Horace Barlow

• MBBCh,MD,ScD
• University of Cambridge

When I first came across William James' dictum that " … this sense of sameness is the very keel and backbone of our thinking," I thought he had foreseen the importance of cross-correlation in the brain, and told myself to find out how he had reached this conclusion. When I finally did this a year or two ago, I slowly came to realize that I had comp...

Distortions of the local spatial-frequency power spectrum caused by motion blur may be used by the visual system to improve motion analysis (e.g., Barlow and Olshausen, 2004 Journal of Vision 4415-426). We tested this hypothesis by measuring the error of perceived motion direction of moving patterns in the presence of random noncoherent motion mani...

Neurons that respond selectively to the orientation of visual stimuli were discovered in V1 more than 50 years ago, but it is still not fully understood how or why this is brought about. We report experiments planned to show whether human observers use cross-correlation or auto-correlation to detect oriented streaks in arrays of randomly positioned...

Distortions of the local spatial frequency spectrum caused by motion blur may be used by the visual system to improve motion analysis (Geisler,1999; Burr, 2000; Barlow & Olshausen, 2004). We tested this hypothesis by measuring the error of perceived motion direction of 50 pairs of dots that moved (motion signal) in the presence of 50 pairs of dots...

In my final year as a pre-clinical medical student at Cambridge in 1943 I became fascinated by the recently published paper by Hecht, Shlaer & Pirenne on the absolute threshold of human vision. It seemed to show, with very little room for doubt, that the human visual system was able to make very good decisions as to whether a weak flash of light ha...

The aim of minimum-entropy coding is to find representative elements that occur as nearly as possible independently of each other, so that the probabilities of all input states can be calculated approximately from the probabilities of these elements. Finding predictive and other associations is an extremely important task for the real nervous syste...

The problem discussed is the relationship between the firing of single neurons in sensory pathways and subjectively experienced sensations. The conclusions are formulated as the following five dogmas: 1. To understand nervous function one needs to look at interactions at a cellular level, rather than either a more macroscopic or microscopic level,...

IntroductionCode-breaking as an Analogy for CognitionThe Possibility of Code-breakingThe Importance of RedundancyRedundancy of ImagesGrammar and MeaningLanguage in the Analogy and in Real LifeCode-breakingGenetic KnowledgePlasticity, Deprivation, and the Cortical CodeGene-selection and Cortical PlasticityConclusions References

Photoreceptors strongly attenuate high temporal frequencies. Hence when an image moves, high spatial frequency components are lost if their direction of modulation coincides with the direction of movement, but not if it is orthogonal. The power spectra of natural images are remarkably consistent in having a 1/f 2 falloff in power in all directions....

this paper we are concerned with one particular aspect of this system, namely the way it acquires, stores, and uses background knowledge of the sensory environment. Probably everyone will agree that visual perception uses knowledge of images and their ways to reach a valid interpretation of the current scene; to illustrate this, ask yourself when y...

Statistical regularities of the environment are important for learning, memory, intelligence, inductive inference, and in fact, for any area of cognitive science where an information-processing brain promotes survival by exploiting them. This has been recognised by many of those interested in cognitive function, starting with Helmholtz, Mach, and P...

Soon after Shannon defined the concept of redundancy it was suggested that it gave insight into mechanisms of sensory processing, perception, intelligence and inference. Can we now judge whether there is anything in this idea, and can we see where it should direct our thinking? This paper argues that the original hypothesis was wrong in over-emphas...

Hoffman is worried that perception itself leaves no time for the computation and compilation of statistics, but this has never been proposed. It is the underlying mechanisms that are thought to have evolved in response to the statistics of sensory stimulation, and which are capable of adjusting their parameters in response to changes in these stati...

Learning about a causal or statistical association depends on comparing frequencies of joint occurrence with frequencies expected from separate occurrences, and to do this, events must somehow be counted. Physiological mechanisms can easily generate the necessary measures if there is a direct, one-to-one relationship between significant events and...

Almost all representations have both distributed and localist aspects, depending upon what properties of the data are being considered. With noisy data, features represented in a localist way can be detected very efficiently, and in binary representations they can be counted more efficiently than those represented in a distributed way. Brains opera...

Guidelines for submitting commentsPolicy: Comments that contribute to the discussion of the article will be posted within approximately three business days. We do not accept anonymous comments. Please include your email address; the address will not be displayed in the posted comment. Cell Press Editors will screen the comments to ensure that they...

A tune is a succession of musical tones that is easily recognizable when repeated. Here we describe an experimental technique and preliminary results that test a simple theoretical idea about tune recognition.

What use can the brain make of the massive flow of sensory information that occurs without any associated rewards or punishments? This question is reviewed in the light of connectionist models of unsupervised learning and some older ideas, namely the cognitive maps and working models of Tolman and Craik, and the idea that redundancy is important fo...

We estimated the sensitivity for detecting a row of collinear target elements (usually dots) by measuring the maximum density of randomly positioned noise elements that allowed 75% correct detection of the orientation of alignment (binary choice: horizontal versus vertical) of the target elements. We varied the number of target elements, their mode...

Prediction, like filling-in, is an example of pattern completion and both are likely to involve processes of statistical inference. Furthermore, there is no incompatibility between inference and neural filling-in, for the neural processes may be mediating the inferential processes. The usefulness of the “bridge locus” is defended, and it is al...

The reductionist approach to the brain shows promise of revolutionizing our ideas about what single neurons can do. A spine on a cortical pyramidal cell is about the size of a single Escherichia coli, and if the internal machinery of a spine is anything like as well organized as that of E. coli, the whole pyramidal cell with its 5000 spines must be...

In the random dot kinematograms used to analyze the detection of coherent motion in the middle temporal visual area (MT) and in psychophysical experiments the exact way that dots are paired between successive presentations is not known by the observer. We show how to calculate the limit to coherence threshold caused by this uncertainty, which we ca...

Knowledge is often thought to be something brought from outside to act upon the visual messages received from the eye in a 'top-down' fashion, but this is a misleadingly narrow view. First, the visual system is a multilevel heterarchy with connections acting in all directions so it has no 'top'; and second, knowledge is provided through innately de...

We tested the hypothesis that neurons in the primary visual cortex adapt selectively to contingencies in the attributes of visual stimuli. We recorded from single neurons in macaque V1 and measured the effects of adaptation either to the sum of two gratings (compound stimulus) or to the individual gratings. According to our hypothesis, there would...

We measured human observers' thresholds for detecting coherent motion in random-dot kinematograms and compared them with theoretical thresholds predicted by the hypothesis that unavoidable correspondence noise in the kinematograms themselves is the limiting factor. In experiments where the stimulus had two fields we varied the dot density, the rati...

Purpose. Even when there are no >'oherently moved dots in a random dot kinematogram, there can still be a non-zrro coherent motion signal as a consequence of chance coincidences between the dot-; in sequential frames. We examined the influence of this correspondence noise i i stimuli consisting of multiple fields. We asked: (1) If this unavoidable...

Observers were asked to report the orientation of alignment (horizontal vs vertical) of a group of target dots embedded in randomly placed noise dots. For this perceptual grouping task, sensitivity (the number of noise dots giving 75% correct performance) was measured as a function of the number of target dots. We report surprising tolerance to pos...

Interacting intracellular signalling pathways can perform computations on a scale that is slower, but more fine-grained, than the interactions between neurons upon which we normally build our computational models of the brain (Bray D 1995 Nature 376 307-12). What computations might these potentially powerful intraneuronal mechanisms be performing?...

Purpose: To test whether correspondence noise is the factor limiting detection of coherent motion in two-frame random dot kinematograms we varied the number of dots (N1 and N2) in each frame, calculated the expected change in correspondence noise, and measured the effect on coherence threshold. Methods: Fixed proportions of the first frame dots wer...

Interacting intracellular signalling pathways can perform computations on a scale that is slower, but more fine-grained, than the interactions between neurons upon which we normally build our computational models of the brain (Bray D 1995 Nature 376 307—12). What computations might these potentially powerful intraneuronal mechanisms be performing?...

Using drifting comb-filtered spectra, rapid determinations of the spectral modulation sensitivity function (SMSF) were performed in two normal trichromats and two deuteranopes. Detection thresholds were measured for combfrequencies (f) varying from 0.5 to 3.5 cycles/300nm with a constant temporal frequency (1 Hz). The dichromatic SMSF shows a corne...

Nature is the international weekly journal of science: a magazine style journal that publishes full-length research papers in all disciplines of science, as well as News and Views, reviews, news, features, commentaries, web focuses and more, covering all branches of science and how science impacts upon all aspects of society and life.

This chapter discusses the computational goals of bug-detecting and object-identification. Objects are claimed to cause constellations of sensory signals that frequently occur in association with each other, that can have names attached, and that not only form the elements of the cognitive world but also are the appropriate elements for associative...

The aim of minimum-entropy coding is to find representative elements that occur as nearly as possible independently of each other, so that the probabilities of all input states can be calculated approximately from the probabilities of these elements. Finding predictive and other associations is an extremely important task for the real nervous syste...

It is a mistake to consider perception and learning separately because what one learns is strongly constrained by what one perceives, and what one perceives depends on what one has experienced. I shall propose the hypothesis that perception is the computation of a representation that enables us to make reliable and versatile inferences about associ...

To determine whether a particular sensory event is a reliable predictor of reward or punishment it is necessary to know the prior probability of that event. If the variables of a sensory representation normally occur independently of each other, then it is possible to derive the prior probability of any logical function of the variables from the pr...

Summary Any small region of the cortex receives input through a large number of afferent fibres, and transmits efferent output to other regions of the brain. If the units interact according to an anti-Hebbian rule, the outputs define a coordinate system in which there are no correlations even when the input fibres show strong correlations. The idea...

A method is developed for quantifying the strength of the moiré effects known as Glass patterns. Unpaired randomly placed dots are added to the pattern while the discriminability d′ of the degraded pattern is determined in a yes–no test. For a given discriminability the number of pairs required increases in direct proportion to the number of random...

The human constrast-discrimination function has a curious shape: In addition to rising for increasing contrasts, both positive and negative, it also rises for very low contrasts on either side of zero. It is shown that this rise near zero contrast is not much affected by procedures that increase or decrease the subject’s knowledge of the stimulus;...

Citation Horace Barlow and Denis G. Peli, "Statistical Efficiency of Natural and Artificial Vision: Introduction," J. Opt. Soc. Am. A 4, 2292- (1987) http://www.opticsinfobase.org/josaa/abstract.cfm?URI=josaa-4-12-2292

The cerebral cortex is supposed to be responsible for humanity’s dominance of the natural world, and in particular for the intellectual pre-eminence that underlies this position.

Image processing requires free access to information about all parts of an image, but a nerve cell in V1 can only interact directly with a tiny fraction of the other cells in V1. The problem this poses might be alleviated by forming secondary "neural images" in which information is re-arranged, and some possible rules of projection for forming such...

I feel deeply honoured by your invitation to give the Bartlett lecture, and am especially glad to do so in Holland, the home of so many distinguished psychologists of sensation and perception. And there is a third reason why it has given me much pleasure, for Sir Frederick Bartlett was one of those who had an important influence on the direction of...

Much comparative research aimed at establishing differences in intelligence among vertebrates has failed to convince the sceptic, because it has concentrated on a single experimental paradigm (such as learning sets), while employing a diverse array of species in the hope of establishing a rank ordering of intelligence. The sceptic can insist that s...

At a famous meeting of the International Medical Congress held in London on August 4, 1881 Goltz of Strassburg (as it was then spelt) confronted Ferrier of London on the subject of the localization of function in the cerebral cortex. In the first part of this paper the events of that meeting are recalled. Goltz was reluctant to accept the idea of l...

Responses of brisk-sustained cat retinal ganglion cells were examined using receiver operating characteristic (ROC) analysis. Stimuli were brief luminance changes superimposed upon a weak steady pedestal ranging from 27 to 47,000 quanta (507 nm) per second at the cornea. Overall quantum efficiencies of cells ranged up to approximately 13% and were...

Sixty years ago, E. G. Boring1 pointed out the sorry fact that, for those who believed in tests, intelligence was simply what the tests tested for. This circular definition could have been escaped by using an objective, external criterion for validating them, but in spite of much discussion about the nature of intelligence by the pioneers of mental...

Our eyes see so much in such varied conditions that one might consider the question posed in the title to be meaningless, but we show here that, within the range that we have been able to test, there is a particular spatiotemporal pattern of light that is detected better than any other. At least two plausible theories of visual detection predict th...

The precision of making discriminations between the numbers of dots in a pair of irregular arrays was measured. The results fit the assumption that the observer adds intrinsic variance to whatever variance is present in the numbers displayed, the errors depending upon the sum of the two. We found no evidence for incomplete use of the sample of info...

Because many of you may not be familiar with the biological structures with which we ourselves process images, I shall start with a brief anatomical tour of the complex and incompletely understood visual system of primates. The biological components which perform the computations may also be unfamiliar, and it is interesting to compare them with ph...

Last year's Nobel Prize winners in physiology and Medicine, David Hubel and Torsten Wiesel (see Fig. 1), published a paper 20 years ago that was a landmark in cortical neurophysiology13. And then a year later, in 196314,43,44, they published their first work on the development of the visual pathway in kittens, and this too opened a completely new f...

We propose a model for the first stage of the cortical transformation of the visual image based on the principle that the cortex encodes the information with the minimum number of channels mathematically needed. We restrict our model to be consistent with the data on size adaptations, the known relationships of acuity and the inverse of magnificati...

For colour vision, the task of the eye is to discriminate different distributions of energy over the spectrum. This is usually treated as a problem in the wavelength domain, analogous to treating spatial resolution in terms of spatial positions in the image. What is attempted here is a treatment of colour vision in terms of the system's responses t...

If one knows where ideas have come from and how a subject has developed one gets a much clearer view of which direction you are pointing and what the next problems may be. I shall therefore start by looking at the past history of feature detectors, then I shall talk a bit about my present work, and I shall conclude with a brief guess at the future.

We have measured the overall statistical efficiency of human subjects discriminating the amplitude of visual pattern signals added to noisy backgrounds. By changing the noise amplitude, the amount of intrinsic noise can be estimated and allowed for. For a target containing a few cycles of a spatial sinusoid of about 5 cycles per degree, the overall...

The main factors limiting the performance of the peripheral parts of the visual system can be specified, and doing this clarifies the nature of the interpretive tasks that must be performed by the central parts of the system. It is argued that the critical factor that hinders development of better resolving power is the difficulty of confining ligh...

The average density of randomly placed dots was modulated according to a spatial sinusoid and estimates were made of the signal/noise ratio (d′) of the sensory representation of the pattern from subjects responses. This was compared with the signal/noise ratio of the targets and the results are expressed as the efficiency (F) with which the statist...

Our perceptions of the world around us are stable and reliable. Is this because the mechanisms that yield them are crude and insensitive and thus immune to false responses? Or is it because a statistical censor that blocks unreliable messages intervenes between the signals from our sense organs and our knowledge of them? This question can be answer...

An object moving in discrete spatial jumps is difficult to distinguish from a continuously moving object, provided the time between jumps is not too great. The extent of this perceived continuity may be measured by probing the perceived spatial location at times between the target jumps, by either a vernier alignment or a stereoscopic technique. As...

In this paper I am going to pose the question “What does the cortex, and especially visual cortex, do?” The problem is more puzzling than many people realize, and it is especially tantalizing not to be able to answer it in spite of the fact that many laboratories throughout the world routinely record from single cortical neurons. As a tentative ans...

The detection of mirror symmetry has been investigated by measuring discriminability (d') between two populations of dot displays that contain mirror pairs and random dots in different proportions. The difficulty of the task was varied by changing the proportions of paired dots in the two populations, and also by changing the accuracy of positionin...

Ralph Freeman asked me, as an introduction to this meeting, to try to give you some kind of review of the present status of knowledge in the field of cortical development. The more I looked into it, the more confused I became. There has been a lot of papers published, and these, of course, all constitute local advances; but overall the situation ha...

The Mindful Brain : Cortical Organization and the Group-Selective Theory of Higher Brain Function. By Gerald M. Edelman and Vernon B. Mountcastle. (MIT Press: Cambridge, Massachusetts and London, 1978.)

The ability of human subjects to detect targets of higher average dot density embedded in backgrounds of random dots was measured. As the average density of dots in the target is raised above the average density in the background the value of d′ for detecting the target rises linearly. The subject's d′ is the signal/noise ratio of the internal repr...

I have a rather complicated story to tell, and, what is worse, it contains statistical arguments that tend to make some people bored and others argumentative. So what I shall do is first describe an experiment, then the result, and then tell you why I did it.

1. Action potentials were recorded from optic nerve fibres of lightly anaesthetized cats while parts of the retina remote from the receptive field were stimulated by a shifting grating. 2. Vigorous responses can be obtained under these conditions, confirming McIlwain (1966), Krüger & Fischer (1973), and others. 3. These 'shift responses' are not ca...

1. The slope of curves relating the log increment threshold to log background luminance in cat retinal ganglion cells is affected by the area and duration of the test stimulus, as it is in human pyschophysical experiments. 2. Using large area, long duration stimuli the slopes average 0-82 and approach close to 1 (Weber's Law) in the steepest cases....

Psychophysical measurements were made following adaptation to gratings in order to test the hypothesis that adaptation to patterned features of visual stimulation serves a function analogous to the changes in retinal function during light and dark adaptation. No improvements in detecting changes of spatial frequency, orientation or contrast were fo...

The S-shaped relation between pupil area and illumination is similar in this preparation to that of the human, with a mid-point just below 1 cd/m2. The range of pupil areas is. however, at least 10 times greater in the cat, ranging from more than 120 mn2 to less than 1 mm2.

The past twelve years have been exciting for those interested in the classical question, ``How much do our visual capacities depend on innate developmental factors, how much on the moulding effects of visual experience?'' Recent papers1-4 have settled some issues, but may start new waves of controversy, and this is the occasion for writing this rev...

The sensory system of animals, the words and language that connect the thoughts of one individual with those of another, and man made communication systems, are all greatly influenced by the way information is coded. It is suggested that the role of inductive reasoning is to improve the efficiency of linguistic communication by changing the code, a...