What Is Computational Theory of Mind?

Direct Answer

The computational theory of mind says that thinking is a kind of computing. When you recognize a face, solve a puzzle, or decide what to have for lunch, your brain is performing computations — manipulating representations according to formal rules. The mind, on this view, is to the brain what software is to hardware: a pattern of information processing that can be described independently of the physical stuff that implements it.

This is a claim about what minds are, not just how we should study them. It's stronger than the methodological suggestion that cognitive science can usefully borrow concepts from computer science. CTM says that mental processes literally are computations. The symbols your mind manipulates when you think "the cat is on the mat" have syntactic structure — they combine and transform according to rules — and those transformations are the thinking. The meaning of the thoughts (semantics) rides on top of the formal structure (syntax), the way the meaning of a sentence rides on top of its grammar.

Hilary Putnam introduced this framework in a series of papers in the 1960s, drawing on the mathematical theory of Turing machines. Jerry Fodor then developed it into the language of thought hypothesis, one of the most influential — and controversial — positions in philosophy of mind. Together, they gave cognitive science its foundational metaphor: the mind as computer.

Putnam's Original Formulation

Putnam's 1960 paper "Minds and Machines" was the first systematic philosophical defense of the idea that mental states could be understood computationally. The key insight was that a Turing machine — a mathematical model of computation that manipulates symbols according to rules — could serve as a model for psychological explanation.

Putnam's argument started with an observation about pain. What makes a state a pain state? It can't be any particular physical property, because creatures with radically different physical constitutions — humans, octopuses, hypothetical silicon-based aliens — could all be in pain. What they share isn't a physical property but a functional organization: pain is the state that's typically caused by tissue damage, that causes avoidance behavior, that causes the formation of beliefs like "something is wrong," and so on. This is the core insight of functionalism: mental states are defined by their causal roles, not by their physical realizers.

If mental states are defined by their functional roles, and functional roles can be specified computationally — as input-output mappings mediated by internal states — then minds can be understood as computational systems. Putnam connected this to the concept of a Turing machine: a system whose behavior can be fully described by the rules governing its transitions between internal states. Any system that implements the right state-transition rules has the relevant mental states, regardless of what it's made of. This is what made CTM so exciting: it offered a way to talk about minds that was precise, scientifically respectable, and compatible with physicalism without collapsing mental states into specific neurological configurations.

Putnam later moved away from functionalism and computationalism, arguing in "Representation and Reality" (1988) that the view faced insuperable difficulties. But his original formulation set the agenda for decades of work in philosophy of mind and cognitive science.

Fodor and the Language of Thought

Jerry Fodor took Putnam's computational framework and made it both more precise and more ambitious. In The Language of Thought (1975), Fodor argued that thinking literally occurs in a mental language — what he called "Mentalese." This language has a syntax (rules for combining symbols into structured expressions) and a semantics (the symbols mean things). When you think "John loves Mary," you're tokening a Mentalese sentence with compositional structure: the symbols JOHN, LOVES, and MARY are combined according to syntactic rules to form a structured representation, and the meaning of the whole depends on the meanings of the parts and how they're put together.

Fodor's argument for the language of thought (often abbreviated LOT) rests on the productivity and systematicity of thought. Productivity: you can think thoughts you've never thought before — "the purple hippopotamus is reciting Hamlet" — which suggests that thoughts are built from reusable components combined by rules. Systematicity: if you can think "John loves Mary," you can think "Mary loves John" — the ability to think one is inseparable from the ability to think the other, which suggests that thoughts have internal structure that gets rearranged, not just learned as unstructured wholes.

On Fodor's view, the computational theory of mind gets its bite from this syntactic structure. Mental processes are operations on the syntactic forms of Mentalese sentences. The mind is a syntax-driven system: it transforms representations according to their formal structure, and because the formal structure mirrors the semantic structure, the transformations preserve rational relations. If you believe "all dogs are animals" and "Rex is a dog," a computational process operating on the syntactic forms of these Mentalese sentences can produce "Rex is an animal." The syntax carries the semantics along with it. This is how CTM explains how physical systems can be rational: by implementing the right syntactic operations, a physical system gets rational relations for free.

Functionalism and Multiple Realizability

CTM is closely tied to functionalism, the view that mental states are individuated by their causal-functional roles rather than by their physical composition. The connection runs through the concept of multiple realizability: if the same mental state can be realized in different physical systems — biological brains, silicon computers, alien biochemistry — then what matters for having that mental state isn't the physical substrate but the computational organization.

This is why CTM has been so important for philosophy of artificial intelligence. If minds are computers, then the question of whether a machine can think becomes the question of whether a machine can implement the right computation. The physical substrate is irrelevant; what matters is the program. This insight underlies the Turing Test: if a computer's verbal behavior is indistinguishable from a human's, then for computational purposes, the computer is thinking.

Daniel Dennett has defended a version of this view, arguing that intentional states — beliefs, desires, intentions — are real patterns in the behavior of systems, detectable from the intentional stance. A system has beliefs if predicting its behavior is facilitated by treating it as if it were a rational agent with beliefs and desires. Dennett's position is more deflationary than Fodor's: where Fodor thinks beliefs are concrete structured representations in a language of thought, Dennett thinks they're useful predictive abstractions. But both agree that the computational level of description is the right level for understanding minds.

Searle and the Chinese Room

The most famous objection to CTM is John Searle's Chinese Room argument, introduced in 1980. Searle asks you to imagine yourself in a room with a rulebook. People outside the room slip in questions written in Chinese characters. You don't understand Chinese — you can't even tell the characters apart from squiggles. But the rulebook tells you exactly which characters to write in response to which input. You follow the rules mechanically, sliding your responses back through the slot. To the people outside, the room appears to understand Chinese perfectly — it gives fluent, appropriate responses to any question. But you don't understand a word of it. You're just manipulating symbols according to syntactic rules.

Searle's point is that this is exactly what a computer does. It manipulates symbols according to formal rules without any understanding of what the symbols mean. If the Chinese Room doesn't understand Chinese — and Searle insists it doesn't — then computation alone is not sufficient for understanding. Syntax is not semantics. No matter how sophisticated the program, a system that merely computes over symbols without any way of connecting those symbols to the world doesn't genuinely understand anything.

The argument has generated an enormous literature. Defenders of CTM have offered several responses. The systems reply argues that while the person in the room doesn't understand Chinese, the whole system — person plus rulebook — does. Searle's response is to invite you to memorize the rulebook: now the system is internalized, but there's still no understanding. The robot reply argues that if the room were connected to a body that interacted with the world — a robot that could see, move, and manipulate objects — the symbols would be grounded in perception and action, and understanding would emerge. Searle's response is that this doesn't help: the internal symbol manipulation is still purely syntactic, no matter what's producing the input.

Critics have argued that Searle's argument is essentially a intuition pump rather than a rigorous proof. The question of whether the room "understands" Chinese may be less clear than Searle assumes — Dennett, in particular, has argued that Searle is relying on intuitions about understanding that are exactly what's in question. If you define understanding functionally — as the capacity to produce appropriate responses to questions — then the room does understand. If you define it phenomenally — as involving a felt sense of comprehension — then the room doesn't, but neither does a computer, and the argument just begs the question against CTM.

Objections and Current Status

Beyond the Chinese Room, CTM faces other challenges. The symbol grounding problem asks how mental symbols get their meaning. If Mentalese symbols are just marks in a formal system, what makes them about anything? Fodor's answer was a causal theory of content: symbols mean what they do because of their causal relations to the world. "DOG" means dog because it's typically caused by dogs. But this faces counterexamples — misrepresentation is a problem, since sometimes "DOG" gets tokened by a cat on a dark night, and we need to explain why it still means dog and not cat.

Connectionism offers an alternative to classical CTM. Rather than treating the mind as a symbol-manipulating system, connectionism models mental processes as the propagation of activation through neural networks. There are no discrete symbols, no rules, no language of thought — just patterns of activation that settle into stable configurations. Some philosophers argue that connectionism replaces CTM rather than refines it. Others, including Fodor, have argued that connectionist models can't explain the systematicity and productivity of thought and that classical symbolic computation is still needed.

Despite these challenges, CTM remains foundational to cognitive science. The information-processing paradigm — the idea that cognitive systems encode, store, retrieve, and transform information — has been enormously productive, generating detailed models of perception, memory, language, and reasoning. Even critics who reject the strong claim that minds are computers often accept the weaker claim that computational modeling is a powerful tool for understanding cognitive systems. The debate has shifted from whether the mind is computational to how, exactly, the computational story should be told — and whether it can be told without leaving out the things that make minds distinctive.

Further Learning

1. Stanford Encyclopedia of Philosophy, Computational Theory of Mind.
2. Hilary Putnam, "Minds and Machines," in Dimensions of Mind, ed. Sidney Hook (Collier, 1960).
3. Jerry Fodor, The Language of Thought (Harvard University Press, 1975).
4. John Searle, "Minds, Brains, and Programs," Behavioral and Brain Sciences 3 (1980): 417-457.
5. Daniel Dennett, Brainstorms (MIT Press, 1978).

To explore related concepts, consider reading about the philosophy of mind, philosophy of artificial intelligence, and the work of Hilary Putnam, John Searle, and Daniel Dennett.