Abstract
Performance on perceptual tasks improves with practice. However, contrast discrimination thresholds show a remarkable stability when a large range of contrasts (0–0.6) is practiced. There are two known exceptions: (a) when the practiced target is surrounded by flankers (Adini et al, Nature 415, 790–793, 2002), (b) when practicing with a single base contrast (Yu et al, VSS 2002). The improvement can be explained by increasing the gain of contrast transduction and/or by optimization of discrimination strategies applied to the specific contrast level(s) used during practice. To separate between the two accounts we measured contrast discrimination thresholds before and after learning in conditions where the observers could not predict the target contrast (contrast uncertainty). Learning effects based on plastic changes in the basic sensory mechanism, but not on contrast specific strategies, are expected to survive such an experimental manipulation. The pre-learning tests (using Gabor signals) showed the expected stable performance with typical threshold vs contrast functions for both the certain and the uncertain contrast conditions. Next, observers were trained with contrast discrimination using a constant base (pedestal) contrast (0.5). Discrimination thresholds were almost halved during practice. However, this improvement was found to be specific to the trained condition and post-training tests with contrast uncertainty showed no improvement. A second group of observers practiced the full contrast range with the target embedded in a chain of flankers, showing the expected improvement in contrast discrimination. This learning effect was found to transfer to the post-learning test with contrast uncertainty. The results imply that contrast transduction is modified when contrast discrimination is practiced with flanked targets. Without flankers, learning may involve improvement of decision strategies, depending on the information available to the observer.