Abstract
Introduction: Spatial Glass patterns are a class of stochastic stimuli whose global shape is defined by lawful positional relationships between local elements. Here, we investigated the dynamics of shape coding for concentric Glass stimuli.
Methods: Stimuli consisted of two populations of 256 dipoles that were randomly repositioned on each frame. Both populations alternated between structured (Glass) and unstructured (noise) patterns on each frame. In the ‘in-phase’ condition, Glass and noise patterns from one population coincided in time with their counterparts from the other population. In the ‘out-of-phase’ condition, Glass patterns from one population coincided with noise patterns from the other population. Observers judged the position (left vs. right) of the Glass center relative to the stimulus' aperture. Dipole orientation was jittered to find the 75%-correct point, and we assessed performance over a range of frame rates (∼1 to 25 Hz).
Results: Observers performed significantly better (x2) in the in-phase than in the out-of-phase condition over the 1-to-5 Hz range. Performance above 5 Hz remained high but did not differ between in-phase and out-of-phase conditions.
Conclusions: Data show that shape coding for Glass patterns can exploit a temporal separation between signal and noise of up to 5 Hz (200 ms) - beyond 5 Hz, signal and noise become perceptually fused. Results suggest that the dynamics of shape coding can be approximated by linear lowpass filtering. We are investigating nonlinear aspects of shape-coding dynamics such as “object-locking” - a form of hysteresis whereby object capture is brisk but object release is sluggish.
Supported by NIH/NCRR Grant P20 RR020151.