Abstract

Reflection symmetry is a pop-out image feature which receives preferential visual processing compared with, say, repetition or rotation symmetry (Julesz, 1971). We sought to examine the mechanisms underlying such processing, since its neurophysiological properties are unknown. Reflection symmetry was imposed on a multilevel random image on a Macintosh II computer. Dynamic noise was generated by random assignments to the lookup table for the random image. For opposite contrast symmetry, bright dots in one half field image were matched with symmetric dark dots in the other half. Forced-choice thresholds for symmetry detection were measured as a function of the width of nonsymmetric masking strip around the symmetry axis, forcing detection by features progressively further from the axis. For dynamic noise images, symmetry could be detected in ~120-msec presentations, while 1-sec presentations allowed detection up to masking widths of ~20 arcmin. On the other hand, in static images (sandwiched between dynamic fields) symmetry could be detected in ~60-msec presentations, while 1-sec presentations allowed detection up to much greater masking widths (~5°). This implies a static symmetry mechanism both more sensitive and operating over much larger cortical distances. Opposite contrast symmetry showed longer detection times in both static and dynamic noise, and its masking width was extremely narrow (~3 arcmin) in dynamic noise, implying a third neural mechanism.

© 1991 Optical Society of America