Abstract
Recent physiological studies show that while neurons in higher visual areas such as IT are highly experience dependent, neurons earlier in processing show much smaller changes in tuning with practice. We report here two studies examining how these differences in neural plasticity at different stages in the visual system correspond to improvements in performance with practice across different tasks. In the first study, we measured learning for spatial frequency discrimination using a simple plaid stimulus in which discrimination could be mediated by low level mechanisms tuned for both spatial frequency and orientation. Observers showed almost no learning. We then added noise components to the simple plaid stimulus to create a ‘wicker’ pattern that required the selective integration of information across a wide range of spatial frequencies and orientations. We found significant learning in this mid level task; over eight sessions of training, observers' thresholds dropped to a third of their initial values. In the second study we compared learning across a wider range of tasks using data compiled across sixteen published psychophysical experiments. The tasks in these experiments ranged from low level spatial frequency and orientation discrimination tasks to high level object and face recognition tasks. Since these experiments used a wide variety of performance measures, we converted the measured learning effects into a standardized learning index based on d′ before and after practice. Consistent with the first study, we found that the amount of learning varies widely between different tasks, with less learning in low level tasks than mid or high level tasks. We hypothesize that a variety of factors affect learning, including the number of perceptual dimensions relevant to the task, external noise, familiarity and task complexity.
Supported by NIH grants R01-EY13149 & EY01711, NSF grant SBR-9870897, & Burroughs Wellcome LJIS.