Quantum theory on the brain

by R. F. Streater.

Critique of Why classical mechanics cannot naturally accommodate conciousness but quantum mechanics can, by Henry P. Stapp.


We argue that Stapp has made three errors, one in each of the subjects of classical mechanics, quantum mechanics and experimental brain dynamics. Together they make his version of brain theory a lost cause.

I. The speed at which concepts are formed

Stapp wants a model of the brain which `can bring the information together from far-apart locations' instantaneously to get the `whole thought'. However, whole thoughts are NOT arrived at instantaneously; this is well known; see for example, Science Through a Looking Glass, by E. B. Davies, OUP, 2004. By measuring brain activity and asking the subject what he is thinking, it can be seen that one collects one thoughts from various parts of the brain, and that this activity starts about 1/2 second before the concious part of the brain is aware of the thought or decision. We do not need any information to flow at the speed of light to explain these facts; we certainly do not need any effects that travel faster than light. Yet this is put as one of the failings of classical mechanics: it does not allow faster than light flows of information.

II. The claimed absence of correlations in classical field theory

He first (in 2.6) tries to model the brain as the aggregate of a large number of spatially localised computers, each in some configuration. The (discrete) dynamics is given by local update rules involving the state at a point and the neighbouring points, with no interaction taking place instantaneously over long distances. Thus his idea is to describe the brain as a field taking any of a set of finite values at each point. He says in (2.7), `...the information stored in any one [original italics] of the logically independent computers ...is ...minimal: it is no more than is needed to compute the local evolution. This is the analogue of the condition that holds in classical physics'. This is a deficient description of classical physics. Firstly, he does not distinguish between the fields (the observables) and the states (the probabilities that a certain configuration is realised). Indeed, it is clear that he does not allow a probabilistic description of the state, even though the brain is very noisy, and all the leading models in the subject of neural nets use (classical) statistical physics. But if his description is non-random, what does he mean by saying that the computers are `logically independent' of each other? Does he mean `statistically independent'? He gives a hint. In (2.4) he uses the analogy of a display of pixels on a TV screen as the state of the computer. He writes that his requirement, termed `minimal information', gives rise to an `intrinsic description without any explicit relationship that might exist between the elements, [such as] that Pixel 1000 has the same colour as Pixel 1256'. Now, in a non-random description by a configuration, the state of each pixel is determined, as is the question whether it has or has not the same value as another specified pixel. If this information is missing, it must be because the writer has lost it, by replacing the (pure) state by a mixed state, the product of probability measures at each site. For such a state, one may indeed say that the computers are statistically independent, and that the state shows no correlation between different sites. It is the presence of such correlations in quantum theory that entices the author to invoke quantum theory. However, the author is quite wrong to suppose that classical field theory cannot have states with correlations. The original description by pure states has got such information contained in it. But even in a random theory, we can have correlations between the sites, and this will occur if causal effects are present: if a cascade of firing neurons causes others in its path to fire, then at a later time, there will be positive correlations between the firing rates at separated points in the brain. These can be read (slowly) by other parts of the brain, and the info they hold summarised in a new register, forming the `whole thought'. Correlations have been studied both theoretically and experimentally by Valeria del Prete.

The concept that the author describes, a pure configuration of the brain containing no information about the relations between different points, does not exist in classical mechanics.

III. Supposed non-localities within quantum theory

I have explained on another page that the alleged non-locality shown by the EPR experiment should be called `contextuality' in classical mechanics, and that quantum theory does not suffer from this. I argue that when a commuting set of observables is measured, the interpretation of any correlations must be the same as in classical probability theory. There, it is well known that a correlation between two observables does not imply a causal relation in either direction. Stapp contradicts this in his Appendix D4a as follows:

"Orthodox quantum field theory in its covariant form...is non-local in the sense that for certain systems...the set of correlations predicted by quantum theory ...is incompatible with the following `locality' condition: the result of any possible measurement M must be independent of any free choice that is to be made later...One cannot assume that there is no faster-than-light influence of any kind.

So Stapp is saying (without mentioning me by name) that most of my web-site is wrong. I think this allows me to say that, no, it is he that is wrong; see my page of good cures.

Again, in Appendix 4b, Stapp writes "It [QM] is non-local in...that any... collapse of the wave-function ...changes expectation values all over the universe". As I have explained in my page on EPR, the act of A = Alice by measuring a local observable (but not yet reading the pointers) has no effect whatever on Bob's observables, if they are space-like to the measuring apparatus. On reading the pointers, Alice conditions her state by using Bayes's theorem. This gives her updated information on her INFORMATION ALGEBRA, which is the quantum version of information set used in game theory. Since this info cannot be passed to B until later, it does not affect his state at all. So Stapp has failed to distinguish information available to Alice but not to Bob. The collapse of Alice's wave function is nothing other than the application of Bayes's theorem. To suggest that Bob's statistical description must change just because Alice has some info, is incorrect, and would give the wrong results also in classical game theory. For example, consider a game of cards between two players, in which each player had a hand unseen by the opponent; it is absurd to say that when one of them, Alice, finds that her hand includes the ace of spaces, so the other, Bob, immediately alters his assessment of the likelihood that this is so. Alice's measurement merely reveals the situation to her; it does not cause any change in Bob's hand or strategy. I maintain that information obtained by measurement in quantum theory cannot be passed faster than light, and that information held in various parts of the brain needs time to form a whole thought, decision or idea, just exactly as is observed (especially at my age).

Go to MY HOME PAGE for references and links to my papers and books on quantum field theory and statistical physics.

© by Ray Streater, 16/3/2004.