University of California, Berkeley, California, U.S.A. (Received 14 June 1965)

Transcription

1 J. Phy8iol. (1965), 181, pp With 8 text-ftgure8 Printed in Great Britain SPATIAL INTERACTION IN THE HUMAN RETINA DURING SCOTOPIC VISION BY G. WESTHEIMER From the Neurosensory Laboratory, School of Optometry, University of California, Berkeley, California, U.S.A. (Received 14 June 1965) The eye's sensitivity to light decreases with increasing background luminances. It was once thought that this could be simply explained by the decrease in the reservoir of available photolabile pigments in the receptors, but recent measurements have shown that changes in sensitivity are far too large to be accounted for by a simple theory involving concentration of visual pigments. A limit to the performance of the eye in detecting differences in brightness is set by quantum variability; but this existence of a theoretical limit tells us how well the eye could do in distinguishing brightness differences, not how it in fact achieves this. Even quite elaborate theories relating chemical, structural or functional changes in the individual light receptors to their sensitivity cannot fully encompass all known phenomena in this field. This applies particularly to the spatial interaction of adjoining regions which has been known for a long time to be contributing to some of the observed phenomena of retinal function. Thus in dark adaptation the changes in the absolute sensitivity of the human rod retina proceed at different rates when large or small areas are stimulated (Craik & Vernon, 1941; Arden & Weale, 1954), and there is evidence that extensive pooling of signals from receptors takes place (Rushton & Westheimer, 1962; Rushton, 1965). Electrophysiological work on receptive fields of retinal ganglion cells (Hartline, 1940) has made us familiar with the concept of pooling, and the findings of Barlow, FitzHugh & Kuffler (1957) show that the nature of these pools may change during adaptation. Earlier theoretical analyses (Broca, 1901; Lythgoe, 1940; Craik & Vernon, 1941; Pirenne & Denton, 1952) had sought to account for various adaptational phenomena by postulating changes in summation areas in the retina, but the studies of Barlow et al. make clear that such summation may involve both excitatory and inhibitory signals. A basic merit of the electrophysiological experiments is that they can show the confluence of excitation and inhibition on a single ganglion cell in the retina and permit the conclusion that light falling on a given spatial region acts in an excitatory or inhibitory capacity on that ganglion cell 56-2

2 882 G. WESTHEIMER in that state of adaptation. On the other hand, most of the research on spatial interaction in the human retina has been carried out by finding the threshold for varying sizes and shapes of the test stimulus. Such a procedure differs from the recording of the activity of a single unit by such complicating factors as inevitable light-spread variations from target to target, and the possible encounter of units with lower thresholds as the test area is enlarged. In the present experiment these have been avoided by changing the parameters not of the stimulus, but of the background upon which an incremental stimulus of constant size is presented. It is then possible to analyse how light falling on and in the immediate neighbourhood of a retinal region influences the sensitivity of that region, raising or lowering it in ways that depend on the distance from the test area, and the state of adaptation. Previous experiments in which the background area was a variable include those of Dreyer (1961), who found all sizes of background in the range chosen by him to be equivalent, and those of Crawford (1940), Ratoosh & Graham (1951), and Battersby & Wagman (1964), who gave evidence that the cone system shows the kind of interaction to be described here, but only in Crawford's studies is there any indication that the scotopic retina may also be subject to it. METHOD When the subject was in position in the instrument looking at the fixation point, there were three superimposed concentric light patches on the temporal retina of his right eye (Fig. 1). Fields II and III, which constitute the background, remained on constantly, and Field I, the test stimulus, was flashed on for a fraction of a millisecond every 1-5 sec. The experiments were designed to observe the influence on the threshold for this stimulus of changes in the luminances and areas of the background, on which it was superimposed. The details of the apparatus are shown in Fig. 2. Channels II and III, giving rise to stimulus areas II and III, were identical, while channel I was very similar to these. Beamsplitting pellicles, P, served to superimpose the three beams which were seen by the subject in Maxwellian view through the same lens field, L. The optical components of beams II and III were in order: a 6 V, 2-75 A compact filament tungsten lamp (E) fed by a d.c. power supply; a lens imaging the filament on a small circular aperture (this aperture was imaged by the succeeding components of the optical system on to the pupil of the subject's eye); the target T imaged by the field lens L and the optics of the eye on the retina; beam splitters and the filters and the field lens common to both beams. The first channel was also fed through this field lens, but differed from the others in that its light source consisted of a General Radio Stroboscope flashing at the rate of once every 1-5 sec. In each beam there were neutral wedges that could be adjusted by the experimenter or the subject. The images of all three sources coincided in the pupil of the eye and occupied about 1 mm2 there. The field stops in fields II and III were adjusted in size and position to suit the experimental need, and extra filters were placed in the three beams to control the luminance levels. Field I remained throughout a circular disk subtending 6 min of arc at the eye. Procedure. The subject had his left eye occluded and was fully dark-adapted. The pupils were usually not dilated with drugs, but all findings here reported were checked at least once

3 SCOTOPIC RETINAL INTERACTION 883 with a subject who had his pupils dilated and accommodation fully relaxed by the instillation of a drop of 1 % cyclogyl about 1 hr before the experiment started. The subject looked steadily at the fixation point and saw with 100 peripheral vision the background fields and, flashing in their middle, the small test stimulus. He then adjusted the wedge which, depending on the experiment, varied either the test flash or the background until the flashing test stimulus was judged to be at threshold. The wedge was coupled to the wiper of a potentiometer, and, on activation by the subject of a switch, the wedge setting in the form of a voltage was recorded on a printer. Several replications were taken of this setting and then the experimental variable changed until a complete run was obtained. 11. Surround (12 diameter) I. Test field (0.10 diameter) X 10 -Oedt feld ) p Fixation point IiL Background (variable size) Fig. 1. Stimulus configuration seen by subject with right eye. X, fixation point, placed about 100 to one side from the centre of the three concentric fields. Field I, 0.10 in diameter, presented for less than a millisecond every 1i sec, constitutes the test flash. Fields II and III, exposed steadily, constitute the backgrounds, whose sizes and brightnesses are separately variable. -T T t L p T x Fig. 2. _ Eye Schematic diagram of optical stimulating apparatus. Description in text. Experimental conditions. The experiment was designed to study properties of human rod vision, and hence conditions were selected to display these maximally at the expense of cone vision. Aguilar & Stiles (1954) have done this in an exemplary fashion. The filters chosen here were similar to those used by them: Ilford 623 passing only a narrow band centred on 500 m,u for the test field (channel I), and a red glass filter that fully transmitted light of all wave-lengths beyond 630 m,u and none below 620 m/t for the background fields in channels II and III. That the conditions in this experimnent were comparable with those of Aguilar

4 884 G. WESTHEIMER and Stiles was attested by the observation of the rod saturation effect described by them at about the same light level. Calibration. The retinal illuminance of the unattenuated beams from each of channels II and III was 200,000 photopic trolands of tungsten light of colour temperature K when the lamp voltage was 6 V. At this colour temperature, each scotopic lumen is equivalent to 1-4 photopic lumens (Stiles, quoted by Rushton (1956a)). Aguilar & Stiles (1954) have given for a scotopic troland the value of 4-46 x 105 quanta (507 mp) per square degree of visual field per second incident on the eye. Each square degree of retina in the area under consideration contains about 12,000 rods, and the proportion of incident quanta absorbed by rhodopsin molecules under conditions similar to those in this experiment is about 0-1 (Rushton, 1956b). For each photopic troland, then, the effective light level is 4-46 x 105 x 1-4 x 0-1/12,000 = approx. 5 quanta (507 mgs) absorbed/rod/sec. The scotopic effectiveness of the red filter inserted in beams II and III was determined as follows. With the eye fully dark-adapted the threshold for the rod system was obtained in the usual manner with the background attenuated by the red filter and an additional neutral filter of 2 density units. The red filter was then removed and the extra neutral filter was found which when substituted for the red glass filter left the threshold unchanged. The extra neutral filter, which was 3-5 density units, thus attenuated the white field from the source of colour temperature K as much as the red filter, so far as the scotopic threshold was concerned. It follows that the scotopic value of the unattenuated background, with the red filter in place, was log (200,000 x 5)- 3-5 = 2-5 log quanta absorbed/rod/sec. = 300 quanta absorbed/rod/sec. By over-running the lamp this could be increased by a factor of 2, and by interposing filters it could be reduced by known amounts. The results shown in this paper have the intensities scaled in units of quanta (507 m,u) absorbed/rod/sec. Image 8pread. Recent work has given information concerning the light distribution in the human retinal image (see Westheimer, 1963 for a review). Data obtained for an eye in good focus with a 6 mm pupil (Westheimer, 1963, Fig. 15) provide a conservative estimate of what may be expected under the present experimental conditions, i.e. the actual light spread is certainly no less concentrated than shown here. Using the experimentally determined spread in a line and in gratings, the expected spread in the image of circular disks of various diameters was calculated by the following procedure Diameter of 08 disk (min of 1220 ~~~~~arc) Retinal distance from centre of disk (min of arc) Fig. 3. Light spread in the retinal images of circular disks of constant luminance and various diameters. The curves were obtained by integration from the estimate of the point-spread of a normal eye in good focus with a 6 mm pupil given by Westheimer (1963). If the eye's modulation transfer function is T(c), where w is spatial frequency in cycles per minute of arc, and the eye is presented with a circular object patch of radius a minutes of arc, the illumination I(r) in the retinal image is given by the equation l(r) = I 1(2aw ) JO(27Trw)T( C) codw, Jo 27Taw

5 SCOTOPIC RETINAL INTERACTION 885 where r is the radial distance, in minutes of arc, from the centre of the image. The integration was performed on a digital computer for disk objects of several diameters. Some results are shown in Fig. 3; they will be helpful in evaluating the findings described in this paper. RESULTS Effect of area of background When the attempt was first made to determine the threshold for a small flashing test field seen against a small background, difficulties were encountered because the subjects were aware of a flashing sensation surrounding the background, much as if scattered light from the test flash had stimulated the surround. It was therefore decided to superimpose both the background and the test flash on a 120 uniform field whose luminance was always 1 log unit below that of the background. The data make clear that the effect of this level of surround, compared with the complete absence of surrounding light, is not significant. In the middle of Test field 0-10 Retinal illuiminance of background quanta absorbed/ rod/sec 4 r_l A * e e O~ _0A A 60 2 / -oq> 3 A _ 0 0 (>0'" 0 0>C-(> C 3 / 0N 0 _ ----O ~~~ ~ ~ ~ ~~~~0 II.I Diameter of background (degrees) Fig. 4. Incremental threshold for a small blue-green test flash seen in the middle of red backgrounds of various diameters and retinal illuminance levels. Zero on the axis of ordinates denotes the absolute threshold for the test flash. Peripheral vision for dark-adapted eye. this large field was the circular background, the threshold in whose centre was measured by the test flash. The threshold was determined as a function of diameter of the background in the range of diameters of 6 min of arc to The results are shown in Fig. 4 for five levels of retinal illuminance of the background, ranging from 0-03 to 600 quanta (507 m,u) absorbed per rod per second. The curves for the lowest and the highest backgrounds show a somewhat different relation between the variables from that in the middle three curves. Taking the middle curves first, we see that for a typical scotopic

6 886 G. WESTHEIMER background light level the influence of size of background on the increment threshold for a small central test field is: (1) an increase in threshold with increasing area in the range of background diameters 6' (the size of the test field) to about 0.750; followed by (2) a lowering of threshold by as much as 1 log unit when the area is further increased to a diameter of 2.50, beyond which there is apparently little change. Summation versus stray light The initial increase in threshold with increase in area of background is found in all curves. With the kind of point-spread function the eye is known to have, an increase in background area necessarily involves an increase in quantity of light falling in the centre of the field where the test flash is situated. It is well known that the threshold increases with increasing quantity of background light: is the observed threshold change of the order of magnitude to be expected on the light-spread hypothesis? Figure 3 shows that, on a conservative estimate, the maximum light level in the region tested increases by a factor of about 1-5 when the background diameter is increased from 6' to 40'. The threshold, however, has risen by a much larger factor and no reasonable Weber fraction could account for this. There is yet another way of demonstrating that this increase of threshold is not due to the spread of light from outlying regions on to the test region with a consequent rise in threshold by the appropriate Weber fraction. The proportion of light falling on this central region from the surround is independent of the light level. If the change of threshold with increasing area of background were merely a result of the light falling on the central region, the curve of threshold versus intensity (i.e. log Al against log I) would be laterally displaced as if, for the larger background area, the I value were increased by a constant proportion (the log I value increased by a constant amount). Figure 5 shows that the curve of log AI against log I for a background of 45' diameter will not superimpose on one for 6' diameter for any single lateral shift and that, in any case, the nearest fit would be for a lateral shift of the order of 1b5 log units, which is not in accord with the light-spread data by a whole order of magnitude. We have to conclude that what increases the threshold in the central test field is not light from the surrounding regions. No such argument is necessary to convince the reader that the subsequent fall in threshold, when the background area is further increased, cannot be due to the spread of light: the effect of increasing background area is to add light to the centre (though the expected quantity, see Fig. 3, is quite small) and the threshold always rises monotonically with increasing background light.

7 SCOTOPIC RETINAL INTERACTION 887 The increase in increment threshold with increasing retinal illuminance is usually ascribed to a change in level of excitation of the region whose sensitivity is being tested by a threshold measurement: the more light in the background, the higher is the level of excitation or activity, and the more energy there has to be in the increment stimulus for it to reach threshold. Extending this concept to the main results in Fig. 4, we can say that surrounding areas of the background when illuminated add to the level of excitation of the rod retina when they are close by, but decrease the level of excitation when they are further removed bo ~~~A -co Background retinal illuminance log quanta absorbed/rod/sec Fig. 5. Incremental threshold (in terms of AIo, the absolute threshold) for a small blue-green test flash (0 10) centred on a small (0.10 diameter, triangles) and somewhat larger (0-750, circles) red background at various retinal illuminance levels (log co-ordinates). Data replotted from Fig. 4. A single lateral shift will not superimpose the two curves and this is evidence that stray light does not explain the difference in their shape. Very high and very low scotopic background intensities It remains to account for the apparently anomalous shape of the lowest and highest curves of Fig. 4. In Fig. 6 an increment threshold curve is shown for the same subject and identical test conditions except that the background is large. The principal features pointed out by previous investigators can be seen in this curve: the relatively shallow slope of the curve characteristic of a small, brief test flash (Barlow, 1957), the rod saturation at high retinal illuminance (Aguilar & Stiles, 1954), and finally the appearance of a cone branch of the curve. Comparison of the vertical scales of Fig. 4 and Fig. 6 shows that the highest curve in Fig. 4 reaches a level of increment threshold at which the signals come from cones and not rods. This was further proved by demonstrating that the spectral effectiveness of threshold stimuli is here that for cones.

8 888 G.WESTHEIMER It is known that phenomena similar to the ones described here occur for the photopic system (Crawford, 1940; Ratoosh & Graham, 1951), but the distances over which the photopic interaction occurs are very much smaller. The flat segment of the top curve of Fig. 4 is, therefore, merely an expression of the fact that for cone increment thresholds the area of background is no longer relevant when the latter exceeds about '. This kind of explanation is, of course, not possible for the bottom curve of Fig. 4, which differs from the middle three in having a flatter beginning segment and in showing little or no drop in threshold with increasing area. The retinal illuminance of the background for this curve is 0 03 quanta absorbed/rod/sec which is in the region where the slope of the log AI against log I curve of Fig. 6 is quite shallow. If the increase of excitation due to spatial summation of background excitation is at all akin to the increase of excitation with increasing retinal illumination, one might readily expect this result. Further evidence for this view is given below oo Background retinal illuminance log quanta absorbed/rod/sec Fig. 6. Increment threshold (in terms of the absolute threshold AIO) for a small brief blue-green test flash 0.10 centred on a 3.50 diameter red peripheral background field of varying retinal illumination (log co-ordinates). Effect on increment threshold of intensity of surround Several possible explanations come to mind to account for the absence in the bottom curve of Fig. 4 of threshold reduction for larger background. It may be, for example, that the excitation in the test region has here not yet reached a level at which the reducing effect of adjoining regions can exert its influence; on the other hand it may be that these neighbouring regions do not begin to influence our test patch until their own excitation has reached a somewhat higher level. To distinguish between the two explanations. the following experiment was carried out.

9 SCOTOPIC RETINAL INTERACTION 889 A background patch 0.60 in diameter was set up and the incremental threshold in its centre was determined in the usual manner. The background patch was given a large surround (extending to a total diameter of 12 ) whose retinal illuminance could either be equal to the central patch or differ from it by + 1-0, + 0*5, - 0*5 and log1o units. The increment threshold was now determined for three retinal illuminance levels of the central patch each time with five surround levels ranging in i log unit steps from being 1 log unit dimmer through equality to being 1 log unit brighter. The results are shown in Fig. 7. When the central field corresponds in retinal illuminance to the level at which the bottom curve of Fig. 4 was determined, its threshold is substantially independent of surround brightness until the surround is 1 log unit brighter. Here the threshold is raised, a finding which may be readily ascribed to the addition of scattered light to the central patch from the large bright surround which it has under these conditions. Retinal illuminance 0 of background (quanta absorbed 0 rod-1 sec-') _ 0-3 0>e ~ ~ ~ ~ ~ ~ ~ 0 bo_ Log surround illuminance relative to background A 0.60 circular patch of peripheral retina had in its centre a small brief Fig. 7. test flash (0-10) whose incremental threshold was determined in terms of its absolute threshold, AlIO. The 0-60 background field was surrounded by a large (120 diameter) field whose retinal illuminance could be separately adjusted. The incremental threshold for the test flash was measured for three representative scotopic retinal illumination levels of the background, each time with the surround 1 and i log unit dimmer, equal, and i and 1 log unit brighter than the 0.60 background. The middle curve of Fig. 7 also demonstrates this last effect, but in addition the threshold is increased, i.e. sensitivity decreased, when the surround is dimmed below the level of the centre. This is the same phenomenon shown in Fig. 4 in the decrease in threshold going from a j background to a larger background, only here it manifests itself as an increase in increment threshold as the surround is dimmed from equality with the central patch to be i and then 1 log unit lower than the central patch. When the surround is made i log unit brighter, there is no change in

10 890 G.WESTHEIMER threshold, and when it is made 1 log unit brighter, the threshold is increased only slightly. The increase in threshold with a 1 log unit brighter surround for the lower curve, i.e. when the retinal illuminance of the central patch is quite low, is double that for the middle curve. The proportion of stray light from the surround falling on the centre is the same in the two cases, and in view of the steeper slope of the log AI against log I curve for the retinal illuminance represented by the middle curve, one would have expected a higher increase of this threshold in the middle curve than in the bottom curve. Again, increasing the surround brightness of a patch of retinal illuminance 0 3 quanta absorbed/rod/sec (middle curve) from 0 03 to 0 3 log quanta absorbed/rod/sec decreases the threshold by nearly 1 log unit. When exactly the same thing is done to a patch of retinal illuminance 0 03 quanta absorbed/rod/sec, the threshold remains unchanged except for the stray-light effect. That it is possible for a surround brighter than the central patch to decrease the threshold is illustrated in the top curve, where a clear decrease in threshold is seen not only on increasing the surround from being 1 log unit dimmer than the centre to equality, but also when increasing it upward from equality. (The fact that in this curve the thresholds are equal for surrounds both 2 and 1 log unit brighter is probably explained by an offsetting of the thresholdlowering due to surround illumination by the threshold-raising due to the stray light from the surround falling on the centre.) The conclusion drawn from the data in Fig. 7 is that the excitation level associated with a given retinal illuminance is lowered by illumination of the surround, an effect which is partly offset by the stray light falling on the centre when the surround is considerably brighter than the centre. The threshold reduction can occur only when the central region has reached a certain level of excitation, for a given surround illuminance which reduces the threshold of a background at a medium scotopic illuminance (0.3 quanta absorbed/rod/sec) fails to do so when the background is lower. Area-intensity relationships for backgrounds In the experiments so far reported, elevation and depression of the threshold for a small brief test flash in the centre of a patch of peripheral retina has been used as an indicator of the excitation level. It is, however, possible to free the findings from even this constraint by designing a null experiment. This has been done in the results shown in Fig. 8. Each curve represents the relation between area of circular patches of rod retina and their retinal illuminance when a constant small test flash presented in their centre has just reached its increment threshold. On the hypothesis that the excitation level of a region of the rod retina is the same whenever the threshold for a small defined flash has a given value, the curves show

11 SCOTOPIC RETINAL INTERACTION 891 the interaction of retinal illuminance and area of background in developing three representative excitation levels. It is seen that up to an area of about 1 degree2, the slope of the curves is about - 1; this reciprocity between area and retinal illuminance indicates that the summation of light in the background is approximately linear at all three levels. For the lowest curve a further increase in area permits a small further decrease in the retinal illuminance level for constancy of threshold, a result which is probably due to stray light, although partial summation of excitation over a larger region has not been ruled out. For higher levels of excitation, a successively more pronounced increase in retinal illuminance is necessary for constancy of threshold as the area is increased. 0E1\ IA -~0 -o Fig Area of background (degrees2) Area-intensity relation for circular backgrounds. The retinal illuminance of backgrounds of various sizes was found, so the increment threshold for a small brief test flash in their centre had a constant value. Three increment threshold levels were used, 0 3 (circles), 1 (squares) and 2 (triangles) log10 units above absolute threshold. Blue-green test flash (0-10), red background, peripheral retina, eye dark-adapted. DISCUSSION It has been demonstrated that a given region of the rod retina, whose sensitivity is being assessed by the small test flash, is affected in a thresholdraising fashion by light falling on it and close by, and in the opposite fashion by light falling further away. Pirenne (1958) in clearly describing the threshold-raising effect of surrounding light in dark-adapted vision has also show-n that this cannot be ascribed to stray light and has invoked for its explanation the concept of neural interaction. Unfortunately he has used the term inhibition in connexion with the phenomenon of spatial summation of excitation with its consequent raising of threshold. Inhibition implies a reduction of an excitation level and, if it is used at all, this term might best be reserved for the observed fall in threshold when light on surrounding areas produces a fall in the excitation level.

12 892 G. WESTHEIMER A boundary is often given special significance in discussions of visual phenomena. One might, therefore, raise the question whether the threshold for the incremental flash is not increased whenever the boundary of the background is in a particular geometrical relation to the flash. That the presence of a boundary, per se, is not responsible for the findings is shown in Fig. 7, where a steady progression of threshold changes, different for each curve, is demonstrated for a range of surrounds including those that are dimmer than the immediate background (dark-to-light boundary), equal (no boundary), and brighter (light-to-dark boundary). Moreover, not all curves in Fig. 4 show the particular effects emphasized here, yet in all of them was the boundary present. The series of curves in Fig. 8 can be related to some other phenomena in the realm of thresholds for scotopic vision. Area-intensity relations have often been studied, most recently and thoroughly by Barlow (1958). Ricco's and Piper's laws, for example, have been derived from such experiments. The lowest curve of Fig. 8 has a resemblance to the curve obtained when measuring the intensity required for a patch of given area to be seen at threshold, either in the completely dark-adapted eye or against a background of retinal illuminance. But the experiments whose results are shown in Fig. 8 are fundamentally different: they show the intensities of various background areas which raise the excitation in their centre to a fixed level, viz. one at which a given small test flash is just at threshold. The concordance of the lowest curve of Fig. 8 with area-intensity relations for threshold seeing of the full area is then a confirmation of a widely accepted concept of threshold seeing for larger areas, viz. that a fixed excitation level in a single retinal location is necessary for threshold seeing and that this excitation level can be developed reciprocally by light intensity and area over the region for which complete summation holds (Ricco's law) and partially so for larger areas. However, the area-intensity relation for backgrounds which have a constant-level small incremental threshold in their centre does not match, for higher scotopic adaptation levels, the area-intensity relation for variable-area incremental thresholds upon a fixed large background. Both curves show an initial summation, Ricco's law, but the surround inhibition makes a steady-state large area have an effectively lower excitatory capacity when it is a background than when it is used as an incremental stimulus. This is not surprising since we have already demonstrated that inhibition does not come into play until a certain level of excitation has been reached. When testing a larger area for its capacity to act as an increment threshold stimulus, we are letting it, by definition, provide only a minimal excitation and it is an acceptable premise that under these conditions inhibition does not enter. When the same large areal configura-

13 SCOTOPIC RETINAL INTERACTION 893 tion of light is presented as a steady-state background it now will exhibit the excitatory state made up of the excitation from the centre and inhibition from the surround, which gives then a proportionately lower total, as compared to a small area in which inhibition plays no part, than when we are dealing with increment thresholds. There is yet an additional phenomenon that would make a larger area appear to demonstrate a proportionally higher excitation capacity (compared to a small area) when it is used as an increment threshold stimulus than when it is used as a background upon which a small increment threshold is presented. The detection of the presence of a large area presumably occurs whenever the threshold is reached for the most sensitive subunit within the whole area covered by the increment stimulus. When many independent units are available, probability summation allows for detection with lower signal strength. It is of considerable interest, on the other hand, to note that the series of curves of Fig. 8 of this paper have an almost exact counterpart in the result of an entirely different experiment: the area-intensity relation for threshold firing of retinal ganglion cells of the cat (Barlow et al. 1957). These workers determined the threshold retinal illuminance necessary for circular patches of different areas to produce impulse discharges in retinal ganglion cells. When the eye was dark-adapted, the curves had the shape of the bottom one of Fig. 8 of this paper, while if the areas were presented as incremental stimuli of relatively long duration (0-38 sec) the intensities necessary for threshold firing of the ganglion cells had the shape of the upper curve in Fig. 8. The evidence they have adduced for the presence of lateral inhibition under these conditions in the cat retina makes it likely that the phenomena described in the present paper are due to a similar lateral inhibition. SUMMARY 1. The influence of various sizes and brightnesses of background on the sensitivity of scotopic vision was assessed by observing their effects on increment threshold for a small brief extrafoveal test flash. 2. Light falling on or near the retinal area tested elevates the increment threshold. 3. Additional light falling outside a zone about 3 in diameter centred on the test flash lowers the increment threshold, but this lowering cannot be demonstrated until the threshold has been raised to a certain level by illumination of the central zone. 4. The observations were confirmed by an experiment in which the size and retinal illuminance of the background were varied so as to keep the increment threshold constant.

14 WESTHEIMIER This research was aided in part by grant NB-3154 from the National Institutes of Health, U.S. Public Health Service and a contract between the Office of Naval Research and the University of California. REFERENCES AGU`ILAR, M. & STILES, W. S. (1954). Saturation of the rod mechanism of the retina at high levels of stimulation. Optica Acta, 1, ARDEN, G. B. & WEALTE, R. A. (1954). Nervous mechanisms and dark adaptation. J. Phy8iol. 125, BARLOW, H. B. (1957). Increment thresholds at low intensities considered as signal/noise discriminations. J. Phy8iol. 136, BARLOW, H. B. (1958). Temporal and spatial summation in human vision at different back. ground intensities. J. Physiol. 141, BARLOW, H. B., FITzEHUGH, R. & KUFFLER, W. S. (1957). Change of organization in the receptive fields of the cat's retina during dark adaptation. J. Physiol. 137, BATTERSBY, W. S. & WAGMAN, I. H. (1964). Light adaptation kinetics: The influence of spatial factors. Science, N.Y., 143, BROCA, A. (1901). Causes r6tiniennes de variation de l'acuit6 visuelle en lumiere blanche. J. Phy8iol. Path. Gen. 3, CRAIK, K. S. W. & VERNON, M. D. (1941). The nature of dark adaptation. Brit. J. P8ychol. 32, CRAWFORD, B. H. (1940). The effect of field size and pattern on the change of visual sensitivity with time. Proc. Roy. Soc. B, 129, DREYER, V. (1961). On visual contrast thresholds: V. The influence of the background area on thresholds, determined by the method of constant stimuli. Acta Ophthal. 39, HARTLINE, H. K. (1940). The receptive field of optic nerve fibers. Amer. J. Phy8iol. 130, LYTHGOE, R. J. (1940). The mechanism of dark adaptation. Brit. J. Ophthal. 24, PIRENNE, M. H. (1958). Some aspects of the sensitivity of the eye. Ann. N.Y. Acad. Sci. 74, PIRENNE, M. H. & DENTON, E. J. (1952). Accuracy and sensitivity of the human eye..nature, Lond., 170, RATOOSH, P. & GRArAM, C. H. (1951). Areal effects in foveal brightness discrimination. J. exp. P8ychol. 42, RUSHTON, W. A. H. (1956a). The difference spectrum and the photosensitivity of rhodopsin in the living human eye. J. Phy8iol. 134, RUSHTON, W. A. H. (1956b). The rhodopsin density in the human rods. J. Phy8iol. 134, RuIJHTON, W. A. H. (1965). Visual adaptation. Proc. Roy. Soc. B, 162, RUSHTON, W. A. H. & WESTHEIMER, G. (1962). The effect upon the rod threshold of bleaching neighbouring rods. J. Physiol. 164, WESTHEIMER, G. (1963). Optical and optomotor factors in the formation of the retinal image. J. opt. Soc. Amer. 53,