Issue #19 January 2019

Some Notes on Foucault on Discourse

Albrecht Dürer — Unterweisung der Messung. 3. Buch. Von den Körpern

Foucault’s aim in The Archaeology of Knowledge (1969/2013) is to pin down a methodology for the studying of discourse. The question of what ‘discourse’ is for The Archaeology of Knowledge is one of the central problems of the book, and thus difficult to summarise. We can perhaps start, provisionally and simply, by saying that a discourse is a wide scale ‘conversation’ being carried out. For example, in recent years a feature of the political discourse in the west (the conversation about politics) has been the ‘return’ or ‘rise’ of extremist right wing nationalist movements. Analyses have been offered, touching on, or quilting together, a wide number of other discourses: economics, anthropology, media studies, history etc. This ‘conversation’ is carried out as much in the mass media, and houses of politics, as it is carried out in its own way between individuals in bars and lounge rooms or on message boards and twitter feeds. What is important is that we can demarcate a particular style, a set of observations, and a way of organizing the facts around this conversation, and thus can speak of ‘a discourse’.

What’s more, insofar as the identification of a phenomenon, emerging or otherwise, is done down the lines of this or that discourse, and the objects that they take to be their targets of inquiry, we need to refer to a discourse for the rules and methods of demarcating phenomenon in the first place. It is because ‘ideology’ is an object for a discourse that we can bring into clear view an ‘emerging’ ideological phenomenon, which can then be traded, or offered to, other discourses (economic, historical, etc).


Now, one way of approaching Foucault’s work is to see it as an analysis of historical discourses, with a specific aim to explain and trace the ‘law of rarity’ (Foucault, 2013, p.134). This phenomenon is only vaguely perceptible in our ‘conversations’ in the present (us being too close to them), but becomes starker as we move back into the past (or as we travel across cultural borders in the present). What becomes remarkable upon the examination of past discourses are the absences — both of objects or properties that seem to us as ‘self-evident’, as well as steps between ideas/premises that we take to be necessary. Thus, past discourse avails itself of strange ‘jumps’ that we feel are ‘obviously’ illegitimate, as well as paying no attention to solutions, paths, and objects that we feel are loudly and ‘self-evidently’ there.

Furthermore, insofar as the field of possible language is vast (though still smaller than the field of ‘possible vocal noises’), it far outstrips what presents itself in the discourses of any age (ours included), even combined. What we find in a discourse is a very modest set of claims, manoeuvres, and sequences of ideas that cling tightly together relative to the expanse of possibilities. We can thus ask, in the style of the Fermi Paradox, when we survey all of historical recorded knowledge, taking into account the incredibly wide limits of what is possible to say, to perceive, and to posit, “Where are all the ideas?”

It may seem attractive to say that ideas build upon ideas, and that certain breakthroughs in ‘science x’ needed to happen for breakthroughs in ‘science y’ to occur, but there are two responses to this:

Firstly, this observation is more based on the staggering complexities of 20th century physics and 21st century technology. The theory of evolution, for example, in its 19th century Darwinian formulation, did not depend on advances in mathematics or microscope technology, or biology, or so on. It was merely an interpretation of the observation (often made throughout history) of the fittedness of living things to each other and their environments. All of the pieces of the theory had been lying around for millennia. There’s nothing intrinsically absurd about the idea of the exact formulation of Darwin’s (hereditable characteristics, mutation, competition) appearing verbatim in Lucretius, or Heraclitus, or Lao Zi, or even the book of Genesis. Yet the fact remains that even though the elements of Darwin’s theory have been on hand for so long, and their connection so simple, their particular stating needed to wait so long. From this it would seem that it’s not the case that there exists an infinite freedom for thought to connect up observations and data points into myriad theories, but seemingly heavy limitations on how these connections are made. The structures of discourse.

Secondly, going in the other direction, this observation about the preconditions of certain ideas requiring other ideas reinforces the notion of rarity. From time to time claims arise that, say, modern particle physics was pre-empted by some ancient (usually exotic) text, the Upanishads for example. However, it is difficult to know what this means. If the text features a statement that can only be translated into the claim that a ‘125 GeV Higgs Boson should decay into a pair of tau leptons’, we would be right to claim forgery just on the grounds that it beggars belief that a 2,500 year old text would have distinct terms for a GeV scale (what did it need it for?), and make discriminations between bosons and leptons. Terms like ‘boson’ or ‘lepton’ have no meaning outside of modern particle physics — not just terminologically speaking, but even in regards to their referents. The Higgs Boson is an entity that emerges from a model, that is inseparable from the other elements and entities of that model (such as leptons), against which it is constrasted. It is then determined in relation (and in opposition) to these other entities in the model. An ancient text, then, can only pre-empt modern particle physics in a meaningful way by demonstrating a near full congruence with all of the entities, and their relations of, say, the Standard Model (of which the Higgs Boson is an element). That is to say that the claim that some ancient discourse pre-empts a modern one requires a more far reaching identity than just the similarity of some one element, mood, or theme. The atom of the Atomists is not the atom of the physicists.

Then, insofar as the entities of modern physics are dependent upon one another and their organization into a theoretical model that defines their relations, and this model is a precondition for the positing of these objects, we can see how the development of future knowledge is beholden to its present and past. But, then, knowledge forms of a kind of tree within a possibility space, like an evolutionary, ‘tree of life’ structure. So, the question returns: what rules and events and principles determined its evolution in just this way that it did, occupying these branches of the possibility space? In the case of species, we can look at the whole ‘tree of life’ and be satisfied with the explanation that it is blind historical contingency all the way through (extant species being just those forms that have not had their cycle of reproduction destroyed yet), but in the case of knowledge the temptation is to say that the forms our developing knowledge take, over time, trace the contours of something really there. But this assumes a liquidity and nimbleness to the institutions of knowledge which runs counter to fact. Historically, these institutions have moved by the slow, plodding along, processes of accumulation. We like to highlight exciting ‘paradigm shifts’ and births of entirely new disciplines, but these are the exception rather than the rule — the rarity of these being due to the fact that sciences possess their own conditions for the verification of statements made within them concerning the objects they have defined as belonging to them, and insofar as new statements concerning these objects satisfy these conditions then the science plods along in a steady trajectory, accumulating, but insofar as statements don’t satisfy these conditions, then they are excluded as false or senseless.

Albrecht Dürer — Unterweisung der Messung. Fol. B2r. Gleiche_Schraubenkurve

Grace (2014) characterizes one of Foucault’s aims as exposing “the way in which experiences are held “captive” by arbitrary distributions” (p.114). In this formulation, with the previous considerations, we can then characterize the ‘law of rarity’ as that which ‘captivates’ the movement and development of knowledge, and the experiences on which it is based, such that it traces such a thin set of branches in a vast possibility space.

Foucault is going to need a middle level of analysis to grasp what is occurring here. When we survey the possibility space (be that the sum of all possible things that could be said in a language, or the conditions of possibility of experience in general) we get no rule or indication as to why the development of knowledge should chart this particular thin set of lines just here. But when we limit ourselves to the empirical facts (the lines) we also get no indication of why it should be this way, we just discover, in a historical analysis of all of the extant texts, that it is this way.

Thus, our unit of analysis for discourses needs to differ from the broad ‘conditions of possibility’ in that they’ll be more immanent, more closely associated with the content of actually existing discourses (and precisely how these are carried out), yet at the same time it cannot become completely identical to these pieces of actually existing discourses (the historical corpus). This unit Foucault terms the ‘statement’, partially intending this to be taken in its common sense as ‘isolatable proposition’, but at the same time contrasting it endlessly with the propositions of logic, the formed or possible sentences of language, and speech acts as expressions of motives or other psychological determinants. The statement as fundamental unit of analysis needs to be:

“Too repeatable to be entirely identifiable with the spatio-temporal coordinates of its birth (it is more than the place and date of its appearance), too bound up with what surrounds it and supports it to be as free as a pure form (it is more than a law of construction governing a group of elements)” (Foucault, 2013, p.117).

So, for example, the statement “All men are mortal” forms a single proposition, and a single sentence of language, and may feature as an utterance in a few of the things that may be ‘done’ with speaking (speech acts). But, on the level of the statement the identity of the instantiations of this proposition/sentence are broken over a few unique statements. The only way of working this out is by investigating where this sentence occurs and what it makes possible relative to the preceding and succeeding statements in texts, as a kind of regularity; that is, its position in a discursive network. Ordinarily the sentence “all men are mortal” is followed by “Socrates is a man”, and so on, and then usually on to a number of other sentences discussing or introducing the form of the syllogism. In these cases the statement remains as constant as the sentence. One may even formulate a joke “All men are mortal, except Socrates” and the statement here remains the same for the first identical part of the sentence, insofar as the addition still enters into a definite relation with all of those introductions to the syllogism. But now imagine a text in which Socrates’ last days are dramatized, and the sentence is attributed to Socrates himself. In this text, the statement opens up new possibilities for further statements, and practically excludes the introduction to the syllogism that traditionally follows, insofar as our trajectory through the narrative thus far up to this point of the text has made it such that a sudden interjection from an introductory logic text book would create the sense of a radically different discourse intervening.

“The associated field is also made up of all the formulations whose subsequent possibility is determined by the statement, and which may follow the statement as its consequence, its natural successor, or its conversational retort” (Foucault, 2013, p.111).

In this firm connection to what precedes and succeeds it, the statements form together into a discursive formation that makes things like, for example, the novel possible as a demarcated, stylistic unit, with its own possibilities. Technically an author may follow any sentence by any other. I adjusted the menu on the table to be flush with the edge, she said. This is the ‘possibility space’ discussed before. But the fact is a) authors don’t, thus demonstrating they are deploying a narrower, more determined set of possibilities in the act of saying novel things, and b) even if an author did haphazardly construct a text in line with the total possibilities of language, disregarding the ‘law of rarity’ that circumscribes a discourse, it is not clear if such a text would succeed in saying anything at all. The cat wasn’t. Thus, discourses are constructed of statements that depend wholly on their progression, on what they do relative to other statements in a discourse, on what possibilities and paths they open up, and the analysis of discourse, then, is interested in bringing these functions to light through the evidence they leave of their functioning (sentences, inferences, speech acts).

Statements belong to a discursive formation the way sentences belong to a text (Foucault, 2013, p.131). There is a key difference though; in order for a collection of sentences to form a text, reference must be made to certain rules and conventions of language. These rules and regularities determine the construction of sentences, provide the grammatical structures in which meaningful word-atoms may be inserted, as well as regulate the combination of the sentences one after the other. But statements, forming a discursive formation, find their rule nowhere else but in their particular form of aggregation. That is their rules of combination are not ‘conditions of possibility’ but rather ‘rules of coexistence’. Furthermore, insofar as the statement differs from the sentences it makes possible, the statement should not be thought of as some unit of ‘meaning’ behind sentences, but the function of their combination. Statements, then, always exist in a formation, and if they are isolatable for analysis, it is only as a rule of transition. Statements are not ‘things’, but interrelated functions.

Albrecht Dürer — Unterweisung der Messung. Fol C5v. Parabel als Brennspiegel

Material Repeatability

If statements are going to prove themselves to be a useful level of analysis, they need to differentiate themselves from other levels, by not perfectly coinciding with them in all instances. This becomes clear if we think about repeatability. On one level, a sentence I write out twice, on two pieces of paper, are each unique. They are, after all, two distinct objects with their own spatio-temporal narratives. On another level, the two sentences are identical, insofar as I’ve just stipulated that the sentences are the same. On the level of highest universality they would be identical even if I translated one of them into code, or another language. These two contrasted levels see in the sentence on paper either an unrepeatable empirical material, or an infinitely repeatable transcendent meaning that is merely instantiated in this or that material. If the statement is as perfectly unrepeatable as the former, or as infinitely repeatable as the latter, it seems to not be a distinct unit. The challenge for Foucault, whose intuition is firmly that statements are a distinct unit in the analysis of discourse, is to show how and where the statement sits in between these two extremes regarding repeatability.

On the one hand he wants to say that all of the copies of a book, including its different editions, are composed of identical statements (it matters not for the analysis of discourse whether one looks at an original printing or a reprinting, a copy borrowed from the library, or a copy bought online), but on the other hand, on the level of the statement, repeatability must have a limit ‘tighter’ than an infinitely repeatable transcendent meaning — an edition of ‘Hamlet’ rendered into modern, vernacular English, then, would be home to some (or many) new statements. Thus, statements can’t be ambivalent to all of their material instantiations (lest they become identical to transcendent meanings), only some of them, but this ‘material repeatability’ requires its a priori principle, lest it becomes ad hoc (we simply say that a new statement has occurred when we feel like one has, or it is expedient to do so).

Foucault’s solution is to anchor the statement in ‘fields of use’ that conditions their repeatability (2013, p.117). Relative to a particular task in a particular domain, can the auxiliary statements inherent in that task persist unmodified in the face of a substitution of a repeated version of the statement under examination, or does it require them, too, to change? For example, if I take language itself as my object, and the English language in particular, to discuss the grammatical form of English sentences, then a statement used as an example and its translated repetition become distinct statements (ibid). If, on the other hand, I translate myself while speaking in a bi-lingual environment so my interlocuter may understand my previous statement better, then the statement is repeated in two different languages (provided my translation achieves my aims). This is not quite an a priori principle for the determination of whether a given statement is novel or a repetition, the level of granularity is ‘fuzzy’. For example, an identical text but in different editions may have different pagination, so statements of the type ‘on page 109…’ are modified by the different physical instantiations of the text in a discourse that discusses it, and we may in some instances want to ignore this discrepancy, and other times make it thematic. However, the key point here is that the limits on repeatability of a statement are measured by its effect on the proliferation of other statements in a discursive formation, rather than the brute, unrepeatable materiality, or an ideal form. Sometimes closer to the former, sometimes closer to the latter, all depending on the role of the statement relative to other statements in the formation.

So, to quickly summarize the above, statements are the basic elements of discourse. However, they are not to be confused with sentences or propositions simpliciter, but are, rather, functions organizing and placing limits upon the succession of sentences in a given discourse. What, then, becomes interesting in the analysis of discourse is how these discourses are structured by these statements in such a way to exclude some (many) possibilities and open up others. In this way, despite the apparent limitlessness of the possibilities in language and our imagination, we end up tracing the contours of the same basic conversations of our little moment and place in history.

John C. Brady is a student of philosophy and educator situated in Beijing. He gets most of his reading done in traffic jams. He is also a co-editor of this magazine, by way of full disclosure.

Works Cited

Foucault, M. (2002). The Archaeology of Knowledge. New York: Routledge Classics.

Grace, W. (2014), Foucault and Deleuze: Making a Difference with Nietzsche. Foucault studies, 17, pp. 99–116.

Webb, D. (2013). Foucault’s Archaeology: Science and Transformation. Edinburgh: Edinburgh University Press.


January 2019


The Continuity of Being: C.S. Peirce’s Philosophy of Synechism

by Brian Kemple

Against Ontology: A Naturalist Critique on Two Varieties of Mathematical Structuralism

by Jio Jeong

Some Notes on Foucault on Discourse

by John C. Brady

What is Vitalism?

by Timofei Gerber