Depth seen with subjective

Transcription

1 Japanese Psvcholog cal Research 1983, Vol.25, No,4, Depth seen with subjective contours1 TAKAO SATO2 Department of Psychology, Faculty of Letters, University of Tokyo, Bunkyo-ku, Tokyo 113 The phenomenal depth effect induced by Kanizsa figure was quantitatively compared with that evoked by real superposition cue using apparent motion as a probe. Experiment 1 was carried out to test this new method by studying the way apparent motion interacts with a figure inserted in the path of motion as a function of the distance-in-depth between the flashes and the inserted figure (relative distance). In Experiment 2, using this new method, the perceived position-in-depth was measured for three types of inserted figures; (A) a simple rectangle,(b) a rectangle with a real superposition cue, and (C) a rectangle consisting of subjective contours. The result indicated that the apparent depth effect induced by subjective contours is larger in magnitude than that evoked by the superposition cue given by real contours. This finding was discussed in relation to the depth theory of subjective contours. Key words: subjective contour, depth perception, apparent motion, Kanizsa figure, superposition. Subjective Contours and Depth Impression It is now a widely accepted phenomenon that the area surrounded by the Kanizsa type subjective contour (Kanizsa, 1955) is seen in depth. A typical example shown in Fig. 1 A and the white triangle at the center is seen above or in front of the surface of paper. This phenomenal depth is an important characteristic of subjective controus which has been pointed out as early as in 1904 by Schumann who originally described this kind of subjective contour. It has been suggested by Coren (1972; see also Rock and Anson, 1979 for a review of various theories) that the existence of a depth cue, namely the superposition, in the stimulus configuration is the critical factor which gives rise to the perception of subjective contour itself. However, when real contours are added to a Kanizsa t A part of this study was presented in a different form at the 4-0th annual Convention of Japanese Psychological Association, Nagoya, September, I am grateful to Professor Hiroto Katori of University of Tokyo for his constant encouragement throughouthe course of this research. I also wish to thank Professor Tadasu Oyama and Dr. Yuko Kimura for their critical reading of earlier drafts of this manuscript. figure to replace the subjective contours, the depth impression is much reduced or even disappears (see Fig. 1 B). This is quite puzzling in that depth impression is stronger when the depth cue is anomalous than is fully provided. Although this paradoxical situation is quite intriguing, no quantitative investigation has ever been attempted. The lack of data might stem from the fact that it is quite difficult to measure the phenomenal depth induced by subjective contours. The conventional method, a direct matching of position-in-depth with a probe which can be moved, is not very effective. The test of this direct matching method was carried out in an earlier pliot study in the author's laboratory, whereby the subject was asked to match the position-in-depth of a small light spot with that of the surface surrounded by subjective contours. The result indicated that the depth impression and very often the perception of contours itself disappears. This probably resulted from the damaging effect of fixation on the perception of subjective contours, which has been pointed out by several authours including Schumann (1904); the subjective contours tend to disappear when they are directly fixated.

2 T. Sam Gregory and his colleagues devised a method for measuring apparent depth in the Muller-Lyer figure (Gregory, 1965). In their method, the subject sees the test gure with only one of his eyes while fihe sees the depth-probe binocularly. This method was also employed in the author's laboratory with the Kanizsa figure as the stimulus. It was found that the monocular viewing of the test figure did not reduce the damaging effect of fixation on the perception of subjective contours, and therefore this method cannot be used reliably to measure the depth effect of subjective contours. A completely new method for measuring the apparent depth of subjective contour without any damaging effect on subjective contour is thus necessary. This paper examines the possibility of using apparent motion as a probe to measure position-in-depth of subjective contours. Apparent Ilotion as a Depth Probe When a rectangle, for example, is inserted in the pathway of an apparent (optimal) motion, the observer usually sees the motion going through behind the inserted figure. The trajectory of the motion is often curved at the center. That is, the center of pathway is seen as being farther from the observer than the two visible ends. This phenomenon is not a novel one and has been known since the beginning of the century. The description, however, has been limited to the case in which the flashes inducing apparent motion (hereafter called ashes) and the figure inserted in the fl motion trajectory (hereafter called inserted gure) are placed within the same fi frontparallel plane. Thus, the question to be posed is: What does an observer see when the flashes are placed closer to him relative to the inserted figure? It is quite plausible in this case that he sees the motion passing in front of the inserted figure. If this happens and the position-in-depth of flashes relative to the inserted figure has a systematic effect on the frequency of seeing A B C A Fig. 1. Subjective contour figures adapted from Kanizsa (1955).(A) Subjective white triangle. (B) Real contours are added to replace the subjective contours. B Fig. 2. Schematic representation of the three types of interaction between apparent motion and a rectangle inserted in the path of motion. Each pannel represents a typical example of each type.(a) Type 1, a motion always passing behind the inserted figure.(b) Type 2, a motion rotating around the inserted figure.(c) Type 3, a motion always going in front of the inserted figure.

3 Depth seen with subjective contours the motion passing behind or in front of the inserted figure, apparent motion may be used as a probe to measure the posidon-in-depth of an object. The obvious advantage of using apparent motion as a probe is that it does not require fixation or concentration of attention to any specific point in the visual field. To examine this possibility, the author and his colleagues undertook a preliminary observation. It was found that apparent motion could actually be seen passing in front of the inserted figure when flashes were placed closer in depth to the observer relative to the inserted figure. As relative distance-in-depth between flashes and inserted figure (hereafter called relative distance) was changed systematically, three types of motion were identified, typical examples of which are shown in Fig. 2. They are (1) type 1, both rightward and leftward motion passes behind inserted gure as shown in Fig. 2 A, (2) fi type 2, one way (either rightward or leftward) passes behind the inserted figure and the other way passes in front of inserted figure, i.e. the motion is perceived as rotating around inserted figure as shown in Fig. 2 B, and (3) both ways of motion pass in front of inserted figure as shown in Fig. 2 C. In the preliminary observation, when relative distance was small, type 1 was almost always seen. As relative distance was increased, however, frequency of seeing either type 2 or type 3 increased. The primary objective of this study is to compare quantitatively apparent depth which accompanies the perception of subjective contours and the depth induced by real contours with superposition cues. In Experiment 1, feasibility of using apparent motion as a probe to measure position-indepth was examined using four naive subjects. A systematic effect of the relative distance in depth between flashes and inserted figure on the type of motion perceived by the observer (i.e. the motion going in front of, or in back of the insert- ed figure) was obtained. Following this, in Experiment 2, the depth induced by subjective and real contours were compared using the new method developed in Experiment 1. Experiment 1 The purpose of this experiment is to examine whether or not apparent motion can be used as a probe for measuring the position in depth of a figure that is inserted in the path of the motion. Method Subjects. Four undergraduate students, two male and two female subjects, participated in this experiment. They were all unaware as to the purpose of this experiment and they all had excellent uncorrected vision with visual acuity better than 1.2. Stimulus and apparatus. The general construction of the visual display is shown in Fig. 3. The subject binocularly saw the display from a viewing distance of 115 cm. The inserted figure was drawn on a black poster board with self-luminescent paint. The flashes were four red light-emitting diodes (LED's). A small apperture for each LED was made on two metal plates on which the LED's were mounted. These Fig. 3. Stimulus display. Inserted figure is painted on a poster board (A). The position-indepth of flashes (B) relative to the inserted figure can be changed by sliding the base (C).

4 T. Sato metal plates were, then, connected by a long metal rod to a sliding base which could be moved on track to adjust the position-in-depth of the flashes relative to that of the inserted figure. On the poster board, there were two holes through which the ashes could change their position-indepth back and forth. fl The configuration of the flashes and the inserted figure in the front parallel plane is shown in Fig. 4 A. The inserted figure was a rectangle which subtended a visual angle of 1 horizontally and 2.5 vertically. The figure was made of aluminum tape on which self-luminescent paint was applied. The width of the lines (tape) was 6' of visual angle. The diameter of the four circular light-sources (flashes) shown as crosses in Fig. 4 A was 2.4' of visual angle and the separation between two adjacent A B C Fig. 4. Stimulus figures. type of inserted figure.(a) Each panel shows a A simple rectangle. (B) A rectangle with real superposition cue.(c) A rectangle consisting of subjective contours. Crosses represent the position of flashes (LED's). light-sources was 1.5. TheLED'sfor flashes were regular red Ag-As types (Toshiba, TLR-109) and a DC current of 2 ma for each was supplied from a regulated powersupply. Four flashes, instead of conventional two, were used to induce more stable impression of apparent motion. The duration and the inter-stimulus-interval were thus not very critical to maintain an optimal motion, and it was possible to employ fixed values of duration and interstimulus-interval for all the subjects. The duration of each flash was 75 ms and the inter-stimulus-interval was 140 ms. Procedure. The experiment was conducted in a completely dark room and the subject was dark-adapted for at least 15 minutes prior to the experiment. Relative distance was changed from - I to +6 cm (a positive value indicates that the flashes were closer to the subject relative to the inserted figure and a negative value means the opposite) by steps of 1 cm, giving rise to eight different values of the relative distance. The method of constant stimuli was used for the measurement. Each value of relative distance was repeated 20 times, so that there were 160 (8 x20) presentations for each subject in total. These 160 trials were broken into four sessions each consisting of 40 presentations. In each session, the order of presentation was randomized. In each trial, the subject was presented with three cycles of an oscillating sequential presentation of the four flashes from the leftmost one to the rightmost, then back to the leftmost. That is, if the flashes are numbered 1, 2, 3, and 4 from the leftmost position, the sequence was 1, 2, 3, 4, 3, 2, 1, 2... and so on. The subject was instureted to give a verbal report immediately after each presentation on the way apparent motion and inserted figure interacted to each other; that is, the impression of the motion trajectory relative to the positon of the inserted figure. To have the subject fully

5 Depth seen with subjective contours understand his task, the three types of motion, which were found in the preliminary observations and described in the introuduction (see Fig. 2) were suggested as examples, but care was taken not to be too specific to avoid any strong bias. The subject was encouraged to report whatever he could see. There \v as no fixation point in the display, but the subject was instructed not to pursue the motion, and to pay attention around the center of the inserted figure. Results The description given by the four subject fell into the three types described above (Fig. 2), and there was a strong effect of relative distance on the frequency of seeing each of these three types of motion. When the relative distance was small (-1 to about +I cm), the subject saw the motion always going behind the inserted figure (type 1). As relative distance se as increased (i.e. the flashes were moved closer to the subject) to approximately +2 cm, the subject started to see the motion rotating around the inserted figure (type 2). When relative distance was increased further to approximately +4 cm, the subject often reported seeing the motion always in front of the inserted figure (type 3). There, was good agreement between the four subjects on this systematic trend, especially the change front the first to the second type at the relative distance of +1 cm which was very clear for all the subjects. However, the border between the latter two types was not very clear and there were individual differences. Two of the subjects were even able to alternate these two percepts at their will. Because of this, these two categories were collapsed and the data were summarized as relative frequency of seeing the following two types; (1) a motion always going behind the inserted figure, and (2) a motion with at least one way going in front of the object. The relative frequency of seeing the type 1 (always behind) for two typical individual subjects and the mean for the four subjects are shown in Fig. 5 as a function of relative distance. These results suggest that relative distance has a strong systematic effect on the type of motion perceived, especially on the frequency of seeing the type 1 motion. Supplementary Experiment 1 To verify the result of Experiment I, a small supplementary experiment was carried out using two subjects. In this experiment, the inserted figure was presented with a crossed disparity of 6' using a SY KU NM Fig. 5. The results of Experiment 1. The frequency of seeing type 1 motion is plotted as a function of relative distance. Left and center panels show typical examples of individual data and each point represents the results of 20 trials. The right panel illustrates the mean for the four subjects.

6 T. Sato A SS TU B Fig. 6. Results of Supplementary Experiment 1. As for Fig. 5 but for the stimuli with (0) and without (0) disparity (left and center panels). The right panel presents the stimuli used: rectangles with (A) and without (B) crossed disparity. conventional Wheatstone type haploscope.a The stimulus configuration was basically the same except for a large surrounding frame (8' x8') which was added to help fusion. The measurements were made under the same procedure as in the main experiment for a figure with disparity and a control figure with no disparity. When there was crossed disparity, the subject had very clear and compelling depth perception such that the rectangle was seen closer to him. This subjective impression was reflected clearly in the result of the experiment shown in Fig. 6. Here, the relative frequency of seeing type 1 motion is plotted in the same way as for Fig. 5, and the function for the disparity condition is shifted to the right relative to one for the control. Subjects. Ten graduate and undergraduate students (six males and four females) served as subjects. They were naive as to the purpose of the experiment, and they all had excellent uncorrected vision with visual acuity better than 1.2. Stimulus and apparatus. The general construction of the visual display was the same as for that used in Experiment 1. The only major difference between this ex- right, left, and center channels using a pair of half-mirrors each placed in front of each eye. Inserted figure was presented by means of a pair of stereograms in the right and left channels and the subject dichoptically saw the images of the stereograms reflected by the half-mirrors. The stereograms are shown schmatically in Fig. G. They ere negatives and lighted from the rear, so wthat the lines of the patterns were light while the back ground was dark. Four flashes, the same kind as those in main experiments were presented in the center channel, which was seen binocularly through the half-mirrors. Based on these results, as well as those from the main experiment, it is concluded that the interaction between apparent motion and object inserted in the path of motion is an useful mean for measuring position-in-depth of the object. Experiment 2 The purpose of this experiment is twofold. The first objective is to examine whether subjective contours interact with apparent motion in the same way as real contours do as shown in Experiment 1. The second objective, the main objective of this study, is to compare the magnitude of phenomenal depth induced by superposition cue with real contours and that which accompanies the perception of subjective contours, using the method developed in Experiment I. Method

7 Depth seen with subjective contours periment and Experiment 1 was in the type of inserted figure that was used. Figure 3 shows the three types of figures used; a simple rectangle (A), a rectangle with real superposition cue (B), and a rectangle consisting of subjective contours (C). The way these figures were made was the same as in Experiment 1. Procedure. The procedure was generally the same as in Experiment 1. For each subject, eight different values of relative distance from 1 to +6 cm by 1 cm steps were used, and each of these values was repeated 20 times for each inserted figure. Therefore, there were 160 (8 x20) presentations for each inserted figure and 480 (160 x3) presentations in total for each subject. These 480 trials were broken into 12 sessions each consisting of 40 presentations. The type of inserted figures presented was randomized among sessions and the value of relative distance was randomized within each session. Results All ten subjects identified the three types of motion (Fig. 2) for all three types of inserted figures. The frequency of seeing each type of motion was systematically affected by relative distance in the same manner as in Experiment 1, and the trend was quite similar for all three kinds of inserted figures. Here again, the perception of type 1 motion was quite stable, and the frequency of seeing this type was a clear function of relative distance. But the perception was not so stable for type 2 and 3. There were large individual differences, and seeing these two types of motion sometimes alternated even within a single presentation. Because of this, as in Experiment 1, the results were summarized as frequency of seeing type 1 motion as a function of relative distance. This is shown in Fig. 7 for two typical individual subjects and for the mean of the ten subjects. As relative distance was increased, the frequency of seeing type 1 motion began to decrease first for inserted figure A, then for inserted figure B, and finally it decreased for inserted figure C. This order was the same for all ten subjects without exception. If we define the threshold relative distance for seeing type 1 motion as the relative distance corresponding to 0.5 relative frequency of seeing type 1 motion (dotted line in Fig. 7), it is cm for inserted figure A, cm for inserted gure B, and cm for inserted fifigure C for the group data. To test the reliability of these differences, the thresholds for inserted figures A, B, and C were obtained for each individual subject using a graphical method SY SS MEAN N:10 Fig. 7. The results of Experiment 2. As for Fig. 5 but for three different types of inserted figures; a simple rectangle ( œ), a rectangle with real superposition ( ), and a rectangle consisting of subjective contours ( ). See Fig. 4 for details of each inserted figure.

8 T. Sate (Bock & Jones, 1970), and the differences between these thresholds were statistically evaluated using correlated 1-tests. The results indicated that both inserted figures B and C were located significantly closer in depth to the subject than A (p<.01, two tails) and the difference between inserted figures B and C was also significant (p<.05, two tails). Non parametric Walshtests was performed to verify the results of the 1-tests. They also indicated that all three differences were significant (p<.01). Supplementary Experiment 2 To establish the finding of supplementary Experiment 2 further, a supplementary experiment with smaller step-size (3 to 5 mm depending on subject) and with more repetitions at each relative distance (54 repetitions) was conducted using five subjects. Among these five subjects, two were new subjects and the other three were those that had participated in the main experiment. The measurement was made using the same material and procedure (except for the step-size and number of repetitions) as in the main experiment. The results for two typical subjects are shown in Fig. 8, and they replicated the important findings of the main experiment. The order in magnitude of threshold relative destance for the three inserted figures was the same for all five subjects; the threshold relative distance was smallest for inserted figure A, the threshold for inserted figure B was next, and that for C was largest. General Discussion First of all, the results indicated that subjective contours have a function equivalent to that of real contours. In this experiment, subjective contours interacted with apparent motion, in a way that was similar to that observed with real contours, as indicated in the psychometric functions in Fig. 6. Previous research has demonstrated that subjective contours may behave in a way similar to real contours, such as in the Poggendorff illusion (Gregory, 1972), tilt aftereffect (Smith & Over, 1976), and apparent motion of subjective figures (von Griznau, 1979). The present study is thus another example in which subjective contours have an equivalent function to that of real contours. Second, present results clearly indicated that apparent depth which accompanies the perception of subjective contours is larger in magnitude than that induced by superposition cue consisting of real contours. The results of Experiment 2 clearly indicated that the threshold relative dis- SU OH MEAN N=4 Fig. 8. The result of supplementary Experiment. As for Fig. 6 but in this experiment smaller step size and more repetitions were employed. Each point represents the results of 54 trials. Note that the step size is not constant across subjects.

9 Depth seen with subjective contours tance for inserted figure C (subjective contour) was larger than that for B (real superposition), and same trend was replicated in Supplementary Experiment 2. All the subjects in these two experiments agreed well in this respect. Thus, it was concluded that apparent depth induced by subjective contours is larger in magnitude than that evoked by real superposition cue. This conclusion might appear to be in agreement with Coren's (1972) depth hypothesis. However, Coren's hypothesis does not predict any difference in the magnitude of depth impression between real superposition and subjective contour, while the present results indicate that there is a difference. Nevertheless, an interesting phenomenon which could be interpreted as supporting the depth hypothesis was found. All the subjects saw apparent motion through the real rectangle while the moving red spot was passing behind the rectangle (dotted parts in Fig. 2). That is, the area surrounded by real contours was perceived as being transparent. On the contrary, while the motion was passing behind the rectangle surrounded by subjective contours, most subjects reported that the spot disappears; that is, the rectangle was this time perceived as being opaque. Coren (1972), in his effort to generalize Kanizsa type subjective contour with the binocular ones (e.g., Julesz, 1964; Kaufman, 1965; Lawson & Gulick, 1967) appealed that whenever clear subjective contours were evoked by the depth cue system, the area surrounded by the contours appeared as a opaque plane overlapping a region in the background. Therefore, this anecdotal finding gives limited support to Coren's hypothesis, but the support is by no means unequivocal. It should be noted that the present study is not one which can critically evaluate the currently proposed theories. Instead, the results should be considered as reporting another phenomenon, for which a sufficient theory of subjective contour must account. Further research is needed to uncover the mechanism underlying the perception of subjective contours. References Bock, R. D., & Jones, L. V Measurement and prediction of judgement and choice. San Francisco: Holden-Day. Coren, S Subjective contrours and apparent depth. Psychological Review, 79, Gregory, R. L.1965 Seeing in depth. Proceedings of Royal Institute, 40, 311. Gregory, R. L Cognitive contours. Nature, 238, GrUnau, M. W. von1979 The involvement of illusory contours in stroboscopic motion. Perception and Psychophysics, 25, Julesz, B.1964 Binocular depth perception without familiarity cues. Science, 145, Kanizsa, G.1955 Marzini quazi-percettiv in campi constimolazione omogenea. Revista di Psicologia, 49, Kaufman, L Some new stereoscopic phenomena and their implications for theories of stereopsis. American Journal of Psychology, 78, Lawson, R. B., & Gulick, W. L.1967 Stereopsis and anomalous contour. Vision Research, 7, Rock, I., & Anson, R Illusory contours as the solution to a problem. Perception, 8, , Schumann, F Einige Bebachtungen uber die Zusammenfassung von Gesichtseindriicken zu Einheiten. Psychologische Studien, 23, Smith, A. T., & Over, R Color-selective tilt aftereffects with subjective contours. Perception and Psychophysics, 20, (Received May 13, 1983; accepted Nov. 19, 1983)