Lateral inhibition between orientation detectors in the human visual system

An experimental study of linear inclination

Two mechanisms have been proposed to account for perceptual distortion of angular subtension. The first implies that errors made when estimating sizes of angles should correspond to inaccuracies of estimating the inclinations of their two component lines. From the second it follows that the two types of judgement are unrelated. This paper considers which of these mechanisms accounts most consistently for apparent angular distortion. Two experiments are reported. In the first, four independent groups of subjects estimated the bearing of a single straight line presented in a series of inclinations throughout the entire circular range. Subjects themselves attempted to place the same line into specified inclinations in the second experiment. The results reveal errors in both tasks. An attempt is made to distinguish between the two proposed mechanisms by predicting angular distortions from those of corresponding linear inclinations. Possible reasons for failure of this prediction are discussed.

Investigation of temporal integration by video sampling

The results from the preliminary set of experiments in which a new video sampling apparatus was used are reported. With the aid of this apparatus experiments were carried out to measure the maximum visual temporal integration time (critical duration) at various background intensities (0·034–34 cd m−2). The aim was to determine to what extent this phenomenon is attributable to either ‘central’ or ‘peripheral’ events. The extended integration period found for the number recognition task is interpreted as evidence of a ‘central’ process; to follow the argument further, an attempt was made to demonstrate information integration using a rotating form in a similar identification–discrimination situation. Monocular, binocular, and dichoptic arrangements were employed, and the amount of dichoptic summation of form information, achieved by both normal and strabismic subjects without stereoscopic depth perception, was used to test two theoretical models of binocular fusion. In addition, stereoscopic depth was generated with uncorrected sampling of the left and right images, which may be due to the action of a ‘fusion hierarchy’. Signal detection theory is suggested as a possible solution to the problem of expectation effects in identification-threshold experiments.

Role of lateral interaction in overestimation of acute angles

The overestimation of acute angles is one component of certain visual illusions, e.g., the Poggendorff. The effects of lateral interaction are discussed and the hypothesis is suggested that overestimation of angular size is best viewed as due to lateral interaction in the perception of the distance between two converging lines. Further evidence is required concerning the form of the function relating the intensity of lateral interaction effects to the parameters controlling them.

The effects of temporal modulation on the orientation channels of the human visual system

The orientation specificity of spatial adaptation was quantified by using an equivalent-contrast transformation. The half-width at half-amplitude of the adaptation effect was 6·5° when the sinusoidal gratings were stationary, but was increased to 11° when the gratings drifted laterally. Another type of temporal modulation, flashing the gratings on and off repetitively, also increased the half-width, but only at low spatial frequencies where these stimuli actually appeared to be flickering back and forth at threshold. At higher spatial frequencies the flashing stimuli appeared to be stationary at threshold and the half-width was as small as that for truly stationary gratings. It is suggested that there are two types of channel in the visual system: they differ in their orientation specificities, in their temporal properties, and in their roles in the analysis of a spatiotemporal stimulus.

Periodic vernier acuity

1. Sinusoidal curvature was produced in a vertical stimulus line generated on an oscilloscope screen. Sensitivity to sinusoidal curvature (periodic vernier acuity) was measured by the method of adjustment as a function of spatial frequency of the curvature.2. Sensitivity was maximal at about 3 c/deg and reduced steeply for both higher and lower spatial frequencies of curvature.3. Sensitivity at 10 degrees in the periphery was greatly reduced but showed very little low frequency reduction.4. The effect of number of cycles present was checked by reduction of the field size with spatial frequency such that a fixed number of cycles of the stimulus were visible. This procedure did not affect the decline in sensitivity at spatial frequencies below 0.3 c/deg but reduced sensitivity at higher spatial frequencies.5. The results were interpreted in terms of cortical orientation detectors. It is suggested that the low frequency sensitivity is limited by a fixed value for the maximum difference in orientation in the stimulus (i.e. 20 and 30' for the present subjects). The high frequency sensitivity was limited by grating acuity and required spatial integration of the stimulus information over about 2.5 degrees .

The effect of body tilt on the Zöllner illusion

The magnitude of the Zöllner illusion was measured for a range of intersect angles in the four combinations of subject, vertical or tilted, and display, horizontal or tilted. When the main lines of the display fall on the horizontal retinal meridian, body tilt per se has no effect; there are however, anomalous results when the body is tilted and the main lines do not fall on the horizontal meridian, and the possible implications are discussed.

Adaptation to square-wave gratings: inhibition between spatial frequency channels in the human visual system

1. The observation that the detection threshold for a square-wave grating depends only on that of its fundamental was confirmed by showing that adapting to the fundamental spatial frequency caused elevation of the square-wave threshold, to the same extent as the fundamental threshold was elevated by the same adapting pattern. Adapting to the third harmonic frequency had no effect on the square-wave threshold.2. Adapting to a square-wave grating should elevate the thresholds for both the fundamental and third harmonic frequencies (Blakemore & Campbell, 1969), and the amount of elevation at each frequency should be predictable from the contrast of that frequency within the square-wave.3. It was found, however, that both the fundamental and third harmonic, when present in the square-wave, were much less effective as suprathreshold adapting stimuli than would be predicted from their effects when viewed in isolation.4. Adapting to a mixture of two sinusoidal gratings (with 3:1 frequency ratio) demonstrated that the fall in adapting power was not due to the higher harmonics of the square-wave nor to nonlinearities in the stimulus display. Similar effects were found when the phase relations of the adapting gratings were changed, showing that the interaction is not a special property of square-waves or of edges.5. It is suggested that the spatial frequency channels subserving the fundamental and third harmonic frequencies inhibit each other when the patterns are some way suprathreshold. At or near threshold, there is no such reciprocal inhibition.

Acute angles and the Poggendorff illusion

Burns and Pritchard's (1971) explanation of the Poggendorff illusion is criticized. An experiment was designed to determine whether the acute angle plays any role in the perception of the illusion. The results showed that (i) an inducing line which crossed a test-line was highly effective in altering the apparent orientation of the test line, (ii) an inducing line forming an acute angle with a test-line had a small effect in changing the apparent orientation of the test-line, and (iii) an acute angle which formed part of the Poggendorff configuration produced an effect opposite to that predicted by the view that acute angles are perceptually enlarged.

Perceived curvature of arcs and dot patterns as a function of curvature, arc length, and instructions

Arcs of circles, with six arc lengths and four radii of curvature, and an equivalent set of figures composed of three dots were used as stimuli. Subjects in Group I imagined the circle from which an arc or dot triplet was taken and indicated the centre of the circle. Group II subjects estimated the location of the point that was equidistant from the middle and ends of an arc, or equidistant from the three dots of a triplet. The results from arcs showed, in Group I, an underestimation of curvature that decreased as a function on the length of the arc. In Group II, however, overestimation of the curvature of most arcs occurred, indicating a strong influence of the difference in the perceptual task on the results. The effect of instructions was similar with the dot figures but, in general, more errors resembling overestimation of curvature occurred with these figures.

Geometrical illusions and the response of neurones in the cat's visual cortex to angle patterns

1. This report describes the responses of thirty-six single neurones in the primary visual area of the cat's neurologically isolated and unanaesthetized forebrain, to movements of thin white lines across the visual field. The experiments were designed to record the effects upon the response to a single test line of an added line, which was either parallel to the test line or joined it, making an angle-pattern of 30 degrees . Unit responses were measured in terms of the peak probability of firing derived from a post-stimulus histogram.2. All of the cortical neurones tested exhibited a preferred orientation for stimulation by the test line, i.e. an orientation of the line which produced a maximal response when the line passed through the centre of the unit's receptive field.3. There was no evidence that the orientation of a single test line preferred by cortical neurones was different from that preferred by the same cell when excited by an angle pattern, one arm of which was the original test line.4. The position of a test line (with preferred orientation) in the visual field that produced a maximal response from cortical neurones, was not always the same as the position for maximal response, when a second line was added to make either an angle pattern or to make a pattern of two parallel lines.5. Where the two lines of these patterns were close together and separated by less than the radius of the receptive field, the position for maximal response to the test line was shifted towards the added line. Where the two lines were further apart than this but separated by less than a receptive field diameter, the optimal position for the test line was displaced away from the added line.6. Some evidence was found of a lateral inhibition in the visual system, sufficient to account for the displacements described in paragraphs 4 and 5 above.7. It is concluded that the tip of an angle pattern of 30 degrees produces a distorted cortical image within the primary visual area.8. This neural distortion of sensory information seems adequate to explain the well known illusions of orientation that are associated with human perception of patterns containing acute angles.