Wittgenstein and Foucault: Two Philosophers of Practice
By Matthew Parish
Tuesday 13 October 2026
Ludwig Wittgenstein and Michel Foucault are not philosophers whom intellectual history naturally places beside one another. Wittgenstein belongs securely within the canon of twentieth-century analytic philosophy, even if he spent much of his career attempting to cure philosophers of some of the habits that analytic philosophy had acquired. Foucault belongs equally securely within the continental tradition, although he disliked labels and his work wandered freely across philosophy, history, psychiatry, criminology and political theory. Their prose styles could scarcely be more different. Wittgenstein wrote compressed remarks, questions and examples, while Foucault constructed sweeping historical narratives populated by prisons, hospitals, schools, soldiers, doctors, criminals and madmen.
Yet beneath these differences lies a striking philosophical affinity. Both thinkers were profoundly suspicious of the idea that philosophical problems are best understood by looking behind ordinary human practices for hidden essences. Both insisted, in radically different ways, that concepts acquire their significance through their use within organised forms of human activity. Both regarded philosophical abstraction as potentially misleading when detached from the practices within which words, classifications and rules actually operate.
The comparison should not be exaggerated. Wittgenstein and Foucault disagreed about too much for either to be comfortably assimilated to the other. Wittgenstein was not a theorist of political power and Foucault was not principally a philosopher of language. Wittgenstein’s philosophical temperament was therapeutic where Foucault’s was genealogical. Nevertheless, reading them together reveals something important about twentieth-century philosophy. Two traditions conventionally presented as intellectual opposites independently arrived at a related conclusion: if we want to understand concepts, we must examine what human beings actually do with them.
Meaning and use
The familiar slogan associated with the later Wittgenstein is that meaning is use. Like most philosophical slogans, it becomes misleading if treated as a complete theory. Wittgenstein’s point was methodological rather than merely definitional. When philosophers become confused about the meaning of a word, they are tempted to search for an object, essence or mental representation corresponding to it. Wittgenstein repeatedly redirects attention towards the circumstances in which the word is actually employed, because language is an activity before it becomes an object of philosophical theorising.
People promise, command, joke, pray, calculate, insult, describe, apologise, ask questions and tell stories. Words operate differently within these different practices and there is no reason to suppose that every meaningful expression performs the same logical function. Hence Wittgenstein’s celebrated notion of a “language-game”. Speaking a language is embedded within patterns of action governed by rules, expectations and shared forms of life. To understand an expression is not merely to associate it with an object. It is to know how to participate competently in the practices in which the expression has a role.
Foucault approaches concepts from an apparently different direction. He asks how societies came to speak about madness, criminality, illness, sexuality and normality in particular ways. His concern is not primarily with the semantics of individual words but with systems of classification. Yet he too refuses to detach concepts from practices. What does it mean, for example, to call somebody mentally ill? The answer cannot be exhausted by consulting a dictionary because the classification operates within medical and institutional practices.
Doctors diagnose, hospitals admit, governments regulate, insurance systems reimburse, courts determine capacity, families alter their behaviour and patients may come to understand themselves through diagnostic categories. The concept lives through these activities. Here Wittgenstein and Foucault converge upon a methodological principle that is deceptively simple but philosophically powerful: look at the practice before constructing the theory.
From language-games to institutional games
The comparison can be sharpened by extending Wittgenstein’s vocabulary. A language-game is embedded within a form of life, while Foucault asks what happens when such practices become institutionalised. Consider the word “criminal”. At one level, its use is governed by legal rules. A person convicted of an offence may appropriately be described as a criminal. Wittgenstein might ask us to examine the circumstances in which the expression is used rather than seek some metaphysical property of criminality.
Foucault asks a further series of questions. Who possesses the authority to apply the classification? What procedures determine whether it applies? What records are created once it has been applied? What consequences follow from the classification? How does the classified person subsequently understand himself? How do police, judges, prison officers, employers and other institutions respond to him? In this way the language-game becomes an institutional game.
This extension matters because classifications have asymmetric consequences. Calling somebody a bad chess player may wound his pride, but calling somebody a dangerous offender may deprive him of liberty. Calling somebody mentally incompetent may remove her legal authority to make decisions, while calling a child educationally deficient may alter the course of the child’s education. Some language-games therefore come attached to machinery, and Foucault’s distinctive contribution is to make that machinery philosophically visible.
Rules and rule-following
The later Wittgenstein’s discussion of rule-following provides another point of contact. No rule contains within itself a complete specification of every possible future application. A finite expression must somehow guide indefinitely many cases, yet another rule interpreting the first would itself require interpretation. We cannot escape into an infinite hierarchy of instructions explaining instructions. Eventually rules operate because human beings possess shared practices concerning what counts as following them correctly.
This insight has enormous institutional significance. Legal systems, bureaucracies and scientific disciplines are saturated with rules, yet written rules never completely determine their own application. Officials acquire habits of interpretation, professions develop conventions, courts accumulate precedents and organisations acquire cultures. What formally appears to be the application of an abstract rule therefore depends upon practices that cannot themselves be reduced to the words in which the rule is expressed.
Foucault’s histories repeatedly examine this territory from the opposite direction. He is interested in how institutional practices stabilise particular ways of classifying and treating people. The rule is never merely a sentence in a book. It becomes embodied in architecture, routines, professional expertise, records, examinations and surveillance. A school timetable is a rule embodied in the organisation of time, a prison cell is a rule embodied in architecture, a medical record is a classification embodied in bureaucracy and an examination is a standard embodied in a procedure. Foucault’s institutions might therefore be understood as vast mechanisms for making rules practically effective.
Forms of life and forms of power
Here the differences between the two philosophers become more important. Wittgenstein’s concept of a form of life is not ordinarily political. Shared human practices provide the background against which linguistic communication becomes possible. At some point justification comes to an end because participants simply act in particular ways. Foucault refuses to leave matters there. Whenever a practice has become established, he wants to know how it became established, who administers it and what effects it produces. What Wittgenstein sometimes treats as the background conditions of intelligibility, Foucault may treat as the outcome of history.
This creates a fascinating philosophical tension. Suppose a society shares a practice governing what counts as normal behaviour. Wittgensteinian analysis might investigate how the word “normal” functions within the relevant language-games. Foucault asks why precisely this practice of normality arose. Which institutions reinforced it? Which forms of expertise legitimised it? What happened to people classified as abnormal? Wittgenstein asks, metaphorically, how the game is played, while Foucault asks who built the board. That distinction captures much of what separates their philosophical temperaments.
The temptation of hidden essences
Both philosophers are nevertheless enemies of a particular philosophical temptation: the belief that beneath messy social practices there must lie some pure essence waiting to be discovered. Wittgenstein’s discussion of games is the obvious example. What do board games, ball games, card games and children’s games all possess in common? Perhaps nothing except overlapping similarities. The demand for a single essence may arise from philosophical expectations rather than from the structure of the phenomenon itself.
Foucault makes an analogous move historically. What is madness? What is sexuality? What is punishment? He suspects that philosophers and social scientists too readily assume that these categories identify stable objects that societies gradually came to understand more accurately. His genealogies instead investigate how the categories themselves changed. This does not entail that reality disappears into language. People suffered psychosis before psychiatry existed, human beings had sexual desires before modern sexology classified them and societies inflicted suffering upon offenders before modern criminology appeared.
The Foucauldian question concerns the conceptual organisation of those realities. This distinction is important because some of Foucault’s followers have converted a sophisticated historical observation into an implausible metaphysics. From the fact that classifications are socially constructed, it does not follow that the phenomena being classified are imaginary. Maps are constructed but mountains are not, and Wittgenstein can help rescue Foucault from this temptation. Different linguistic practices can organise our understanding of reality without creating reality by linguistic fiat.
Philosophy as therapy and genealogy
Their methods diverge most sharply in what they think philosophy should accomplish. Wittgenstein frequently describes philosophical problems as arising when language “goes on holiday”: words are removed from their ordinary uses and transformed into metaphysical puzzles. Philosophy should return them to their everyday employment. The resulting activity is therapeutic. The philosopher does not necessarily construct a better theory but instead dissolves the confusion that generated the demand for a theory.
Foucault is far less peaceful. His genealogical method does not merely clarify existing practices but destabilises them. By showing that institutions have histories, genealogy undermines the impression that their current form is necessary or inevitable. A society may imagine that prisons are the natural form of punishment, so Foucault asks how imprisonment became dominant. A society may imagine that psychiatric categories simply correspond to timeless divisions in nature, so Foucault examines the institutions through which those categories developed. A society may regard contemporary sexual identities as natural kinds, so Foucault investigates the historical emergence of the vocabulary through which sexuality became an object of scientific and administrative knowledge.
There is nevertheless a deeper structural similarity between the two projects. Wittgenstein wants to release us from philosophical confusion, while Foucault wants to release us from historical inevitability. Both strategies involve showing that something we considered compulsory was compulsory only because we had misunderstood the source of the compulsion. Wittgenstein exposes the grammatical habits that generate philosophical necessity, while Foucault exposes the historical processes that make contingent institutions appear natural.
The politics Wittgenstein leaves behind
The strongest Foucauldian criticism of Wittgenstein is that descriptions of linguistic practice may conceal asymmetries within the practice itself. Not everyone has equal authority to determine how words are used. A judge’s interpretation of “reasonable” matters more than a defendant’s, a psychiatrist’s classification of a patient may matter more than the patient’s classification of herself and a teacher’s judgement of competence may determine a child’s opportunities. A government’s definition of national security may justify secrecy or coercion. Language-games therefore take place within structures of authority.
This does not invalidate Wittgenstein’s philosophy of language. It reveals a dimension that was rarely his concern. Shared practices make meaning possible, but practices can simultaneously distribute power unequally among participants. Foucault asks us to notice who is authorised to speak and, just as importantly, whose classifications have institutional consequences.
This is particularly important in institutions organised around expertise. Modern societies delegate extraordinary authority to specialists. Doctors determine illness, psychiatrists determine pathology, economists advise governments, lawyers classify conduct according to legal categories and intelligence agencies determine which information constitutes a threat. Increasingly, algorithms participate in these classificatory processes. Expertise is indispensable to complex societies, yet expertise is also a form of power because classifications produce consequences. Foucault’s achievement was to insist that epistemology and institutional authority cannot always be studied separately.
The Wittgensteinian criticism of Foucault
The criticism can also run in the opposite direction. Foucault sometimes writes as though the discovery that a practice involves power has exposed something hidden and sinister about it. A Wittgensteinian might respond that this is another philosophical temptation generated by abstraction. Before drawing conclusions from the proposition that a relationship involves power, we should ask what “power” means in the particular practice under examination.
A parent preventing a child from walking into traffic exercises power. A conductor directing an orchestra exercises power. A court enforcing a contract exercises power. A university requiring students to pass examinations exercises power. A surgeon anaesthetising a patient exercises extraordinary physical power over another human being. Nothing follows morally from the abstract description alone. We must return to the particular practice and ask what the participants are doing, what rules govern them, what purposes the activity serves, what alternatives exist and what counts within the practice as abuse.
The Wittgensteinian demand for attention to particulars therefore provides an antidote to one of Foucault’s weaknesses. “Power” can become a metaphysical abstraction every bit as misleading as the abstractions Wittgenstein attacked. Foucault is strongest when he describes mechanisms precisely and weakest when power becomes an all-purpose explanatory substance.
Artificial intelligence and rule-following
The comparison between Wittgenstein and Foucault becomes especially interesting in the age of artificial intelligence. Large language models participate in linguistic practices without participating in human forms of life in the ordinary biological sense. They learn statistical regularities from immense bodies of human-produced language and can reproduce patterns of rule-following with astonishing competence. This creates a recognisably Wittgensteinian puzzle.
If correct linguistic behaviour consists substantially in competent participation in practices, what should we say about a machine that participates successfully? Must understanding require some private mental accompaniment or does the ability to navigate language-games provide at least part of what we mean by understanding? Wittgenstein offers no simple answer because contemporary artificial intelligence lay far beyond his technological horizon. Yet his attack upon private and purely mentalistic theories of meaning makes the question unusually uncomfortable for anyone who insists that machines obviously cannot understand language.
Foucault introduces a different question: who controls the language-games in which artificial intelligence participates? Large language models do not merely reproduce language. They increasingly mediate it. They summarise documents, answer questions, recommend actions, rank information, draft correspondence and interpret complex bodies of knowledge. Their classifications may influence employment, finance, medicine, education, policing and government. The machinery of discourse is therefore becoming partly computational.
The two philosophers consequently direct our attention towards complementary problems. A Wittgensteinian asks whether the machine follows the rule and what it would mean for us to say that it does. A Foucauldian asks who designed the system, what institutional interests surround it and what consequences follow when its classifications acquire authority. In a world increasingly organised through artificial intelligence, both questions matter.
Algorithmic forms of life
The synthesis becomes still more interesting when artificial intelligence systems begin interacting primarily with one another. Human language-games historically presupposed communities of human participants, but autonomous systems may develop protocols, conventions and patterns of interaction optimised for machine-to-machine communication. Financial algorithms already transact at speeds incomprehensible to human traders, while software agents increasingly exchange information and coordinate actions. We may therefore eventually confront something at least metaphorically resembling artificial forms of life: stable systems of interaction in which rules acquire significance from repeated computational practice.
Here Wittgenstein and Foucault point towards different dimensions of the same future. Wittgenstein helps us ask how meaning might emerge from rule-governed interaction without requiring every meaningful operation to be explicitly represented in advance. Foucault helps us ask how such systems acquire authority over human beings once embedded within institutions. An algorithmic classification may eventually matter not because any human understands or endorses it but because other systems respond to it.
A credit model might assign a risk category, another system might adjust an interest rate, another might limit a transaction and yet another might modify an insurance premium. No individual human need have intended the complete chain of consequences. Power in such circumstances becomes procedural rather than personal. That prospect would surely have fascinated Foucault because it represents an extreme version of his insight that social power need not be reducible to the intentions of a sovereign individual.
Two paths away from metaphysics
There is ultimately a deeper reason to read Wittgenstein and Foucault together. Both represent attempts to bring philosophy back into the world. Traditional philosophy repeatedly searches behind appearances for foundations: essences, representations, transcendental structures, ideal languages or universal theories. Wittgenstein and Foucault travel in the opposite direction. Wittgenstein brings philosophy back towards ordinary practices, while Foucault brings philosophy back towards historical institutions.
One asks us to examine what people do with words and the other asks us to examine what institutions do with people once those words have become classifications. The combination is powerful because each corrects the other’s characteristic weakness. Wittgenstein supplies conceptual restraint to Foucault: not every pattern is domination, not every classification conceals an ideology and not every social practice requires genealogical suspicion. Foucault supplies political depth to Wittgenstein: practices do not simply exist but have histories, distribute authority, create winners and losers and remain capable of change.
Between them emerges a philosophy of practice more compelling than either thinker quite developed independently. Human beings inhabit worlds of rules and classifications, but these are simultaneously linguistic and institutional phenomena. We cannot fully understand how a classification works without understanding its use, yet neither can we fully understand its significance without asking who has the authority to apply it and what happens when they do.
After the analytic-continental divide
The most interesting consequence may concern philosophy itself. The analytic-continental division now looks increasingly like an artefact of twentieth-century academic history. Contemporary problems do not respect it. Artificial intelligence, algorithmic governance, digital surveillance, biotechnology, online identity and automated decision-making simultaneously raise questions about language, knowledge, classification, agency, institutions and power. No single philosophical tradition possesses all the necessary tools.
Analytic philosophy offers extraordinary resources for conceptual precision, logic, epistemology and the philosophy of language. Foucault’s tradition offers extraordinary sensitivity to history, institutions, classification and the political consequences of knowledge. Wittgenstein and Foucault provide an unusually productive bridge because neither ultimately believes that philosophy can understand concepts by examining them in isolation from practice.
There is a lesson here for philosophers themselves. The interesting question about a concept is not merely what it means. We should also ask how people use it, what rules govern its application, which institutions administer those rules, who possesses authority to apply them and what happens to human beings once the classification has been made. Wittgenstein teaches us to look at the game, while Foucault teaches us to look at the institution surrounding the game.
In the twenty-first century there is one further complication neither man could have anticipated. Machines are increasingly becoming participants in our games, interpreters of their rules and arbiters of classifications that affect human lives. They may eventually become designers of some of the games themselves. Philosophy will therefore need Wittgenstein’s attention to rules, meaning and practice alongside Foucault’s attention to institutions, knowledge and power. The old disciplinary border between Oxford and Paris looks rather less important when both traditions are staring at the same machine.
