Linear perspective and the Ponzo illusion: a comparison between rhesus monkeys and humans'

Transcription

1 Japanese Psychological Research 1996, Volume 38, No. 3, Special Issue: Cognition and behavior of nonhuman primates Linear perspective and the Ponzo illusion: a comparison between rhesus monkeys and humans' KAZUO FUJITA Department of Behavioral and Brain Sciences, Primate Research Institute, Kyoto University, Inuyama, Aichi 484, Japan Abstract: Rhesus monkeys trained to classify bars of different lengths as "long" or "short" were tested for their bias of this classification when photographic linear perspective was superimposed on the bar. The photograph was either an upright or an upside down picture of the landscape that provided good linear perspective to human eyes. In Experiment 1, these photographs appeared together with the upwardconverging lines. The rhesus monkeys reported "long" more often when the target was located above the center of the stimulus, that is, near the apex of the converging line, than when it was located below the center, irrespective of the orientation of the photograph. In Experiment 2, the photographs appeared without the converging lines. The subjects showed a bias similar to, but less than, Experiment 1 only for the upright photograph. No bias was found for the upside down photograph. It was concluded that in nonhuman primates photographic linear perspective induces the Ponzo illusion, but the effect is so small that it is overshadowed by the converging lines typically used to produce the Ponzo illusion. In Experiment 3, humans tested in a comparable procedure perceived the illusion as long as the upright picture was there, irrespective of the presence or absence of the converging lines. This suggested that linear perspective provided by natural photographs strongly induces the illusion in humans. Key words: Ponzo illusion, visual perception, linear perspective, rhesus monkeys, chimpanzees, humans. Geometric illusions have been a major field of study among psychologists of human perception (e.g., Coren & Girgus, 1978 Frisby, 1979 Imai, 1984), because they provide useful materials with which to investigate human visual information processing. Several different theories of geometric illusions have been proposed: misapplied constancy scaling theory (Gregory, 1%3), assimilation theory (Pressey, 1971; Pressey & Epp, 1992), blur theory (Chiang, 1968; Coren, 1969), and contour proximity theory (Fisher, 1969,1970). However, none seems to provide a general account of illusions. Part of the reason seems to be that multiple mechanisms are involved in many of the illusions and small modifications to the figure could enhance some of the mechanisms while inhibiting others. One interesting strategy to address possible mechanisms of visual illusions is to test illusory perception of nonhuman animals. Nonhuman animals are likely to have less complicated perceptual mechanisms than humans. Therefore it ' This study was supported by the grant-in-aid for Scientific Research, Ministry of Education, Science, and Culture, Japan, Nos and I wish to thank Sumiharu Nagumo of the Primate Research Institute, Kyoto University, for his technical support Correspondence may be addressed to the author, who is now at the Department of Psychology, Faculty of Letters, Kyoto University, Yoshida-honmachi. Sakyo-ku. Kyoto 606, Japan, or to fulita@kupsy kyoto-u ac 1p via Japanese Psychological Association Published by Blackwell Publishers Ltd, 108 Cowley Road, Oxford OX4 ljf, UK and 238 Main Street, Cambridge, MA 02142, USA

2 Ponzo illusion in nonhuman primates 137 is expected that nonhuman perceptual processes should give us insight into the fundamental aspects of human perception, which could be too complicated to analyze in itself. Several nonhuman animals have been shown to be susceptible to geometric illusions. For example, Bayne and Davis (1983) demonstrated that rhesus monkeys experienced a version of the Ponzo illusion, in which horizontal bars appear longer as they approach the apex of inverted-v context lines. Benhar and Samuel (1982) showed that anubis baboons perceived the Zoellner illusion, in which parallel lines look non-parallel because of the transverse short diagonal lines. Dominguez (1954) reported that monkeys perceived a horizontalvertical illusion, in which vertical lines appear longer than horizontal lines However, these analyses are little more than a demonstration, and fall short of ascertaining whether the observed illusory perception is homological to that of humans. Fujita and his colleagues analyzed perception of the Ponzo illusion by pigeons, rhesus monkeys, chimpanzees, and humans in much more detail (Fujita, 1994; Fujita, in press; Fujita, Blough, & Blough, 1991, 1993). In this series of experiments, some aspects of the illusion were found to be common among these species, including humans, but others were not. For instance, all the species tested perceived the Ponzo illusion, but the magnitude of the illusion was overwhelmingly greater in pigeons than in primates. Additional linear perspective provided by multiple converging lines enhanced the illusion in no species. On the other hand, two short vertical lines instead of the converging lines induced very little illusion in rhesus monkeys, whereas in humans they induced illusion greater than with converging lines. Clearly, more analysis of nonhuman illusory perception is needed to discuss the relationship of human and nonhuman illusory perception. In the present study, I tested the effects of photographic linear perspective on the perception of the Ponzo illusion by nonhuman primates. According to Gregory s (1963) theory of misapplied constancy scaling, or a perspective theory, stronger linear perspective ought to enhance the illusion, because the visual system tries to adjust the small retinal image to maintain (actually misapplied) size constancy. This was demonstrated in some human studies (Leibowitz, Brislin, Perlmutter, & Hennessy, 1969), though there are also studies that do not support this theory (e.g., Fineman & Carlson, 1973; Georgeson & Blakemore, 1973; Humphrey & Morgan, 1965; Newman & Newman, 1974). Perception of perspective from drawings seems to differ across cultures and educational levels in humans, and corresponds to the magnitude of the Ponzo illusion (Brislin, 1974; Brislin & Keating, 1976; Kilbride & Leibowitz, 1975). Thus previous failure in animals to find the effect of linear perspective provided by geometric figures could be due to difficulty in abstracting the perspective in the figure. Cerella (1977) demonstrated that at least pigeons have great difficulty in perceiving perspective from line drawings. Therefore it is necessary to test nonhumans using pictures with natural perspective, which is likely to provide easier clues of depth. Experiment 1 In this experiment, upright and topbottom reversed pictures providing a good linear perspective to human eyes were compared for their ability to enhance the Ponzo illusion induced by upward-converging context lines. Method Subjects. The subjects were two young adult rhesus monkeys (Macaca mulatta). One was an 8-year-old male named Gonta, and the other was a 7-year-old female named Jusco. The subjects did not have their food restricted during the experimental period except that occasionally a mild deprivation of food was imposed on them when their performance deteriorated considerably. The use of these subjects adhered to the Guide for the care and use of laboratory primates (1986) of the Primate Research Institute, Kyoto University. Apparatus. The experimental box was a monkey cage sized 70 cm x 70 cm x 70 cm. A Q Japanese Psychological Association 1996

3 138 K. Fujita Upright picture + lines Upside down picture + lines Low context Middle context (training stimuli) High context Figure 1. Examples of stimuli used in Experiment 1 14-inch CRT monitor (KX-I4HD, Sony) was installed on one wall of the box. A touch sensor (Hypertouch, Nissha Intersystems) was mounted on the monitor. A universal feeder (S-100, Sanso Electronics) could deliver pieces of food - a mixture of sweet potatoes and monkey chow - into a food cup at the bottom of one panel of the experimental box. An electronic doorbell installed behind the panel signaled the delivery of food. A house light at the top illuminated the experimental box during the experimental session. A personal computer (PC286VE, Epson) controlled all the equipment. Stimuli. Stimuli were graphic patterns on the computer monitor. The spatial resolution was pixels (100 pixels covered 44 mm on the monitor). The viewing distance was about 30 cm. Viewed from this distance, the visual angle of 100 pixels was 8.41 min. Figure 1 shows examples of the stimuli used. Each stimulus had one horizontal target bar and a context. The target bar was 40, 48, 56, 64, 72, or 80 pixels long and 10 pixels thick. Contexts consisted of two black upwardconverging lines and a colored photograph of an upright or upside down landscape. The upward-converging lines were 140 pixels high and 3 pixels wide. They were slanted at 45" and so met at 90". The size of the photograph was 512 pixels horizontally and 260 pixels vertically. The remaining area of the display was white. In baseline training, described below, the context lines were placed so that the target bar appeared in the center of them (middle-context condition, see Figure 1). For tests, the context appeared 20 pixels above (high-context condition) or below (low-context condition) the middle-context condition. The upwardconverging lines and the photographic context always moved together. Procedure. Figure 2 outlines the trial. Trials started with a stimulus at the center of the monitor after intertrial intervals of 5 s. Five touches on the target bar produced two choice locations at both bottom corners of the monitor. The two choice locations were assigned to 0 Japanese Psychological Association 1996

4 Figure 2. - Intertrial interval Timeout Ponzo illusion in nonhuman primates 139 Choice response A schematic representation of the task. long or short. The assignment was counterbalanced across subjects. The subjects were required to choose short choice locations for the three shorter bars (40, 48, and 56 pixels) and to choose long for the remaining three longer bars (64, 72, and 80 pixels). In training trials, a single response to the choice location corresponding to the length of the bar was reinforced either by food (primary reinforcer) accompanied by a 0.5 s doorbell sound or a doorbell sound only (conditional reinforcer). The probability of the primary reinforcer was 25%. A single response to the other choice location was followed by 5 s of timeout, during which the monitor went dark. Both subjects had mastered this discrimination of the length of the horizontal bar presented without context patterns before this experiment (Fujita, 1994, in press). Therefore the subjects started with the baseline discrimination training, in which the target bar was presented together with the upward-converging lines and the upright and upside down photographs. The number of trials per session was 384. A series of test sessions followed after overall accuracy was more than 80% correct, the accuracies for the two extreme lengths were above 90% correct, and those for the two intermediate lengths were more than 50% correct for two consecutive sessions. The test sessions consisted of 432 trials, 288 of which were baseline (training) trials and 144 were test trials. Test trials consisted of all 36 combinations of two types of contexts (picture upright and upside down), three context locations (high-, middle-, and low-context conditions), and six bar lengths. Thus each combination appeared four times per session in a random order. In the baseline trials, subjects responses were differentially reinforced as in training sessions. In the test trials, all the responses, either long or short, were non-differentially reinforced either primarily or conditionally; that is, an all-reinforced-probe procedure was followed. The test sessions were repeated 10 times, each separated by at least one training session. At the end of the series of the test sessions, one original training session in which target bars appeared without context patterns was conducted to ensure that the subjects performances were based on the length of the target bar. In this testing procedure, perception of the illusion is demonstrated by the bias in the proportion of long and short choices as a function of the location of the context stimuli, low, middle, and high; that is, in conditions in which the subject overestimates the length of the target bar, the proportion of long choices tends to be higher, and vice versa. Results and discussion Both monkeys maintained accuracy as high as baseline sessions when they performed with isolated target bars (no context stimuli). This confirmed that the subjects were responding to the length of the target bar as a cue. Figure 3 shows the result of this experiment. The left panel is for the upright picture and the upward-converging lines, and the right panel is for the upside down picture and the upwardconverging lines. The horizontal axis gives the length of the target bar in pixels and the vertical axis the proportion of the subjects choices of long. As is clear from the figure, the proportion of long reports was highest for low-context conditions, intermediate for middle-context conditions. and the lowest for 8 Japanese Psychological Association 1996

5 140 K. Fujita Upright picture Upside down picture + upward-converging lines C Bar length in pixels Figure 3. The results of Experiment 1. The vertical axis gives the proportion of the subjects reports that are long and the horizontal axis gives the length of the target bars in pixels. Data are the average of the two rhesus monkeys. The left panel is for the upright picture plus inverted-\/ context and the right panel is for the upside down picture plus inverted-\/ context. Solid lines are for the lowcontext, broken lines are for the middlecontext. and dotted lines are for the highcontext condition. high-context conditions. A four-way ANOVA of context type x context location x bar length x subject, with N = 1, found a statistically significant main effect of the context location (F(2, 10) = 11.76,~ =.002). This consistent bias in the choice proportion of long and short demonstrated that rhesus monkeys perceived the Ponzo illusion when the natural photograph was superimposed on the converging lines. However, there seems to be no difference in this response bias between the upright picture and the upside down picture. The same ANOVA found no significant interaction of context location and context type (F(2, 10) = 0.737, ns). This result suggested that natural linear perspective, which ought to be stronger in the upright picture than the upside down one, did not enhance the Ponzo illusion in rhesus monkeys. This was consistent with previous results in pigeons and nonhuman primates using a line-drawn linear perspective (Fujita, in press; Fujita et al., 1991). Experiment 2 Experiment 1 failed to find that natural linear perspective enhanced the illusion provided by converging lines. In Experiment 2, a pure effect of natural pictures was tested using stimuli without converging lines. Method The subjects and the apparatus were the same as in Experiment 1, and the stimuli were exactly the same except that there were no converging lines (Figure 4). The same procedure as in Experiment 1 was followed. Results and discussion Both monkeys maintained accuracy as high as baseline sessions when they performed with isolated target bars without context stimuli. This confirmed once again that the subjects were responding using the length of the target bar as a cue. Figure 5 shows the proportion of long reports as a function of bar length in Experiment 2. The left panel is for the upright picture and the right panel is for the upside down picture. The data are the average of the two subjects. For the upright picture, the proportion of long reports was the highest for the lowcontext condition, intermediate for the middlecontext condition, and the lowest for the 0 Japanese Psychological Association 1996

6 Ponzo illusion in nonhuman primates 141 Upright picture only Upside down picture only Figure 4. Examples of stimuli used in Experiment g ao u) C 60 E! 'g M Bar length in pixels Figure 5. The results of Experiment 2. The left panel is for the upright picture without line contexts and the right panel is for the upside down picture without line contexts. Key as in Figure 3. high-context condition, to a somewhat lesser degree than in Experiment 1. On the other hand, this effect was absent with the upside down picture. A four-way ANOVA of context type x context location x bar length x subject, with N = 1, was conducted as in Experiment 1. This failed to find a statistically significant main effect of the context location (F(2, 10) = 2.50, p =.132) because of apparent lack of the effect for the upside down picture. On the other hand, the interaction of context location and context type was almost significant (F(2, 10) = 3.916, p =.055). In separate three-way ANOVAs of context location x bar length x 6 Japanese Psychological Assoclatlon 19%

7 142 K. Fujita subject, with N = 1, for each of the two orientations of the pictures, the main effect of context location was significant for the upright picture (F(2, 10) = 8.89, p =.006) but not for the upside down picture (F(2, 10) = 0.10, ns). These analyses demonstrated that rhesus monkeys perceive the illusion similar to that in Experiment 1 for the upright natural photograph, which provides good perspective. mere are two potential causes of this effect; one is the perceived perspective in the upright picture and the other is the contrast with small objects in the upper half of the picture where the scenery is subjectively far away. However, the latter effect seems inappropriate in this case because the upside down picture, in which there are exactly the same small objects in the lower half, had no symmetrical effect. It seems clear that rhesus monkeys perceive a Ponzolike illusion induced solely by the natural linear perspective. However, this effect does not seem strong enough to enhance the Ponzo illusion induced by line contexts because the orientation of the same picture had no effect in Experiment 1. Thus the contribution of the linear perspective to the Ponzo illusion by nonhuman primates seems small and, at best, supplementary. Experiment 3 In this experiment, human subjects were tested using a comparable procedure. Method Subjects. lko female and four male adult humans (Homo sapiens) aged between 23 and 42 years served as subjects. All had normal or corrected-to-normal visual acuity. The subjects received Y1,OOO after the session. Apparatus. Human subjects were tested using a mouse instead of a touch sensor. A 14-inch CRT monitor (PC-KD862S, NEC) and a personal computer (PC386GS, Epson) were used. Stimuli. The four types of stimuli from Experiments 1 and 2, those illustrated in Figures 1 and 4, and another type having only inverted-v context lines were used. The sizes of the stimuli were exactly the same as in Experiments 1 and 2. The viewing distance was about 50 cm. Viewed from this distance, the visual angle of 100 pixels was 5.04 min. Procedure. Human subjects were tested for one session using a titration of the bar lengths. The task was the same as that for monkeys and the subjects were instructed to choose one of the two choice locations depending upon the length of the target bar. Intertrial intervals lasted 1 s and the number of responses required for each target bar was one. The test session consisted of 407 trials The first 32 trials were training trials in which a target bar 59 or 61 pixels long appeared without context patterns. Correct responses ( short for 59 pixels and long for 61 pixels) simply advanced the trials, but incorrect responses were followed by 1 s of timeout, during which the monitor went dark. The remaining 375 trials were titration trials in which the five context patterns mentioned above appeared and all responses simply advanced the trial. In these test trials, 15 independent titration schedules (five types of context x three locations of context lines) ran simultaneously. The initial length of the target bar was 60 pixels. The titration followed 2-up 2-down step-1 procedure; namely the bar was shortened by 1 pixel when the subject reported long in two consecutive trials of the same stimulus patterns, and vice versa. The number of the titration trials was 25 per pattern. In this procedure, perception of the illusion is demonstrated by the difference in the bar length during the titration trials; for example, in conditions in which the subject overestimates the length of the target bar, the bar is made smaller, and vice versa. Results and discussion Figure 6 shows the magnitude of the illusion as percent overestimation as a function of the type of context. these values were calculated as follows. First, the length of the target bar during the last 10 titration trials of each stimulus pattern was averaged. Suppose these values for one context pattern are V,, V,, and V, for high, middle, and low context, respectively. The magnitude of the illusion for this context 0 Japanese Psychological Association 1996

8 Ponzo illusion in nonhuman primates 143 A- D-o- C+ F+ + lines picture + lines only picture Type of context Figure 6. The results of Experiment 3. The vertical axis gives the percent overestimation in lowcontext conditions compared with highcontext conditions and the horizontal axis gives the types of context. The data for individual subjects are shown as line graphs and the average appears as the histogram. pattern is (V, - V,) / V, x 100. The data for the individual subjects appear as line graphs and the averages appear in the histogram. Human subjects overestimated the length of the target bar in the low-context condition compared with the high-context condition for inverted-\! inverted-v plus upright picture, and upright picture without inverted-v contexts. A two-way ANOVA of context type x subject, with N = 1 (or one-way ANOVA of context type with subject as a block), revealed a significant main effect of context type (F(4,20) = 24.44, p =.O00). A post-hoc Tukey pairwise comparison suggested significant differences for all the combinations except inverted-v and upside down picture only, upright picture plus lines and upright picture only, and upside down picture plus lines and upside down picture only. This suggested that the magnitude of illusion was larger for the two stimuli with the upright picture than for the other three. Clearly, human subjects perceive the illusion more readily with the upright picture with natural perspective. As in the case of monkeys, this effect cannot be attributed to the contrast with small objects in the picture because the upside down picture had exactly the same small objects in it but produced no symmetrical 0 Japanese Psychological Association 1996

9 144 K. Fujita effect. This finding for the significance of linear perspective contrasted with monkey data. Another interesting point unique to the human data was that when the picture was turned. upside down it worked to reduce the illusion provided by converging lines to a level where they even had an opposite effect. This was not evident for monkeys in Experiment 1. General discussion Experiments 1 and 2 clearly demonstrated once again that rhesus monkeys perceive the Ponzo illusion. In Experiment 1, the magnitude of the illusion was comparable irrespective of the orientation of a picture having natural linear perspective superimposed on the upwardconverging lines. On the other hand, the same upright picture induced the illusion when it was presented without the lines. The fact that the upside down picture had no such effect confirmed that the contrast with small objects in the picture was not the source of this illusion. Thus we can conclude that in rhesus monkeys the natural linear perspective provided by photographs of real scenery induces the illusion. However, this effect seems to be so small that it is overshadowed by the effect of converging lines (Experiment 1). In Experiment 3, humans perceived a big illusion whenever the upright picture was presented, irrespective of presence or absence of the converging lines. Apparently humans receive a big effect from the linear perspective. This is consistent with some other human data with natural pictures as stimuli (Leibowitz et al., 1969). As in the monkeys, this effect was not due to a contrast effect with small objects in the picture because the upside down picture did not induce the illusion by itself. This effect of linear perspective in rhesus monkeys and humans may seem to contradict previous findings showing no enhanced effect with line drawings that provide perspective (Fujita, in press; Fujita et al., 1991). The animals failure could have been due to the inability to abstract linear perspective from line drawings (Cerella, 1977), as noted before. However. this does not account for the human data. Humans clearly are able to perceive depth from line drawings yet they received no advantage from the perspective provided by line drawings in previous studies (Fujita, in press; Newman & Newman, 1974). It is likely that the illusions induced by natural perspective are different from the Ponzo illusion. From a factor analysis of a variety of geometric illusions, Imai (1982) suggested that illusions induced by perspective drawings are to be classified into a group different from typical Ponzo illusions. The illusion induced here by an upright picture in natural perspective may be misapplied constancy scaling (Gregory, 1963). At the least, the illusion found here in both humans and monkeys cannot be accounted for by either an assimilation effect or a contrast effect. One puzzling phenomenon found in this study was that in humans an upside down picture superimposed on converging lines tended to induce a negative illusion, whereas the same upside down picture by itself did not. So far we have been unable to find a plausible explanation for this. The cause of this needs to be studied. In conclusion, this study demonstrated that: (1) linear perspective is a source of visual illusion that is common to humans and rhesus monkeys; (2) the strength of this factor is different between these two species; (3) in either species, the Ponzo illusion is not likely to be accounted for by this factor. References Bayne, K. A. L., & Davis R. T. (1983). Susceptibility of rhesus monkeys (Macaca mulatto) to the Ponzo illusion. Bulletin of the Psychonomic Society. 21, Benhar, E., & Samuel, D. (1982). Visual illusions in the baboon (Papio anubb). Animal Learning & Behavior, 10, Brislin, R. W. (1974). The Ponzo illusion: additional cues age, orientation, and culture. Journal of Cross Cultural Psychology, 5, 13%161. Brislin, R. W., & Keating. C. F. (1976). Cultural differences in the perception of a three-dimensional Ponzo illusion. Journal of Cross Cultural Psychology, 7, Q Japanese Psychological Association 1996

10 Ponzo illusion in nonhuman primates 145 Cerella, J. (1977). Absence of perspective processing in the pigeon. Pattern Recognition, 9,6568. Chiang, C. (1968). A new theory to explain geometrical illusions produced by crossing lines. Perception & Psychophysics, 3, Coren, S. (1969). The influence of optical aberration on the magnitude of the Poggendorff illusion. Perception & Psychophysics, 6, Coren, S., & Girgus J. S. (1978). Seeing is deceiving: The psychology of visual illusions. Hillsdale, NJ: Lawrence Erlbaum Associates. Dominguez, K. E. (1954). A study of visual illusions in the monkey. Journal of Genetic Psychology, 85, Fineman, M. B., & Carlson, J. (1973). A comparison of the Ponzo illusion with a textural analogue. Perception & Psychophysics, 14, Fisher, G. H. (1%9). Towards a new explanation for the geometrical illusions. I: The properties of contours which induce illusory distortion. British Journal of Psychology, 60, Fisher, G. H. (1970). Towards a new explanation for the geometrical illusions. 11: Apparent depth or contour proximity? British Journal of Psychology, 64, Frisby, J. P. (1979). Seeing: illusion, brain, and mind. Oxford: Oxford University Press. Fujita, K. (1994). Perception of the Ponzo illusion by rhesus monkeys and chimpanzees. Research report of Grant-in-aid for Scientific Research, Ministry of Education, Science, and Culture, Japan, No (pp ) (in Japanese). Fujita, K. (in press) Perception of the Ponzo illusion by rhesus monkeys, chimpanzees, and humans: Similarity and difference in the three primate species. Perception & Psychophysics. Fujita, K., Blough, D. S., & Blough, P. M. (1991). Pigeons see the Ponzo illusion. Animal Learning & Behavior, 19, Fujita, K., Blough, D. S., & Blough, P. M. (1993). Effects of the inclination of context lines on perception of the Ponzo illusion by pigeons. Animal Learning & Behavior, 21, Georgeson, M. A., & Blakemore, C. (1973). Apparent depth and the Mueller-Lyer illusion. Perception, 2, Gregory, R. L. (1963). Distortion of visual space as inappropriate constancy scaling. Nature, 199, Humphrey, N. K., & Morgan, M. J. (1%5). Constancy and the geometric illusions. Nature, u)6, Imai, S. (1982). A factor analysis of visual illusions. Journal of Social Sciences and Humanities (Jimbun-Gakuho), 152, 1-18 (in Japanese with English summary). Imai, S. (1984). Figures of optical illusions. Tokyo: Science-sha (in Japanese) Kilbride, P. L., & Leibowitz, H. (1975). Factors affecting the magnitude of the Ponzo perspective illusion among the Baganda. Perception & Psychophysics, 17, Leibowitz, H., Brislin, R., Perlmutter, L., & Hennessy, R. (1969). Ponzo perspective illusions as a manifestation of space perception. Science, 166, Newman, C. V., & Newman, B. M. (1974). The Ponzo illusion in pictures with and without suggested depth. American Journal of Psychology, 87, Pressey, A. (1971). An extension of assimilation theory to illusions of size, area, and direction. Perception & Psychophysics, 9, Pressey, A., & Epp, D. (1992). Spatial attention in Porno-like patterns. Perception & Psychophysics, (Received Nov. 10,1995; accepted March 16,1996) 0 Japanese Psychological Association 1996