, immediate plenitude, which we usually regard as the primitive law of a discourse that has become divorced from it through error, oblivion, illusion, ignorance, or the inertia of beliefs and traditions, or even the perhaps unconscious desire not to see and not to speak. To substitute for the enigmatic treasure of 'things' anterior to discourse, the regular formation of objects that emerge only in discourse. To define these objects without reference to the ground, the foundation of things, but by relating them to the body of rules that enable them to form as objects of a discourse and thus constitute the conditions of their historical appearance. To write a history of discursive objects that does not plunge them into the common depth of a primal soil, but deploys the nexus of regularities that govern their dispersion.

However, to suppress the stage of 'things themselves' is not necessarily to return to the linguistic analysis of meaning. When one describes the formation of the objects of a discourse, one tries to locate the relations that characterise a discursive practice, one determines neither a lexical organisation, nor the scansions of a semantic field: one does not question the meaning given at a particular period to such words as 'melancholia' or madness without delirium', nor the opposition of content between psychosis' and 'neurosis'. Not, I repeat, that such analyses are regarded as illegitimate or impossible; but they are not relevant when we are trying to discover, for example, how criminality could become an object of medical expertise, or sexual deviation a possible object of psychiatric discourse. The analysis of lexical contents defines either the elements of meaning at the disposal of speaking subjects in a given period, or the semantic structure that appears on the surface of a discourse that has already been spoken; it does not concern discursive practice as a place in which a tangled plurality - at once superposed and incomplete - of objects is formed and deformed, appears and disappears.

The sagacity of the commentators is not mistaken: from the kind of analysis that I have undertaken, words are as deliberately absent as things themselves; any description of a vocabulary is as lacking as any reference to the living plenitude of experience. We shall not return to the state anterior to discourse - in which nothing has yet been said, and in which things are only just beginning to emerge out of the grey light; and we shall not pass beyond discourse in order to rediscover the forms that it has created and left behind it; we shall remain, or try to remain, at the level of discourse itself. Since it is sometimes necessary to dot the 'i's of even the most obvious absences, I will say that in all these searches, in which I have still progressed so little, I would like to show that 'discourses', in the form in which they can be heard or read, are not, as one might expect, a mere intersection of things and words: an obscure web of things, and a manifest, visible, Coloured chain of words; I would like to show that discourse is not a slender surface of contact, or confrontation, between a reality and a language (langue), the intrication of a lexicon and an experience; I would like to show with precise examples that in analysing discourses themselves, one sees the loosening of the embrace, apparently so tight, of words and things, and the emergence of a group of rules proper to discursive practice. These rules define not the dumb existence of a reality, nor the canonical use of a vocabulary, but the ordering of objects. 'Words and things' is the entirely serious title of a problem; it is the ironic title of a work that modifies its own form, displaces its own data, and reveals, at the end of the day, a quite different task. A task that consists of not - of no longer treating discourses as groups of signs (signifying elements referring to contents or representations) but as practices that systematically form the objects of which they speak. Of course, discourses are composed of signs; but what they do is more than use these signs to designate things. It is this more that renders them irreducible to the language (langue) and to speech. it is this 'more' that we must reveal and describe.