Abstract
Previous studies have shown that practice lowers identification thresholds for textures embedded in noise, and that such perceptual learning is stimulus-specific and long lasting. Here we ask whether better detection of the relevant signal is sufficient to improve identification, or whether experience in distinguishing the textures is necessary. In other words, is perceptual learning of texture identification task specific, or does it simply reflect improved stimulus detectability? Separate groups of subjects practiced texture detection or identification on Day 1; on Day 2 they performed either the same task as on Day 1, or transferred to the untrained task. Stimuli were 10 briefly (200 ms) presented band-limited noise textures embedded in static noise. Detection performance was measured with a Yes/No task, and identification performance was measured with a 10-AFC task. Texture contrast was varied across 7 levels using the method of constant stimuli; feedback was provided on each trial. We calculated d′ at each contrast level to compare psychometric functions across tasks and days. On Day 1, psychometric functions spanned the full performance range for both tasks. On Day 2, after practice with the same task, texture identification improved substantially at all contrasts, whereas texture detection improved only slightly. There was no transfer of learning from detection to identification, and some evidence for transfer in the opposite direction. Therefore, better detection of the stimulus is not sufficient to improve identification of noisy patterns. Learning requires telling the patterns apart - perceptual learning of texture identification is task-specific as well as stimulus-specific.
Natural Sciences and Engineering Research Council of Canada (NSERC).