QuestionsCan a machine think?The Chinese room
Can a machine think?Understanding is missing

The Chinese room

The Chinese room argument (Searle)

John Searle 1932 · 2025Consciousness is a real and irreducible fact of inner life; to deny it is to overlook what is in front of us.
See profile
What it claims

Running a program is not enough to understand. No program, by itself, makes a mind: something else would be needed, such as the causal powers of a brain.

Why

A program manipulates symbols by their shape alone, which is not enough to understand what they mean.

The same idea through history
Gottfried Wilhelm Leibniz1714Leibniz imagines a mill: if we walked inside a machine that thinks, we would see only parts pushing one another, never a perception. The idea is close: looking at the mechanism does not reveal understanding.
John Searle1980In the journal Behavioral and Brain Sciences, Searle imagines a man locked in a room who receives Chinese characters and a rule book in English, and sends back answers without understanding Chinese.
John Searle1990Ten years later, in Scientific American, Searle sums up the reasoning in a few axioms: programs are formal, minds have content, syntax is not enough to produce meaning. This is the form we present.
In short

"The Chinese room" is an argument by John Searle ("Minds, Brains, and Programs", 1980; "Is the Brain's Mind a Computer Program?", 1990, 1980) on the question "Can a machine think?" (answer: understanding is missing). Its claim: Running a program is not enough to understand. No program, by itself, makes a mind: something else would be needed, such as the causal powers of a brain. According to AskPhilosophy's logic check (Z3 solver), the conclusion does follow from its 3 premises. AskPhilosophy presents 6 objections, 5 of them with a documented reply, and the original text behind each premise, checked word for word.

"Minds, Brains, and Programs", 1980; "Is the Brain's Mind a Computer Program?", 1990
Valid

Imagine you are locked in a room. Through a slot, you are handed texts in Chinese, which you cannot read: to you they are just drawings. You have a big rule book, written in English, that says: "when you receive this string of signs, send back that string". You follow the rules and send back signs. From outside, your answers are those of a real Chinese speaker. But you have not understood a word. John Searle (1980) concludes that even a computer that passed the Turing test would understand nothing: it would do as you do, following rules about shapes. Meaning does not come from the rules alone. The argument targets the "strong AI" thesis (to think is to run the right program), not the idea that a machine could one day think. Searle himself says the brain is a machine that thinks.

P1A program is purely formal: it says which signs to produce from which signs, according to their shape. Whoever runs it does not need to know the meaning of any sign.
P2A mind that understands has meaning: when I understand a sentence, I know what it means. This is what separates my reading of English from my reading of the Chinese signs.
P3Manipulating signs by their shape is not enough to understand their meaning (syntax is not enough for semantics). In the room I have everything a program can give me, and yet I understand nothing.
So running a program is not enough to understand. No program, by itself, makes a mind: something else would be needed, such as the causal powers of a brain.

click a premise to reject it

If you accept every premise, logic forces you to accept the conclusion. So the debate is about the premises.
Checked by a logic solver: the premises cannot all be true while the conclusion is false.
The argument is valid: if a program alone is formal (P1), if the formal does not give meaning (P3) and if a mind that understands has meaning (P2), then a program alone is not a mind. The debate is about P3. Searle proves it through the thought experiment; his opponents say the experiment proves nothing, or that it does not target the right system.
Analogy. Someone who copies a score perfectly without being able to read music does not play music. Copying shapes is not understanding. What remains is whether this analogy suits an immense, fast system.

Objections

6

Each objection is a conversation: click to see who replies and where the debate stands.

The systems reply (Berkeley) 1980 · attacks premise 3

The man alone does not understand Chinese, true. But he is only one part of a whole: the man, the manual, the paper, the lists of signs. It is this whole that understands, just as a neuron understands nothing yet a brain does.

Searle replies
John Searle 1980
Searle answers: let the man memorise the manual and the lists, and do all the calculations in his head. He then contains the whole system. He still understands nothing of Chinese, so the system does not either, since there is nothing in the system that is not in him.
Daniel Dennett and Douglas Hofstadter 1981
Dennett and Hofstadter answer that the experiment misleads our intuition. We identify with the man, who is only a small part, like the processor of a computer, and we picture a small manual. But a program able to converse like a Chinese speaker would be immensely complex. That the man does not understand does not tell us what the system he runs does.
See the rest of the debate (1)
John Searle 1980
Searle had anticipated the idea of a subsystem that would understand Chinese without the man knowing it. He answers that the only reason to believe it is that the system passes the Turing test, which is exactly the point in dispute: the systems reply assumes what had to be proved.
Where the debate stands (our summary)The point of disagreement: can understanding be a property of a system none of whose parts understands? For the systems reply, yes, as with a brain. For Searle, one must show that such a system understands, not assume it. Neither side has won.
T
The robot reply (Yale) 1980 · attacks premise 2

Put the computer inside a robot, with a camera, arms and legs. The robot sees, acts, lives in the world. Meaning comes from the link to things, not from signs alone: the robot would understand.

Searle replies
John Searle 1980
Searle first notes that this reply already admits that thinking is not only manipulating signs. Then he puts the man back in the room: some signs come from the camera, others drive the motors, but he does not know it. These capacities add signs, not understanding.
Where the debate stands (our summary)The point of disagreement: does the link to the world make meaning? Searle holds that this link also arrives as signs. Others, such as embodied cognition researchers, think a body and a life make a real difference. Open point: no widely accepted answer.
The brain simulator reply (Berkeley and MIT) 1980 · attacks premise 3

Write a program that simulates, neuron by neuron, the brain of a Chinese speaker. It takes Chinese as input, imitates the structure of its synapses and gives Chinese as output. To refuse to say it understands would be to refuse to say it of the brain itself.

Searle replies
John Searle 1980
Searle answers with water pipes: the man operates valves that imitate synapses. The right answers come out, but neither the man nor the pipes understand. The simulation copies the wrong thing about the brain: its form, not its causal properties, which produce intentionality.
Paul and Patricia Churchland 1990
The Churchlands answer that Searle's room is a calculator that handles signs one at a time. A brain works with millions of connections in parallel, and so do neural networks. This is not the machine the experiment puts to the test: it proves nothing against a parallel system.
See the rest of the debate (1)
John Searle 1990
Searle answers with a "Chinese gym": a large number of men, each playing the role of a neuron, work in parallel. None understands Chinese, and the gym does not either. Parallelism changes nothing, since the gym still performs only formal operations.
Where the debate stands (our summary)The point of disagreement: does a perfect simulation of a brain reproduce what matters, or only its form? Searle says the simulation does not reproduce the brain's causal powers, as a simulated storm wets no one. His opponents say only organisation matters. Open point.
T
The other minds reply (Yale) 1980 · attacks the conclusion

How do you know that other people understand Chinese, or anything? Only by their behaviour. The computer can pass the same tests. If we attribute thought to others on that basis, we must attribute it to the computer too. This is Turing's position: the behaviour test is our only criterion.

Searle replies
see the argument
John Searle 1980
Searle answers that the problem is not how we know that others think, but what we attribute to them when we say they think. A computation and its output can exist without a thought: so behaviour cannot be what we attribute.
Where the debate stands (our summary)The point of disagreement: is behaviour all we can know about another's thought, or only its sign? Searle says the question is not how we attribute thought, but what we attribute. Open point.
Paul and Patricia Churchland (the luminous room) 1990 · attacks premise 3

Imagine someone waving a magnet in a dark room: no light appears. Should we conclude that light is not made of electromagnetic waves? No: the room is too slow. Likewise, Searle's slow experiment does not show that syntax can never produce meaning.

Searle replies
John Searle 1990
Searle answers that the analogy fails: electromagnetic waves have physical powers, whereas symbols have no physical property that produces meaning. For him, syntax does not exist in nature: it is the observer who sees a computer as computing.
Where the debate stands (our summary)The point of disagreement: is P3 an established fact or an intuition? Searle holds it a truth of logic, the Churchlands an empirical matter. Open point: no widely accepted answer.
Daniel Dennett 1980 · attacks the conclusion

The idea of a special power of the brain, which would produce "intentionality" like milk, is not explained: Searle does not say what this power would be. For Dennett, the brain is a machine too, and nothing forbids another machine from doing the same.

no reply
Searle never answered this objection.
Where the debate stands (our summary)The point of disagreement: must Searle say what the brain's "causal powers" are made of? He answers that this is a question for biology, not philosophy. His opponents think that without an answer his conclusion remains incomplete. Open point.
What today's voices say
Defend it
John LennoxHe uses the Chinese room to say that passing the Turing test does not show consciousness.
Criticize it
Daniel DennettHe says Searle misrepresents the kind of program a computer would run: a program that passes the Turing test would take trillions of sheets of paper to simulate by hand, so imagining a trivial one misleads.