Yehouda Harpaz yh@maldoo.com 6 Jan 96 [last updated 20 APR 2012] related texts
In the motor areas, the neurons that control specific muscles cluster
as well. This is a special case, which is not applicable to the rest
of the cortex. In either case, the 'columns' do not seem to do anything
with the internal organization of the cortex.
In many cases, the idea of "columns" is used even if the evidence
does not show columns at all. In general, neuroscientists use
"columnar organization" to describe cases where it seem that some
property (e.g. direction selectivity) varies along the sheet of the
cortex, but not in the orthogonal direction. That is clearly not
"columnar", as it doesn't show columns (district vertical units), but
that doesn't stop neuroscientists.
A good example is given in
Neuroscience: exploring the brain (Bear, Connors & Paradiso,
Second edition, Williams & Wilkins, 2001). In their figure 10.25, they
show the change in direction sensitivity along the cortex. This change
is clearly continuous. In figure 10.26, they show a "module" according
to Hubel & Wiesel suggestion. In this figure, there are vertical lines
dividing the "module" to columns according to sensitivity to
direction. However, this division is based on the data in Figure
10.25, in which there is no such divisions, because it is continuous.
([18 Apr 2002] The two figures are available here,
p.16, title "Cortical Module". The graph of dots is 10.25, the box is
10.26; [1Aug2003] Scanned version ).
In this case, at least the authors show the original data, so the reader can figure out that the lines in Figure 10.26 are fictitious. However, it will require a very alert reader to figure out the discrepency. A person using the book as a reference, rather than reading through it sequentially, is even less likely to identify it. Most of other books don't actually show the original data, so the reader doesn't have a chance to figure out that the lines are fictitious. It should also be noted that the "borders" of the "module" that is present in figure 10.26 are completely fictitious too.
In this introduction to special issue of Cerebral cortex about "computation in cortical columns" (Mountcastle, Cerebral Cortex, Vol. 13, No. 1, 2-4, January 2003 ), mountcatsle, who first advanced the concept of columns, mentions in the first section ".. horizontal disjunctions in functional properties .." and ".. abrupt transitions in functional properties which separate one column from the next." However, these statements are not accompanied by any kind of reference, and he seems to forget them in the following sections. Obviously, none of the articles in the issue even tries to discuss "disjunctions" or "abrupt transitions", as they all know that there aren't such disjunctions or transitions.
In the preface to the special issue (Cereb. Cortex 2003 13: 1) above, the authors start by saying:
The six layered neocortical columnar microcircuit, implemented most extensively in human brain...So they don't have a doubt that there is a "neocortical columnar microcircuit". In the list of question they have, they have two questions that express some doubts: "Does a column function as a discrete computational unit?" and "to what extent does a functionally defined column reflect a distinct anatomical entity?". Even in those question they don't doubt the concept of columns. As I wrote above the first question is not discussed in the following articles.
The second question is touched to some extent. For example, Lund et al (Cereb. Cortex 2003 13: 15-24;full text) say:
There are many reasons to reject the concept that a column of cortical issue is functional entity, on both physiological and anatomical grounds.They then list these grounds (not mentioning explicitly not being discrete). However, they still refrain from explicitly rejecting the column concept. In fact, their following discussion is written with implict acceptance of the concept of columns, even when they bring an evidence that clearly contradict it. It is most obvious when they write (p.20, top right):
Thus, while the columns of terminals have definite boundaries, they are defined by what connects to them and what they connect to tangentially across the cortex. They do not have unique occupancy of a 250-300 µm diameter column of cortical tissue - many other columns share portions of the same territory.Clearly, if they share territory with each other, the entitities they discuss are not columns.
In some cases, it is claimed that the fact that the activity of neurons correlates better with close neurons shows columnar organization, but that is plain nonsense, as that is what we should expect if the cortex is a continuous sheet. The important question is whether there is anything that separates the cortex to small units (columns), and nothing like this is seen in the cortex.
The only observation that could be called 'evidence' for the existence of columns is that the density of neurons is not homogenuous in the cortex, and neurons tend to clusters in small regions. These may be called 'columns', but their dimensions (20-50 microns) are much smaller than the dimensions of the 'columns' which are normally discussed (several 100s of microns). Thus this observation does not support to "columnar organization" the way it is promoted. Normally this clustering is ignored, because most people cannot think of any funcional significance to attach to it.
[18Jul2003] Just found this article (Edward Jones, Proc. Natl. Acad. Sci. USA. 2000 May 9; 97 (10): 5019-5021). It is interesting because while he is not accepting all the nonsense that goes around, is quite skeptical about the significance of the columns, and does bring substantial evidence against it, he still ignores the most important counter-facts. For example, he discusses the bundles of apical-dendrites of pyramidal cells (approximately middle of the text, search for "bundling of apical dendrites"). He then says in the next paragraph that the dendrites of the cells extend horizontally for hundreds of microns, which undermines the idea that the bundles are a functional unit. But he does not mention that the axons of these cells spread much further. In the figure (Fig. 2) the axon of the pyramidal cell appears as an insignificant stump. The only mention of the axons of pyramidal cells is in sentence that says that "for a time it seemed...[that axons].. would eventually yield a basic circuit diagram", and there is no mention of their actual spread. Thus even this article, which looks quite even-handed on first reading, actually ignores the main relevant fact for evaluating the idea of columns (spread of pyramidal axons, which makes the concept totally untenable), and is in fact quite misleading because of that.
Other "evidence" for columnar organization is that in many cases the synapses of an axonal process, or even a bundle of them, are concentrated in restricted areas ("columns"). That would have make sense if the division to "columns" by the input processes into a region was the same for all the input processes. Since it is not, it is simply daft to consider this as evidence for columnar organization.
That doesn't bother many researchers to refer to "evidence" for columns and columnar organization even when the "evidence" does not actually support it. It can be regarded as an example of the confusing and ambiguous usage of terms error, where "columnar organization" means "organization in columns" when it is already assumed, and "topographic organization" when evidence is presented for it.
As an example, here is the research summary of some laboratory, where it says:
This radial stack of neurons forms the cortical column, a minimal unit of neurons and connections required to analyze a small portion of the sensory world.This, of course, has no basis whatsoever in experimental evidence and is a totally spurious statement. It is an interesting question if the person that wrote it actually believes this statement, but it is diffcult to find alternative interpretation ([16Jan2002] Her reply to my query. She seems to refer to the hypercolumns ).
The "columnar organization" is by now very pervasive, and this kind of fictitious ideas are being taught in many places. for example, in an online turotial for psychology students Athabasca Universirty, the last senetnce says:
Experiments reveal that these vertical columns form mini circuits that perform unique functions.Which is plain false (unless "mini circuits" and "unique function" are stripped of any meaning).
This article (Kazushige et al, Nature Neuroscience, V 4, No 8, august 2001) (full text) gives a good example of hyping columns. There is nothing in their data that is even related to the concept of columns, but that doesn't stop the authors from discussing their findings if they show columns, "each of which represents a visual feature". It seems these authors don't have the ability to consider the cortex without thinking about columns. The interesting observation is that such skewed thinking doesn't bother the reviewers and editor of Nature Neuroscience.
[4Jan2003] A stunning example appears in the new edition of the
textbook "Neurophysiology" by R.H.S Carpenter(2003, Fourth edition,
ISBN 0 340 80872 1). On page 373 it says:
This statement and figure are obviously totally false. The author does
not give any reference, and the comment in p.403 just says (after some
comment on drinks etc.) "see chapter 14" (next chapter), but this
chapter has nothing that is relevant.
The stunning thing is that this blatantly false statement and figure
comes from a neuroscientist ( "A reader in Oculomotor neurophysiology,
University of Cambridge". In the UK, "Reader" is the second most
senior academic title.). Presumably he is not telling us lies, so he
actually believes what he writes. Thus the rhetoric of researchers
confuses not only lay people, but also quite senior neuroscientists.
This
encyclopedia article (Cortical Columns, Carreira-Perpinan &
Goodhill, Encyclopedia of Cognitive Science, Macmillan Publishers
Ltd., 2002) also declares, without any reference to supporting evidence:
In this course
material at the University of Massachusetts Medical School they say:
[6 Apr 2004] This
article (Casanova et al, The Neuroscientist V. 9, No 6, 496
- 507, Disruption in the Inhibitory Architecture of the Cell
Minicolumn: Implications for Autism) says:
[30Apr2004] Just to add to the confusion, in this
article (Petersen et al, Science, V 304, 5671,
739-742, 30 April 2004) they write "column" when they mean "barrel",
but you cannot figure this out from the abstract only. To make it
worse, the abstract makes a strong impression that the "columns" (i.e.
barrels) are distinct units, thus gives the reader a strong
impression of distinct columns.
Kathleen
Rockland, who researches connectivity in the cortex, doesn't find
anything that fit "columns", but she seems to have a problem saying
that. For example, in this
article (Rockland, Journal of Neurocytology 31 (3-5): 247-253,
2002) she writes:
[15 Feb 2005] The latest Nature contains an article with the
title "Functional imaging with cellular resolution reveals precise
micro-architecture in visual cortex" (Ohki, Chung, Ch'NG, Kara & Reid,
Nature 433, 597 - 603 (10 February 2005)). First thing to note is
that, as usual, "precise" doesn't actually mean precise in the usual
sense. They didn't find anything that corresponds to anything else
precisely. The second thing is that they write about "columnar
borders", even though the transitions they show do not correspond to
the concept of columns. They see pretty sharp transitions between
cells tuned to opposite directions (180 degrees change). The concept
of columns is about change in small steps, but that doesn't bother the
authors or the reviewers. They go as far as saying: "The strict
dichotomy between opposite directions offered am ideal opportunity to
study the precision of a columnar border in the cerebral cortex",
which is clearly nonsense. This article shows how the concepts of
"precision" and "columns" dominate the way that scientists in the
domain think.
Figures 1 and 5 in this
article (Hirsch et al, Nature Neuroscience, 2003, 6: 1300-1308;
full text) show
examples of neurons in the cortex, and it obvious that they are not
restricted to any kind of columns.
In this
review (The cortical column: a structure without a function,
Horton & Adams, Philos Trans R Soc Lond B Biol Sci. 2005 Apr
29;360(1456):837-62) they explicitly state that the column concept is
useless. See also the home page of
the more "cognitive" of the authors (the other author is more
"clinical").
I would have thought that the review above would kill the column concept. It
seems to have helped, but the conept is still alive. For example,
there is The Blue Brain
Project, which list as their main achievement "simulating a rat
cortical column".
In their Glossary page,
they list nine terms, of which seven (Axon, Ion Channel, Cortex,
Dendrite, Neuron, Neurotransmitters, Synapse) correspond to real
physical entities, one (Neuroinformatics) is a branch of science, and
the ninth one is cortical column. According to them, it "...
constitutes a unit associated with a specific function." They
obviously didn't read the review above.
Here is some email exchange with their
leader, Markram. Obviously, he is not going to change his mind
whatever the evidence. The interesting thing is that he still gets
substantial funding and support.
It is important to emphasize that cells in any one column talk a great
deal to each other, but essentially turn their backs on their
neighbours, like people gossiping at different tables in pubs (Fig
13.10).
(Italics in the source).
Fig 13.10 shows few cylinders with arrows inside them, and one arrow
connecting two cylinders, and the caption says: "Columns debate within
themselves, then send their conclusions to other columns."
The majority of intracortical circuits are local, connecting neurons
within the same columns, with only a minority of connections being
between columns.
In principle, if we translate "within the same column" to "close to
each other", this statement is true, but this is not a translation
that readers can be reasonably expected to make. All readers will
interpret it to mean that there are distinguishable columns, with higher
connectivity inside, which is false.
This is specially odd, because otherwise this article is more balanced
than other articles.
Migrating cells from the ventricular zone travel radially, hence most
connectivity in mature cortex is vertical, laying the foundation for
the cortical columns which were discussed in detail in the visual
system.
The middle bit ("most connectivity in mature cortex is vertical") is
trivially false, but that doesn't stop them from telling it to their
students.
The modular arrangement of the neocortex is based on the cell
minicolumn: a self-contained ecosystem of neurons and their afferent,
efferent, and interneuronal connections.
The mini-column is clearly not self-contained, because all the neurons
in it have more connections to neurons outside the mini-column than
to neurons inside it. In the case of the pyramidal cells (which are
the majority of cells), the ratio is tens or hundreds to one. That
doesn't bother these researchers. Note that the point is completely
irrelevant to their article, but they make it anyway.
The viewpoint adopted will be that "columnarity" in this context is
not a matter of "repetitive" identical units, but rather may be a
mechanism for selective re-combinations of inputs through the
3-dimensional cortex.
It is an interesting question whether that is because she really thinks
"columnarity" is a useful concept, or because she feels she will not
be able to publish papers if she says otherwise.
Online articles that show actual neurons in the cortex
Somebody actually figured it out
[5 Nov 2005]
Not dead yet
[20 Apr 2012]