INTEGRATION OF TEXTURE AND MOTION CUES TO DEPTH IS ADAPTABLE.
((R. A. Jacobs and I. Fine)) Center for Visual Science, University of Rochester, Rochester, NY 14627.
Purpose: Several
investigators have suggested that observers' percepts of depth are a linear
combination of the perceived depths based on each of the individual cues in a
scene (e.g., Landy, Maloney, Johnston, and Young, 1995). These linear coefficients have been
estimated psychophysically using cues corrupted by noise (e.g., Johnston,
Cumming, and Parker, 1992). We examined
whether observers could adapt their linear coefficients during the course of
training in order to improve performance on a depth judgment task. Results suggest that observers learn to
adapt their cue combination strategies in an experience-dependent manner. Methods:
Observers judged which of two sequentially presented cylinders was greater in
depth using a two-alternative forced choice procedure with auditory
feedback. The shapes of the cylinders
were indicated by motion and texture cues.
Using a graphics manipulation, the motion and texture cues could
indicate different shapes simultaneously; e.g., the motion cue might indicate a
cylinder with a circular horizontal cross-section while the texture cue
indicated a cylinder of identical height and width but with an elliptical
horizontal cross-section that was more deep than wide. Half the observers were first trained with
motion as the reliable cue (only motion distinguished the depths of the two
cylinders; the texture cue indicated a circular horizontal cross-section for
both cylinders). After training,
observers were run on test trials designed to estimate the relative degree to
which their overall perception of depth was based upon motion and texture. These observers were then retrained with
texture as the reliable cue, and their linear coefficients were
re-estimated. The other half of the
observers were trained and tested in the reverse order. Results
and Conclusions: When the preceding training session used motion as the
reliable cue, observers' linear coefficient on the depth percept based upon the
motion cue was significantly larger than when the preceding training session
used texture as the reliable cue. Thus,
observers learn to adapt their cue combination strategies in an
experience-dependent manner. This work
was supported by NIH grant R29-MH54770. None.