Purchase this article with an account.
Adrian Garcia, Shu-Guang Kuai, Zoe Kourtzi; Distinct mechanisms for visual shape learning at different time scales. Journal of Vision 2011;11(11):1030. doi: https://doi.org/10.1167/11.11.1030.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Learning is known to facilitate our ability to perform complex perceptual tasks. Interestingly, behavioural improvements after training are observed not only after practice with highly similar stimuli that are difficult to discriminate, but also after exposure to dissimilar stimuli that can be easily discriminated. However, little is known about the mechanisms that mediate learning with difficult compared to easy stimuli. Here we investigate the time-course of learning when training involves highly similar vs. dissimilar visual shapes. Observers were asked to discriminate between radial and concentric Glass patterns, comprising oriented dipoles defined by dot pairs. We manipulated the similarity between these patterns by varying the angle of the local dipoles. Each observer participated in three sessions, each comprising test runs without feedback and training runs with feedback. Half of the observers were trained with highly similar stimuli that were difficult to discriminate (angles of 30° and 60°), while the rest of the participants were trained with easily discriminable stimuli (angles of 10° and 80°). All observers were tested with the highly similar stimuli (angles of 30° and 60°). Our results showed a distinct time-course for learning with difficult compared to easy stimuli. In particular, when observers trained with difficult stimuli, we observed long-term behavioural improvement across sessions, but no significant improvement within session. In contrast, when observers trained with easy stimuli, we observed short-term training improvement within the first session but no further significant improvement in subsequent sessions. These results suggest that training with difficult stimuli entails extensive practice that may optimize bottom-up feature learning mechanisms, while training with easy stimuli is achieved at much shorter time scales and may relate to top-down facilitation through feedback mechanisms. Thus, our findings propose dissociable mechanisms for long- vs. short-term shape learning that depend on the context of the training task.
This PDF is available to Subscribers Only