Book review Brain (2005), 128, 1226–1229 Measuring sight by sound In proposing a vote of thanks for the three Special University Lectures in Physiology that David Hubel delivered at University College, London, in February 1965, the late Professor Joseph Barcroft said that they belonged in the same league as those he had been privileged to have heard from Ramon y Cajal, Sir Charles Sherrington, Lord Adrian and Sir John Eccles, each one of whom had given major insights into the nervous system. The accolade was nothing if not deserved. The lectures were brilliantly delivered, and remain in my memory as amongst the finest that I have heard on any subject. They described the work that David Hubel and Torsten Wiesel had begun in 1958, a collaboration that was to last for 25 years and to revolutionize our knowledge of the workings of the visual brain. Brain and Visual Perception is a record of that collaboration, most of it in the form of collected papers, with a foreword and afterword to each. Although almost all of the latter are written by David Hubel, each author has contributed an autobiographical article. The entire book is an inspiration to read. The original papers and the additional chapters are beautifully written—which means that they are stylistically elegant, free from jargon and cliché and, above all, devoid of the current, vulgar, craze for acronyms and abbreviations and of other devices that serve to make science even more inaccessible. In fact, a re-reading of these papers in the chronological order in which they are presented here serves to correct the impression given by the authors that their work was nothing more than a sort of fact-finding mission, devoid of any hypothesis. For, collectively, they give a powerful sense of the logical development of work that culminated in a detailed understanding of the functions and functioning of area 17 of the visual cortex. When the authors began their work on the visual cortex, and more specifically on the primary visual cortex or area V1 (also known as the striate cortex or area 17), not a great deal was known. The visual cortex had been established as the recipient of fibres from the retinas through the optic radiation. The work of Salomon Henschen in Sweden, Tatsuji Inouye in Japan and Sir Gordon Holmes in England had established that there is a topographic map of the retina, and hence of the visual field, within V1 which, in the mid-twentieth century, was considered to be the sole visual area in the brain. By the time their collaboration was finished, Hubel and Wiesel had shown that the ordering of V1 is much more precise and elegant; they had demonstrated the presence and shape of ocular dominance and orientation columns within it; established the details of connections between the geniculate nucleus and V1; and asked daring questions about the role that nature and nurture play in its development. Their work on V1 also led them to develop what has come to be known as the hierarchical model of visual BRAIN AND VISUAL PERCEPTION The story of a 25-year collaboration By David H. Hubel and Torsten N. Wiesel 2004. New York: Oxford University Press. Price £29.99 ISBN 0-19-517618-9 perception, which supposes that the same details of the visual world are reanalysed at increasing levels of complexity. In this, they were substantially wrong, since it is now generally agreed that the simultaneous, parallel, processing of different attributes of the visual world is the modus operandi of the visual brain although a hierarchical strategy may characterize each of the parallel systems. Read today, some 50 years after the initial work was published, the papers still retain their freshness and their capacity to arouse wonder, not only at the way in which nature has elaborated such an impressive organ, but also at the tenacity and the powerful conceptual thinking that was behind their collected work. Several factors went into making this collaboration one of the most outstanding in neuroscience. There was, of course, the element of chance behind many of the great leaps that came from their collaboration, starting with the fact that both happened to be in Steve Kuffler’s laboratory at Johns Hopkins. The epoch-making discovery of orientation-selective cells in the cortex was itself a chance occurrence, arising from the fortuitous insertion of a slide into the projector. The imaginative work that led to a direct anatomical demonstration of ocular dominance columns in the striate cortex was made possible because Janet Chen, an expert histology technician, had by chance moved to Boston; the failure to notice any distribution of autoradiographic material in striate cortex after eye injections was corrected when a chance visit to Ray Guillery alerted Hubel to the fact that its distribution is best viewed in dark-field illumination. But chance was merely a catalyst, and to them it was ‘more a matter of bull-headed persistence, a refusal to give up when we seemed to be getting nowhere’. The truth—which they themselves cannot utter—is that what distinguished their work from that of others was a conceptual # The Author (2005). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: journals.permissions@oupjournals.org Book review David H. Hubel Torsten N. Wiesel brilliance; and with that kind of intuitive brilliance fancy methodology was unnecessary. Before Hubel and Wiesel started their work, most believed that the visual cortex would be activated by light, as such. Failure to achieve results with diffuse light stimulation was a trigger for the construction of more sophisticated apparatus. In Germany, Richard Jung and his colleagues spent much time and effort constructing ‘elaborate machines’ to activate the cells with diffuse light. Jung was later to recount that he might have discovered the orientation-selective cells during his 5 years’ work on the visual cortex if he had used a stick ‘instead of the quantifying machine’. Hubel and Wiesel had themselves used a now outmoded approach to stimulate the visual system, when that chance insertion of a slide into the projector elicited a wild response from a cell. Their concern from then on was to simplify the machinery and make the cells a good deal easier to study. For them, ‘just as important as stubbornness, in getting results, was almost certainly the simplicity, the looseness, of our methods of stimulation. The incredible crudeness of our first slide projectors . . . and our refusal to waste time bothering with measuring intensities, rates of movement and so on . . . all worked in our favor’. The term ‘loose’ is curiously out of place, given how much valuable information was obtained with it, and how little from sophisticated stimuli. From then on they seemed to have developed a lasting contempt for fancy 1227 machinery, Hubel once boasting that the best use they could make of their computer was to heat the laboratory. How could one avoid this disdain when papers, published as late as 1977, using sophisticated methodology and advanced computer analysis, did not improve much on the precision with which the properties of cells were described, and provided no new insights whatsoever? How can one admire the detailed and tedious measurement of the orientational preferences of cells that are not orientation-selective, as some have tried to do with the directionally selective cells of V5? There is perhaps a little overmodesty in the description of their work as being hypothesis-free, or what many would today describe pejoratively as a sort of fishing expedition. Even if true, this was a major fishing expedition with an ineluctable and beautifully exploited logic; and the catch was astonishing. It is no wonder that every chance was seized avidly. Once orientation-selective cells were discovered (by chance) in the visual cortex of an animal (the cat) which has two eyes and a retinotopically organized primary visual cortex, the rest followed easily, or so this collection of papers makes it seem. The succession of questions can be summarized as follows: How many orientations are represented within any given retinotopically defined position of the visual field? How are cells responding to different orientations organized with respect to one another? Are all orientations represented for each eye? How are the orientation columns and the ocular dominance columns organized with respect to each other? What is the overall organization of the two sets of columns in the context of the topographic map within V1? Given that visual deprivation early in life leads to lifelong blindness—a topic widely discussed since the time of John Locke but put on a scientific and clinical basis by the publication of Marius von Senden’s book Space and Sight (1962)—it was natural to want to study the effects of visual deprivation on the specificity of visual cells in the cortex. Here, two eyes were better than one, since it was possible to deprive one eye, or both, or neither. It would, of course, be stretching things too much to claim that the decision to work on a two-eyed animal was taken by chance but, even if this accident is allowed, they still used the two eyes in a most imaginative way. Buried within this set of questions is some kind of implicit hypothesis. And the authors were well aware, right from the start, that their undertaking was a major one and that they were on to some striking findings. Why else, as they tell us, would they project to write a book as early as the 1960s, well before their work came to maturity? Why else, too, would they have been so aware of potential competitors, jealously guarding their results and projected experiments? It is likely that their apprenticeship prepared Hubel and Wiesel, if not for the accidents and lucky breaks, then at least for how to proceed once the significance of the accident had been appreciated. The biographical sketches detail their brilliant intellectual surroundings. The catalogue of names is a veritable aristocratic roll call: Bernard Katz, William Rushton, E. D. Adrian, John Eccles, Alan Hodgkin, Andrew Huxley, Edwin Land, Michael Fuortes and Francis Crick. Indeed they regard themselves as the scientific great-grandsons of Charles 1228 Book review Sherrington, a lineage of which the latter would no doubt have approved. No room here for the more modest in ability, who might nevertheless come up with an interesting question or two. Towering above all in inspiration to them was Steve Kuffler, the grand maestro of neurobiology in the 1950s and the 1960s, who had surrounded himself with brilliant investigators. Clearly, the atmosphere at Johns Hopkins, and later at Harvard Medical School, was highly congenial and deeply inspiring. Clearly, too, the focus was on the problem rather than on thoughtless measurement. Hubel recounts how he was once advised by Kuffler to stick in some quantification, so as to soothe the scientific conscience of the mindless measurers, making the paper seem more scientific and therefore better suited for publication. He laments that their work would be unpublishable today, since referees, under the guise of a scientific rigour that is both coarse and imperceptive, would demand a host of detailed measurements. In one paper, they write of the long time taken to produce a graph using a computer, adding: ‘We concluded that for both speed and for precision it is hard to beat judgments based on the human ear. Certainly [the curves] could not have been obtained with computer averaging methods before the authors reached the age of mandatory retirement’. It is difficult to deny that their work would have suffered greatly if they had been working in the present era of white-coated, rubber-gloved and dispassionate scientific honesty, one that, distrustful of human judgement, confides all measurements to a computer that supposedly does not suffer from human bias. Yet it is also interesting and instructive to compare the results that Hubel and Wiesel obtained when they judged the response specificities of cells by ear (itself not a negligible measuring device) with the host of papers that make generous use of the apparently neutral computer-generated ‘orientation selectivity index’, the ‘colour selectivity index’, and many other indices. After a quarter of a century of rigorous application, the latter have revealed nothing or little that is interesting or new. The authors themselves make the point, perhaps unwittingly, in the commentaries that follow each paper. Advances there have been, they seem to be saying, but nothing to compare with the original discoveries. It is, no doubt, this poverty of results, compared to the richness of the original findings, that has led David Hubel to regard modern technologies, and the mathematical pitchforks associated with them, with such disdain. The Epilogue of the book has him wandering sadly through the debris left by the muchvaunted computational approach which, one cannot but agree, has contributed little that is of importance to understanding how the brain works. He disapproves, as indeed do I, of linear systems analysis and apparently shares my horror for the current craze for ‘multiplexing’, the belief that few if any cells in the visual brain are highly specialized and that most of them are multipurpose. He has nothing but praise for the fact that the major textbook of molecular biology, The Molecular Biology of the Gene, contains no equations and that ‘no one has tried to fit a protein molecule to a Gabor function’. Before anyone dismisses these views too hastily, it is worth considering how much information the ‘crude’ techniques used by Hubel and Wiesel gave, and how little, by contrast, these fancy approaches have provided. The evidence is all there in this book. How did such a collaboration last for so many years and why did it eventually end? The first question is relatively easy to answer. One successful set of experiments led to another, generating an unstoppable momentum. Both authors were intoxicated enough by their results to persevere with one experiment and then plan and push on with the next. Several factors aided them in this. Steve Kuffler was evidently some kind of Lorenzo de’ Medici, content to see brilliant work progress in his department, dispensing advice in an avuncular way and not anxious, as so many are today, to add his name to the papers and thus get his cut of the glory. The relationships established in the laboratory were evidently friendly enough for the authors to say nothing more than that they were ‘sometimes mixed with resentments and varieties of complexes’. Crucially, both were spared the curse of administration— one that so many academics, in spite of their protestations, love. After one year administering the Physiology Department at Harvard, Hubel had had more than he could take when the installation of a pencil sharpener necessitated at least two elaborate meetings. They were also working at a time when cats were even more plentiful than monkeys and not much more expensive. Editors were more relaxed and referees kinder. Nor were the grant-giving committees quite as forbidding as today. The authors state that their kind of work could not be done today. Sadly, they are right. Their laboratory, small in size and shared with others, encouraged continual dialogue and discussion—one of the characteristics of great laboratories such as the Laboratory of Molecular Biology in Cambridge, UK, or the Bell Laboratories, USA. One gets the impression that there was also a feeling of self-sufficiency. One of the authors’ great strengths—but in the longer term also a weakness—was their understandable indifference to work done by others, perhaps encouraged by Steve Kuffler’s oftrepeated question: ‘Do you want to be a consumer or a producer?’ For example, even in this book, and after so many years of discussion, they still believe, erroneously, that I mapped area V5 of the owl monkey in 1980, thus getting both the species and the date wrong, though I should be grateful that they got my name right! They are not even aware of where V5 is, describing it as lying in the ‘anterior lunate sulcus’, whereas it is actually located some distance away, in the posterior bank of the superior temporal sulcus. This is surprising, since V5 is now one of the most studied visual areas after V1 and is coming close to surpassing the latter. But all this presumably matters little to David Hubel and Torsten Wiesel, since the work was not undertaken by them! It is said that on one occasion, when asked whether he intended to respond to criticisms made of their work in a paper, Hubel replied that, to do so, he would have to read the paper, which was more than he was prepared to do. By the time they parted company in 1980, the authors had succeeded in charting the anatomy and physiology of area V1 of the brain in greater detail than ever before, and making it the best understood area of the cerebral cortex. They had used Book review almost all techniques available to them; charted ocular dominance and orientation columns and looked at their anatomy; established with greater precision than ever before the details of how the fibres from the lateral geniculate nucleus terminate in the visual cortex; established the concept of the column in the visual cortex firmly; shown that selectivities are established at birth and are susceptible during a critical period after that; and, above all, with their ‘loose’ methods had introduced new and very high standards of evidence into cortical neurobiology which few have managed to emulate. Why, then, did this flourishing scientific collaboration end? Their work had been mainly on area V1, or area 17, and with so much achieved it seemed as if V1 was well enough understood, and the work thus began to run out of steam. Indeed, as early as 1968 they had thought of V1 as sufficiently well characterized that ‘despite the large areas still unexplored, in broad outline the function of area 17 is probably now relatively well understood’. It was to areas beyond V1 that one should turn one’s attention, or so they implied when they wrote in the same paper ‘How this information [in area 17] is used at later stages in the visual path is far from clear, and represents one of the most tantalizing problems for the future’. In his autobiographical sketch, Torsten Wiesel writes with disarming honesty that ‘there may have been a more profound reason that our partnership came to an end. The additional demands on our time [came] when we were investigating the properties of cells of higher visual areas beyond the primary visual cortex, an exploration that each of us eventually regarded as a failure . . . The two naturalists, who for so long had journeyed together with a seemingly inexhaustible sense of wonder, were unaccustomed to the frustrations that are the daily bread of so much scientific research . . . what developed between us, our special bond and private dialogues, took place while we carried out our experiments. When these explorations stalled, when the wonder faded, so did our collaboration. Perhaps it was necessary for each of us to turn to new partners and questions’. It is indeed through the exploration of the higher visual areas that their concept of hierarchy began to be questioned. That concept was largely the result of concentrating on 1229 orientation-selective cells and ignoring other categories of cell. They managed to identify different levels of complexity in the functional properties of orientation-selective cells, and it is this that led them to the hierarchical concept of visual physiology. Ironically, it was adherence to this hierarchical doctrine that made them misinterpret, or not interpret, the functions of a visual area in the cat that is equivalent to area V5 in the monkey, the Clare–Bishop area, in an article that they describe as ‘not one of our favorite papers’. Hubel was later to repudiate this exclusive hierarchical concept implicitly, but never explicitly. The implicit repudiation came with the work he did with Margaret Livingstone, which showed that their earlier model of V1 was incorrect; that there was a functional segregation within V1, much as had been predicted from the demonstration of functional specialization among the higher visual areas of the brain, beyond V1, and which receive their input from V1. It is a pity that Hubel and Wiesel ignored the concepts of functional specialization and parallelism, established by work in the higher visual areas. It is a pity, too, that in gazing into the future through the past David Hubel (the main author of the commentaries of this book) has nothing to say about this functional specialization—perhaps because it does not figure in their collaboration. It is, in my view, a major oversight, but it is an oversight that one might want to forgive, much as one would forgive Einstein for reputedly ignoring quantum mechanics. When one has finished rereading, or reading for the first time, the extraordinary output that has resulted from this legendary collaboration, one decides pretty much to forgive such blemishes. Neuroscience should rejoice that, during a mere 25 years, its world was enriched not only by a wealth of knowledge but also by new standards of evidence and elegance of methodology which have left a permanent imprint. Semir Zeki Laboratory of Neurobiology University College London London, UK doi:10.1093/brain/awh507