Selective changes of sensitivity after adaptation to simple geometrical figures*

Transcription

1 Perception & Psychophysics Vol. 13. So Selective changes of sensitivity after adaptation to simple geometrical figures* ANGEL VASSILEV+ Institu te of Physiology. Bulgarian Academy of Sciences. Sofia Bulgaria Using simple geometrical figures. the effect of adaptation upon sensitivity to figures of the same or different form was investigated. The adapting stimuli (blank white field. contour triangles. or circles) were presented for I or 60 sec. The test stimuli were always triangles. presented for a short exposure time at and 1.2 sec after the offset of the adapting pattern. The threshold duration of the test stimuli was measured. It was found that both adapting patterns raised the threshold above that measured for adaptation with the blank white field. But it was only for the pair "adapting triangles-test triangles" that an increase of the time of adaptation beyond 5 sec resulted in an increase of this effect. This suggests that selective lowering in sensitivity during prolonged viewing of some figures may develop. The effect disappeared after about 1.2 sec following an exponential curve. Recently. it has been suggested that the process of pattern recognition is performed in three consecutive stages: (1) extraction of the signal from the noise, e.g., such processes as temporal and spatial summation and detection of changes in light intensity and color; (2) extraction of the simple features of the images, as lines of definite orientation, curves, corners, etc.; (3) decision process based on the features or on a combination of features-form recognition is the most important process at this stage (Glezer, 1966). Adaptation plays an essential role in the first stage, permitting the visual system to work over an extremely wide range of light intensities. Some recent investigations have shown that adaptation also exists at the second stage (Gilinsky. 1968; Blakemore & Campbell. 1969: and others). This naturally raises the question: "Can adaptation occur at the third stage of pattern recognition?" There are various examples of sensory adaptation, the most impressive of which were recently reviewed by Gregory (1966). It seems, however, that adaptation at the level of form recognition has not yet been shown-at least 'in the "direct way"-as a selective lowering in sensitivity to some figures as a result of their prolonged viewing. Evidence of such adaptation is more difficult to obtain than it may seem at first glance. As far as figures of equal form consist of similar elements, it is difficult to distinguish between adaptation at the level of form recognition and adaptation at lower levels of visual performance. But if once adaptation to some figures is found, then additional experiments could help to specify *This paper was partly reported at the Third Symposium on Biocybernetics, Leipzig, August tthe author thanks J. Nachmias and A. Penchev for their critical reading of the early versions of the manuscript, as well as N. Yakirnoff and S. Mateeff for their participation in the experiment and valuable discussion of the experimental conditions. the level at which it occurs. Some possible experiments of this kind are pointed out below. This paper is concerned with an initial step in searching for adaptation at the level of form recognition. namely, to check out the possibility of selectively influencing the sensitivity to some figures. METHODS The experiments were performed using a three-channel tachistoscope, Scientific Prototype GB-320. This apparatus allows the presentation of three different stimuli in the same region of the visual field. in sequence and of durations chosen by the operator. The overall paradigm of the experiments was as follows. The S was presented with an adapting pattern for duration t 1: then. after an interval of duration t.. a test pattern was presented for duration t 3 The temporal sequence of stimulation is shown in Fig. 1. During the interval t 2 and the interval between trials. the field was uniformly illuminated, without objects. with a luminance equal to the mean luminance of both adapting and test fields. namely, 22 cd/rn". The inter trial interval was 10 sec. Three adapting stimuli were used: (a) "blank" card, containing only two small crosses to facilitate accommodation: (b) card containing contour triangles with sides each 20 min of arc long (Fig. 2. Adapting Pattern b): and (c) card containing circles instead of the triangles. The triangles and circles were drawn with India ink and were equated for contrast. thickness. and length of the contours. Two test patterns were employed in order to introduce a forced choice procedure. One of the test cards contained the same triangles as Adapting Pattern b. while the other contained identical triangles but with interruptions of about 3 min of arc on each of the sides (Fig. 3. Test Stimuli a and b). The test patterns were presented in a random sequence. The S's task was to indicate after each trial which of the test stimuli had been presented. The minimum duration t 3, necessary for three consecutive correct answers was measured. using the "up-and-down transformed response rule," as described by Wetherill and Levitt (1965). In every daily session, the procedure continued until three reversals of t 3 for each condition were obtained. The magnitude of each step of t, was 0.04 log units. The test duration measured in the way described will be referred to as a "duration threshold." The duration threshold was the most convenient measure to obtain under the conditions of the experiment. Since its values did not exceed 12 msec, i.e. were in 356

2 SELECTIVE CHANGES OF SENSITIVITY AFTER ADAPTATION 357 ter trial Adapting Interf'igure Test Intertrial White) Pattern ("bite) Stimulus (White) uration: t1 tz t. 10 Sec Fig. 1. Diagram of the temporal order of stimulation. for each condition in order to maintain the levels of luminance as close as possible. Nevertheless, a detectabl shift in the luminance occurred at the moments of change of the channels. The value of this shift was not measured. but it was small enough not to have any influence on the threshold. All the Ss had normal vision and were between the ages of 22 and 35 years. RESULTS In the first part of the experiment, the card containing triangles and the blank card were presented as adapting stimuli. The interval t 2 was held constant at 0.4 sec, and the time of adaptation was varied (I, 5, 30, or 60 sec). The thresholds for the three Ss are shown in Fig.4 as a function of the time of inspection of the adapting pattern. For the sake of clarity, the data for S S.M. are shifted upwards by 2 msec, and that for N.Y. by 4 msec. The means and ± 1 standard error are c Fig. 2. Adapting Patterns b and c (Adapting Card a was a blank white card with only two small crosses to facilitate the accommodation). a the range of temporal summation, it can be considered to be a satisfactory measure of the contrast sensitivity, The stimulation was binocular, without artificial pupils. In order to avoid the appearance of afterimages, the Ss were instructed to move their gaze around the central part of the adapting pattern while looking at it. When the time of adaptation was more than 5 sec, a warning signal was given 3-5 sec before the test stimulus. The durations t, and t z were constant in a block of trials. and the Ss knew their values. Each block finished after three reversals of t. as described above. The order of blocks was randomly selected and changed in each daily session. Each session, Adapting Pattern a was used to obtain the reference level. The order of adapting patterns-a and one of the remaining two-was balanced out over the sessions. In each trial, after the S's report. the test stimulus was presented again. but for a longer time-for about 0.2 sec. This was to give the 5 feedback about the correctness of the answer and to keep the lamps of the test channel warmed up. The intensity of the lamps of the other two channels was readjusted Fig. 3. Test cards.

3 358 VASSILEV presented. each value being estimated from nine reversals of t,. The points for zeroinspel'tion time were obtained in the trials with the blank adapting card. (To check on the possibility of some artifacts. such as attention shift. etc.. this card was also presented for and 60 sec. However. no dependence of the threshold on this duration was found. As far as these results show the threshold without adaptation to a patterned stimulus, they were combined and presented at zero inspection time.) It is evident from the figure that the increase of the time of adaptation raised the threshold. e.g.. the threshold after 30 or 60 sec adaptation was always higher than after 1 or 5 sec adaptation. The curves are quite similar for NY. and AV. (the author). but a maximum at 30 sec with S.M. was obtained, for which I have no explanation. The next figure (Fig. 5) shows the time course of the recovery of sensitivity after the offset of the adapting pattern. The adaptation time was always 30 sec, and the interval t, was varied (.1..2,.4,.8. or 1.2 sec). The points at the right-hand side. connected with the curves by dashed lines. represent the threshold after adaptation to the blank field. After adaptation to triangles, the threshold exceeded 2 3 times the control value and fell approximately along an exponential curve. After an interval of 1.2 sec. it no longer differed from the control value. An experiment analogous to that presented in Fig. 5, but with the card containing circles being used as the 12 N.Y.I '""I I! I 10 I 3 J /I Q I r "Y. I i, i I- y----i- o 20 INSPECTION TIWE -1 Fig. 4. Duration threshold of discrimination of whole from interrupted triangles as a function of time of adaptation (inspection time). Adapting pattern-triangles: t, =.4 sec. The curves for Ss S.M. and N.Y. are replaced along the ordinate with 2 and 4 msec, respectively. Vertical bars show ± 1 standard error. 40 SECONDS '1. '"Q z 12 0 u L!J <f) ::::; -' ;[ 0 -' 0 :I:.J.J '" :l:: z 2 4er:: is 1J N.Y. ("".-nsj.. i S "2",,5i..., A... _-._} Tl ME AP:R ADAP' AIIO 'Ij - S ::CO,..: s Fig. 5. Duration threshold of discrimination of whole from interrupted triangles as a function of the interval between adapting and test patterns. Adapting pattern-triangles: time of adaptation = 30 sec: the points on the right (above the mark "Pr, white" on the abscissa) are the thresholds after adaptation to the blank field. adapting card, was performed with the same Ss. The results are presented in Fig. 6. When compared with Fig, 5, this shows that adaptation to circles increases the threshold for triangles to a lower degree and for a shorter time interval than does adaptation to triangles. In addition, Fig. 6 shows that if one wanted to see how the threshold depends on the duration of adaptation to circles (i.e., to perform an experiment like that presented in Fig. 4), the interval t, would have to be set at a lower value than the.4 sec used in the experiment with triangles. Thus. the influence of the time of adaptation to circles upon the sensitivity to triangles was investigated with an interfigure interval of.1 sec. For the purpose ofcomparison. this time interval was also investigated with adaptation to triangles. The results are presented in Fig. 7 with separate coordinates for each S. part of the data from Fig. 4 also being o.s 2

4 SELECTIVE CHANGES OF SENSITIVITY AFTER ADAPTATION 359 DISCUSSION III 10 c u W III :3 i 8 I C 5 w a: NY (., ms i [--_.1. S'" (-2 ms) _ A.V....! TIME AFTER ADAPTATION-SECOND Pro wtllte Fig. 6. Same function as in Fig.4 but after adaptation to circles. shown. Unlike Figs. 4-6, no curves are shifted. The conditions under which each curve was obtained are shown in the figure. The word "same" means adaptation to the same figures as the test ones (i.e.. triangles), and "different" means adaptation to circles. The interfigure interval is also given. For all Ss, only the data with adaptation to triangles and the interfigure interval of.4 sec show significant increase of the threshold with the increase of the time of adaptation. The other consistent result seems to be a decrease of this threshold when the adapting patterns werecircles after 30 sec adaptation, in comparison with the same threshold for 5 sec adaptation. The upper curves (adaptation to triangles and interfigure interval of.1 sec) show some increase of the threshold with inspection time for N.Y. and SM., and no changes for A.Y The data presented in Figs. 4-7 show that a prolonged viewing of some figures (triangles in this case) has an effect on the sensitivity to the same figures which differs from the effect 01 other figures: it is stronger, lasts longer, and is positively related to the time of adaptation. This rules out one possible supposition, namely, that the observed decrease in sensitivity is a nonspecific one, due simply to the presence of a nonuniformly illuminated adapting field. The results do not show clearly, however, if this change in sensitivity is connected with some process of adaptation. In the experiments, the significance of the time of adaptation and of the interval between the figures were systematically investigated because these factors were expected to characterize the processes underlying the raising of the threshold. The influence of the time of adaptation on the threshold suggests the involvement of slow processes such as adaptation or fatigue. This seems to be the only explanation of the fact that 30- and 60 sec viewing resulted in a higher threshold to the same figures than did 1 and 5-sec viewing. Such a suggestion is also supported by the finding that the increase of time of adaptation raised much more the threshold measured at t 2 =.4 sec than the threshold at t 2 =.1 sec. But, on the other hand, the sensitivity recovered too quickly to be accepted as being influenced by adaptation. In Blakemore and Campbell's experiment with adaptation to gratings, a 60 sec adaptation keeps the sensitivity low for about 30 sec. In the experiments reported here, it is only after 1.2 sec that the threshold does not differ significantly from the control level (Fig. 5). Such a short-lasting effect is possibly connected with transient responses to the disappearance of the adapting stimulus. But, if so, then this response must depend on the time of adaptation, which also means the involvement of slowly developing changes of sensitivity. The most probable explanation for the short-lasting change in sensitivity seems to be that the 30 sec adaptation with the l G-sec intertrial N.Y. S,M, A,V. 10 IL----' :.l., :3 i, 8 o en '" rr i! Fig. 7. Same function as in Fig. 3. but measured under three conditions: adapting pattern-triangles and t z =.1 or.4 sec (the curves labeled "same"): and adapting pattern-circles: r, =.1 sec (the curves labeled "different"). The data obtained with each S are presented with separate coordinates. and no curve is shifted from the real position., J 20 3C JC 3C

5 360 VASSILEV interval was insufficient to produce enough effect. Other explanations. however. are also possible. It may be that ad apt a t ion to figures differs in its temporal characteristics from the known kinds ofadaptation. Two tasks still remain before answering the question, "Can adaptation occur at the third stage of pattern recognition?" The first is to look for experimental conditions which provide more prolonged changes in sensitivity. The second is to find the localization of these changes. The results rule out the supposition that the decrease in sensitivity occurs at the first of the three stages of visual information processing mentioned in the introduction (as far as features of this decrease were found which were specific only for the pair "adapting triangles-test triangles"). However. such data do not necessarily prove the existence of adaptation at the third stage or at high "logical" level. a hypothetical level of invariant response to different triangles. The test stimuli used consisted of lines of three orientations, and the observed change in sensitivity could well be a result of adaptation to lines rather than to the triangles. This alternative can be checked in experiments with different orientation of the adapting and test stimuli. For example, it is known that some transformations are invariant at the level of pattern recognition and some are not. A figure rotated about its horizontal axis is a new object for the visual system (Nevskaya, 1966). But in the case of triangles used in this experiment. such transformation does not change the slope of the lines fanning a triangle. Thus, adaptation at the second stage should not depend on this transformation, and adaptation at the third stage would "feel" it. REFERENCES Blakemore, G., & Campbell, F. W. On the existence of neurones in the human visual system selectively sensitive to the orientation and size of retinal images. Journal of Physiology (London) Gilinsky, A. S. Orientation-specific effects of patterns of adapting light on visual acuity. Journal of the Optical Society of America Glezer, V. D. Mechanisms of visual image identification. (In Russian) Leningrad: Nauka, Gregory, R. 1. Eye and brain: The psychology ofseeing. New York: McGraw-Hill, Nevskaya, A. A. On the invariance of human visual recognition. (In Russian) Problemi Fiziologicheskoi Optiki, 1966, 13, Wetherill. G. B., & Levitt, H. Sequential estimation of points on a psychometric function. British Journal of Mathematical & Statistical Psychology, 1965, 18 (Part 1) (Received for publication June 6, 1972; revision accepted and returned to author for language revision September 20, 1972; final acceptance, December 20, 1972.)