Abstract
Perceptual learning (PL) has been shown to improve depth discrimination thresholds in random-dot stereograms (RDS). A previous study demonstrated that PL of RDS is consistent with a narrower tuning of neurons underlying the response and not with internal noise reduction, but effects of subjects diverting attention to a specific target location were not considered. This study examined location specific effects with PL of depth discrimination in RDS. Two vertically separated 4.3 × 4.3 ardeg, 500 msec duration, dynamic RDS patterns of 50% density, containing 6.7′ × 6.7′ dot elements, were offset by 0.43 deg from a central fixation spot. Six normal subjects, assigned a training side (right or left), judged the depth of the bottom test relative to the upper reference stimulus, with audio feedback. Pre- and post-learning at the untrained location thresholds were determined for 100 trials. Practice blocks of 200 trials were repeated in the trained location until thresholds leveled off. Untrained and trained normalized pre- and post-learning threshold ratios were compared. Stereothresholds and standard errors decreased systematically with training. Rapid learning occurred within 2000 trials, leveling off thereafter. Pre- and post- learning threshold ratios in trained and untrained locations were not statistically significant (Paired T-Test, P=0.4), excluding a location-specific learning effect. The absence of a location-specific PL effect implies that neural tuning does not underlie the learning. Because previous findings excluded internal noise reduction, the improvement in depth discrimination is most likely due to an improvement in task performance at higher levels of visual processing.
Supported by NIH grants R01 EY01139, R01 EY12810 and P30 EY07751.