Mechanisms of perceptual learning for vernier acuity

Abstract

Purpose. To understand the neural mechanisms underlying perceptual learning in Vernier acuity. Methods. Our training protocol was divided into three phases: namely pre-training, training and post-training. A three-alternative forced-choice interleaved staircase paradigm was used to measure vernier acuity in noise in ten normal adult observers. The Vernier stimulus was comprised of two segments with a 12 minute gap between the two segments. Each segment consisted of eight discrete Gabor patches (carrier SF, 10 cpd). Positional noise was produced by distributing each Gabor patch around the mean line position according to a Gaussian probability function. Observers were tested and trained at five noise levels (including zero). Vernier threshold was defined as the offset at which 66% correct responses were obtained. Trial-by-trial feedback was provided. Each observer had given more than 10000 responses at the end of the experiment. Results. Vernier acuity gradually improved across training sessions. With no noise, it improved, on average, by about 35%. Vernier acuity also gradually improved across training sessions at all spatial noise settings. We used several models (including the LAM model and the PTM model) to fit the data. All of the models were consistent in showing no statistically significant changes in additive internal noise across training sessions, but a significant increase in efficiency. Our observers showed no significant transfer of learning either across eyes or orientations. Conclusions. Vernier performance in noise can be improved after repetitive training. The improvement in Vernier acuity with practice can be mainly attributed to the improved efficiency. In our laboratory, we are currently studying the mechanisms of perceptual learning in amblyopes.