Saturday, October 16, 2010

Brain Mechanisms of Vision (Hubel & Wiesel)

With the proverbial appeal to authority, it might prove useful at this juncture to review a few key passages from an article entitled "Brain Mechanisms of Vision" written in 1979 by Nobel laureates David Hubel and Torsten Wiesel for Scientific American in a special issue devoted to the brain (Vol. 241, No. 3), two years prior to receiving their Nobel Prize, which they shared with Caltech biologist, Roger Sperry, in 1981:

Understanding of this large and indispensable organ [the brain] is still woefully deficient. This is partly because it is very complex, not only structurally but also in its functions, and partly because neurobiologists' intuitions about the functions have so often been wrong (p. 150).

In the primary visual cortex the map is uncomplicated by breaks and discontinuities except for the remarkable split of the visual world down the exact middle, with the left half projected to the right cerebral cortex and the right half projected to the left cortex (p. 150).

Important as the advances in mapping cortical projections were, they tended for some time to divert thought from the real problem of just how the brain analyzes information. It was as though the representation could be an end in itself instead of serving a more subtle purpose--as though to cater to some little green man who sat inside the head and surveyed the images playing across the cortex. [F]or vision, at least, the world is represented in a far more distorted way; any little green man trying to glean information from the cortical projection would be puzzled indeed (p. 152).

It follows that this [the visual cortex] cannot by any stretch of the imagination be the place where actual perception is enshrined. Whatever these cortical areas are doing, it must be some kind of local analysis of the sensory world. One can only assume that as the information on vision or touch or sound is relayed from one cortical area to the next the map becomes progressively more blurred and the information more abstract (p. 152).

It is important to realize that this part of the cortex [the visual areas] is operating only locally, on bits of the form; how the entire form is analyzed or handled by the brain--how this information is worked on and synthesized at later stages, if indeed it is--is still not known (p. 155).
From their subsequent book, published 26 years later, Brain and Visual Perception: The Story of a 25-year Collaboration (OUP), it would seem that this general characterization remains largely the same, in spite of further discoveries of theirs and those made by other researchers, such as Ramachandran.


  1. I remember seeing this when it first came out, and I have always been very impressed by Hubel and Wiesel's work since they were actually getting empirical results from a laboratory here. One does have to be a little bit careful with regress arguments like theirs though, since it may not take another little man to espy what is going on. My own theory is that an embedded two dimensional space may do the trick, if not here perhaps elsewhere in the brain, and that there is at most a topologic isomorphism of close points to close points; perhaps with the events in visual space only being sensitive to neurons having special properties, possibly firing at certain frequencies or possibly something else. In any event I think that there are replies that can be made to this type of regress argument.

  2. On the face of it one might take H & W's "little green man" to imply a regress (argument), but I am not sure that was their intention. Rather they were going back to an idea at least as old as Newton, that the eyes conveyed images to the brain (or sensorium), a more contemporary version being John's analogy of the TV screen, the point being to get the visual image *into* the organism (to put it in biological terms), analogous to remote "televisual" sensing as a general principle, rather than sensing by contact as in the case of touch in the most primitive organisms that have no optics--in other words, the whole seeming purpose of the exteroceptive senses, in which objects do not touch the organism, but an intermediate medium does (light, air, odorants).

    Basic to a geometrical account is congruence relations and isomorphism is not sufficient to meet the equivalent of rigid bodies required by a metric space, if sufficient for a topological relationship. The fact that the visual areas are divided in half between the two hemispheres creates an almost insuperable topological problem, because it involves tearing relative to VS.

    The most interesting point H & W make IMO is that it is not clear the brain is actually engaged in anything like figural synthesis, which creates a major problem for any kind of congruence relation between VS and brain structure/activity. Rather they seem to be saying that the pieces of Humpty Dumpty are not reassembled further upstream in visual "processing."

  3. It is kind of a case of "pick your poison" here, any position has its share of paradoxes, and thus I am satisfied that the truth here is rather odd, at least from the standpoint of common sense.
    I do recall that I once did discuss some of these matters with Torsten Wiesel and that he referred me to the work of Wilder Penfield who claimed that consciousness probably resides in the thalamus. But again, I think that the best that we can do is to identify the implications of various hypotheses. One other point is that it would seem that the neural correlates of visual experiences must involve in some way inputs from both eyes due to the existence of stereopsis.

  4. One alternative is that rather than opening a can of worms and getting involved with debates about the validity of causality (by which I mean classical causality, i.e., action by contact), perhaps we might consider the possibility that the causal theory of perception is wrong--not the whole of it necessarily, but the link to conscious visual experience which, after all, for our discussion is the most important link.

    One of the critics of brain mapping was British neurologist William Gooddy (d. 2006), who saw the brain as being something like a timing mechanism more than anything else. In a quadruple review of memoirs by neurologists that appeared in the journal "Brain" five years ago, the author of the review summarized the essentials of Gooddy's thinking on these matters:

    "Gooddy was influenced much by the writings of Walther Reise, and a school of thought quite unpopular in the 1950s rejected even now by those who trade in today's neo-phrenological neuroimaging-based neurological syntheses. To quote: ‘the three words, “function”, “represented” and “areas” have no valid neurological (or other) meaning. None of us really knows what functions represent or what represents a function’.

    "A follower of Hughlings Jackson, he reminds us that the human brain has been several billion years evolving, that it functions in time as well as in space, and that ‘the highest achievement of nervous activity is to detect and obliterate every threat to our survival’. He draws attention to the neglected fact that we spend our time talking about neurons and cerebral activity, when the former comprise ‘only 8% of the neural network of fibres in the brain’, and the granule cells of the cerebellar cortex account for nearly 7/8 of all neurons! Yet their state changes with chronons (the shortest known unit of time), not our more manageable seconds or even milliseconds, giving 4339 potential different brain states in any individual in the course of a day.

    "His bêtes noires include Paul Broca, Carl Wernicke and all succeeding localizers, even those of today who use modern brain imaging techniques but not their own brains. The methods ‘which now assist us in localising cerebral (and other) structural defects, are also still being used by misguided Luftmensch to “prove” the existence of “centres” and “functions”, in further support of one of the greatest fallacies of all scientific theory’.

    "William Gooddy explains his own theory of a ‘new kind of nervous activity’, which has a bit, but not much, in common with today's connectionist theories, and is based on concepts of continuous cycles of nervous activity (sensory–motor–sensory–motor and so on), which hint at Antonio Damasio's latest ideas. Such cycles become blocked by lesions, after which ‘the whole nervous system is altered. The entire person is never the same again.’"

    (Michael Trimble, “The Tales Neurologists Tell,” Brain, Vol. 128, No. 3, p. 690)

  5. To show that the late Dr. Gooddy would be quite at home on this blog, I quote the abstract from a paper of his published in 1966, "Disorders of Orientation in Space-Time":

    "This paper suggests a unitary hypothesis to link neuro-psychiatric modes of thought with those of modern high science. The scientist analyses and objectifies in his great works of art those which the artist, knowing the same breadth of perception, memory and intuitive rightness, makes into the brief elegant or poetic phrase of speech, of literature, of music. The scientist and the artist each has his highest abstraction, his formula.

    "The neurologist should find it his duty to harmonize the ways of the scientist and the artist; for when they (or he) become sick in mind or brain, they must turn to a creative neurological experience for remedy or alleviation.

    "We may end with a quotation illustrative of the theme of this work, again taken from Einstein's 'Relativity': 'I wished to show that space-time is not necessarily something to which one can ascribe a separate existence independently of the actual objects of physical reality.'"
    (British J. of Psychiatry, 112: 661-670)

  6. Another holistic theory you might have mentioned here Bill is Pribram's model of memory as being like a hologram in critism of Lashley's concept of localized engrams, and even Marr speaks of ganglion cells in the retina as performing something like Fourier transforms. My question though how is anything like this supposed to result (in a causal or any other sense) in specific conscious visual experiences.

  7. The problem with the Bohm-Pribram model is how you get the "explicate order" out of the "implicate order," i.e., how you get the equivalent of the hologram (the image) out of the holographic matrix, like the image-less swirls on a holographic plate. It is somewhat analogous to the problem of the "sections" that John speaks of: There needs to be something that literally makes the sections, much like conical sections--they must be made from a cone. Something must do the slicing. If VS is a lower dimensional section of a space of some embedding space, something *makes* it that way, otherwise there would seem to be nothing to curtail its extension into the embedding space. These are the "chains" that both John and Saul-Paul Sirag have talked about.

    I don't recall Marr's claim about ganglion cells. In conjunction with his holographic theory Pribram invoked the work of Fergus Campbell and Russell DeValois, who proposed that the visual cortex is performing a spatial frequency analysis on the visual input. That ultimately seemed to get nowhere, though, and has been abandoned by visual neuroscience. Rereading an old chapter of Pribram's ("Holonomy and Structure in the Organization of Perception") from 1977, I am considerably less impressed by the ideas than I was when I first read it. Roger Sperry at Caltech thought both Pribram's and Bohm's ideas were for the birds--as did Bela Julesz, during his stint in our lab there--he even went so far as to ask Pribram to stop quoting him in his papers!

    I read a paper by Gooddy many years ago that impressed me in which he basically debunked the whole notion of cerebral "representation," and he evidently stuck by that view for the balance of his career. I have not seen his two monographs "Time and the Nervous System" and "Neurological Cosmology" in which he apparently addresses that topic again, but I know that his main claim to fame is research on disorders of time perception (see his entry in Vol. 3 of the "Handbook of Clinical Neurology," viz. "Disorders of the Time Sense").

  8. Having a chance yesterday to chat with John at UCSD about the neurophysiological account of so-called "saccadic suppression," which actually makes a good deal of sense to me (at least how he explained it), I was struck again by the overwhelming impression that what he was describing had nothing to do with generating VS, but with indeed some kind of "processing" of visual input, largely for control of eye movements. Already we have found a great deal of brain activity associated with eye movement control, and it makes me wonder if this is not primarily what the visual system is doing rather than perfecting a visual image seen by a homunculus? (Just now I am thinking of J.B. Watson's iconoclastic "motor theory of thought" in this regard, which ultimately came under the theoretical rubric of "peripheralism").

    As John has explained here, he would like to interpret saccadic suppression and the filling in of scotomata through a process comparable to ICT as evidence of how the visual system is like a television system, ultimately generating VS. But if one looks at the "diagnosis" given by H & W now over 30 years ago, the neurophysiology does not support that interpretation. Yet in spite of that much vision research seems to proceed on the assumption that it is the visual system's job to create a high quality visual image for the observer, rather like television engineers labored over the problems of signal transmission and display in creating the TV we have now long enjoyed and come to take for granted.

  9. I tend to agree with you Bill about the merits of Pribram's theory, although it is at least an attempt at describing a holistic method of encoding information. In any event I think I have a better idea now of what your theory is not, but perhaps you could at least attempt to restate its positive content.

  10. I suppose the best way to describe my own work is as being similar to what Friedrich Paulsen (whom I have quoted here) referred to as "critical realism," specifically as pertains to current theory on visual system function, and perception more generally.

    As soon as I can retrieve the computer file for it, my intention has been to post a "working paper" of something I wrote many years ago on basic topological properties of VS, with the strategy of working backwards from them to how they might arise. In this regard I used as a model Graham Nerlich's account of physical geometry in his book "The Shape of Space."

    It is my contention that once we have a better formal description of VS and its properties, we may be in a much better position in attempting to systematically relate them to physical correlates (or the lack thereof). But as I see it, right now all that has been geometrized are *objects* in VS, not the space itself, as Einstein did with space-time by attributing its curvature to gravity. We have no corresponding explanation for curvature in VS, and mostly rely on the optics of perspective projection to explain the structure of what we see. For example, the various "constancies" might be compared to the action of specific forces, rather than being just vague "psychological" processes.

    My feeling is that the blog should function as a "think tank," with the eventual goal of submitting some of the fruits of it as a paper to a journal, either individually or as a group.

  11. Mea culpa--it was not Paulsen's book that discussed critical realism but the "Introduction to Philosophy" by Oswald Külpe:
    (My confusion was due to both books having virtually the same title!)

  12. I look forward to seeing your working paper Bill, but in the meantime it might help if you said a bit more about how your position is a version of "critical realism." I have heard my own position described as "critical realism" although I prefer the term "representational realism" and I am not sure that everyone uses "critical realism" the same way. Does Kulpe have a definition?

  13. Somehow I completely missed my unintentional pun (culpa and Külpe!), but Külpe discusses critical realism on pp. 195 & 202f of his book.

    My "critical" stance on realism is perhaps closest to that of Richard Avenarius, who sought to dispel the excess (and often baseless) mentalizing and subjectivizing of phenomena with his principle of the "elimination of introjection" as expounded in his magnum opus, "Kritik der reinen Erfahrung" (1888-90). One might say it is the converse of Churchland's "eliminative materialism."

    I believe we have been led astray by a kind of neoplatonic view of the world that attributes greater reality to things we cannot experience than to our own experience. In this regard, scientific realism strikes me as somewhat arbitrary, if well reasoned (and justified). My own approach is often historical because sometimes a misstep in thinking has been made long ago and has not been challenged or corrected, and has thus been perpetuated.

  14. Here is an entry on Avenarius:
    But unlike him, I do not feel the need to commit to monism (neutral or other), nor dualism for that matter--thus my invoking Newton (above)!

  15. One reason that I never particularly cared for the term "critical realism" is that it is possible to be "critical" about various things that one may be a realist about. I also have never cared for eliminative materialism, but it seems to me that its opposite would have to be some version of idealism (where one eliminates material entities apart from our perceptions of them). Perhaps you just wish to be an agnostic on this last issue though, or don't wish to be further classified. On this last issue, one thing that always bothers me about John Searle (who I admire in some respects) is that he keeps rejecting any attempts to classify his position, while it seems to me to be clear that he holds a version of property dualism. So rejecting being classified can have its problems as well.

  16. If NDEs are as they appear to be, literally involving some form of survival of bodily death and transcendence into another existence, then "reality" is far stranger than physics currently realizes, even with so-called "quantum weirdness."

    Alternatively nothing could be more banal and unconvincing than Susan Blackmore's crude neuroreductionist (and unprovable) attempts to explain away NDEs as representing the activity (read: malfunctioning) of a dying brain, any more than Michael Persinger's pathetic mission to show that all religious experience and psi is due to malfunctioning temporal lobes. Yet these pedestrian and unimaginative physicalistic thinkers will take anything over the paranormal, even if it involves cutting off their noses to spite their faces. It never seems to even occur to them that science itself may be both incomplete and in fundamental senses, just wrong.

    For these reasons my inclination is to believe in something like body-soul dualism, simply on phenomenological grounds alone, not being convinced of the skeptic's dismissive arguments. But what I may choose to believe, unfortunately, may not help serve the purpose of elucidating the reality of the matter!

  17. With respect to "critical realism," I also had in mind a passage from another introductory philosophy text, "A Brief History of Modern Philosophy," by Danish philosopher Harald Høffding (1843-1931), who was Niels Bohr's philosophy teacher. Perhaps significantly Høffding brings in critical realism in the context of Wilhelm Wundt, erstwhile chair of Leipzig, who founded the first experimental laboratory of psychology in the 1860s:

    In fact, Wundt published in 1896 a whole paper devoted to the very topic of naive vs. critical realism:
    It apparently has not been translated into English. Perhaps not by coincidence, the paper was very critical of Avenarius, so much so in fact that many having read it did not give Avenarius and his "empiriocriticism" his due in response(this may be a better term for my own stance, for that matter, than critical realism). This was a mistake according to Adam Wiegner, author of
    "Observation, Hypothesis, Introspection" (1964; Engl. trans. 2005), because Wundt misunderstood some of his basic premises:

  18. Alas, I still need to convert my old WordStar file to a Word file in order to post my working paper here, and am investigating ways to accomplish that (short of purchasing some expensive conversion software for that purpose).

    Meanwhile I found an interesting little paper by Crick and Koch that appeared in "Nature" in 1996, "Are we aware of neural activity in primary visual cortex?" in which the authors close by saying: "We have yet to track down the location and nature of the neural correlates of visual awareness," having ruled out V1 as a candidate. That is already 17 years after H & W's paper which had reached the same conclusion and it lends support to our general consensus of opinion here that such "correlates" are so elusive as to be virtually non-existent!