Foucault for Analytic Philosophers
Matthew Parish
Tuesday 15 September 2026
Michel Foucault is one of those philosophers whom analytic philosophers are often taught, implicitly if not explicitly, not to read. This is an unfortunate state of affairs. The reasons have less to do with the merits of Foucault’s work than with the peculiar intellectual division that developed in twentieth-century philosophy between the so-called analytic and continental traditions. Foucault wrote in a style unfamiliar to philosophers trained upon Frege, Russell, Wittgenstein, Quine or Kripke. He seldom stated propositions and then defended them through explicit sequences of argument. He moved promiscuously between philosophy, history, sociology, psychiatry, criminology and political theory. He coined expressions whose meanings shifted between books and he was suspicious of precisely the kind of conceptual architecture that analytic philosophers traditionally prize.
Nevertheless, there is a perfectly intelligible Foucault hiding underneath the unfamiliar vocabulary. Once his principal claims are separated from some of his historical rhetoric, a surprising amount of his philosophy can be reconstructed as a collection of relatively precise hypotheses about institutions, concepts, knowledge and human behaviour. Some of those hypotheses are false, some are exaggerated and others are extraordinarily illuminating. The mistake is to imagine that because Foucault does not argue like an analytic philosopher, there is nothing there for an analytic philosopher to analyse. Indeed, Foucault’s philosophical project can usefully be understood as asking what apparently necessary features of human institutions and categories might instead be historically contingent.
The most useful place to begin is with a distinction that analytic philosophers will immediately recognise: the difference between a thing and the criteria by which we classify something as that thing. Foucault was fascinated by classifications. What counts as madness? What counts as illness? What counts as criminality? What counts as sexual abnormality? What counts as knowledge? These questions might superficially resemble ordinary problems of conceptual analysis. Foucault’s distinctive move was to insist that the criteria by which societies answer them are neither timeless nor politically innocent.
Consider madness. There are plainly human beings suffering from conditions that contemporary medicine classifies as schizophrenia, bipolar disorder or severe depression. Foucault’s more interesting claim is not that suffering of this kind is imaginary but that the institutional category of “the mad” cannot simply be read off a collection of biological facts. A society develops practices determining which kinds of unusual behaviour require treatment, tolerance, supervision, confinement or exclusion. Those practices change over time and, once institutions embody them, classifications acquire consequences for the people to whom they are applied.
An analytic philosopher might reconstruct the point in deliberately austere terms. Suppose a society possesses a classificatory predicate M(x), meaning that x is mentally disordered. The extension of M depends partly upon facts about individuals but also upon rules determining which facts count as relevant to the classification. Those rules affect what happens to people falling within the predicate’s extension, while the resulting institutional practices may in turn affect how the rules are understood. Foucault’s histories are, among other things, investigations into how predicates of this kind acquire their institutional extensions and social consequences. Put in those terms, the allegedly mysterious Foucault begins to look considerably less mysterious.
Power without a sovereign
The concept for which Foucault is most famous is power, and here analytic readers are particularly liable to misunderstand him. An ordinary conception of power is agent-centred. A has power over B if A can cause B to do something that B would otherwise not do. Governments possess power because they can imprison people. Employers possess power because they can dismiss employees. Generals possess power because they can issue orders to soldiers. Foucault thought this model captured an important phenomenon but only part of the phenomenon.
Imagine a prison in which every prisoner knows that a guard might be watching at any moment but cannot tell when observation is actually taking place. Jeremy Bentham’s Panopticon supplied Foucault with his famous example. The ingenious feature of such an institution is that continuous coercion becomes unnecessary. Prisoners regulate their own behaviour because they know that observation is possible. Foucault’s wider account of disciplinary power similarly emphasises observation, normalising judgement and examination as techniques that can shape behaviour without requiring the constant intervention of a sovereign commander.
The important philosophical point is therefore not really about prisons. It concerns the possibility of decentralised power. A system of rules, expectations, measurements and surveillance may produce highly regular behaviour even though no particular individual intends the resulting pattern. In contemporary terminology, we might describe some such patterns as emergent properties of institutional systems. There is nothing especially alien to analytic philosophy about that proposition. Economics routinely analyses outcomes that nobody individually intended, game theory explains equilibria arising from strategic interaction without requiring a central planner and philosophy of language recognises conventions that no individual speaker created. Foucault’s account of power can often be understood in analogous terms: institutions create incentive structures, informational asymmetries and norms that constrain behaviour without requiring a sovereign continuously issuing commands.
His stronger and more interesting claim is that individuals eventually participate in these structures themselves. The prisoner watches himself, the employee evaluates his own productivity, the pupil compares herself with educational standards and the patient learns to describe himself through medical categories. Power becomes more efficient precisely when its subjects participate in its exercise. This is part of what Foucault means when he treats power as productive rather than merely repressive. Institutions do not simply prevent people from doing things. They create categories, expectations, identities and forms of behaviour, thereby helping to produce the kinds of persons upon whom they subsequently operate.
Knowledge and power
Foucault’s expression “power/knowledge” has caused enormous confusion because it is easily interpreted as the absurd proposition that truth is simply whatever powerful people say it is. If that were his argument, analytic philosophers would be right to dismiss it. Whether water consists of hydrogen and oxygen does not depend upon the political influence of chemists. The interesting Foucauldian thesis is subtler and considerably more plausible.
Knowledge in organised societies requires institutions. Scientific research requires universities, laboratories, journals, funding bodies, professional qualifications and procedures for deciding who counts as an expert. Medicine requires hospitals, licensing systems, diagnostic conventions and systems of record-keeping. Law requires courts, legislatures, lawyers and archives. These institutions influence which questions are investigated, which observations are recorded, which classifications become authoritative and which people acquire the standing to make recognised knowledge claims. Foucault’s analysis of the examination is particularly revealing because examination simultaneously produces information about individuals and subjects them to standards by which their conduct can be assessed.
None of this requires relativism about truth. Suppose proposition p is true independently of whether anyone believes it. There remains a separate question: under what institutional circumstances will p be discovered, accepted, disseminated and acted upon? Foucault is primarily interested in the second question. A society’s mechanisms for producing accepted knowledge are inevitably intertwined with its institutions of authority. Truth and socially authorised knowledge therefore need not be identical. The former may be objective while the mechanisms determining recognition of the latter remain thoroughly social.
Contemporary artificial intelligence makes the distinction unusually vivid. A large language model may generate a true proposition, but whether human beings trust that proposition depends upon systems of verification, reputation and authority. Conversely, institutions may continue to certify propositions that are false. Foucault’s question is therefore not merely “what is true?” but “how does a society organise the production and recognition of statements that it treats as true?” These are different questions and analytic philosophers ought to be comfortable recognising the legitimacy of both.
Genealogy as philosophical method
Foucault’s historical method is usually called genealogy, a term inherited from Nietzsche. Analytic philosophers may reasonably wonder what genealogy contributes to philosophy. If we want to know what justice means, why should it matter how prisons operated in eighteenth-century France? Foucault’s answer is that genealogy challenges claims of necessity by demonstrating contingency. His genealogical method seeks to show that systems of thought which appear inevitable are products of contingent historical developments rather than the conclusions of an inexorable process of rational progress.
Suppose somebody claims that institution I must have characteristic C because C follows from the essential nature of I. One way of challenging the claim is conceptual: construct a possible institution possessing the relevant identity conditions without C. Another is historical: demonstrate that institutions recognisably belonging to the same family previously existed without C. Genealogy can therefore perform something resembling the work of a counterexample, although its evidential materials are historical rather than purely conceptual.
Foucault repeatedly takes institutions that appear natural and demonstrates that they are historically peculiar. The modern prison, psychiatric hospital, clinic and apparatus of sexual classification did not simply emerge because humanity gradually discovered the uniquely rational way of organising these matters. They arose through complicated historical processes involving administrative convenience, economic change, scientific development, moral fashions and political conflict. History thereby becomes a weapon against conceptual complacency.
This does not prove that contemporary institutions are wrong. The genetic fallacy remains a fallacy and showing that an institution arose for dubious reasons does not establish that there are no good reasons for retaining it. Foucault sometimes invites readings that slide dangerously close to this inference. Nevertheless, genealogy can perform an important philosophical function by revealing that arrangements we mistake for necessities may actually be contingent. Once contingency has been established, the genuinely difficult normative argument can begin.
The missing normative theory
Here we reach one of the strongest analytic objections to Foucault. If institutions are saturated with power, why should we resist one form of power rather than another? If psychiatric classifications exercise power, does that make them objectionable? If prisons discipline people, is discipline necessarily bad? If educational institutions normalise behaviour, should they stop doing so? Foucault frequently diagnoses power relations while refusing to provide a systematic account of the values according to which they should be judged.
This is a serious problem because power cannot itself be the criterion of injustice. Virtually every stable social institution involves power. A traffic light constrains behaviour. A university examination establishes standards and ranks people according to them. A criminal court imposes classifications with enormous consequences. Parents exercise power over children and language itself constrains the categories through which we communicate. If all these things instantiate power, then discovering power tells us remarkably little about whether something is wrong.
Foucault’s defenders may respond that he deliberately refused universal normative foundations because those foundations can themselves become mechanisms of domination. Yet the objection simply reappears one level higher. Why should domination concern us? Why prefer resistance to submission or plurality to uniformity? At some point evaluative premises are unavoidable. Analytic philosophy can therefore improve Foucault by supplying what he often declines to provide: distinctions between legitimate and illegitimate exercises of power. Consent, welfare, autonomy, equality, rights and procedural fairness are obvious candidates for doing this work. One need not accept any particular theory of these values to recognise that something beyond the mere identification of power relations is required.
Foucault and ordinary language
There is another unexpected bridge between Foucault and analytic philosophy. Both traditions, at their best, distrust philosophical abstractions detached from actual practices. The later Wittgenstein famously redirected philosophical attention towards the uses of words, language-games and the forms of life within which linguistic activities acquire their significance. Foucault asks us to examine the practices through which concepts acquire institutional lives. Their methods and ambitions are radically different but there is a family resemblance: meaning and classification do not float free from practice.
Take the concept of the criminal. An abstract philosophical definition may specify necessary and sufficient conditions for criminal liability. Foucault asks what happens after the concept enters the world. Police identify suspects, courts create records, prisons classify inmates, criminologists construct statistical categories, employers ask about convictions and governments calculate recidivism rates. Individuals may eventually begin to understand themselves as offenders or ex-offenders. The concept has become embedded in a network of practices.
This is a useful corrective to a certain temptation in analytic philosophy: the assumption that once we have clarified the logical structure of a concept, the philosophical work is complete. Foucault reminds us that concepts have biographies as well as definitions. A concept may possess a perfectly respectable semantic analysis while its institutional application generates consequences that cannot be discovered through semantic analysis alone.
The algorithmic Panopticon
Foucault may actually be easier to understand in the twenty-first century than he was when Discipline and Punishappeared in 1975. The digital world has constructed systems of observation far exceeding Bentham’s imagination. Smartphones can record locations, platforms record preferences, financial institutions model risk, employers measure performance and algorithms classify individuals into categories that may influence which advertisements they see, which transactions are flagged and which opportunities are presented to them.
No single tyrant needs to control this machinery and nobody may completely understand its aggregate operation. Millions of people create data, thousands of organisations process it and automated systems generate classifications whose effects emerge without central direction. Power becomes distributed across networks of software, institutions, commercial incentives and human decisions. This is profoundly Foucauldian because the important question is not whether a particular technology executive or government minister personally intends to control a particular individual. The interesting question is how architectures of observation alter behaviour.
People modify what they write because digital statements may be permanent. Employees alter their conduct because performance is measured. Consumers modify choices because platforms structure the options placed before them. Institutions modify decisions because algorithms assign scores. The Panopticon no longer requires a guard in the tower because the tower can be software. Foucault’s account of disciplinary power, in which observation and normalisation operate through dispersed institutional mechanisms rather than a single sovereign centre, lends itself remarkably well to this technological environment.
Artificial intelligence pushes the argument further. Traditional bureaucracies classified people according to relatively crude categories because human administrators had limited processing capacity. Machine learning permits classifications of extraordinary granularity. Every individual can potentially become a category of one. Foucault’s concept of normalisation therefore acquires an unexpected new dimension. Industrial societies compared individuals with populations: the normal pupil, normal patient, normal worker or normal citizen. Algorithmic societies can compare individuals against dynamically generated predictive cohorts. The norm need no longer be a simple average. It can become a continuously recalculated probability distribution.
Foucault could scarcely have imagined the computational machinery, but he would have recognised the institutional logic. Information about human beings is accumulated, classifications are generated from that information and those classifications are then used to shape the circumstances in which human beings act. Knowledge and administration become progressively harder to separate.
Where Foucault goes wrong
None of this means analytic philosophers should become Foucauldians. Foucault’s historical claims require scrutiny by historians and many have been contested. His terminology is frequently unstable. He sometimes turns illuminating tendencies into sweeping generalisations. His suspicion of universal explanation can itself become suspiciously universal. Above all, his reluctance to distinguish systematically between benign and malign power leaves his political philosophy radically incomplete.
There is also a persistent danger of what might be called the hermeneutics of omnipotence. Once every institution is interpreted as a manifestation of power, every piece of contrary evidence can itself be redescribed as another manifestation of power. A patient who accepts a diagnosis demonstrates medical power because she has internalised it, while a patient who rejects the diagnosis may be described as demonstrating the same power because she is resisting it. Agreement and disagreement can thereby appear to confirm the same theory. Analytic philosophers are right to be suspicious of explanatory structures that threaten to become unfalsifiable.
The remedy, however, is not to discard Foucault but to disaggregate him. His work contains historical claims, causal hypotheses, conceptual proposals and political evaluations. These should be separated and tested independently. Some will survive and others will not. That is not a hostile way to read Foucault. It is what philosophers normally do with philosophers.
An analytic Foucault
There is therefore a productive way for analytic philosophers to read Foucault without adopting either his vocabulary or his intellectual temperament. Translate “power” into specific mechanisms of incentives, constraints, information and institutional authority. Translate “discourse” into socially organised practices governing the production, classification and acceptance of statements. Translate “normalisation” into processes through which populations are measured against standards and individuals modify behaviour in response. Translate “genealogy” into historical investigation designed to test claims of conceptual or institutional necessity. Translate “power/knowledge” into hypotheses concerning the interaction between epistemic institutions and structures of authority.
Once this translation has been performed, Foucault becomes available for ordinary philosophical criticism. We can ask whether a proposition is true, what evidence supports it, what would falsify it, whether the conclusion follows from the premises, whether a descriptive claim has been confused with a normative one and whether alternative causal explanations are available. We can ask whether historical evidence genuinely establishes contingency or merely demonstrates variation. Foucault need not be protected from these questions. Asking them is the most respectful way to read him.
There is also a benefit running in the opposite direction. Foucault forces analytic philosophers to notice that concepts, once clarified, do not remain peacefully inside philosophical journals. They enter bureaucracies, hospitals, courts, schools, prisons, corporations and computer systems. There they become criteria for decisions about actual people. Analytic precision is indispensable to understanding those concepts but institutional analysis is indispensable to understanding what happens when they are used.
Two philosophical cultures
The larger lesson concerns the strange division of modern Western philosophy into analytic and continental camps. At its worst, each tradition developed caricatures of the other. Continental philosophers portrayed analytic philosophy as sterile manipulation of symbols, indifferent to history, politics and human suffering. Analytic philosophers portrayed continental philosophy as literary obscurantism in which metaphor substitutes for argument. Both caricatures contain enough truth to sustain the prejudice and nowhere near enough truth to justify it.
Foucault presents an unusually good test case because his weaknesses are precisely those analytic philosophy is well equipped to expose, while his strengths concern subjects analytic philosophy has sometimes neglected. He is exceptionally good at noticing that philosophical concepts become institutional machinery. He is less good at specifying exactly when the operation of that machinery is justified. The traditions therefore have something to offer one another.
An analytic reconstruction of Foucault strips away some unnecessary mystification, but Foucault also poses a challenge to analytic philosophy. Philosophers can spend enormous energy determining the necessary and sufficient conditions governing a concept while overlooking the institutions that decide how that concept is applied to actual human beings. Yet the latter question may have vastly greater consequences. A perfect analysis of insanity tells us remarkably little about what happens when somebody is taken through the doors of a psychiatric hospital. A perfect analysis of punishment tells us remarkably little about what imprisonment does to a human personality. A perfect theory of knowledge tells us remarkably little about which institutions society chooses to believe.
Foucault’s enduring insight is that philosophy does not end when concepts have been clarified. Concepts enter institutions, institutions organise people and people learn to understand themselves through the categories institutions provide. For analytic philosophers, this proposition should not sound outrageous or even particularly continental. It should sound like the beginning of an argument.
