related texts

Dale Purves is an interesting case. His main idea, that human vision is based on matching the statistical structure of natural light patterns, is clearly true, because it is an immediate conclusion from the fact that humans learn to see, and humans learning is based on pattern matching. Dale Purves first stands out by figuring out that humans vision is based on this matching, but even more interesting is that this is still coupled to a religious belief in the innateness of vision. he believes that the matching to the statistical is coded in the genome by evolution.

Below is a correspondence between Dale Purves and me. The article I refer to is available here. His three arguments (I should write "arguments") are really dumb. (1) is just religious belief and (2) is simply false. (3) is simply demagoguery, because the question is (lack of ) similarity across animals, not within the same animal. His only reference is to his own book. Apparently, the best example he can think of against stochastic connectivity is the work on spinal reflexes, and his apparent belief that "receptive field properties (wiring) of visual cortical neurons prior to experience" show that the wiring is the same across individuals, which is clearly very broken logic.

Dale Prves is ostensibly a neuroscientist, enough that he edits textbooks, though I don't see any sign that he actually does research on neurons or neural tissue (his site is all "behavioural" studies, i.e. just looking at human behaviour and reports of perception). That somebody so close to the actual research can still get it so wrong shows how misleading is the language that is used by the researchers that actually do the research.



----- Original Message ----- 
From: "Yehouda Harpaz" 
To: "Zhiyong Yang"
Cc: Dale Purves
Sent: Saturday, September 18, 2004 3:22 PM
Subject: How the bain knows the statistical structure of natural light



Hi,

Just read your article "The statistical structure of natural light patterns
determines perceived light intensity" (PNAS, June 8, 2004, vol. 101,
no. 23, 8745-8750).  In it you repeatedly state that the statistical
structure was acquired "during evolution".

Why not assume that it is acquired during ontology, i.e. that
each individual learn it?

The reason not to believe that it was acquired "during evolution"
would require that it is the same circuits in all the individuals of
the species. however, in the cortex, including the visual cortex, the
connectivity is stochastic, i.e. varies random across individuals.
Therefore nothing like the circuits that are speculated in the
discussion of your article could be implemented in the cortex
by evolution.

Thanks,

Yehouda Harpaz
http://human-brain.org 



=====================================
His answer:
=====================================
From: "Dale Purves" 
To: "Yehouda Harpaz" 
Sent: Saturday, September 18, 2004 3:39 PM
Subject: Re: How the bain knows the statistical structure of natural light


>Hi,
>>
>>Just read your article "The statistical structure of natural light patterns
>>determines perceived light intensity" (PNAS, June 8, 2004, vol. 101,
>>no. 23, 8745-8750).  In it you repeatedly state that the statistical
>>structure was acquired "during evolution".
>>
>>Why not assume that it is acquired during ontology, i.e. that
>>each individual learn it?

Several reasons: 1) for the most part,evolution has determined 
the wiring of the visual system; individual developoment only refines 
it. 2) newborn animals without significant postnatal experience 
behave as if they see fine (e.g., newborn rhesus monkeys); 3) 
electrophysiology in kittens has shown that visual cortical wiring is 
configured like the adult before the eyes open.

>>
>>The reason not to believe that it was acquired "during evolution"
>>would require that it is the same circuits in all the individuals of
>>the species. however, in the cortex, including the visual cortex, the
>>connectivity is stochastic, i.e. varies random across individuals.

This is not true, as many experimental observations have shown.

>>Therefore nothing like the circuits that are speculated in the
>>discussion of your article could be implemented in the cortex
>>by evolution.

You need to read a  text on neural development; try Purves and 
Licttman, Principles of Neural Development, which details why the way 
you are thinking about this is on the wrong track

Hope this is helpful.

dp

>>
>>Thanks,
>>
>>Yehouda Harpaz
>>http://human-brain.org


-- 
---
Dale Purves, M.D.
Director, Center for Cognitive Neuroscience,
Levine Sciences Research Center (Rm B203)
Box 90999
Duke University
Durham NC 27708


phone =919-684-6122
fax=919-681-0815
website=
http://www.purveslab.net


=====================================
My reply :
=====================================

----- Original Message ----- 
From: "Yehouda Harpaz" 
To: "Dale Purves" 
Sent: Saturday, September 18, 2004 3:58 PM
Subject: Re: How the bain knows the statistical structure of natural light


 
>>The reason not to believe that it was acquired "during evolution"
>>would require that it is the same circuits in all the individuals of
>>the species. however, in the cortex, including the visual cortex, the
>>connectivity is stochastic, i.e. varies random across individuals.
> 
> This is not true, as many experimental observations have shown.
 
Really? Can you give me a reference for such an observation?
I don't mean a text book, I mean primary research data. 
 
 


=====================================
His first reply:
=====================================
 
>>>The reason not to believe that it was acquired "during evolution"
>>>would require that it is the same circuits in all the individuals of
>>>the species. however, in the cortex, including the visual cortex, the
>>>connectivity is stochastic, i.e. varies random across individuals.
>>
>>This is not true, as many experimental observations have shown.
>
>Really? Can you give me a reference for such an observation?
>I don't mean a text book, I mean primary research data.

In the book I referred you to, many chapters and references are 
pertinent to appreciating the highly determined way neural 
development proceeds in the absence of experience (Hamburger's work 
on the development of spinal reflexes in embryos, for example). If 
you want some specific references to evidence showing that wiring is 
in place prior to experience (which then refines what early 
development determined by a bunch of developmental rules that 
evolved), take a look at the description of the classical work of 
Hubel and Wiesel described in Chapter 14.

dp



-- 
---
Dale Purves, M.D.
Director, Center for Cognitive Neuroscience,
Levine Sciences Research Center (Rm B203)
Box 90999
Duke University
Durham NC 27708


phone =919-684-6122
fax=919-681-0815
website=
http://www.purveslab.net


=====================================
My second reply:
=====================================
 

----- Original Message ----- 
From: "Yehouda Harpaz" 
To: "Dale Purves" 
Sent: Saturday, September 18, 2004 5:51 PM
Subject: Re: How the bain knows the statistical structure of natural light


 
> In the book I referred you to, many chapters and references are 
> pertinent to appreciating the highly determined way neural 
> development proceeds in the absence of experience (Hamburger's work 
> on the development of spinal reflexes in embryos, for example). If 
> you want some specific references to evidence showing that wiring is 
> in place prior to experience (which then refines what early 
> development determined by a bunch of developmental rules that 
> evolved), take a look at the description of the classical work of 
> Hubel and Wiesel described in Chapter 14.
> 
 
That is a typical example of a bogus reference, 
Hubel and Wiesel certainly have never shown that wiring 
is in place prior to experience, except in an a very gross
patterns. 

Meanwhile I looked up in the "Neuroscience" textbook that you 
have edited (http://www.sinauer.com/detail.php?id=7250). At least
in this book you didn't dare to make these claims, which makes it
better than other textbooks. 

While I was there I noticed that as usual you show the reader in
figures 23.3 and 23.6 the ocular dominance pattern from one 
animal, and don't tell the reader that the pattern varies across
individuals animals, thus giving the impression that it is fixed. To
your defence, the only one that I have ever seen make this point
is Francis Crick in the Astonishing Hypothesis.

The caption of 23.3 also erroneously claims that the pattern for 
humans is shown in figure 12.10.

Returning to the actual issue: the connectivity in the cortex 
varies randomly across individuals. There are some gross 
patterns, but are far from being wiring patterns. In numerical 
values, if specifying the connectivity of an individual cortex 
(compared to another cortex) would have require 10**14 bits if
they were both random,  the gross patterns that are seen may 
reduce the number of bits maybe to 10**13,  but not more. 
They are still different enough that nothing of interest can be 
actually be coded in it.

=====================================
His second reply:
=====================================
Yehouda

I'm not sure why you are so upset; I'm just trying to indicate to you 
the literature that you asked about, and don't have any axe to grind 
here. Basically, what you are asking about here is not an issue that 
anyone debates. The structure of the brain is fundamentally 
inherited, just like the structure of the rest of us, and bodies and 
brains are further shaped post-natally. Obviously, we are as 
individualistic in brain structure as in anything else, and no two 
people are alike in either body or brain. Moreover, developmental 
processes are plenty noisy. The reasons for all these statements are 
straightforward.

Again I hope this helps you get to where you want to go.

dp

PS I think you misunderstood why I was pointing you to the 
Hubel/Wiesel stuff. The pertinent evidence concerns the receptive 
field properties (wiring) of visual cortical neurons prior to 
experience, not ocular dominance columns (although those patterns are 
also in place before experience). They found the neuronal properties 
to be adult-like before the eyes opened.


>>In the book I referred you to, many chapters and references are 
>>pertinent to appreciating the highly determined way neural 
>>development proceeds in the absence of experience (Hamburger's work 
>>on the development of spinal reflexes in embryos, for example). If 
>>you want some specific references to evidence showing that wiring 
>>is in place prior to experience (which then refines what early 
>>development determined by a bunch of developmental rules that 
>>evolved), take a look at the description of the classical work of 
>>Hubel and Wiesel described in Chapter 14.
>>
>
>That is a typical example of a bogus reference, Hubel and Wiesel 
>certainly have never shown that wiring is in place prior to 
>experience, except in an a very gross
>patterns.
>Meanwhile I looked up in the "Neuroscience" textbook that you have 
>edited (http://www.sinauer.com/detail.php?id=7250). At least
>in this book you didn't dare to make these claims, which makes it
>better than other textbooks.
>While I was there I noticed that as usual you show the reader in
>figures 23.3 and 23.6 the ocular dominance pattern from one animal, 
>and don't tell the reader that the pattern varies across
>individuals animals, thus giving the impression that it is fixed. To
>your defence, the only one that I have ever seen make this point
>is Francis Crick in the Astonishing Hypothesis.
>
>The caption of 23.3 also erroneously claims that the pattern for 
>humans is shown in figure 12.10.
>
>Returning to the actual issue: the connectivity in the cortex varies 
>randomly across individuals. There are some gross patterns, but are 
>far from being wiring patterns. In numerical values, if specifying 
>the connectivity of an individual cortex (compared to another 
>cortex) would have require 10**14 bits if
>they were both random,  the gross patterns that are seen may reduce 
>the number of bits maybe to 10**13,  but not more. They are still 
>different enough that nothing of interest can be actually be coded 
>in it.


-- 
---
Dale Purves, M.D.
Director, Center for Cognitive Neuroscience,
Levine Sciences Research Center (Rm B203)
Box 90999
Duke University
Durham NC 27708


phone =919-684-6122
fax=919-681-0815
website=
http://www.purveslab.net