Perceptual & Motor Skills: Perception 2014, 118, 3, 850-862. © Perceptual & Motor Skills 2014
VISUAL ATTENTION TO REFERENCE FRAMES AFFECTS PERCEPTIONS OF SHAPE FROM SHADING1 HIROYUKI SASAKI Department of Child Education Niigata Chuoh Junior College, Japan Summary.—Perception of shape from shading is processed locally in a bottomup manner, but is also influenced by global or contextual factors. This study examined the influence of attention to the reference frame on the perception of shape from shading. In a visual search task, participants were asked to identify the location of a target relative to the reference frame. The results showed that shaded targets were more quickly and accurately detected when the shading gradient was parallel, rather than orthogonal, to the orientation of the environmental reference frame. This was further supported by a second experiment with a masking paradigm. Consequently, the perceptual process of shape from shading may be a flexible mechanism in which the representation of gradient orientation is calibrated by topdown processing in visual attention.
The appearance of a visual scene is fundamentally ambiguous because information about the three-dimensional world is projected onto a two-dimensional retinal surface. To resolve ambiguity, the visual system utilizes prior constraints on the way visual scenes are structured. One of the constraints comes from the assumption that scenes are lit from above our heads (Ramachandran, 1988a, 1988b; Mamassian & Goutcher, 2001). For example, even in the absence of a clear source of illumination, we typically perceive a light-above-dark disk to be a dome-shaped mound, whereas a dark-above-light disk is perceived to be a bowl-shaped depression. This convex or concave percept, however, becomes ambiguous and unstable when a horizontal light-to-dark gradient is provided. These percepts suggest that the visual system incorporates an overhead-lighting constraint in its interpretation of shape-from-shading (Ramachandran, 1988a, 1988b; Mamassian & Goutcher, 2001). Some visual search studies have explored whether shape from shading is interpreted during preattentive or higher cognitive processing. Kleffner and Ramachandran (1992; see also Aks & Enns, 1992; Symons, Cuddy, & Humphrey, 2000) found that vertically shaded targets were immediately detected regardless of the number of distractors; however, horizontally shaded targets were detected less efficiently. Because it is possible to preattentively search for a vertically shaded target, the authors argued that Address correspondence to Hiroyuki Sasaki, 16–18, Gakko-cho, Kamo, Niigata 959–1322, Japan or e-mail ([email protected]). 1
DOI 10.2466/24.22.PMS.118k26w9
13-PMS_Sasaki_130168.indd 850
ISSN 0031-5125
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
851
the overhead-lighting constraint is implemented in a hard-wired, early visual process. In a dual task paradigm (Braun, 1993), performance in a shape from shading search task was not compromised by a concurrent letter-discrimination task. Braun therefore argued that perception of shape from shading does not significantly depend on visual attention. A different line of evidence for the overhead-lighting constraint was provided by studies on whether the visual system uses retinal or gravitational coordinates. These experiments examined changes in the perception of shape from shading caused by rotation of the head; this rotation caused a discrepancy between the retinal and gravitational coordinates when the object was lit from above. Kleffner and Ramachandran (1992) and Wenderoth and Hickey (1993) found that the segregation of texture patterns made from shaded disks was impaired when the direction of the shading gradient was parallel to gravity and orthogonal to the vertical axis of the head (for instance, when participants lay on their sides; see also Howard, Bergstrom, & Ohmi, 1990). Thus, the perception of shape from shading was found to be dependent on the retinal coordinate system. This also supports the claim that the overhead-lighting constraint is invoked during an early stage of visual processing that preserves retinotopy. Although these findings suggest that shape from shading is processed locally in a stimulus-driven or bottom-up manner, some studies have shown a role for global or contextual factors in the perception of shape from shading. The hollow-face illusion is a striking example of this. The expectation that certain objects will be convex outweighs the assumption that light is coming from above (Gregory, 1973); a picture of a hollow face lit from above looks convex and its light source is perceived as coming from below (see also Liu & Todd, 2004 for a similar convex bias taking precedence over the overhead-lighting constraint). Similarly, the shading cues attached to a familiar object, which imply the position of the light source, were found to overshadow the perception of shape from shading of a stimulus presented adjacent to the object (Berbaum, Bever, & Chung, 1984; Ramachandran, 1988a; Sun & Perona, 1996). For example, the concave/convex ambiguity of shape from shading depends on the light source information from a shaded face rather than the assumption of overhead lighting. Jenkin, Jenkin, Dyde, and Harris (2004) measured the perceived convexity of a shaded disk in terms of environmental, gravitational, and retinal frames of reference by independently rotating the visual context and the head. As a result, the perception of shape from shading was found to depend not only on the retinal coordinates but also on gravitational and environmental coordinates. This finding indicates that the visual system
13-PMS_Sasaki_130168.indd 851
24/05/14 9:31 AM
852
H. SASAKI
determines which direction is up from the sum of the vectors represented in the retinal and the other frames. In the present study, another type of global interaction effect was examined, namely, the influence of attention to the environmental reference frame on the perception of shape from shading. It was assumed that focusing attention on a task-relevant stimulus indicating the environment's upright orientation could modulate the perception of shape from shading. To manipulate the observer's reference frame, an experimental paradigm was employed in which the participants responded to the target location relative to the orientation of the rotated reference frame against which the shaded disks were mapped (see Fig. 1 and its caption). Such manipulation is consistent with a previous suggestion that viewer rotations are easier to imagine than object rotations (Wraga, Creem, & Proffitt, 1999). The vertical axis of the reference frame was tilted 45° clockwise (CW) or 45° counterclockwise (CCW) from the retinal and gravitational vertical axes. Participants detected a shaded target among distractors located around the reference cue, which was displayed at the center of the monitor. The direction of the light-to-dark gradient of the shaded stimuli was also tilted 45° CW or 45° CCW; however, this tilt varied independently from the orientation of the reference frame. A shaded disk with an oblique luminance gradient allowed the perception of three-dimensional shape, based on the implicit assumption that light is coming from overhead (Symons, et al., 2000). It was expected that the perception of shape from shading depended on the direction of the light source derived from the environmental reference frame. Generally, it is known that a dark-above-light target among light-above-dark distractors is more discriminable than vice versa (Kleffner & Ramachandran, 1992; Sun & Perona, 1997). Thus, in this study the search asymmetry should occur more clearly under a concordant condition than under a discordant condition. Hypothesis 1. Shaded targets will be detected more quickly and/or accurately when the direction of the shading gradient is parallel to the orientation of the vertical axis of the reference frame; for instance, when a CW-tilted target was located in a CW-tilted reference frame (the concordant condition) than when the target and reference frame have different orientations, such as a CW-tilted target within a CCW-tilted reference frame (the discordant condition). Hypothesis 2. The search asymmetry between light-above-dark and dark-above-light targets will be stronger under the concordant condition than under the discordant condition.
13-PMS_Sasaki_130168.indd 852
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
853
Experiment 1 METHOD Participants Twelve university students participated in the experiment. Their ages ranged from 23 to 33 years. All participants had normal or corrected-tonormal vision and were naïve regarding the purpose of the experiment. Participation was voluntary and anonymous. After the data were analyzed and summarized, the findings were distributed to the participants. There was no requirement for ethical review, but the experiment followed the guidelines in the Declaration of Helsinki. Apparatus and Stimuli Stimuli were generated by an IBM PC-compatible computer and presented on a CRT monitor with a frame rate of 60 Hz and a resolution of 800 × 600 pixels. The pixel size was approximately 2.4 min. of arc at a viewing distance of 57 cm. Targets and distracters were constructed from 256 luminance levels ranging from 5.3 to 25.2 cd/m2 and were presented on a background of 13.0 cd/m2. For the trials, each participant sat upright in a darkened room with his or her head fixed on a chin and head rest. A biological-motion display (a point-light walker) was used as an environmental reference cue because it was expected that the walker pattern would provide reliable information about upright orientation to the viewer. The biological-motion display was generated according to Cutting's algorithm (Cutting, 1978) and consisted of 11 black dots positioned at the head and main joints of the profile of a walker. The walking figure was centered in the middle of the monitor and subtended 5.2° in height and 2.4° in width at the most extended point of the gait cycle. A full gait cycle was completed in 60 frames, with a frame duration of 16.6 msec. The direction of walking (left or right) varied randomly from trial to trial. The walker was either upright or tilted at either 45° CW or 45° CCW. Although biological-motion perception is known to depend on the orientation of the display with respect to the retinal coordinates (Troje, 2003), a display oriented 45° CW or 45° CCW is easily perceived as a walker (Pavlova & Sokolov, 2000). Four shaded disks were presented on an imaginary circle with a radius of 4° of visual angle around the reference cue. Each disk subtended 1.5°. One of the disks, which served as the target, differed from the others in the luminance polarity of its light-to-dark gradient. The target and distractor items were located at the 45° CW, 135° CW, 45° CCW, and 135° CCW positions from the upright within the reference frame as defined by the walker. The position of each item randomly varied within a radius of 0.5° to prevent judgments from being made based on item collinearity. Thus, when the reference frame was tilted 45° CW, for instance, the 135° CCW target
13-PMS_Sasaki_130168.indd 853
24/05/14 9:31 AM
854
H. SASAKI
FIG. 1. Search display, including the dark-above-light target, around the reference cue. The left panel shows the discordant condition (the shading gradient tilted CCW and the walking figure tilted CW), whereas the right panel shows the concordant condition (both the shading gradient and the walking figure tilted CCW). The middle panel shows the neutral condition. To choose the 135° CCW target position relative to the orientation of the walking figure tilted CW (left panel), the F key was pressed with the left index finger. The 45° CCW target position relative to the walking figure tilted CCW (right panel) was selected by pressing the E key with the left middle finger.
was located at the same retinal 9 o'clock position as the 45° CCW target when the reference frame was tilted 45° CCW (Fig. 1). The direction of the light-to-dark gradient of the shaded disks was tilted either 45° CW or 45° CCW, and the type of targets was either light-above-dark disk or darkabove-light disk. Design and Procedure The within-subjects design included three independent variables: orientation of the rotation axis of the reference frame (upright, 45° CW or 45° CCW), direction of the shading gradient (45° CW or 45° CCW), and target type (light-above-dark or dark-above-light). An experimental session consisted of nine blocks (three blocks for each of the three orientations of the reference frame). Each block contained a total of 48 trials (12 trials for each of the two possible directions of the shading gradient × the two types of target). Thus, each session contained a total of 432 trials for each participant. The order of the blocks (i.e., the presentation order of the three orientations of the reference frame) was randomized within each session, and the order of the trials (i.e., the presentation order of the two directions of the shading gradient and the two types of target) was randomized within each block. The participants engaged in two experimental sessions. Therefore, each data point was based on 72 trials per stimulus condition (12 trials × 3 blocks × 2 sessions). A practice session comprising at least three blocks was carried out before each session. At the beginning of each block, the biological-motion display was presented for 1 sec. to inform participants about the orientation of the en-
13-PMS_Sasaki_130168.indd 854
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
855
vironmental frame of reference. The reference frame orientation was fixed for a block consisting of 48 trials. Each trial began with the appearance of the reference cue (the biological-motion display) for 1.5 sec. before the search display was introduced. The search display and the reference cue remained visible until the participant responded. An acoustic signal was provided following an incorrect response. The next trial followed 1.5 sec. after the response. The search display of four shaded disks was presented around the reference cue. The participants were told that the target, which could be distinguished from the distractors by its shape (convex vs concave), would always appear in the search display. They were then instructed to identify the target position relative to the reference cue, which was either upright or tilted either CW or CCW from the retinal and gravitational vertical axes, as quickly and accurately as possible by pressing the corresponding key on a computer keyboard. Responses for the 45° CCW, 135° CCW, 45° CW, and 135° CW target positions from the upright within the reference frame were assigned to the E, F, I, and J keys on the keyboard, which were pressed by the left middle, left index, right middle, and right index fingers, respectively. The keyboard was aligned parallel to the frontal plane of the body and the monitor. The participants were free to move their eyes while searching for the targets. Eye movements were not monitored. Reaction time (RT) was measured as the time between the onset of the search display and the button-press response. RTs in error trials were omitted from the data analysis. RTs under 150 msec. or over 3,000 msec. were also excluded. RESULTS Figure 2 shows the mean RTs and mean error rates. Repeated-measures analyses of variance (ANOVAs) were conducted on the RT and error rate data with reference frame orientation (CW, upright, or CCW), lightto-dark gradient orientation (CW or CCW), and target type (light-abovedark or dark-above-light) as factors. Reaction Times The main effect of the light-to-dark gradient orientation was significant (F1, 11 = 20.17, p < .001, ηp2 = 0.65), indicating that the participants responded faster when the gradient was tilted CCW than when it was tilted CW. The main effect of target type was also statistically significant (F1,11 = 8.35, p < .05, ηp2 = 0.43). This replicates the previously reported search asymmetry between light-above-dark and dark-above-light targets; darkabove-light targets are easier to detect than light-above-dark targets (Kleffner & Ramachandran, 1992). On the other hand, reference frame orientation only marginally affected the reaction time (F2, 22 = 3.08, p = .07, η2p = 0.12).
13-PMS_Sasaki_130168.indd 855
24/05/14 9:31 AM
856
H. SASAKI 1500
20
15
1300
Error Rate (%)
RT (msec.)
1400
1200 1100 1000
CW - light-above-dark CW - dark-above-light CCW - light-above-dark CCW - dark-above-light
10
5
900 800
0 CCW
Upright
CW
Reference Frame Orientation
CCW
Upright
CW
Reference Frame Orientation
FIG. 2. Experiment 1: mean RTs (left panel) and mean error rates (right panel) for the light-to-dark gradient orientation and the target type as a function of the orientation of the reference frame. Open symbols and filled symbols denote the conditions in which the gradient was tilted CW and CCW, respectively. Circles denote conditions with a light-above-dark target and triangles denote conditions with a dark-above-light target. Error bars denote standard errors of the mean.
The interaction between the light-to-dark gradient orientation and the target type was significant (F2, 22 = 9.35, p < .05, ηp2 = 0.46). More importantly, the interaction between the reference frame and the light-to-dark gradient orientations was significant (F2, 22 = 3.73, p < .05, ηp2 = 0.15); there were significant differences in the gradient direction when the reference frame orientation was upright (F1, 33 = 18.99, p < .001 ηp2 = 0.37), and tilted CCW (F1, 33 = 17.46, p < .001, ηp2 = 0.35); however, there was no significant difference when the reference frame orientation was tilted CW (F1, 33 = 2.43, p > .1, ηp2 = 0.07). Other interactions were not significant (Fs2, 22 < 0.6, ps > .5). Error Rates The main effect of the gradient orientation was only marginally significant (F1, 11 = 4.24, p = .06, ηp2 = 0.28), but it seems to indicate that the participants responded more accurately when the gradient was tilted CCW than when it was tilted CW. The main effects of reference frame orientation and target type were not significant (F2, 22 = 2.30, p > .1, ηp2 = 0.09 and F1, 11 = 1.02, p > .1, ηp2 = 0.09, respectively). The interaction between the reference frame and the gradient orientations was significant (F2, 22 = 5.35, p < .05, ηp2 = 0.20); there were significant differences in the gradient orientation when the reference frame orientation was upright (F1, 33 = 7.15, ηp2 = 0.18, p < .05) or tilted CCW (F1, 33 = 6.47, p < .05, ηp2 = 0.16). On the other hand, there was no significant difference when the reference frame orientation was tilted CW (F1, 33 = 0.002, p > .5, ηp2 < 0.01). The other interactions were not significant (Fs2, 22 < 0.4, ps > .5).
13-PMS_Sasaki_130168.indd 856
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
857
DISCUSSION The results showed that performance in the concordant condition was better than performance in the discordant condition when the reference frame was tilted CCW, although there was no difference in performance between the two conditions when the reference frame was oriented CW. In considering this result, performance in the upright frame of reference should be taken into account. The upright reference frame functioned as a neutral cue for the detection of shaded targets with shading gradients tilted either CW or CCW. Thus, less cognitive effort should be needed to align the mental coordinate system with the reference frame when responding to a target presented in the upright reference frame. Still, the shaded target was detected faster when the shading gradient was tilted CCW than when it was tilted CW. This bias towards the CCW direction (i.e., the upper-left direction) is consistent with the findings reported by Sun and Perona (1998) and by Mamassian and Goutcher (2001), who maintained that the visual system prefers above-left lighting to direct overhead lighting. The preference for above-left lighting is probably independent of reference frame manipulation. In the case of the CCW-tilted frame of reference, the above-left lighting situation (specifically, the shading gradient tilted CCW) corresponded to the concordant condition. Consequently, the significant effect of the CCWtilted frame of reference can plausibly be interpreted as the combined result of two biases: the overhead-lightning constraint within the CCW-tilted reference frame and the preference for above-left lightning. The results for error rates were comparable to those of RTs. Error rates increased in conditions in which RTs were slower. Thus, variations in RTs did not reflect a speed-accuracy trade-off. These results seem to offer some support for Hypothesis 1, that the perception of shaded targets would be influenced by whether the direction of the shading gradient corresponded to the reference frame orientation. However, there remains a possibility that the reference cue (the biological-motion display) itself may have been sufficient to trigger the differences (cf. the context effect; Sun & Perona, 1996; Jenkin, et al., 2004) without the participants focusing attention on the reference cue. Therefore, to confirm the need for attention to the reference cue, a supplementary control experiment was conducted in which the participants were not required to attend to the reference cue tilted either CW or CCW, and the search display of four shaded disks (tilted either CW or CCW) was placed with respect to the upright retinal coordinate. The results of this control experiment showed that the performance difference between the concordant and discordant conditions disappeared when participants were not required to obtain the reference frame from the biological-motion display. Although the walker pattern explicitly provided upright information as
13-PMS_Sasaki_130168.indd 857
24/05/14 9:31 AM
858
H. SASAKI
a reference frame, it had no effect on the performance of the detection of shape from shading targets. Therefore, this implies that not only the presentation of the reference frame but also attention to it was necessary for the results obtained in Study 1. On the other hand, a search asymmetry between the light-above-dark and dark-above-light targets was found in this experiment. This occurred under the discordant as well as the concordant condition, contradicting Hypothesis 2, that the search asymmetry would occur more clearly under the concordant condition than under the discordant condition. Experiment 2 In Experiment 1, the search display was presented until the participants responded (mean RTs from the onset of the search display were longer than 1 sec.). This leaves the possibility that the results may have reflected higher cognitive processing rather than perceptual processing. Namely, attention to the reference frame may not have directly affected the shape perception from shading, and the poor performance in the discordant condition may have been due to a cognitive conflict caused by the difference in the orientations of the reference frame and the light source. Thus, Experiment 2 examined whether attention to the reference frame showing the environmental upright orientation modulates the perception of the shaded disks or generates a cognitive conflict with the upright direction derived from the light-to-dark gradient orientation (i.e., overheadlighting assumption). For this purpose, a masking paradigm was used (cf. Sun & Perona, 1996, 1997). Sun and Perona (1996) suggested that experiments involving shaded stimuli followed by backward masking confirm that shape from shading can be processed in early stages of visual processing. Therefore, if the findings of Experiment 1 could be replicated with the masking paradigm, it would appear that attention to the reference frame could affect the early processing of shape from shading. METHOD Participants Ten university students participated in the experiment. Their ages ranged from 23 to 33 years. All participants had normal or corrected-tonormal vision and were naïve regarding the purpose of the experiments. Nine of these students had participated in Experiment 1. Apparatus and Stimuli The apparatus was the same as that used in Experiment 1. The reference frame indicated by the biological motion display was tilted either 45° CW or 45° CCW. The light-to-dark gradient of the shaded disks was tilted either 45° CW or 45° CCW. The targets were either the light-above-
13-PMS_Sasaki_130168.indd 858
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
859
dark or dark-above-light disks. Random-dot disks were used as the mask stimuli. Each disk replacing each shaded disk comprised 50% density bright/dark dots, and subtended 2°. The dot size was 12 min. of arc. Design and Procedure The design and procedure were similar to Experiment 1 with a few exceptions. An experimental session consisted of six blocks (three blocks for each of the two orientations of the reference frame). A practice session comprising at least two blocks was carried out before each of the two sessions. As in Experiment 1, the search display was presented 1.5 sec. after the onset of the reference cue. The search items were then replaced by mask stimuli, which were presented for 1 sec. Before the main experiment, asynchrony between the onset of the disks and the onset of the masks, which ranged from 33 to 333 msec., was adjusted for each participant using an abbreviated psychological staircase method. This procedure was designed to prevent floor and ceiling effects on performance. After the termination of the reference cue and the mask stimuli, the participants responded to the target location relative to the orientation of the reference frame by pressing the corresponding key on a computer keyboard. The target-response assignment of the keys was identical to that used in Experiment 1. Unlike in Experiment 1, participants were not required to respond quickly. RESULTS Figure 3 shows the percentage of error responses. Repeated-measures ANOVAs were conducted with reference frame orientation (CW or CCW), light-to-dark gradient orientation (CW or CCW), and target type (lightabove-dark or dark-above-light) as factors. The main effect of light-to-dark gradient direction was statistically significant (F1, 9 = 16.58, p < .005, ηp2 = 0.65), indicating that the participants located the targets more accurately when the gradient was tilted CCW than when it was tilted CW. The main effect of target type was also significant (F1, 9 = 19.70, p < .005, ηp2 = 0.69), indicating that the participants located the light-above-dark targets more accurately than the dark-abovelight targets. On the other hand, the main effect of the reference frame orientation was not significant (F1, 9 = 0.25, p > .5, ηp2 = 0.03). The interaction between the reference frame and the gradient orientations was significant (F1, 9 = 20.61, p < .005, η2p= 0.70); there was a significant difference in the gradient direction when the reference frame orientation was tilted CCW (F1, 18 = 33.68, p < .001, η2p = 0.65). There was, however, no significant difference when it was tilted CW (F1, 18 = 1.52, p > .1, ηp2 = 0.08). The other interactions were not significant (Fs1, 9 < 0.8, ps > .1).
13-PMS_Sasaki_130168.indd 859
24/05/14 9:31 AM
860
H. SASAKI
60
CW - light-above-dark CW - dark-above-light CCW - light-above-dark CCW - dark-above-light
Error Rate (%)
50 40 30 20 10 CCW CW Reference Frame Orientation FIG. 3. Experiment 2: mean error rates for the light-to-dark gradient orientations and target types as a function of the orientation of the reference frame.
DISCUSSION In Experiment 2, perceptual sensitivity to shape from shading was examined with a backward masking paradigm. This paradigm did not require quick reactions from the participants, and demanded only accurate judgments. Under such circumstances, it is unlikely that the participants' failure to locate the target reflected a cognitive conflict caused by the difference in the orientations of the reference frame and the light source. Nonetheless, performance in target discrimination differed between the concordant and discordant conditions. Thus, the mismatch between the reference frame and the gradient orientations should not have resulted in interference in a later cognitive stage. The response accuracy data in Experiment 2 showed a pattern similar to that of the reaction times and response accuracy in Experiment 1. These results suggest that the perception of shape from shading was modulated by attention to the reference frame. The effect of focusing attention on the upright orientation in the environment seems to have occurred during a perceptual process rather than in a higher cognitive process. GENERAL DISCUSSION In the present study, the authors investigated the effect of attention to the orientation of the rotated reference frame on the perception of shape from shading. Evidence from visual search and masking paradigms basically supported Hypothesis 1, that shaded targets are detected faster and /or more accurately when the light source is located in a position defined as overhead in terms of the frame of reference. This result was not due to
13-PMS_Sasaki_130168.indd 860
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
861
cognitive conflict at a higher stage of processing or a contextual effect. In other words, the early visual processing of shape from shading was affected by attention to the reference frame. The effect of attention to the reference frame has been measured by a voluntary attention paradigm similar to that used in the current study (Logan, 1996; see also Hommel & Lippa, 1995). In Logan's (1996) study, the response to a target color was determined by the location of the target relative to a visual context (a rotated face) that served as a frame of reference. Logan argued that directing attention from reference cues to targets implies an understanding of the spatial relationship between the two and that the reference frame orients attention to space, just as a spotlight does in traditional theories of attention. Correspondingly, reference to the environmental frame in the present study affected the perception of visual features in a manner similar to feature-based attention (e.g., Saenz, Buracas, & Boynton, 2003). The mental coordinate system aligned with the environmental reference frame could work to modulate shape perception from shading. These facts thus suggest that the effects of attention to the reference frame influence both the discernment of spatial relationships and the modulation of perceptual shape, in contrast to the well-known distinction between spatial orientation and feature analysis (Goodale & Milner, 1992; LaBerge, 2002). In conclusion, the visual process that implements the overhead-lighting constraint seems to be a flexible mechanism that can be modulated by attentional control of the mental coordinate system. Although the lightsource information is represented implicitly by the early visual system, enabling the detection of shaded targets without focal attention (Braun, 1993), the representation of gradient-orientation tuning could be calibrated by top-down processes. A similar calibration mechanism has been observed in early visual processing of other features (e.g., Rossi & Paradiso, 1995; Alais & Blake, 1999). Further investigation is needed to determine whether the top-down calibration produced by attention to the reference frame is achieved with perceptual processing of the other visual features. REFERENCES
AKS, D. J., & ENNS, J. T. (1992) Visual search for direction of shading is influenced by apparent depth. Perception & Psychophysics, 52, 63-74. ALAIS, D., & BLAKE, R. (1999) Neural strength of visual attention gauged by motion adaptation. Nature Neuroscience, 2, 1015-1018. BERBAUM, K., BEVER, T., & CHUNG, C. S. (1984) Extending the perception of shape from known to unknown shading. Perception, 13, 479-488. BRAUN, J. (1993) Shape-from-shading is independent of visual attention and may be a texton. Spatial Vision, 7, 311-322. CUTTING, J. E. (1978) A program to generate synthetic walkers as dynamic point-light displays. Behavior Research Methods & Instrumentation, 10, 91-94.
13-PMS_Sasaki_130168.indd 861
24/05/14 9:31 AM
862
H. SASAKI
GOODALE, M. A., & MILNER, A. D. (1992) Separate visual pathways for perception and action. Trends in Neurosciences, 15, 20-25. GREGORY, R. L. (1973) The confounded eye. In R. L. Gregory & E. H. Gombrich (Eds.), Illusion in nature and art. London, UK: Duckworth. Pp. 49-95. HOMMEL, B., & LIPPA, Y. (1995) S–R compatibility effects due to context-dependent spatial stimulus coding. Psychonomic Bulletin & Review, 2, 370-374. HOWARD, I. P., BERGSTROM, S. S., & OHMI, M. (1990) Shape from shading in different frames of reference. Perception, 19, 523-530. JENKIN, H. L., JENKIN, M. R., DYDE, R. T., & HARRIS, L. R. (2004) Shape-from-shading depends on visual, gravitational, and body-orientation cues. Perception, 33, 1453-1461. KLEFFNER, D. A., & RAMACHANDRAN, V. S. (1992) On the perception of shape from shading. Perception & Psychophysics, 52, 18-36. LABERGE, D. (2002) Attentional control: brief and prolonged. Psychological Research/Psychologische Forschung, 66, 220-233. LIU, B., & TODD, J. T. (2004) Perceptual biases in the interpretation of 3D shape from shading. Vision Research, 44, 2135-2145. LOGAN, G. D. (1996) Top-down control of reference frame alignment in directing attention from cue to target. In A. F. Kramer, M. G. H. Coles, & G. D. Logan (Eds.), Converging operations in the study of visual selective attention. Washington, DC: American Psychological Association. Pp. 415-437. MAMASSIAN, P., & GOUTCHER, R. (2001) Prior knowledge on the illumination position. Cognition, 81, B1-B9. PAVLOVA, M., & SOKOLOV, A. (2000) Orientation specificity in biological motion perception. Perception & Psychophysics, 62, 889-899. RAMACHANDRAN, V. S. (1988a) Perceiving shape from shading. Scientific American, 259, 58-65. RAMACHANDRAN, V. S. (1988b) Perception of shape from shading. Nature, 331, 163-166. ROSSI, A. F., & PARADISO, M. A. (1995) Feature-specific effects of selective visual attention. Vision Research, 35, 621-634. SAENZ, M., BURACAS, G. T., & BOYNTON, G. M. (2003) Global feature-based attention for motion and color. Vision Research, 43, 629-637. SUN, J., & PERONA, P. (1996) Preattentive perception of elementary three-dimensional shapes. Vision Research, 36, 2515-2529. SUN, J., & PERONA, P. (1997) Shading and stereo in early perception of shape and reflectance. Perception, 26, 519-529. SUN, J., & PERONA, P. (1998) Where is the sun? Nature Neuroscience, 1, 183-184. SYMONS, L. A., CUDDY, F., & HUMPHREY, K. (2000) Orientation tuning of shape from shading. Perception & Psychophysics, 62, 557-568. TROJE, N. F. (2003) Reference frames for orientation anisotropies in face recognition and biological-motion perception. Perception, 32, 201-210. WENDEROTH, P., & HICKEY, N. (1993) Object and head orientation effects on symmetry perception defined by shape from shading. Perception, 22, 1121-1130. WRAGA, M., CREEM, S. H., & PROFFITT, D. R. (1999) The influence of spatial reference frames on imagined object and viewer rotations. Acta Psychologica, 102, 247-264. Accepted April 3, 2014.
13-PMS_Sasaki_130168.indd 862
24/05/14 9:31 AM
Copyright of Perceptual & Motor Skills is the property of Ammons Scientific, Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
VISUAL ATTENTION TO REFERENCE FRAMES AFFECTS PERCEPTIONS OF SHAPE FROM SHADING1 HIROYUKI SASAKI Department of Child Education Niigata Chuoh Junior College, Japan Summary.—Perception of shape from shading is processed locally in a bottomup manner, but is also influenced by global or contextual factors. This study examined the influence of attention to the reference frame on the perception of shape from shading. In a visual search task, participants were asked to identify the location of a target relative to the reference frame. The results showed that shaded targets were more quickly and accurately detected when the shading gradient was parallel, rather than orthogonal, to the orientation of the environmental reference frame. This was further supported by a second experiment with a masking paradigm. Consequently, the perceptual process of shape from shading may be a flexible mechanism in which the representation of gradient orientation is calibrated by topdown processing in visual attention.
The appearance of a visual scene is fundamentally ambiguous because information about the three-dimensional world is projected onto a two-dimensional retinal surface. To resolve ambiguity, the visual system utilizes prior constraints on the way visual scenes are structured. One of the constraints comes from the assumption that scenes are lit from above our heads (Ramachandran, 1988a, 1988b; Mamassian & Goutcher, 2001). For example, even in the absence of a clear source of illumination, we typically perceive a light-above-dark disk to be a dome-shaped mound, whereas a dark-above-light disk is perceived to be a bowl-shaped depression. This convex or concave percept, however, becomes ambiguous and unstable when a horizontal light-to-dark gradient is provided. These percepts suggest that the visual system incorporates an overhead-lighting constraint in its interpretation of shape-from-shading (Ramachandran, 1988a, 1988b; Mamassian & Goutcher, 2001). Some visual search studies have explored whether shape from shading is interpreted during preattentive or higher cognitive processing. Kleffner and Ramachandran (1992; see also Aks & Enns, 1992; Symons, Cuddy, & Humphrey, 2000) found that vertically shaded targets were immediately detected regardless of the number of distractors; however, horizontally shaded targets were detected less efficiently. Because it is possible to preattentively search for a vertically shaded target, the authors argued that Address correspondence to Hiroyuki Sasaki, 16–18, Gakko-cho, Kamo, Niigata 959–1322, Japan or e-mail ([email protected]). 1
DOI 10.2466/24.22.PMS.118k26w9
13-PMS_Sasaki_130168.indd 850
ISSN 0031-5125
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
851
the overhead-lighting constraint is implemented in a hard-wired, early visual process. In a dual task paradigm (Braun, 1993), performance in a shape from shading search task was not compromised by a concurrent letter-discrimination task. Braun therefore argued that perception of shape from shading does not significantly depend on visual attention. A different line of evidence for the overhead-lighting constraint was provided by studies on whether the visual system uses retinal or gravitational coordinates. These experiments examined changes in the perception of shape from shading caused by rotation of the head; this rotation caused a discrepancy between the retinal and gravitational coordinates when the object was lit from above. Kleffner and Ramachandran (1992) and Wenderoth and Hickey (1993) found that the segregation of texture patterns made from shaded disks was impaired when the direction of the shading gradient was parallel to gravity and orthogonal to the vertical axis of the head (for instance, when participants lay on their sides; see also Howard, Bergstrom, & Ohmi, 1990). Thus, the perception of shape from shading was found to be dependent on the retinal coordinate system. This also supports the claim that the overhead-lighting constraint is invoked during an early stage of visual processing that preserves retinotopy. Although these findings suggest that shape from shading is processed locally in a stimulus-driven or bottom-up manner, some studies have shown a role for global or contextual factors in the perception of shape from shading. The hollow-face illusion is a striking example of this. The expectation that certain objects will be convex outweighs the assumption that light is coming from above (Gregory, 1973); a picture of a hollow face lit from above looks convex and its light source is perceived as coming from below (see also Liu & Todd, 2004 for a similar convex bias taking precedence over the overhead-lighting constraint). Similarly, the shading cues attached to a familiar object, which imply the position of the light source, were found to overshadow the perception of shape from shading of a stimulus presented adjacent to the object (Berbaum, Bever, & Chung, 1984; Ramachandran, 1988a; Sun & Perona, 1996). For example, the concave/convex ambiguity of shape from shading depends on the light source information from a shaded face rather than the assumption of overhead lighting. Jenkin, Jenkin, Dyde, and Harris (2004) measured the perceived convexity of a shaded disk in terms of environmental, gravitational, and retinal frames of reference by independently rotating the visual context and the head. As a result, the perception of shape from shading was found to depend not only on the retinal coordinates but also on gravitational and environmental coordinates. This finding indicates that the visual system
13-PMS_Sasaki_130168.indd 851
24/05/14 9:31 AM
852
H. SASAKI
determines which direction is up from the sum of the vectors represented in the retinal and the other frames. In the present study, another type of global interaction effect was examined, namely, the influence of attention to the environmental reference frame on the perception of shape from shading. It was assumed that focusing attention on a task-relevant stimulus indicating the environment's upright orientation could modulate the perception of shape from shading. To manipulate the observer's reference frame, an experimental paradigm was employed in which the participants responded to the target location relative to the orientation of the rotated reference frame against which the shaded disks were mapped (see Fig. 1 and its caption). Such manipulation is consistent with a previous suggestion that viewer rotations are easier to imagine than object rotations (Wraga, Creem, & Proffitt, 1999). The vertical axis of the reference frame was tilted 45° clockwise (CW) or 45° counterclockwise (CCW) from the retinal and gravitational vertical axes. Participants detected a shaded target among distractors located around the reference cue, which was displayed at the center of the monitor. The direction of the light-to-dark gradient of the shaded stimuli was also tilted 45° CW or 45° CCW; however, this tilt varied independently from the orientation of the reference frame. A shaded disk with an oblique luminance gradient allowed the perception of three-dimensional shape, based on the implicit assumption that light is coming from overhead (Symons, et al., 2000). It was expected that the perception of shape from shading depended on the direction of the light source derived from the environmental reference frame. Generally, it is known that a dark-above-light target among light-above-dark distractors is more discriminable than vice versa (Kleffner & Ramachandran, 1992; Sun & Perona, 1997). Thus, in this study the search asymmetry should occur more clearly under a concordant condition than under a discordant condition. Hypothesis 1. Shaded targets will be detected more quickly and/or accurately when the direction of the shading gradient is parallel to the orientation of the vertical axis of the reference frame; for instance, when a CW-tilted target was located in a CW-tilted reference frame (the concordant condition) than when the target and reference frame have different orientations, such as a CW-tilted target within a CCW-tilted reference frame (the discordant condition). Hypothesis 2. The search asymmetry between light-above-dark and dark-above-light targets will be stronger under the concordant condition than under the discordant condition.
13-PMS_Sasaki_130168.indd 852
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
853
Experiment 1 METHOD Participants Twelve university students participated in the experiment. Their ages ranged from 23 to 33 years. All participants had normal or corrected-tonormal vision and were naïve regarding the purpose of the experiment. Participation was voluntary and anonymous. After the data were analyzed and summarized, the findings were distributed to the participants. There was no requirement for ethical review, but the experiment followed the guidelines in the Declaration of Helsinki. Apparatus and Stimuli Stimuli were generated by an IBM PC-compatible computer and presented on a CRT monitor with a frame rate of 60 Hz and a resolution of 800 × 600 pixels. The pixel size was approximately 2.4 min. of arc at a viewing distance of 57 cm. Targets and distracters were constructed from 256 luminance levels ranging from 5.3 to 25.2 cd/m2 and were presented on a background of 13.0 cd/m2. For the trials, each participant sat upright in a darkened room with his or her head fixed on a chin and head rest. A biological-motion display (a point-light walker) was used as an environmental reference cue because it was expected that the walker pattern would provide reliable information about upright orientation to the viewer. The biological-motion display was generated according to Cutting's algorithm (Cutting, 1978) and consisted of 11 black dots positioned at the head and main joints of the profile of a walker. The walking figure was centered in the middle of the monitor and subtended 5.2° in height and 2.4° in width at the most extended point of the gait cycle. A full gait cycle was completed in 60 frames, with a frame duration of 16.6 msec. The direction of walking (left or right) varied randomly from trial to trial. The walker was either upright or tilted at either 45° CW or 45° CCW. Although biological-motion perception is known to depend on the orientation of the display with respect to the retinal coordinates (Troje, 2003), a display oriented 45° CW or 45° CCW is easily perceived as a walker (Pavlova & Sokolov, 2000). Four shaded disks were presented on an imaginary circle with a radius of 4° of visual angle around the reference cue. Each disk subtended 1.5°. One of the disks, which served as the target, differed from the others in the luminance polarity of its light-to-dark gradient. The target and distractor items were located at the 45° CW, 135° CW, 45° CCW, and 135° CCW positions from the upright within the reference frame as defined by the walker. The position of each item randomly varied within a radius of 0.5° to prevent judgments from being made based on item collinearity. Thus, when the reference frame was tilted 45° CW, for instance, the 135° CCW target
13-PMS_Sasaki_130168.indd 853
24/05/14 9:31 AM
854
H. SASAKI
FIG. 1. Search display, including the dark-above-light target, around the reference cue. The left panel shows the discordant condition (the shading gradient tilted CCW and the walking figure tilted CW), whereas the right panel shows the concordant condition (both the shading gradient and the walking figure tilted CCW). The middle panel shows the neutral condition. To choose the 135° CCW target position relative to the orientation of the walking figure tilted CW (left panel), the F key was pressed with the left index finger. The 45° CCW target position relative to the walking figure tilted CCW (right panel) was selected by pressing the E key with the left middle finger.
was located at the same retinal 9 o'clock position as the 45° CCW target when the reference frame was tilted 45° CCW (Fig. 1). The direction of the light-to-dark gradient of the shaded disks was tilted either 45° CW or 45° CCW, and the type of targets was either light-above-dark disk or darkabove-light disk. Design and Procedure The within-subjects design included three independent variables: orientation of the rotation axis of the reference frame (upright, 45° CW or 45° CCW), direction of the shading gradient (45° CW or 45° CCW), and target type (light-above-dark or dark-above-light). An experimental session consisted of nine blocks (three blocks for each of the three orientations of the reference frame). Each block contained a total of 48 trials (12 trials for each of the two possible directions of the shading gradient × the two types of target). Thus, each session contained a total of 432 trials for each participant. The order of the blocks (i.e., the presentation order of the three orientations of the reference frame) was randomized within each session, and the order of the trials (i.e., the presentation order of the two directions of the shading gradient and the two types of target) was randomized within each block. The participants engaged in two experimental sessions. Therefore, each data point was based on 72 trials per stimulus condition (12 trials × 3 blocks × 2 sessions). A practice session comprising at least three blocks was carried out before each session. At the beginning of each block, the biological-motion display was presented for 1 sec. to inform participants about the orientation of the en-
13-PMS_Sasaki_130168.indd 854
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
855
vironmental frame of reference. The reference frame orientation was fixed for a block consisting of 48 trials. Each trial began with the appearance of the reference cue (the biological-motion display) for 1.5 sec. before the search display was introduced. The search display and the reference cue remained visible until the participant responded. An acoustic signal was provided following an incorrect response. The next trial followed 1.5 sec. after the response. The search display of four shaded disks was presented around the reference cue. The participants were told that the target, which could be distinguished from the distractors by its shape (convex vs concave), would always appear in the search display. They were then instructed to identify the target position relative to the reference cue, which was either upright or tilted either CW or CCW from the retinal and gravitational vertical axes, as quickly and accurately as possible by pressing the corresponding key on a computer keyboard. Responses for the 45° CCW, 135° CCW, 45° CW, and 135° CW target positions from the upright within the reference frame were assigned to the E, F, I, and J keys on the keyboard, which were pressed by the left middle, left index, right middle, and right index fingers, respectively. The keyboard was aligned parallel to the frontal plane of the body and the monitor. The participants were free to move their eyes while searching for the targets. Eye movements were not monitored. Reaction time (RT) was measured as the time between the onset of the search display and the button-press response. RTs in error trials were omitted from the data analysis. RTs under 150 msec. or over 3,000 msec. were also excluded. RESULTS Figure 2 shows the mean RTs and mean error rates. Repeated-measures analyses of variance (ANOVAs) were conducted on the RT and error rate data with reference frame orientation (CW, upright, or CCW), lightto-dark gradient orientation (CW or CCW), and target type (light-abovedark or dark-above-light) as factors. Reaction Times The main effect of the light-to-dark gradient orientation was significant (F1, 11 = 20.17, p < .001, ηp2 = 0.65), indicating that the participants responded faster when the gradient was tilted CCW than when it was tilted CW. The main effect of target type was also statistically significant (F1,11 = 8.35, p < .05, ηp2 = 0.43). This replicates the previously reported search asymmetry between light-above-dark and dark-above-light targets; darkabove-light targets are easier to detect than light-above-dark targets (Kleffner & Ramachandran, 1992). On the other hand, reference frame orientation only marginally affected the reaction time (F2, 22 = 3.08, p = .07, η2p = 0.12).
13-PMS_Sasaki_130168.indd 855
24/05/14 9:31 AM
856
H. SASAKI 1500
20
15
1300
Error Rate (%)
RT (msec.)
1400
1200 1100 1000
CW - light-above-dark CW - dark-above-light CCW - light-above-dark CCW - dark-above-light
10
5
900 800
0 CCW
Upright
CW
Reference Frame Orientation
CCW
Upright
CW
Reference Frame Orientation
FIG. 2. Experiment 1: mean RTs (left panel) and mean error rates (right panel) for the light-to-dark gradient orientation and the target type as a function of the orientation of the reference frame. Open symbols and filled symbols denote the conditions in which the gradient was tilted CW and CCW, respectively. Circles denote conditions with a light-above-dark target and triangles denote conditions with a dark-above-light target. Error bars denote standard errors of the mean.
The interaction between the light-to-dark gradient orientation and the target type was significant (F2, 22 = 9.35, p < .05, ηp2 = 0.46). More importantly, the interaction between the reference frame and the light-to-dark gradient orientations was significant (F2, 22 = 3.73, p < .05, ηp2 = 0.15); there were significant differences in the gradient direction when the reference frame orientation was upright (F1, 33 = 18.99, p < .001 ηp2 = 0.37), and tilted CCW (F1, 33 = 17.46, p < .001, ηp2 = 0.35); however, there was no significant difference when the reference frame orientation was tilted CW (F1, 33 = 2.43, p > .1, ηp2 = 0.07). Other interactions were not significant (Fs2, 22 < 0.6, ps > .5). Error Rates The main effect of the gradient orientation was only marginally significant (F1, 11 = 4.24, p = .06, ηp2 = 0.28), but it seems to indicate that the participants responded more accurately when the gradient was tilted CCW than when it was tilted CW. The main effects of reference frame orientation and target type were not significant (F2, 22 = 2.30, p > .1, ηp2 = 0.09 and F1, 11 = 1.02, p > .1, ηp2 = 0.09, respectively). The interaction between the reference frame and the gradient orientations was significant (F2, 22 = 5.35, p < .05, ηp2 = 0.20); there were significant differences in the gradient orientation when the reference frame orientation was upright (F1, 33 = 7.15, ηp2 = 0.18, p < .05) or tilted CCW (F1, 33 = 6.47, p < .05, ηp2 = 0.16). On the other hand, there was no significant difference when the reference frame orientation was tilted CW (F1, 33 = 0.002, p > .5, ηp2 < 0.01). The other interactions were not significant (Fs2, 22 < 0.4, ps > .5).
13-PMS_Sasaki_130168.indd 856
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
857
DISCUSSION The results showed that performance in the concordant condition was better than performance in the discordant condition when the reference frame was tilted CCW, although there was no difference in performance between the two conditions when the reference frame was oriented CW. In considering this result, performance in the upright frame of reference should be taken into account. The upright reference frame functioned as a neutral cue for the detection of shaded targets with shading gradients tilted either CW or CCW. Thus, less cognitive effort should be needed to align the mental coordinate system with the reference frame when responding to a target presented in the upright reference frame. Still, the shaded target was detected faster when the shading gradient was tilted CCW than when it was tilted CW. This bias towards the CCW direction (i.e., the upper-left direction) is consistent with the findings reported by Sun and Perona (1998) and by Mamassian and Goutcher (2001), who maintained that the visual system prefers above-left lighting to direct overhead lighting. The preference for above-left lighting is probably independent of reference frame manipulation. In the case of the CCW-tilted frame of reference, the above-left lighting situation (specifically, the shading gradient tilted CCW) corresponded to the concordant condition. Consequently, the significant effect of the CCWtilted frame of reference can plausibly be interpreted as the combined result of two biases: the overhead-lightning constraint within the CCW-tilted reference frame and the preference for above-left lightning. The results for error rates were comparable to those of RTs. Error rates increased in conditions in which RTs were slower. Thus, variations in RTs did not reflect a speed-accuracy trade-off. These results seem to offer some support for Hypothesis 1, that the perception of shaded targets would be influenced by whether the direction of the shading gradient corresponded to the reference frame orientation. However, there remains a possibility that the reference cue (the biological-motion display) itself may have been sufficient to trigger the differences (cf. the context effect; Sun & Perona, 1996; Jenkin, et al., 2004) without the participants focusing attention on the reference cue. Therefore, to confirm the need for attention to the reference cue, a supplementary control experiment was conducted in which the participants were not required to attend to the reference cue tilted either CW or CCW, and the search display of four shaded disks (tilted either CW or CCW) was placed with respect to the upright retinal coordinate. The results of this control experiment showed that the performance difference between the concordant and discordant conditions disappeared when participants were not required to obtain the reference frame from the biological-motion display. Although the walker pattern explicitly provided upright information as
13-PMS_Sasaki_130168.indd 857
24/05/14 9:31 AM
858
H. SASAKI
a reference frame, it had no effect on the performance of the detection of shape from shading targets. Therefore, this implies that not only the presentation of the reference frame but also attention to it was necessary for the results obtained in Study 1. On the other hand, a search asymmetry between the light-above-dark and dark-above-light targets was found in this experiment. This occurred under the discordant as well as the concordant condition, contradicting Hypothesis 2, that the search asymmetry would occur more clearly under the concordant condition than under the discordant condition. Experiment 2 In Experiment 1, the search display was presented until the participants responded (mean RTs from the onset of the search display were longer than 1 sec.). This leaves the possibility that the results may have reflected higher cognitive processing rather than perceptual processing. Namely, attention to the reference frame may not have directly affected the shape perception from shading, and the poor performance in the discordant condition may have been due to a cognitive conflict caused by the difference in the orientations of the reference frame and the light source. Thus, Experiment 2 examined whether attention to the reference frame showing the environmental upright orientation modulates the perception of the shaded disks or generates a cognitive conflict with the upright direction derived from the light-to-dark gradient orientation (i.e., overheadlighting assumption). For this purpose, a masking paradigm was used (cf. Sun & Perona, 1996, 1997). Sun and Perona (1996) suggested that experiments involving shaded stimuli followed by backward masking confirm that shape from shading can be processed in early stages of visual processing. Therefore, if the findings of Experiment 1 could be replicated with the masking paradigm, it would appear that attention to the reference frame could affect the early processing of shape from shading. METHOD Participants Ten university students participated in the experiment. Their ages ranged from 23 to 33 years. All participants had normal or corrected-tonormal vision and were naïve regarding the purpose of the experiments. Nine of these students had participated in Experiment 1. Apparatus and Stimuli The apparatus was the same as that used in Experiment 1. The reference frame indicated by the biological motion display was tilted either 45° CW or 45° CCW. The light-to-dark gradient of the shaded disks was tilted either 45° CW or 45° CCW. The targets were either the light-above-
13-PMS_Sasaki_130168.indd 858
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
859
dark or dark-above-light disks. Random-dot disks were used as the mask stimuli. Each disk replacing each shaded disk comprised 50% density bright/dark dots, and subtended 2°. The dot size was 12 min. of arc. Design and Procedure The design and procedure were similar to Experiment 1 with a few exceptions. An experimental session consisted of six blocks (three blocks for each of the two orientations of the reference frame). A practice session comprising at least two blocks was carried out before each of the two sessions. As in Experiment 1, the search display was presented 1.5 sec. after the onset of the reference cue. The search items were then replaced by mask stimuli, which were presented for 1 sec. Before the main experiment, asynchrony between the onset of the disks and the onset of the masks, which ranged from 33 to 333 msec., was adjusted for each participant using an abbreviated psychological staircase method. This procedure was designed to prevent floor and ceiling effects on performance. After the termination of the reference cue and the mask stimuli, the participants responded to the target location relative to the orientation of the reference frame by pressing the corresponding key on a computer keyboard. The target-response assignment of the keys was identical to that used in Experiment 1. Unlike in Experiment 1, participants were not required to respond quickly. RESULTS Figure 3 shows the percentage of error responses. Repeated-measures ANOVAs were conducted with reference frame orientation (CW or CCW), light-to-dark gradient orientation (CW or CCW), and target type (lightabove-dark or dark-above-light) as factors. The main effect of light-to-dark gradient direction was statistically significant (F1, 9 = 16.58, p < .005, ηp2 = 0.65), indicating that the participants located the targets more accurately when the gradient was tilted CCW than when it was tilted CW. The main effect of target type was also significant (F1, 9 = 19.70, p < .005, ηp2 = 0.69), indicating that the participants located the light-above-dark targets more accurately than the dark-abovelight targets. On the other hand, the main effect of the reference frame orientation was not significant (F1, 9 = 0.25, p > .5, ηp2 = 0.03). The interaction between the reference frame and the gradient orientations was significant (F1, 9 = 20.61, p < .005, η2p= 0.70); there was a significant difference in the gradient direction when the reference frame orientation was tilted CCW (F1, 18 = 33.68, p < .001, η2p = 0.65). There was, however, no significant difference when it was tilted CW (F1, 18 = 1.52, p > .1, ηp2 = 0.08). The other interactions were not significant (Fs1, 9 < 0.8, ps > .1).
13-PMS_Sasaki_130168.indd 859
24/05/14 9:31 AM
860
H. SASAKI
60
CW - light-above-dark CW - dark-above-light CCW - light-above-dark CCW - dark-above-light
Error Rate (%)
50 40 30 20 10 CCW CW Reference Frame Orientation FIG. 3. Experiment 2: mean error rates for the light-to-dark gradient orientations and target types as a function of the orientation of the reference frame.
DISCUSSION In Experiment 2, perceptual sensitivity to shape from shading was examined with a backward masking paradigm. This paradigm did not require quick reactions from the participants, and demanded only accurate judgments. Under such circumstances, it is unlikely that the participants' failure to locate the target reflected a cognitive conflict caused by the difference in the orientations of the reference frame and the light source. Nonetheless, performance in target discrimination differed between the concordant and discordant conditions. Thus, the mismatch between the reference frame and the gradient orientations should not have resulted in interference in a later cognitive stage. The response accuracy data in Experiment 2 showed a pattern similar to that of the reaction times and response accuracy in Experiment 1. These results suggest that the perception of shape from shading was modulated by attention to the reference frame. The effect of focusing attention on the upright orientation in the environment seems to have occurred during a perceptual process rather than in a higher cognitive process. GENERAL DISCUSSION In the present study, the authors investigated the effect of attention to the orientation of the rotated reference frame on the perception of shape from shading. Evidence from visual search and masking paradigms basically supported Hypothesis 1, that shaded targets are detected faster and /or more accurately when the light source is located in a position defined as overhead in terms of the frame of reference. This result was not due to
13-PMS_Sasaki_130168.indd 860
24/05/14 9:31 AM
ATTENTIONAL EFFECTS ON SHAPE FROM SHADING
861
cognitive conflict at a higher stage of processing or a contextual effect. In other words, the early visual processing of shape from shading was affected by attention to the reference frame. The effect of attention to the reference frame has been measured by a voluntary attention paradigm similar to that used in the current study (Logan, 1996; see also Hommel & Lippa, 1995). In Logan's (1996) study, the response to a target color was determined by the location of the target relative to a visual context (a rotated face) that served as a frame of reference. Logan argued that directing attention from reference cues to targets implies an understanding of the spatial relationship between the two and that the reference frame orients attention to space, just as a spotlight does in traditional theories of attention. Correspondingly, reference to the environmental frame in the present study affected the perception of visual features in a manner similar to feature-based attention (e.g., Saenz, Buracas, & Boynton, 2003). The mental coordinate system aligned with the environmental reference frame could work to modulate shape perception from shading. These facts thus suggest that the effects of attention to the reference frame influence both the discernment of spatial relationships and the modulation of perceptual shape, in contrast to the well-known distinction between spatial orientation and feature analysis (Goodale & Milner, 1992; LaBerge, 2002). In conclusion, the visual process that implements the overhead-lighting constraint seems to be a flexible mechanism that can be modulated by attentional control of the mental coordinate system. Although the lightsource information is represented implicitly by the early visual system, enabling the detection of shaded targets without focal attention (Braun, 1993), the representation of gradient-orientation tuning could be calibrated by top-down processes. A similar calibration mechanism has been observed in early visual processing of other features (e.g., Rossi & Paradiso, 1995; Alais & Blake, 1999). Further investigation is needed to determine whether the top-down calibration produced by attention to the reference frame is achieved with perceptual processing of the other visual features. REFERENCES
AKS, D. J., & ENNS, J. T. (1992) Visual search for direction of shading is influenced by apparent depth. Perception & Psychophysics, 52, 63-74. ALAIS, D., & BLAKE, R. (1999) Neural strength of visual attention gauged by motion adaptation. Nature Neuroscience, 2, 1015-1018. BERBAUM, K., BEVER, T., & CHUNG, C. S. (1984) Extending the perception of shape from known to unknown shading. Perception, 13, 479-488. BRAUN, J. (1993) Shape-from-shading is independent of visual attention and may be a texton. Spatial Vision, 7, 311-322. CUTTING, J. E. (1978) A program to generate synthetic walkers as dynamic point-light displays. Behavior Research Methods & Instrumentation, 10, 91-94.
13-PMS_Sasaki_130168.indd 861
24/05/14 9:31 AM
862
H. SASAKI
GOODALE, M. A., & MILNER, A. D. (1992) Separate visual pathways for perception and action. Trends in Neurosciences, 15, 20-25. GREGORY, R. L. (1973) The confounded eye. In R. L. Gregory & E. H. Gombrich (Eds.), Illusion in nature and art. London, UK: Duckworth. Pp. 49-95. HOMMEL, B., & LIPPA, Y. (1995) S–R compatibility effects due to context-dependent spatial stimulus coding. Psychonomic Bulletin & Review, 2, 370-374. HOWARD, I. P., BERGSTROM, S. S., & OHMI, M. (1990) Shape from shading in different frames of reference. Perception, 19, 523-530. JENKIN, H. L., JENKIN, M. R., DYDE, R. T., & HARRIS, L. R. (2004) Shape-from-shading depends on visual, gravitational, and body-orientation cues. Perception, 33, 1453-1461. KLEFFNER, D. A., & RAMACHANDRAN, V. S. (1992) On the perception of shape from shading. Perception & Psychophysics, 52, 18-36. LABERGE, D. (2002) Attentional control: brief and prolonged. Psychological Research/Psychologische Forschung, 66, 220-233. LIU, B., & TODD, J. T. (2004) Perceptual biases in the interpretation of 3D shape from shading. Vision Research, 44, 2135-2145. LOGAN, G. D. (1996) Top-down control of reference frame alignment in directing attention from cue to target. In A. F. Kramer, M. G. H. Coles, & G. D. Logan (Eds.), Converging operations in the study of visual selective attention. Washington, DC: American Psychological Association. Pp. 415-437. MAMASSIAN, P., & GOUTCHER, R. (2001) Prior knowledge on the illumination position. Cognition, 81, B1-B9. PAVLOVA, M., & SOKOLOV, A. (2000) Orientation specificity in biological motion perception. Perception & Psychophysics, 62, 889-899. RAMACHANDRAN, V. S. (1988a) Perceiving shape from shading. Scientific American, 259, 58-65. RAMACHANDRAN, V. S. (1988b) Perception of shape from shading. Nature, 331, 163-166. ROSSI, A. F., & PARADISO, M. A. (1995) Feature-specific effects of selective visual attention. Vision Research, 35, 621-634. SAENZ, M., BURACAS, G. T., & BOYNTON, G. M. (2003) Global feature-based attention for motion and color. Vision Research, 43, 629-637. SUN, J., & PERONA, P. (1996) Preattentive perception of elementary three-dimensional shapes. Vision Research, 36, 2515-2529. SUN, J., & PERONA, P. (1997) Shading and stereo in early perception of shape and reflectance. Perception, 26, 519-529. SUN, J., & PERONA, P. (1998) Where is the sun? Nature Neuroscience, 1, 183-184. SYMONS, L. A., CUDDY, F., & HUMPHREY, K. (2000) Orientation tuning of shape from shading. Perception & Psychophysics, 62, 557-568. TROJE, N. F. (2003) Reference frames for orientation anisotropies in face recognition and biological-motion perception. Perception, 32, 201-210. WENDEROTH, P., & HICKEY, N. (1993) Object and head orientation effects on symmetry perception defined by shape from shading. Perception, 22, 1121-1130. WRAGA, M., CREEM, S. H., & PROFFITT, D. R. (1999) The influence of spatial reference frames on imagined object and viewer rotations. Acta Psychologica, 102, 247-264. Accepted April 3, 2014.
13-PMS_Sasaki_130168.indd 862
24/05/14 9:31 AM
Copyright of Perceptual & Motor Skills is the property of Ammons Scientific, Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
