Dear Jacques Mahler,

I just got your letter, rejecting my paper (#19970224001). I was surprised to see the review that you enclosed. As I show in the attached comments, this is an amazing piece of nonsense, which is a shame to cognition and to yourself. The main points are:

1) The reviewer does not analyse my argument at all, and has nothing to say about the working of the brain.

2) The reviewer does not understand what `stochastic' means, and did not bother to read carefully section 4 of my paper, which explains it.

3) The reviewer ignores the distinction between a designed system and a stochastic one. I explicitly highlight this point in my paper.

4) The reviewer restricts possible arguments to "mathematical demonstrations," even though, by definition, the details of the stochastic connectivity cannot be modelled mathematically.

I think by now you know that even though many people don't like the argument in this paper, they find it very difficult to refute, and that the neurobiological side of the argument is sound. That shows that the paper highlights a significant point which hasn't been thought about, which should be enough to justify publication.

I enclose the full text of the review and my comments (indented).

Yehouda Harpaz


First, it should be noted that the reviewer does not analyse the argument that I present on section 5, and says nothing about the brain.

This paper purports to dismiss the symbolic theory of mind via a consideration of the microcircuitry of the brain. In particular, the author argues that the microcircuitry is "stochastic", in that particular neurons do not target particular other neurons, but rather target populations of neurons. Thus any transformation of the neural firing patterns are "stochastic", and a series of such transformations loses the precise, deterministic characteristics that are required for symbolic computing (e.g., addressing).

Almost accurate, but the stochastic transformation lose more than precision. They lose the signal itself. It seems the reviewer does not understand what "stochastic" means, and does not try to understand it, even though I dedicate a complete section to explain it.

I find the argument completely unconvincing. The author seems to equate "stochastic" with "uniformly probable."

This sentence is unrelated to the rest of the text, but suggests that the reviewer did not understand the term `stochastic'.

Moreover, the author seems to be unaware of modern information and communications theory.

"modern information and communications theory" didn't come up yet with a way in which a stochastic transformation of the signal can keep it, simply because there is no way to do that. The reviewer is either ignorant of this fact, or does not understand that the connectivity in the brain is actually stochastic. In my paper, I have highlighted the difference between designed circuits and the stochastic connectivity of the cortex, in the fifth paragraph of section 5: "The stochastic propagation of patterns of activity is the most crucial point to grasp in the whole argument..." The reviewer seems not to bother to grasp this point.

In modern information theory, the major methodology behind the design of codes is "random coding," which was in fact introduced by shannon and is now in common use. Using random codes, it is possible to introduce a coding system that can send signals deterministically through a channel that has arbitrary levels of noise,

The `random codes' cannot be stochastic, i.e. selected arbitrarily. They have to be selected from a very restricted set of codes, that keeps a relation of (at least) many-to-one between the encoded message and the original message. Stochastic transformation are not going to find these codes, because they are stochastic.

e.g, it is possible to send clear images from jupiter, even though the channel between Earth to Jupiter is a highly stochastic.

Plain nonsense. The `channel' between Jupiter and Earth is electromagnetic waves, which are highly deterministic. There is a lot of noise on the way, but the signal itself is not transformed at all.

The same thing can be said of analog telephony.

Nonsense as above.

In fact, the same can be said of digital information processing, as implemented in analog hardware. Each analog transformation introduces certain stochasticity, but the coding procedure ensures that the probability of the error is vanishly small.

Yes, but in the brain the coding procedure itself is stochastic. As I said above, the reviewer simply does not bother to grasp the argument itself.

Another serious conceptual error in the paper is the confounding of "separate signalling" with "unconnected."

My paper does not contain either of the terms "unconnected" or "separate signalling," so this `confounding' is pure imagination of the reviewer.

It is entirely possible to communicate signals separate through a channel even though they are distributed across the same set wires. This is called "orthogonal coding."

It is not obvious against what the reviewer argues here, but it is clearly irrelevant to the brain. For an "orthogonal coding," the sender must distribute the signal in a well-defined manner, so the receiver can know how to assemble it back. If the distribution is stochastic, as in the cortex, the receiver cannot find how to put it together. Again, the reviewer simply ignores the stochastic connectivity of the cortex.

To make any kind of argument of the sweeping generality that the author is trying to make would require careful argumentation,

Since the reviewer did not bother to analyse my argument, shhe cannot say whether it is careful or not.

In particular a mathematical demonstration that it is not possible to do a certain kind of operation.

This is straightforward narrow-mindness. Any good argument is good, mathematical or not. By restriction the acceptable arguments to "mathematical demonstration," the reviewer completely eliminates the possibility of discussing the working of the brain, because stochastic connectivity cannot be modelled mathematically. These models can catch only the statistical characteristics of the system, not the details, and in the brain, it is the details that are important.

One would want to set up a mathematical model of the purported "stochasticity" of neural connectivity and the study the properties of the channel that this defines.

(a) This is the closest that the reviewer gets to saying that shhe does not accept that the connectivity of the cortex is stochastic, by saying `purported "stochasticity"'. (b) The reviewer shows that shhe does not understand what `stochastic' means, because a `channel' based on stochastic connectivity will not have any fixed properties that can be studied. That is the point of my paper.

One would want to define a family of codes and study their properties.

This is a conclusive demonstration that the reviewer did not understand what `stochastic' means. There is no `family of codes' in a stochastic system.

Only if no reasonable codes are able to give one the properties that one requires (in terms of e.g., space and time) could one draw any kind of conclusion.

That is plain nonsense, because it means that you cannot draw any conclusion without traversing all the possible codes, and there is infinite number of possible codes. Anyway, as I said before, there is no defined codes in a stochastic system.

Such mathematical argumentation is, however, entirely absent from the paper.

And the reviewer does not bother to analyse the argument that is presented in the paper.