Episteme

The riverbed beneath our debates: the historical system that decides what an era can think, say and hold to be true. When a machine produces plausible texts at will, this riverbed may be shifting — and naming it allows us to arbitrate without dogma.

Origin: The Order of Things: An Archaeology of the Human Sciences (Les Mots et les Choses. Une archéologie des sciences humaines), 1966

An ordinary scene from our daily life: you are preparing a talk and you need to check a fact. Twenty-five years ago, you would have opened an encyclopaedia; ten years ago, typed three words into a search engine; today, one part of the room puts the question to a chatbot — and the other part is outraged by it. The most unsettling thing is not knowing who is right. It is that we no longer, collectively, grant the same trust to the same sources: what counts as “reliable knowledge” is shifting before our eyes. To think this kind of shift, there is a concept cut to measure.

The riverbed

Imagine the ideas of an era as the water of a river: noisy, restless, ever-changing. Beneath this water there is a bed — silent, shaped to a wholly different rhythm — that decides where it can flow. It is this riverbed that Michel Foucault named the episteme: the historical, largely unconscious system that defines what an era can think, say and hold to be true. Our most widely shared certainties are not the bottom of things: they have a riverbed — and therefore a date of manufacture, and therefore the possibility, one day, of shifting.

Two distinctions are enough to handle the concept without confusing it with its neighbours. The episteme is not epistemology, that reflection which asks what makes a piece of knowledge valid: the episteme does not judge validity, it describes what an era can so much as conceive of knowing. Nor is it ontology, which bears on what exists. A doctrine, one defends; an episteme, one inhabits — most often without knowing it.

Three riverbeds, then ours

Foucault described three of them for Western thought. In the Renaissance, to know is to decipher resemblances: if the walnut resembles the brain, it must therefore heal the head — perfectly serious reasoning at the time. In the classical age, to know is to order: to classify beings into tables, to measure, to set in order. In the modern age, knowledge becomes historicised — life, labour and language acquire a temporal depth, and “Man” becomes at once an object of science and the subject who produces it. One and the same thing — a plant, a coin, a text — is not better known from one age to the next: it is known otherwise. Between the ages, no continuous progress: the river has changed its bed.

Now return to your meeting room. Around the table, one person will hold as established only what a controlled study demonstrates; another brandishes a benchmark; a third cites what a chatbot answered them; the fourth counters with twenty years on the ground. Four ideas of proof in the same room. If the discussion gets bogged down, it is not that one person is rigorous and the others careless: it is that several riverbeds coexist — and that no one sees them, since each flows in their own.

The objection deserves to be heard in its full force. Historians have reproached Foucault for ruptures that are too clean-cut: real eras are more mingled than his three ages. And there is something more unsettling still: if all thought flows in a riverbed, so does Foucault's — from which riverbed does he claim to describe the riverbeds?

The concept survives these criticisms if it is taken for what it is: an instrument, not a dogma. It requires neither that ruptures be surgical, nor that one survey history from nowhere; it asks only that we admit that the thinkable has a history. And the tool cuts both ways, including towards the one who wields it — that is precisely its integrity.

One further word of honesty. The episteme is a dated concept, including for its own author: forged in his “archaeological” period of the 1960s, it was later largely set aside in favour of a more supple notion, the “apparatus” (dispositif). To speak today of a “digital episteme”, or of a riverbed that would shift under the effect of AI, is therefore to extend the concept well beyond the three ages it described. This extension may be fertile — but it is our reading, not a thesis of Foucault's.

What this helps us think

When machines produce plausible texts at will, a question becomes visible again, one that lay dormant beneath the self-evident: who decides the true, from now on, and how? Many of our debates about AI then cease to be technical debates: they become the symptoms of a riverbed that is, perhaps, shifting. To name this shift does not settle it — but it allows us to arbitrate without dogma: neither “adopting” by reflex, nor refusing on principle, but asking first which idea of knowledge each camp takes for granted.

And the riverbed is not the lowest storey. Further down there is the bedrock: what exists, the way a civilisation carves the real into beings. It is the ontology entry that digs to that depth.

The next time a source strikes you as obviously reliable — or obviously suspect — grant yourself a second of geology: which riverbed traced that self-evidence, and how long has the water been flowing there?

What this concept is not

It is not epistemology. Epistemology is a theory of the validity and methods of knowledge; the episteme designates the historical conditions that make a piece of knowledge possible and thinkable in a given era. The one evaluates, the other describes the ground on which one evaluates.
It is not a paradigm in Kuhn's sense. The Kuhnian paradigm is specific to a discipline and known to its practitioners; the episteme runs through all the knowledge of an era and remains, for the most part, invisible to those who think through it.
It is not an ideology or a “spirit of the age”. It is not a set of shared opinions or values, but the impersonal structure that delimits what can be said and known — beneath conscious beliefs.

Examples

The passage from natural history to biology

Foucault shows how the mutation of the classical episteme into the modern episteme transforms natural history (the classification of visible resemblances) into biology (the search for hidden organic functions and structures) — one and the same object of study radically changes its status as knowledge.

Velázquez's Las Meninas, the opening of The Order of Things

Foucault opens the book with an analysis of the painting Las Meninas, which he reads as a staging of classical representation itself — a privileged illustration of how the classical episteme functions.

The digital as a “new episteme” (a contemporary extension)

Intellectuals after Foucault (notably Michel Serres, Milad Doueihi, Jean-Guy Meunier and Bruno Bachimont) describe the digital revolution as a new dominant episteme since the end of the twentieth century. This usage is an interpretive EXTENSION of the concept beyond the three epistemes identified by Foucault himself, and should not be attributed to Foucault as such.

Other perspectives

The trans-epistemic objection (Merquior): fundamental ideas and principles run through several eras, which casts doubt on the thesis of a total rupture between successive epistemes.
The objection of epistemic lag: theories from the preceding episteme persist into the following one.
The objection of the return of earlier ideas, incompatible with the notion of a radical and definitive rupture.
The internal heterogeneity documented by historians of science within periods that Foucault presents as relatively homogeneous.
Foucault's refusal to give a causal explanation of the passage from one episteme to another, judged unsatisfactory by traditional historians.
A three-term distinction never to be flattened: episteme (Foucault, the historical conditions of what can be known and said) ≠ epistemology (the philosophical theory of the validity and methods of knowledge) ≠ ontology (the categories of what exists, the structure of the real). These three notions overlap lexically but operate on distinct planes — and to confuse them is a common error.
The status of the concept after 1969: did Foucault abandon it in favour of the apparatus (dispositif), or did he subsume it as a particular discursive case of the apparatus? Commentators (including Judith Revel) do not settle the matter uniformly.
The legitimacy of broadened contemporary uses of the episteme (the “digital episteme” in Michel Serres, Milad Doueihi and others): an owned and fertile extension of the concept, or a betrayal of Foucault's strictly historical archaeological gesture?

The concept in detail

The Renaissance episteme

The age of resemblance and similitude (sixteenth century): knowledge proceeds by analogy, and language is an autonomous reality in direct, quasi-magical grip on the world.

The classical episteme

The age of representation (seventeenth century to Kant): to think is to represent the object through ideas; knowledge is organised into tables, taxonomies, orders and categories of identity/difference (mathesis, taxinomia).

The modern episteme

The post-Kantian age (nineteenth century to the present day, for Foucault writing in 1966): the emergence of “Man” as an empirico-transcendental doublet, the historicisation of knowledge, the forceful return of language as an object in its own right — giving rise to modern biology, political economy and philology/linguistics.

Rupture and discontinuity (the archaeological method)

The passage from one episteme to another is conceived as a structural rupture, not as cumulative progress; the archaeology of knowledge describes discursive transformations without providing a causal explanation of them.

Unconscious and non-subjective status

The episteme operates beneath the consciousness of speaking subjects: it is formulated by no one, yet it makes possible the statements, theories and practices of a period — in this it differs from a doctrine or an ideology knowingly defended.

Explicit distinction from the Weltanschauung (Dilthey)

Foucault refuses to have the episteme assimilated to a conscious, shared world-view; it is a discursive structure, not a subjective or cultural conception of the world.