Daniel McConnell and Geoffrey P. Bingham Perception-Action Laboratory, Dept. Of Psychology, Indiana University, Bloomington, IN
email: dsmcconn@indiana.edu gbingham@indiana.edu
Spatial metrics are lost in the projection from events into optics, yet temporal scale is preserved. Temporal scale is linked to spatial scale specific to the event considered, via a natural law. This is a potential source of information about scale. We examined observersŐ ability to perceive depth in displays which eliminated other sources of information about distance. We hypothesized that observers rely on event specific information to attune to the appropriate scaling relation. Observers were able to estimate scale, but exhibited a large amount of between-subject variability. After feedback on one event, observers performed better, and generalized training to other events. We conclude that observers know the general form of the scaling relation, but require feedback or experience to attune to the definite scaling.
Previous authors have suggested that observers used event duration as the basis of judgment (Pittenger, 1985; 1990). It is true that the period of a pendulum is proportional to its length, but are observers merely judging time? In studies examining only one event type, event duration is confounded with scale, hence making it impossible to determine the basis of observers' judgment. In the current study, we unconfound these factors by presenting observers with multiple events: Freefall (FF), Slope (with three gradations: Steep, Medium, and Shallow), and Pendulum). The value of the scaling constant varies across events, thus, equal durations in different events will map to different actual distances. This allows us to answer the question: are observers judging time or scale? If observers judge time, then when judgments are compared with event duration, we should find that the slopes of the regression lines for the various events to be the same. However, if differences are found, then we can conclude that observers used a different scaling constant for each event, as appropriate for judging actual distance.
Furthermore, we provide feedback on one event. If the training generalizes to other events, we can conclude that observers are attuned not only to the form of the scaling relation, but have a means of relating one event to another. Thus, we expect to see no differences in performance between events after training, when judgments are compared against actual distance. As before, if differences in performance are found between events when judgments are compared to event duration instead of actual distance, we can conclude that observers were not using time as basis of judgment.
Finally, previous studies of this nature have used computer simulations to present the stimuli (Hecht, Kaiser, & Banks, in press; Muchisky & Bingham, 1992). However, simulations are an ideal depiction that ignore the effects of friction and air resistance on objects in the event. In real life, these forces affect the trajectory of a moving object as a function of mass. Hecht, Kaiser, and Banks (in press) suggest that observers cannot use trajectory information optimally because of the perturbatory affect of these forces. In the current study, we examined this in more detail by including balls of different material composition (wood and styrofoam), and thus of different masses.
Stimuli.
Balls of six different sizes were filmed in patch-light conditions in five events: free-fall, rolling down a slope (with three gradations), and as the bob on a pendulum. Ball size was co-varied with distance to maintain a constant image size in the displays. For two of the ball sizes, there were two different balls, one made of styrofoam, and a much heavier one made of wood, thus making eight balls used in the displays.
Procedure.
The displays were presented to four observers via a monocular head-mounted video display. The task was to judge the distance of the events. After the observer viewed the patch-light display, the video monitor was switched to a head-mounted camera that allowed the observer to see a sphere under patch-light conditions. They used a pulley mounted at eye level to move the sphere to the appropriate distance. Ideally, the observers could apprehend the distance of the sphere through the motion of their arm (and head), combined with the change in image size occurring as the sphere moved along the pulley. The common viewing conditions for the display and response apparatus (with a spatial response task) should reduce the variability prevalent in previous similar studies (Muchisky & Bingham, 1992; Pittenger, 1985).
On Day 1, observers were presented with the steepest slope event only. They began with a (pre-training) block of 24 trials without feedback. This was followed by 15 training trials with feedback. Only the middle 3 of the 5 ball sizes were used in training. The wooden balls were excluded from the training session. Thus, feedback was provided only for the lighter balls. Finally, observers judged 24 post-training trials. Day 2 tested the stability of training to this event and generalization of training to other events. Observers judged 50 trials that included all five events.
Results and Discussion. Distance judgments were regressed against actual distance for the pre-training session of Day 1 for all observers. As shown in Figure 1, overall performance during pre-training was poor, with a slope = .26, y-intercept = 165.47, and r2=.02. This low r2 reflects, in part, a large amount of between subject variability. The mean individual r2 was higher at .25 (sd = .28). Distance judgments were also regressed against actual distance for the post-training session of Day 1. Overall performance improved significantly from the pre-training block, with a slope = 1.33, y-intercept = 44.04, and r2=.54. Examination of individual performance again shows within-subject consistency, with a mean individual r2=.79 (sd = .08). The results of a multiple regression comparing performance for pre- and post-training show that the post-training slope was significantly higher than the pre-training slope (partial F = 48.8, p < .001), and the y intercept was significantly lower (partial F = 27.3, p <.001). The Actual Distance factor was significant also (partial F = 108.2, p < .001), indicating that as actual distance increased, judgments increased as expected. Figure 1 illustrates the improved performance after feedback.
A simple regression of the StS judgments from the multi-event block of Day 2 against actual distance demonstrates that the improved performance on Day 1 was stable over time, with a slope = 1.25, y-intercept = 101.44, and r2=.46. A multiple regression comparing post-training to the Day 2 StS event found no significant differences for the slope or y-intercept. Performance on the other events from the Day 2 multi-event block was similar to the StS performance. A multiple regression found no significant slope or y-intercept differences between the various events. We next examined the effects of duration on the judgments. We compared events two at a time using multiple regression. Out of ten comparisons (five events in comparisons of two), there were five significant slope differences. Free fall differed from three events: MP (partial F = 13.72, p < .001), MS (partial F = 9.29, p < .01), and ShS (partial F = 8.75, p < .01). The StS event differed from MP (partial F = 9.43, p < .01) and MS (partial F = 4.35, p < .05). There were no significant differences in intercept. Within events, judgments increased with duration, as did actual size, but across events, multiple durations gave rise to similar judgments, and any single duration gave rise to many different judgments.
Finally, we examined the effects of varying the mass of the ball. At two of the distances, there were two balls, one of styrofoam and one of wood. We performed an ANOVA on these judgments with actual distance (2 levels) and material (2 levels) as factors. As expected, there was a main effect for distance, F(1, 59) = 65.53, p<.0001. Also, the heavier wooden balls were judged as nearer, F(1, 59) = 10.7, p<.01. The two-way interaction was also significant, F(1, 59) = 6.63, p<.05. The effect of material was stronger at the larger of the two distances. Observers did not alter their scaling strategy in the presence of a mass variation.
General Discussion. We know of two possible solutions to the scaling problem in vision. One is that definite scale might be obtained via somatosensory information accompanying self-motion generated optic flow. (Bingham & Pagano, 1995). The other solution is to achieve spatial scaling via temporal scaling when the two are nomologically linked. Previous results have provided only weak support for the latter solution. However, when viewing patch-light displays of real events, our observers performed nearly optimally after sufficient training. Thus, Hecht, Kaiser, and Bank's (in press) argument that observers cannot use this information optimally because of the effects of friction and air resistance is weakened. We suspect that the sub-par performance observed in their study may have been the result of their information impoverished computer simulations. However, our observers were unable to deal with mass variations. These results indicate that observers are sensitive to the nomological stucture and form of events but are not precisely tuned to the relevant scaling constants outside of particular tasks. However, sensitivity to the interrelated dynamical structures of different events allows observers to attune rapidly to the scaling constants given a modicum of task-specific feedback. Further, the task specific tuning is fairly stable. This is an efficient way to achieve effective spatial scaling in behavior.
References
Bingham, G.P. & Pagano, C.C. (1995). Monocular perception of egocentric distance via head movement towards a target: Verbal vs. Action measures. In (B.G. Bardy, R.J. Bootsma, Y. Guiard, Eds) Studies in Perception and Action III: Eighth Annual Conference on Perception and Action, Marseilles, France.
Hecht, H., Kaiser, M.K., & Banks, M.S. (In press). Gravitational acceleration as a cue for absolute size and distance. Perception and Psychophysics.
Muchisky, M.M. & Bingham, G.P. (1992). Perceiving size in events via kinematic form. In J.Kruschke, (Ed), Proceedings of the 14th Annual Conference of the Cognitive Science Society, pp. 1002-1007, Hillsdale, N.J.: Erlbaum.
Pittenger, J.B. (1985). Estimation of pendulum length from information in motion. Perception, 14, 247-256.
Pittenger, J.B. (1990). Detection of violations of the law of pendulum motion: ObserversŐ sensitivity to the relation between period and length. Ecological Psychology, 2(1), 55-81
Saxberg, B.V.H. (1987). Projected free fall trajectories: II. Human experiments. Biological Cybernetics, 56, 177-184.
Stappers, P.T., & Waller, P.E. (1993). Using the free fall of objects under gravity for visual depth estimation. Bulletin of the Psychonomic Society, 31, 125-127.