GENERAL COMMENTS
The revised version of the manuscript #AM98029 titled "Can the neurons in the brain implement symbolic system?" is a considerable improvement over the previous version. Indeed, many of the grammatical and typographical errors have been corrected. On the other hand, the main argument of this manuscript remains weak. In general I must say that I would expect the author to write his paper himself instead of submitting half-baked ideas in bad English in the expectation that somebody else will do his job for him.
The original manuscript certainly did not contain half-baked ideas. This is a spurious comment. As to the English, the original contained much less grammatical and typographical errors than the impression this reviewer tries to give, and the only comment about the writing from the other two reviewers was "the argument is clear" from the third reviewer.SUGGESTIONS FOR REVISION
1) The main argument of this manuscript remains weak. To give an example, consider the claim "in the next four sections I will show that our knowledge at the neurobiological level is in conflict with the symbol hypothesis" which the author utters in page 7 of his manuscript. This claims is patently false. There is no neurobiological piece of evidence in the manuscript to refute the symbol hypothesis.
Note the manoeuvre: I write about 'knowledge', the reviewer writes about 'evidence'. In the paper I wrote (p.7, middle): "It is important to note that the characteristics listed here are `textbook' knowledge, supported by a large body of consistent experimental evidence, accumulated in over than 100 years of research." The reviewer is not going to challenge this assertion. He simply ignores it. Instead, he tries to mislead the reader to think that I claim that I bring detailed evidence, and that the 'knowledge' is my knowledge, rather than agreed by all neurobiologists. Note that during all the discussion of point (1), the reviewer completely ignores the 'controversial' paragraphs in section 4.Instead this manuscript convincingly make the point that whatever neurobiological evidence is presently (including the pieces of it described in the manuscript) does not support the symbol hypothesis. Refuting a hypothesis and failing to support it is not the same thing. It would certainly be easy to delete this and similar unsupported claims from the manuscript (this MUST be done). But the problem is worse and is an implicit one. Suppose that I make the following claim: "The neurons in the brain are not stochastic, in the sense that they are completely specified so that pyramidal neuron #123456788 in V1 of the brain of human being A is everywhere identical to pyramidal neuron #123456788 in V1 of the brain of human being B; further they are not stochastic in the sense that they are completely specified so pyramidal neuron #123456789 in V2 of the brain of human being A differs from pyramidal neuron #123456789 in V2 of the brain of human being B in specific ways ABCDEF for reasons related to well known difference between these two human beings". There is actually no evidence against this claim.
As is, this claim is nonsensical, because 'neuron #12345678' does not refer to any specific neuron, so there is no way to confirm/refute this claim. To make the claim real, the reviewer would have to give a specification that identifies the neuron (e.g. its location in standard brain coordinates). In this case, it is much more likely than not that there is already evidence to refute the claim.Now, if I were to publish a paper that uses this argument the onus would be on me to either: 1) Provide proof supportive of the claim, or 2) Assume its truth (and thus it would be a very different sort of proposition, namely an assumption) and proceed logically from it.
When advancing new hypothesis (and specially when it is as ridiculous as the one the reviewer offers), that is true. However, the fact that the connectivity in the brain is stochastic (in the sense that I define it in the bottom of page 7) is an established fact, as I say in the paper, and therefore does not need to be established.Since the author of #AM98029 can not prove that neurons are stochastic (if he has the relevance evidence now is the time to produce it) he should state exactly his assumption (i.e., that the neurons in the brain are stochastic) and call it an assumption.
Note how the reviewer drives in the impression that it is 'an assumption', without any reference to what I actually wrote.He is certainly free to also state that to his knowledge there is no evidence to negate the truth content of his cherished assumption.
And to complete the manoeuvre, the reviewer does mention 'knowledge', but only 'my (the author's) knowledge', to try to make the reader forget that it is a well-established knowledge.2) The last sentence of the previous paragraph could replace most of section 4 of #AM98029. This section must be reduced by a lot as it does not strengthen the author's assumptions. For example, the "evidence" alluded to in Paragraph 3 of Page 9 is no evidence at all.
It is enough to convince all neurobiologists. It does not seem convincing to the reviewer because he doesn't know neurobiology.Nobody has tried to compare stellate neuron #123456789 in area 17 of the brain of cat A to that of stellate neuron #123456789 in area 17 of the brain of cat B.
As I wrote above, 'neuron #123456789' is not a meaningful term. When it is replaced with a concrete specification, then the assertion here is false, because many people tried to find matches between different brains, in all its parts. The reviewer himself brings an example below (about the superior colliculus).Proof does not pop out of experiments for free. It only arises in experiments designed to test hypotheses (albeit not necessarily the one the experiments was designed for).
These paternalistic comments would have been more convincing if they were not based on simply ignoring the research that is done by neurobiologists.However, I must admit that assumptions can be rendered less counterintuitive if there is no evidence to refute them so a check of the literature is seldom a waste of time. A couple of paragraphs of evidence of this sort (say paragraph 3 and 4 of P 11) together with a coherent presentation of the argument (say the bottom paragraph of P 9, and the paragraphs of P 10) would suffice for a new Section 4.
That will make the paper non-sensical.3) It is no accident that two referees thought the author's argument applied to the brain and the spinal cord.
The reviewer tries to back-pedal on his comments in his first review where he used synapses in the spinal cord as 'the best studied synapse in the brain'. First, he claims that two reviewers got confused, which is simply false. Only this reviewer made this 'confusion'. Whether it was an accident or not is another question.It never crossed Eccles mind that whatever he found in the spinal cord would not apply to the rest of the CNS. The same is true of almost everybody working in the spinal cord.
Whatever hasn't crossed Eccles mind is irrelevant to the discussion. The question is whether we can assume that patterns of connectivity in the brain are the same as the ones in the spinal cord. I am sure that to this question, all the researchers will give a negative answer.The author is certainly free to limit his argument to whatever part of the nervous system he wishes to consider. But he should be upfront about it.It should be noted that in his review, the reviewer did not say that he think that we can project from the spinal connectivity to the brain connectivity. What he did was to claim that the synapses that he referred to, which are in the spinal cord, are in the brain.
Since I state which part I am talking about (the brain) in the title of the article, I cannot be much more 'upfront' about it. This idea is even more ridiculous coming from this reviewer, because he suggested removing the word 'brain' from the title in his first review.The easiest way to go about it is to modify paragraph 3 of Page 7, as follows: ...the term "brain" in the strict sense to mean the vertebrate telencephalon and brain stem".
Everybody knows that this is what the term 'vertebrate brain' means.4) P 14,15 (Section 6): I still have trouble with the argument regarding coherent connectivity. The author should write his text in such a way that I (and other potential readers) can understand exactly what it is that he has in mind.
The following comments suggest that the reviewer did understand the argument.It could help if he provided definitions.
That is ridiculous, because the discussion is in plain English, and I use words in their standard meaning. The only difficult term is 'coherent connectivity', and I define this in the last paragraph of p. 14. If the reviewer has a problem with any of the terms, he should say which term.Also, it could if he provided an example. To this end he could substitute the word "element" by layer, column or area, etc., and explain himself concretely.
As the reviewer demonstrates, the readers can easily do that themselves, if they find it more convincing.Then if he wished to be general he could again convert to "element". To get back to the argument though, does the author imply that because of their spatial proximity the X neurons of layer 4 of a column in the superior colliculus and the I neurons of layer 4 in the same column of the superior colliculus must receive the same input from the frontal eye fields or the pars reticulata of the substantia nigra?
The reviewer demonstrates again his lack of knowledge of neuroanatomy. The superior colliculus is not part of the cortex, and the discussion he refers to is about the cortex.If this is what he means he has his facts wrong. If this is the case I must insist that he has failed to prove that symbol tokens can not be handled by larger scale structures in the brain.I explicitly write in the third paragraph on p.15 that "A more coherent connectivity is seen outside the cortex" and then "However, these connections clearly are not coherent enough to transfer specific activity across the cortex ". This statement is clearly applicable to the superior colliculus, because the low-level connectivity in the superior colliculus is clearly not replicable between individuals. As to the 'X' and 'I' distinction, it would have been better if the reviewer gave a reference, because textbooks of neurobiology do not mention this distinction.
The superior colliculus is even less likely to take part in a symbolic system (or the thinking system), because it does not have output to the cortex, and seems to be involved only in control of various reflexes and automatic actions like saccades.
b) minor
1. Title: I suggest "Can Brain Neurons Implement Symbolic Systems?"
'Brain neurons' is a confusing term. The current title is much clearer.1. Abstract, lin2 (1) 2: This notion is based on the often implicit assumption that there is a symbolic system in the brain itself.
The text in the paper expresses what I want to say more accurately.2 Page (P) 4, paragraph (p) 1, 1 2-4: the phrase "...framework, which they say is agreed as the appropriate way to study..." remains awkward. It is the English that is awkward not the argument. It is the author's responsibility (and not the referee's) to express the author's ideas in good prose. In any case, something like "...framework which is generally thought to reflect human..." would probably do.
The part that the reviewer wants to change is actually a quote from the book, as the text says. I will make it clearer by adding double quotes around it.3 P 4, p 1, 1 10: In English, no sentence is allowed 1) to start with a parenthesis and 2) with a parenthesis followed by a small letter. The author should modify the statement " ... world. they do mention connectionism later)" in a way that respects grammar good style.
4 P 4, p 4, 15 & 7: The author does not provide a list of the characteristics that he thinks are characteristic of neurons. Thus the pronoun "these" in the sentence "neuron with these characteristics" is meaningless in the strict grammatical sense pertaining to the job of pronouns in English syntax.
This is plain nonsense. The word "these" is part of the term "these characteristics", which clearly refers to "the characteristics of the neurons in the brain" from the previous sentence. It seems that by now the reviewer became desperate to find mistakes in the paper.I suggest that he uses "realistic neurons" as follows: ...implementable by realistic neurons. In this article I will show that symbolic systems are in principle unimplementable with realistic neurons, and hence.... If he wishes he can define "realistic" to mean "neurons endowed with the properties of the real neurons that populate real brains" either before or after the first appearance of realistic in the text.
A nonsensical suggestion that would make the text much more confusing, because there are "realistic neurons" outside the brain.5. My objection regarding "consistent" is that the author means "self-consistent" .
Note again that the reviewer asks "how do YOU know", rather than "how do NEUROBIOLOGISTS know". He refuses to accept that what I am saying is known to anybody except me.7. P 10, p 5, 13, fiber (instead if fibre).
"Fibre" is the British spelling, as any reasonable dictionary would tell the reviewer. This is another example of desperate effort to dig out errors. the previous sentence, and therefore I use "Thus".8. P 14, p 5, 14: ...distributed. On the other hand, the use of higher levels of organization to implement symbol tokens would need to be based on primitive elements of...
This changes the meaning of the sentence, and wrongly. The assertion in this sentence follows from the previous sentence, and therefore I use "Thus".9. P 18, p 5, 11: Does the author mean the following? "We don't know enough about neurons to build symbolic models with them"
There was a mistake here. I will change the text to say what I mean (We don't know enough about neurons to build models based on this knowledge).10. P 22, p 4, 12: It shows that the notion that the brain implements symbolic systems is not well supported by what we currently know about it.