I’m starting to dive into scholarly articles about cortical structures and I found an interesting, albeit old, article claiming the cortical column may not serve a particular function in a general animal’s cortex and may not be the “basic functional entity of the cortex”. I’m not sure of their conclusions, although the evidence presented that some animals have cortical columns in certain regions while others don’t is interesting. However, I did find the article had a lot of detailed history and references on how cortical columns and minicolumns were discovered. Thoughts?
We are aware of this article. It has been brought up several times on the mailinglist. @subutai have written a nice summary of this article in our annotated bibliography
I copied and pasted his summary below
"1.4 Horton: minicolumn naysayer
This paper  is a good summary of the arguments against the importance of columns. We should point out that the term ”column” is very ambiguous - this article reviews all the variations but is also a bit muddled about them. In HTM the term refers strictly to mini-columns. Phenomena like macro columns, hyper columns, ocular dominance columns, etc. are emergent phenomena that group together lots of mini columns. They probably arise as a function of learning, and not really required in a strict sense. You can get lots of variations in these macro-column structures and that would be fine. So most of the article does not contradict HTM theory.
Horton completely disagrees with the notion that minicolumns have any functional significant what- soever. He states, ”dubious is the concept that minicolumns are basic modular units of the adult cortex, rather than simply remnants of foetal development”.
Again we need to be precise about language. Most people debate whether the mini-column is THE basic unit of cortical function, as if the connections between cells in a mini-column compute some- thing. In HTM theory we dont make this claim, its false, and it leads to a lot of confusion. The cellular layer, not the mini-column, is a unit of processing. All we claim is that cells in a mini- column share similar feedforward response properties and that they mutually inhibit each other. We dont require this inhibition be local to just the mini-column, in fact it may very well extend to sur- rounding columns. Horton admits that mini-columns exist physically (pretty much everyone does), but he argues that they arent functionally important.
One of his arguments is that there is a profuse connectivity across columns. This argues against the idea that the mini-column is the unit of computation but it is what we need in HTM layers.
The only claim really under contention, vis–vis the Horton paper, is whether cells in a mini-column share feedforward receptive fields. As the paper points out it is extremely difficult to detect this, as mini-columns are very narrow (35-50um wide and 2500um tall). It is also likely that the cells in a mini-column are not exactly linear. They can move about as much as they want as long as the connections are maintained. Neocortical tissue is more like tapioca and less like wood.
We propose that cells in a mini-column will exhibit very different responses during learned se- quences. This exact behavior has been documented. However, some people argue this proves that the cells in the mini-column do not share common receptive fields! So you have to be careful when reading claims. Was the animal awake, was it a brain slice, what exact stimulus is being used?
HTM theory provides an elegant explanation for the conflicting data and confusion about mini- columns."
Oh, I must have missed the annotated bibliography. Thanks for the link @ycui and the informative response! I apologies for the repeat (it must get slightly annoying).