Perception & Psychophysics 1996,58 (8), JJ68-JJ76
Perceiving self-motion in depth: The role of stereoscopic motion and changing-size cues
STEPHEN PALMISANO University ofNew South Wales, Kensington, New South Wales, Australia
During self-motions, different patterns of optic flow are presented to the left and right eyes. Previ ous research has, however, focused mainly on the self-motion information contained in a single pattern of optic flow. The present experiments investigated the role that binocular disparity plays in the visual perception of self-motion, showing that the addition of stereoscopic cues to optic flow significantly im proves forward linear vection in central vision. Improvements were also achieved by adding changing size cues to sparse (but not dense) flow patterns. These findings showedthat assumptions in the head ing literature that stereoscopic cues facilitate self-motion only when the optic flow has ambiguous depth ordering do not apply to vection. Rather, it was concluded that both stereoscopic and changing size cues provide additional motion-in-depth information that is used in perceiving self-motion.
Ofall the senses known to be involved in self-motion During self-motions, different patterns of optic flow perception-the vestibular, auditory, somatosensory, pro are presented to the left and right eyes (due to the sepa prioceptive, and visual systems-visionappears to play the ration ofthe eyes and their different angles ofregard; see dominant role (Benson, 1990; Howard, 1982). This is Figure 1). Theorists have, however, generally focused only demonstrated by the fact that compelling illusions of on the motion-perspective information contained in a sin self-motion can be induced by visual information alone. gle pattern, assuming that this is sufficient to accurately For example, when subjects are placed in a "swinging perceive self-motion (e.g., Gibson, 1950; Gibson et a1., room," where the walls and ceiling swing back and forth, 1955; Gordon, 1965; Heeger & Jepson, 1990; Koen they soon experience the illusion that they themselves are derink, 1990; Koenderink & van Doorn, 1981, 1987; Lee, swaying (Lee & Aronson, 1974; Lee & Lishman, 1975; 1980; Longuet-Higgins & Prazdny, 1980; Nakayama & Lishman & Lee, 1973). Similarly, when subjects are placed Loomis, 1974). Accordingly, the role that stereoscopic in inside a "rotating drum," a rotating cylinder with a pat formation plays in self-motion perception has received lit terned inner wall, they quickly experience an illusion of tle attention. self-rotation (Brandt, Dichgans, & Koenig, 1973; Mach, Recently, however, van den Berg and Brenner (1994b) 1875). These illusions occur because the swinging room have shown that, in some situations, heading perception and the rotating drum duplicate the visual stimulation (one aspect of self-motion perception) can be improved that normally occurs during real self-motions. by the addition ofstereoscopic cues. Their earlier research A major visual stimulus for self-motion perception is had found that heading estimates were error prone in the optic flow or the temporal change in the pattern of light presence of motion noise, when displays simulated ob intensities at the moving point of observation (Gibson, server motion through a cloud ofdots (van den Berg, 1992; 1966; Warren, Morris, & Kalish, 1988). Gradients ofop van den Berg & Brenner, 1994a). They subsequently dis tical velocity contain several potential sources ofinforma covered that when binocular disparities were added to tion about observer motion through three-dimensional these "cloud" displays, heading estimates became up to space (Gibson, Olum, & Rosenblatt, 1955). There is the four times more resistant to noise. Changing disparity was perspective change in location ofobjects in the optic array not essential for this improved heading performance: Sub (which shall be referred to as motion perspective) and jects performed just as well when each dot had a fixed their optical expansion/contraction (which shall be re retinal disparity for the duration of the display (in this ferred to as changing size). 1 case, only motion perspective simulated self-motion in depth). Van den Berg and Brenner concluded that stereo scopic vision improves heading perception indirectly by Portions of this research were presented at the 1995 Association for providing the depth order ofthe objects in the flow (rather Research in Vision and Ophthalmology Conference held in Fort Lau derdale, FL. The author thanks Barbara Gillam and Shane Blackburn than by providing additional motion in depth informa for their feedback and suggestions regarding this article and Keith tion). Furthermore, they argued that other depth cues, Llewellyn for the loan of his equipment. The author is also grateful to such as occlusion or texture gradients, might improve Myron Braunstein, George Andersen, and two anonymous reviewers for heading judgments in a similar fashion. their insightful comments on an earlier version of this article. Corre spondence should be addressed to S. Palmisano, School ofPsychology, There is reason to believe that stereoscopic informa University ofNew South Wales, POB I, Kensington, NSW 2033, Aus tion might also enhance the subjective experience ofself tralia (e-mail: [email protected]). motion, known as vection. In their study, Andersen and
Copyright 1996 Psychonomic Society, Inc. 1168 PERCEIVING SELF-MOTION IN DEPTH 1169 • • •• • • ••• • •• • iii·.·· • • ...... ',. • ...... • .. .." .. ' " . '...... •... ·;····~::r·~·.·· ' ' .. " -,' .. ' . ... " .. •• " ...... : ~ ...... -. '...... ' • •• • •••• '. •.. . .. '. . .. : ...... ' . , _ '.' .•. " ... .•...... • • • ..- •
Figure 1. A stereogram representing the different optic arrays presented to the left and right eyes at anyone point in time (converge in order to fuse the two images).
Braunstein (1985) simulated forward self-motion through z-axis toward the observer (the projection plane and the observer's a three-dimensional cloud of dots. The perceived three viewpoint remained fixed). A constant density was maintained by dimensionality of these displays was manipulated by al replacing each object as it disappeared from view at the opposite tering the motion-based cues to self-motion in depth. They end of space (a simulated distance of 20 m). All displays had a frame rate of60 Hz. found that the more three-dimensional the inducing dis Stereoscopic displays presented horizontally disparate patterns plays appeared, the stronger the vection in central vision. of optic flow to the two eyes. This was achieved by presenting the Although Andersen and Braunstein did not investigate disparate views in different colors on a single display, which was the role ofstereoscopic cues on vection, an argument can then viewed through red--eyan anaglyph glasses. To fuse these dis be mounted on the basis oftheir data. Ifit is assumed that plays, the subjects needed to verge behind the screen. Thus, prior to adding stereoscopic information to inducing displays their presentation, the subjects were shown a pair ofvertically dis placed nonius lines (one red, one cyan) separated by a disparity rep makes them appear more three-dimensional, it follows resenting the farthest distance simulated by the display? (Hebbard, that such displays might produce stronger vection (in 1962; Mitchell & Ellerbrock, 1955). They then had to alter their central vision) than those with motion perspective alone. convergence until the nonius targets were aligned, before triggering The present experiments investigated whether the ad the stereoscopic display. dition ofstereoscopic information to optic flow would in Nonstereoscopic displays were oftwo types. Monocularly viewed crease forward linear vection in central vision. They were displays presented a single pattern ofoptic flow to one eye. Binoc ularly viewed nonstereoscopic displays presented the same pattern designed to determine whether any such increases were of moving objects to both eyes (producing slightly different flow due to improved depth ordering or to additional motion patterns due to the different positions ofthe two eyes). Prior to the in depth information. In addition, a further two experi monocularly viewed displays, the subjects were told to lower an eye ments examined whether another source of motion in patch over the right eye. Before binocularly viewed nonstereoscopic depth information (changing size) could also produce an displays, the subjects were presented with a pair ofnonius lines set advantage for vection. at zero disparity (since the subjects had to verge on the screen to view these displays). After lowering the eye patch or verging on the screen, the subjects then triggered the nonstereoscopic display. GENERAL METHOD EXPERIMENT 1 Subjects The subjects were students in an introductory psychology course who received course credit for their participation. All had normal Experiment I compared the vection induced by stereo or corrected-to-normal vision and had no previous laboratory ex scopic and monocularly viewed displays simulating self perience with illusions ofself-motion. Different subjects were used motion in depth. In the case of stereoscopic displays, in each ofthe four experiments. vergence was consistent with the presence of three dimensional, virtual space behind the screen. Nonstereo Visual Displays All displays simulated forward self-motion through a cloud of scopic displays were viewed monocularly to remove any randomly positioned stationary objects. The objects were either vergence-based flatness information (Gogel & Sturm, filled-in squares or dots that moved at a constant rate along the 1972; Richards & Miller, 1969). 1170 PALMISANO
Both stereoscopic and monocular displays had chang Results and Discussion ing-size and motion-perspective information about self Self-motion was reported in 184 ofthe 192 trials (16 motion in depth. Since each display was relatively free of subjects responding to 12 stimuli). Separate repeated motion noise.' (unlike those of van den Berg and Bren measures analyses of variance (ANOYAs)wereperformed ner) and extraretinal information accompanied flow due on the onset and duration data. The means are shown in to eye movements, the depth order should have already Figures 2A and 28. Stereoscopic displays were found to been provided by these two monocular cues. That is, ob produce significantly faster onsets [F(1, 15) = 9.803,p < jects with larger relative sizes and faster relative veloci .007] and longer durations ofvection [F(1, 15) = 18.00, ties should have appeared to be nearer to the observer P < .0007] than monocularly viewed displays. In all dis (Braunstein & Andersen, 1981; Hochberg & Hochberg, plays, the depth order was unambiguously specified by 1952). Accordingly, any stereoscopic advantage could be relative size and motion cues. Thus, adding consistent assumed to be the result of additional motion-in-depth stereoscopic information should not have affected the information.
Method 70 Subjects. Seven male and 9 female subjects (between 17 and 32 years ofage) participated in the experiment. • STEREO A Design. Three independent variables were examined. (I) Display 60 D MONO type: Displays were either stereoscopic or nonstereoscopic optic flow patterns. Both stereoscopic and nonstereoscopic displays had 50 motion perspective and changing-size cues consistent with self motion in depth. Square size ranged from the single pixel size of I 0.07° to that of 1.2°. (2) Display speed: Each display simulated one 1 40 of three speeds of self-motion (2.4, 4.8, or 7.2m/sec). (3) Display o density: Each display had one oftwo object densities (20 or 30 vis ible objects per eye). 30 Apparatus. Displays were generated on a 486-DX personal com rl,I puter and presented on a superVGA monitor (with a 1,024 hori 20 zontal X 768 vertical pixel resolution). The screen of this monitor subtended a visual angle of 30° H X 24° V when viewed from a chinrest 50 ern away. Since vection perception has been found to be to dominated by the motion of the perceived background (Ohmi & Howard, 1988; Ohmi, Howard, & Landolt, 1987; Telford, Spratley, o & Frost, 1992), inducing displays were presented 20 em behind a 2.4 ·Ul 7.2 large cardboard mask. Kinetic occlusion and stereoscopic (when present) depth cues always indicated that the display was in the Simulated Speed of Self-motion (mlsec) background, while the mask was in the foreground. This mask was placed in front ofthe subject, and two large partitions were placed on either side to restrict his/her vision. Only the monitor could be 180 seen through a square window at the far end ofthis black "viewing • STEREO B 160 booth" (I m wide X 2 m deep X 2 m high). To view all the displays, D MONO the subjects wore anaglyph glasses, the lenses of which were red T and cyan colored camera filters. During monocular trials, the sub 140 ject wore an eye patch under these glasses to ensure that luminance and contrast were constant across viewing conditions. Procedure. Prior to the experiment, the subjects were given the 100 Randot stereovision test to ensure that they could perceive static c stereoscopic depth (the criterion was a stereoacuity of20 sec ofarc .S!.. or better at a distance of40 ern). They were then given practice using .. 80 ":I the nonius lines to alter their convergence. On completing this prac ~ tice, they were told that they would be shown displays of moving 60 objects and that "sometimes the objects may appear to be moving towards you, at other times you may feel as if you are moving to 40 wards the objects. Your task is to press the mouse button down when you feel as if you are moving and hold it down as long as the expe 20 rience continues. Ifyou don't feel as if you are moving, then don't press the mouse button" (instructions modified from Andersen & o Braunstein, 1985). The subjects were also informed that each dis 2.4 4.8 7.2 play had a fixed duration of 3 min and an intertrial interval of Simulated Speed of Self-motion (mlsec) 20 sec. Furthermore, they were instructed that if they experienced double vision during a display, they were to press any key on the Figure 2. The effect of display speed on (A) vection onsets and keyboard, and this would register that they had trouble with that (8) durations for stereoscopic (STEREO) and nonstereoscopic trial. After two practice trials, the experimental displays were pre (MONO) displays in Experiment I. Error bars represent stan sented in a random order. dard errors of the mean. PERCEIVING SELF-MOTION IN DEPTH 1171 perceived depth order. However, the changing disparities tial confound ofaccommodation by seating subjects at a were providing additional, purely binocular information distance of 1.5 m from the screen. about each object's motion in depth. So, it appears that the stereoscopic cues were improving vection by provid Method ing extra motion in depth information. Subjects. Four male and 5 female subjects (between 18 and 47 Overall, faster simulated speeds of self-motion pro years ofage) participated in this experiment. duced faster onsets [F(2,30) = 9.411, P < .0007] and Design. Two independent variables were examined. (I) Viewing longer durations ofvection [F(2,30) = 15.570,p < .0001]. type: Displays were viewed either binocularly or monocularly. Binocular conditions were either stereoscopic or nonstereoscopic. However, it was found that display density did not sig As in Experiment I, both stereoscopic and nonstereoscopic displays nificantly affect vection onset [F(1,15) = 1.436, p > .05] had motion perspective and changing-size cues consistent with self or duration [F(l,15) = 0.379,p > .05]. These speed and motion in depth. Square size ranged from the single pixel size of density findings are consistent with previous studies using 0.12° to that of 1.5°.(2) Display speed: Each display simulated one time-based measures ofself-motion perception (Ander of two speeds of self-motion (2.7 or 4 mlsec). All displays consisted sen & Braunstein, 1985; Telford & Frost, 1993). of 50 objects that moved along the z-axis toward the observer. Apparatus. Displays were generated on an IBM 486-DX per It is interesting to note the two-way interaction be sonal computer and projected onto a white Mylar screen (151 X tween display type and display speed for vection onset 113 ern) by a Sony VideoGraphic projection TV (with 1,024 hori [F(2,30) = 3.970,p < .03]. As the speed ofsimulated self zontal X 768 vertical pixel resolution ). The screen subtended a vi motion increased, the magnitude ofthe stereoscopic ad sual angle of 54° horizontally and 41° vertically when viewed from vantage decreased. This pattern also appears to be present a head-and-chin rest 1.5 m away.A viewing tube was attached to the in duration data; however, in this case, the interaction head-and-chin rest that occluded the rest of the room from sight. Procedure. Prior to the experiment, the subjects were given the failed to reach significance [F(2,30) = 1.676, p > .05]. Randot test and practiced using nonius lines to alter their conver gence. Since the method of magnitude estimation was used, the first EXPERIMENT 2 display ofeach testing session was used to set the modulus for the subjects' strength ratings (Stevens, 1957). This standard stimulus, Previous research has shown that the larger the retinal assumed to be the optimal vection display, had stereoscopic depth area of motion stimulation, the stronger the resulting self cues and simulated the fastest speed ofself-motion (4 mlsec). After motion perception (Brandt et aI., 1973; Held, Dichgans, a period of70 sec had elapsed, the subjects were asked whether they felt "as if they were moving or stationary." If the subjects responded & Bauer, 1975). Similarly, the more moving elements that they were moving, they were told that the strength of their feel there are in the optic flow, the stronger the self-motion ing ofself-motion corresponded to a value of 70 (with zero repre perception (Held et aI., 1975). Thus, it was possible that senting stationary). This number (the modulus) was entered on a the stereoscopic advantage found previously occurred bar chart that appeared directly after the display timed out. Two only because weak or impaired self-motion stimuli were practice trials then followed. Prior to the first of these, the subjects used. The present experiment attempted to determine were told to (I) press the mouse button as soon as they felt as if they were moving and (2) to rate the strength of their feeling of self whether the stereoscopic advantage would persist when motion (with respect to the modulus 70) on the bar chart following observers were presented with larger patterns of optic each trial. The experimental trials were then presented in a random flow with more moving objects (i.e., more compelling order; each had a duration of 90 sec and an intertrial interval of self-motion displays). 30 sec. Following the first testing session, there was a 5-min break In Experiment I, stereoscopic and nonstereoscopic before the second testing session was run. displays involved different viewing conditions: binocu lar and monocular, respectively. It was possible that Results and Discussion stereoscopic displays might have induced stronger vee As expected, the larger, denser inducing stimuli used tion merely because they had binocular viewing. To in this experiment appeared to be more compelling than overcome this potential confound, the vection induced those used previously. Vection was reported on every by binocularly viewed nonstereoscopic flow was also as trial by all 9 subjects. Furthermore, vection onsets were sessed in the present experiment. This condition pre generally much faster than those found in Experiment I sented a single pattern ofmoving objects to both eyes.' (15.66 vs. 24.08 sec). Ifpresent, vergence information about the display's flat Separate repeated measures ANOVAs were performed ness would have been weak, given the large viewing dis on the onset and rating data. Each of the ANOVAs ana tance of 1.5 m. lyzed families of planned contrasts and controlled the Also, in Experiment I, accommodation would have familywise error rate at .05. The means are shown in Fig indicated that both stereoscopic and monocularly viewed ures 3A and 3B. Stereoscopic displays were found to pro displays were two-dimensional, which might have im duce faster vection onsets [F(l,8) = 8.74, p < .05] and paired vection in central vision (Andersen & Braunstein, stronger vection ratings [F(1,8) = 13.49, p < .05] than 1985). The effectiveness of accommodation as a cue to monocularly viewed displays. Similarly, stereoscopic depth rapidly diminishes as the distance of an object displays produced faster onsets [F(l,8) = 5.45,p < .05] from the observer increases (Fisher & Ciuffreda, 1988). and stronger ratings [F(1,8) = 19.00,p < .05] than binoc Accordingly, the present experiment reduced the poten- ularly viewed nonstereoscopic displays. However, there 1172 PALMISANO
70 Consistent with Experiment 1, displays simulating STEREO A faster self-motions were found to produce significantly • faster onsets [F(l,8) = 8.508,p < .05] and significantly 60 ISJ BIN N-S stronger ratings of vection [F(l,8) = 35.246, P < .05]. OJ 0 MONO N-S Due to the other manipulations performed in this ex .. 50 fIl '-' periment (i.e., increasing the area of stimulation and the .. number ofmoving contrasts), it is difficult to determine the ..fIl 1:1 40 effect ofreduced accommodation on self-motion percep 0 It 0 tion. is possible that the weakening ofaccommodation .. based depth cues helped produce the more compelling ;0., 30
(3) Display density: Each display had one of two object densities been provided by the motion perspective. Thus, it ap (20 or 30 visible objects per eye). pears that changing-size cues can also increase vection in central vision by providing additional information Results and Discussion about each object's motion in depth. Self-motion was reported in 223 ofthe 240 trials (20 There are, however, results that conflict with this ar subjects responding to 12stimuli). Separate repeated mea gument. A study by Telford and Frost (1993) found that sures ANOVAs were performed on the onset and dura the addition of changing-size cues to optic flow did not tion data. The means are shown in Figures 4A and 4B. increase subjects' perception of self-motion in depth. There Displays with changing-size cues to motion in depth pro are a number of possible explanations for this discrep duced significantly faster onsets [F(I, 19) = 13.719, P < ancy with the present finding. The first is that each sub .002] and longer durations ofvection [F( 1,19) = 21.667, ject in the Telford and Frost experiment saw only one of p < .0002] than did displays without these cues. In all tri the two depth conditions tested (changing-size or same als, unambiguous depth order information should have size optic flow). This between-subjects design might not have provided a sensitive enough estimate of the effect that changing-size cues have on vection (especially if this effect was small in magnitude). The second is that Telford 70 A and Frost used a smaller range ofpossible sizes (0.15°• • CHANGING·SIZE 0.45° vs. 0.06°-1.21°). Finally, it is possible that their 60 null finding reflects a ceiling effect. The larger number of 1 o SA\IE-SIZE moving objects in their displays (500 vs. the 20 or 30 ob 5t1 jects used here) might have produced optimal vection ... when only motion-perspective information was present. ~ '"c Unlike in previous studies, display speed did not have o an overall effect on vection onset [F(2,38) = 1.443,p > m = > .05] or duration [F(2,38) 3.035, p .05]. However, there 1 was a significant effect ofdisplay density on vection on set. Denser, 30-object displays produced significantly 2t1 faster vection onsets than did the sparser, 20-object dis plays [F(1,19) = 6.3435,p < .02]. This suggests that dis I() play density might have been the critical difference be tween the present findings and those ofTelford and Frost. n EXPERIMENT 4 Number of Objects Experiment 4 explored one of the possible causes of the discrepant findings of Telford and Frost (1993). It 1110 reinvestigated the effect of changing-size cues on vee B tion using denser patterns ofoptic flow than those ofEx 160 r- periment 3 (50 or 100 objects). All displays were viewed iD SAME·SIZE Uti monocularly and had motion-perspective information con sistent with self-motion in depth. 120 Method c 100 The design, equipment, and procedure were identical to those in .S:... Experiment 3, with the exception that two higher object densities ."... 110 T were used (50 or 100 objects vs. 20 or 30 objects). The maximum ~= density of 100 objects (400 less than in Telford and Frost's experi 60 1 ment) was chosen to maintain the frame rate at 60 Hz (the same frame rate used for the displays in Experiments 1-3). ~o In Experiment 4, 10 male and II female subjects (between the ages of 17 and 29 years) participated. 20 Results and Discussion n Self-motion was reported in 243 of the 252 trials (21 20 30 subjects responding to 12stimuli). Separate repeated mea Number of Objects sures ANOVAs were performed on the onset and dura tion data. The means are shown in Figures 5A and 58. Figure 4. The effect ofobject density on the (A) vection onsets and (B) durations for displays with and without changing-size For the denser displays used in this experiment, the ad cues to motion in depth in Experiment 3. Error bars represent dition of changing-size cues to motion in depth did not standard errors of the mean. produce significantly faster vection onsets [F( I,20) = 1174 PALMISANO
3.5273,p> .05]. Nor did these cues lead to significantly those produced by the 50-object displays. Thus, ifa ceil longervection durations [F(1,20) = 0.015,p > .05]. The ing effect was responsible for the decreased effective fact that the changing-size advantage, found in Experi ness ofchanging-size cues, then vection was at maximal ment 3, was eliminated by increasing the display density levels for the 50-object displays. suggests that changing-size cues have a less robust ef One resolution ofthe findings ofExperiments 3 and 4 fect on vection than does stereoscopic information (since may be that sparse optic flow is analyzed in a different stereo still improved the vection induced by very com (but not necessarily less effective) manner than is dense pelling self-motion stimuli in Experiment 2). flow, Since sparse flow has fewer moving elements than It is also ofinterest to note that display density did not dense flow, its motion-perspective information about have a significant effect on vection in this experiment. self-motion is often less reliable. So it would be adap The 100-object displays did not produce significantly tive, in the case of sparse flow, for the visual system to different vection onsets [F(1,20) = 0.278, P > .05] or extract all the available information about self-motion vection durations [F(1,20) = 0.463, p > .05] relative to (to compensate for the less reliable motion perspective). This may account for the changing-size advantage found for sparse flow in Experiment 3. In the case of dense flow, however, motion-perspective information may be 70 sufficient to determine the nature ofself-motion. In such • CHANGING-SIZE A a situation (e.g., Experiment 4), it is possible that only 60 D SAME-SIZE motion-based information about self-motion is ex tracted. A related possibility, suggested by one reviewer, is
....Q> that with higher density same-size displays, dots in cer tain proximity might have been perceived as a configu ='" 40 o ration (i.e., the vertices ofan invisible object). This would also explain the equivalent vection induced by the high 30 density same-size and changing-size displays used in the present experiment (since both would contain similar op 20 1 tical expansion information). One potential criticism of this experiment is that the 10 denser displays used would have led to an increased prob ability of objects overlapping. Without differences in o color and contrast, these objects would appear to merge 50 100 (as opposed to one occluding the other), and this might have impaired (albeit briefly) relative depth perception. Number of Objects There are several counters to this criticism. First, such overlaps were rare even for the densest displays (which ISO B had one fifth of the dots used in the Telford and Frost • CHANGING·SIZE study). Second, this account would predict that the 100 160 D SAME·SIZE object displays should produce a greater number ofover 140 laps, and thus weaker self-motion perception, than should the 50-object displays. This prediction was not, however, 120 supported by the data.
= 100 GENERAL DISCUSSION .g.... T E SO 1 It is possible that, in the case of heading perception, ~= 60 stereoscopic depth cues are useful only for disambiguat ing impaired self-motion information (i.e., optic flow pat 40 terns complicated by head or eye movements). In such situations, stereoscopic information could be used to pro 20 vide the depth order of objects in these flow patterns. Vanden Berg and Brenner (1994b) argue that depth order o is important in heading perception because the most dis 50 100 tant points in the flow provide the most reliable estimates Number of Objects of head and eye rotations. Using such estimates, flow components due to head or eye movements can be sub Figure 5. The effect of object density on the (A) vection onsets and (B) durations for displays with and without changing-size tracted, leaving optic flow based solely on self-motion. cues to motion in depth in Experiment 4. Error bars represent The focus ofexpansion of this untainted flow can then be standard errors ofthe mean. used to determine the direction ofself-motion. PERCEIVING SELF-MOTION IN DEPTH 1175
In the case ofself-motion perception, however, stereo ities of an object in each eye rather than its changing scopic information appears to play an additional role. In binocular disparity (Regan et al., 1979a). (2) Motion per Experiment 1, where depth order was already unambig spective is encoded as the relative velocities ofdifferent uously provided by relative size and motion, vection was objects in the environment. Ifboth cues are encoded on the still improved by the addition ofstereoscopic motion cues. basis ofrelative velocity, it seems likely that they might be It appears that this "stereoscopic advantage" was due to processed in a similar manner. Thus, ifmotion-perspective the extra, purely binocular information about motion in information is preferred in optimal stimulus conditions, depth. it might be more difficult to disregard stereoscopic motion Changing-size cues to motion in depth were also found cues than to disregard changing-size cues. to increase perceptions of self-motion in depth. In Ex The alternative explanation is that the display charac periment 3, where depth order should have been unam teristics might have favored motion-in-depth perception biguously provided by motion perspective, the addition based on stereoscopic motion. Regan and his colleagues of changing-size cues further improved vection. These have shown that changing disparity produces more ef findings sit well with the idea that changing-size and fective motion-in-depth perception than does changing stereoscopic motion channels converge at the same mo size for fast-moving objects observed for a reasonable tion-in-depth stage of the visual system (Regan & Bev period oftime (e.g., 1 sec; Regan & Beverley, 1979). The erley, 1979; Regan et al., 1979a, 1979b). reverse was found for briefly glimpsed, slow-moving ob It appears then that accounts based solely on monoc jects. Thus, the fast display speeds and long observation ular motion perspective are incomplete. Stereoscopic times used in the present experiments might have led to and changing-size cues provide additional motion-in the more compelling stereoscopic perceptions of self depth information that is used in perceiving self-motion. motion in depth. This motion in depth information might improve vection Finally, the present research has shown that the addi directly by providing more accurate estimates ofheading tion ofconsistent stereoscopic motion and changing-size and egospeed. Alternatively, the improvement might be cues can improve vection in central vision. These results achieved indirectly by making self-motion displays ap further highlight the differences between central and pe pear more three-dimensional. For example, stereoscopic ripheral self-motion perception. Previous research sug and changing-size cues might produce an apparent ex gests that central vision is specialized for the perception pansion of the depth axis. Since the subjects were trav ofself-motion in depth, whereas peripheral vision has no eling through virtual space, the larger the perceived ex such specialization (Andersen & Braunstein, 1985; Stoff tent of that space, the greater the perceived change in regen, 1985; Telford & Frost, 1993). Since central vision location per unit oftime and, thus, the stronger the self is stereoscopic, it seems ideally suited for this specialized motion perception. role. It is possible that stereopsis may have even played a Of these two explanations, the "direct" account has role in the evolution ofthe central-peripheral differences the least empirical support. In many conditions, heading in self-motion perception. estimates based on motion alone are very precise, leaving little room for improvement by stereoscopic or changing REFERENCES size cues (e.g., Warren & Hannon, 1988). Similarly, Mo nen and Brenner (1994) have found that subjects are ANDERSEN, G. J., & BRAUNSTEIN, M. L. (1985). Induced self-motion in central vision. Journal ofExperimental Psychology: Human Percep actually worse at detecting simulated changes in ego tion & Performance, 11,122-132. velocity when stereoscopic information is available. How BENSON, A. J. (1990). Sensory functions and limitations ofthe vestibu ever, the latter is not strong evidence against the direct lar system. In R. Warren & A. H. Wertheim (Eds.), The perception account, since a fair test should produce, at worst, equal and control ofself-motion (pp. 145-170). Hillsdale, NJ: Erlbaum. performance in stereoscopic and control conditions. BRANDT, T., DICHGANS, J., & KOENIG, E. (1973). Differential effects of central versus peripheral vision on egocentric and exocentric motion Although the effects of stereoscopic and changing perception. Experimental Brain Research, 16,476-491. size-based information on vection were similar, the BRAUNSTEIN, M. L., & ANDERSEN, G. J. (1981). Velocity gradients and stereoscopic advantage appeared to be more robust than relative depth perception. Perception & Psychophysics, 29,145-155. that produced by changing size. Experiment 2 showed FISHER, S. K., & CIUFFREDA, K. J. (1988). Accommodation and appar ent distance. Perception, 17,609-621. that the vection induced by compelling self-motion dis GIBSON, J. J. (1950). The perception of the visual world. Boston: plays was still improved by the addition of stereoscopic Houghton Mifflin. information. However, changing-size cues were not al GIBSON, J. J. (1966). The senses considered as perceptual systems. ways effective in improving vection. In Experiment 4, Boston: Houghton Mifflin. which used dense displays, the addition ofchanging-size GIBSON, J. J., GLUM, P., & ROSENBLATT, F. (1955). Parallax and per spective during aircraft landings. American Journal ofPsychology, cues was found to have no effect on vection. 68, 372-385. What might underlie the differences between these GOGEL, W. C, & STURM, R. D. (1972). A comparison ofaccommoda two advantages? One possibility is that the link between tive and fusional convergence as cues to distance. Perception & Psy stereoscopic motion and motion perspective is stronger chophysics, 11, 166-168. GORDON, D. A. (1965). Static and dynamic visual fields in human space than the link between changing size and motion perspec perception. Journal ofthe Optical Society ofAmerica, 55, 1296 tive. This explanation rests on two assumptions: (1) Stereo 1303. scopic motion is encoded as the different relative veloc- HEBBARD, F. W. (1962). Comparison of subjective and objective mea- 1176 PALMISANO
surements of fixation disparity. Journal ofthe Optical Society of subsystems for position in depth and for motion in depth. Proceed America, 52, 706-712. ings ofthe Royal Society ofLondon B, 204, 485-501. HEEGER, D. 1.,& JEPSON, A. (1990). Visual perception ofthree-dimensional REGAN, D., BEVERLEY, K. I., & CYNADER, M. (1979b). The visual per motion. Neural Computation, 2, 129-137. ception ofmotion in depth. Scientific American, 241,136-151. HELD,R., DICHGANS, J., & BAUER, J. (1975). Characteristics ofmoving RICHARDS, W., & MILLER, 1. E, JR. (1969). Convergence as a cue to visual scenes influencing spatial orientation. Vision Research, 15, depth. Perception & Psychophysics,S, 317-320. 357-365. STEVENS, S. S. (1957). On the psychophysical law. Psychological Re HOCHBERG, C. B., & HOCHBERG, J. E. (1952). Familiar size and the per view,64,153-181. ception ofrelative depth. Journal ofPsychology, 34,107-114. STOFFREGEN, T. A. (1985). Flow structure versus retinal location in the HOWARD, I. P. (1982). Human visual orientation. Chichester, U.K.: optical control of stance. Journal ofExperimental Psychology: Wiley. Human Perception & Performance, 11, 554-565. KOENDERINK, J. J. (1990). Some theoretical aspects of optic flow. In TELFORD, L., & FROST, B. J. (1993). Factors affecting the onset and mag R. Warren & A. H. Wertheim (Eds.), The perception and control of' nitude of linear vection. Perception & Psychophysics, 53, 682-692. self-motion (pp. 53-68). Hillsdale, NJ: Erlbaum. TELFORD, L., SPRATLEY, J., & FROST, B. 1. (1992). Linear vection in the KOENDERINK,1. J., & VAN DOORN, A. 1. (1981). Exterospecific compo central visual field facilitated by kinetic depth cues. Perception, 21, nent of the motion parallax field. Journal ofthe Optical Society of 337-349. America, 71, 953-957. VAN DEN BERG,A. V.(1992). Robustness ofperception ofheading from KOENDERINK, J. 1., & VAN DooRN, A. 1. (1987). Facts on optic flow. Bio optic flow. Vision Research, 32, 1285-1296. logical Cybernetics, 56, 247-254. VAN DEN BERG, A. V., & BRENNER, E. (I994a). Humans combine the LEE, D. N. (1980). The optic flow field: The foundation of vision. optic flow with static depth cues for robust perception ofheading. Vi Philosophical Transactions ofthe Royal Society ofLondon: Series B, sion Research, 34, 2153-2167. 290, 169-179. VAN DEN BERG, A. v., & BRENNER, E. (l994b). Why two eyes are bet LEE, D. N., & ARONSON, E. (1974). Visual proprioceptive control of ter than one for judgements ofheading. Nature, 371, 700-702. standing in human infants. Perception & Psychophysics, 15, 529-532. WARREN, W. H., & HANNON, D. J. (1988). Direction of self-motion is LEE, D. N., & LISHMAN, J. R. (1975). Visual proprioceptive control of perceived from optical flow. Nature, 336,162-163. stance. Journal ofHuman Movement Studies, 1,87-95. WARREN, W. H., MORRIS, M. w., & KALISH, M. (1988). Perception of LISHMAN, J. R., & LEE, D. N. (1973). The autonomy of visual kinaes translational heading from optical flow. Journal ofExperimental Psy thesis. Perception, 2, 287-294. chology: Human Perception & Performance, 14,646-660. LONGUET-HIGGINS, H. c., & PRAZDNY, K. (1980). The interpretation ofa moving retinal image. Proceedings ofthe Royal Society ofLon NOTES don B, 208, 385-397. MACH, E. (1875). Grundlinien der Lehre von den Bewegungsempfin I. This distinction was based on the fact that most computer-generated dungen [Basic principles for the study ofmotion perception]. Leip vection stimuli consist ofa moving pattern ofdots, with each dot's size zig: Engelmann. remaining constant regardless ofits simulated location in depth. Motion MITCHELL, A. M., & ELLERBROCK, V.J. (1955). Fixational disparity and perspective information, as defined above, must therefore be responsi the maintenance offusion in the horizontal meridian. American Jour ble for the illusion ofself-motion in these situations. Whether changing nal ofOptometry, 32, 520-534. size information can produce a similar effect is yet to be ascertained. MONEN, 1., & BRENNER, E. (1994). Detecting changes in one's own ve 2. Since nothing else was visible during the nonius displays, this dis locity from the optic flow. Perception, 23, 681-690. parity was relative to the screen border. NAKAYAMA, K., & LOOMIS, J. M. (1974). Optical velocity patterns, 3. It has been argued correctly that the display resolution used might velocity-sensitive neurons, and space perception: A hypothesis. Per have created a limited amount ofmotion noise. Heading perception ap ception, 3, 63-80. pears quite robust in the presence ofmoderate amounts ofmotion noise OHMI,M., & HOWARD, I. P. (1988). Effect ofstationary objects on illu and breaks down only when substantial individual differences are intro sory forward self-motion induced by a looming display. Perception, duced (van den Berg, 1992). It seems unlikely then that stereoscopic cues 17,5-12. improved vection merely by overcoming this small amount of noise. OHMI, M., HOWARD, I. P., & LANDOLT, J. P. (1987). Circularvection as 4. This should not be confused with a synoptic display, where identi a function of foreground-background relationships. Perception, 16, cal patterns of optic flow are presented to each eye (which potentially 17-22. provides stereoscopic information that all the objects in the visual field REGAN, D., & BEVERLEY, K. I. (1979). Binocular and monocular stimuli are infinitely distant). for motion in depth: Changing-disparity and changing-size feed into the same motion-in-depth stage. Vision Research, 19, 1331-1342. (Manuscript received July 5, 1995; REGAN, D., BEVERLEY, K. I., & CYNADER, M. (l979a). Stereoscopic revision accepted for publication January 22, 1996.)