Mind, Matter, and Meaning

Computation and the Church-Turing Thesis

The Church-Turing thesis is one of the foundational results in the theory of computability. In its standard formulation, it identifies the intuitive notion of effective calculability (what can be computed by a systematic procedure) with what a Turing machine can compute. Proposed independently by Alonzo Church and Alan Turing in 1936, it has shaped computer science, mathematics, and philosophy of mind ever since.

The thesis gets invoked constantly in arguments about the nature of mind. If the brain is a physical system, and physical processes are in principle computable, then, the inference goes, the brain is a Turing machine, and any sufficiently complex computer running the right program would have a mind. That inference moves through several steps, and each deserves separate examination: the Church-Turing thesis itself, the further claim that minds are computational systems (computationalism), and the claim that computational processes can generate semantic content (the syntax-semantics question).

Keeping these three apart matters for evaluating any argument about artificial intelligence, machine understanding, or the computational theory of mind.

Three Distinct Claims

1. The Church-Turing thesis (CTT)

The CTT, in its standard form, identifies the intuitive notion of effective calculability with what a Turing machine can compute: any function computable by an effective procedure is Turing-computable. It is a claim about the extension of “computable,” not a claim about minds, brains, or meaning. A separate physical Church-Turing thesis (Cotogno, Galton) asks whether every physically realizable process is Turing-computable — a stronger and more contestable claim, bearing on hypercomputation and the limits of physical realizability. Even granted in full, physical version included, the thesis says nothing about understanding on its own. It tells you a function can be computed; it stays silent on whether computing it amounts to thinking it.

2. Computationalism

Computationalism is the further thesis that cognition consists in computation — that the mind operates as, or gets implemented by, a computational system. Computational functionalism places that thesis inside the broader functionalist framework: functionalism identifies mental states by their causal roles, while computational functionalism adds that the relevant roles are computational states. Strong computationalism goes further again and claims that running the right program suffices for mentality. None of those additions follows from functionalism alone. The computational picture earned its long run at the center of cognitive science honestly. Turing had shown how a purely mechanical process can carry out any effective procedure, which made it possible, for the first time, to see how a physical mechanism might reason without a little person inside doing the reasoning. Hilary Putnam built machine functionalism on the idea; Jerry Fodor built a theory of thought on it; Ned Block’s “The Mind as the Software of the Brain” gives the picture its clearest statement.

The crucial logical point, stressed by Gualtiero Piccinini among others: computationalism does not follow from the CTT. “The brain’s processes are computable” and “the brain is a computer (and the mind its software)” are different claims; the first can be true while the second is false. Critics argue that running the two together lends computationalism a borrowed air of mathematical security — the security of a thesis that was never about minds.

3. Syntax and semantics

The third claim is Searle’s: a computational process is defined over syntax, formal symbol manipulation, and syntax does not, by itself, yield semantics. Running the right syntax is not thereby meaning anything. Searle sharpens this with the observation (Negru) that syntax itself is observer-relative: being a “symbol” or a “computation” is not an intrinsic physical feature of a system but something assigned to it by an interpreter — which is exactly what a mind with original intentionality would have to supply, and so cannot be what computation provides. See Searle and Intentionality.

One further distinction runs through this territory. A perfect computational simulation of a process is not always an instance of that process: a simulated hurricane makes nothing wet, and a simulated digestion digests nothing. Whether a faithful simulation of a brain would realize what the brain realizes, or merely model it, is a separate question from anything the CTT settles. See Simulation and Realization.

The Live Debate

The most direct recent challenge to Searle’s side comes from Vladimír Havlík’s “Meaning and Understanding in Large Language Models” (2024), which argues that the assumption of a complete gap between syntax and semantics is unjustified — that the strong claim (“no syntactic process can produce semantics”) does not follow from the Chinese Room alone, and that “syntactic semantics” (Rapaport) may bridge the gap. The strong/weak distinction marks the live edge of the debate: whether the gap is total, and what argument could show it.

My View

The inference from “the brain is computable” to “a computer running the right program would understand” loses each of its joints once the three claims are kept separate. Computationalism does not follow from the CTT. And even granting computationalism for argument’s sake, an abstract program or formal symbol manipulation considered by itself does not constitute meaning. A running implementation may realize content-fixing history, producer-consumer use, embodiment, and integration that the abstract program leaves unspecified; that wider possibility must be judged on its actual organization.

I don’t deny that brain processes are, in the relevant sense, computable. I deny that the concession establishes program sufficiency, because the inference from computability to understanding runs through three joints: CTT ≠ computationalism (computability of brain processes does not entail that minds are computers); simulation ≠ realization (calling something a simulation does not settle which relevant powers its physical implementation realizes); and syntax ≠ semantics (formal operations, taken by themselves, do not constitute meaning). The Chinese Room pressures the third joint at the level of the operator and the bare formal description; it does not independently settle what a suitably organized implemented whole could understand. That is why Havlík’s challenge deserves an answer rather than a dismissal, and why the wider case turns on the implementation’s content-fixing relations and coordinated capacities.

Understanding requires content-guided capacities to constrain one another within an answerable system; it does not require every content involved to be original. Formal structure alone supplies neither content-fixing history nor that integration. A computational implementation may realize both, and its actual history and organization must decide.

  • The Church Turing Deutsch Principle — the computationalist’s strongest fallback: CTD grounds simulability in physics, and still fails at the simulation/realization seam
  • Simulation and Realization — the middle layer; simulation of a process is not the process
  • The Chinese Room — Searle’s vehicle for the syntax/semantics gap (used as one strand, not the lead)
  • Searle — syntax as observer-relative; the original-intentionality requirement
  • Functionalism — computational functionalism as one species of the broader role theory
  • Derived and Original Intentionality — original vs derived intentionality, the crux of the syntax/semantics layer; also why original makes no claim about novelty