Abstract
Earlier work (Sinha, Heindel and Welch, 2000) has shown that learning a simple category (nine-dot configuration) was possible in the near periphery but did not transfer to other retinotopic locations. This was surprising because category learning had been considered a cognitive task, not a perceptual one. But if the category learning paradigm were included in our understanding of perceptual learning then the results are less surprising. In our experiments, we examined whether using the same task but a different visual stimulus would also show no learning transfer. Our category consisted of line drawings of a novel 3D object, comprising a four-sided pyramid with three curved cylinders attached to its sides. Category and non-category shapes were defined by the cylinders' locations along the pyramid's sides A category learning task has two phases: a training phase when subjects are familiarized with several category members, and a testing phase when subjects are asked to distinguish between category members and non-members. We trained subjects in one retinal location and tested them in three locations, including the training location. Performance in all locations was indistinguishable. Category learning transferred to all locations tested. This indicated that the category learning paradigm alone could not dictate what level of visual processing actually occurred; stimulus characteristics could influence depth of processing. In this case, the pyramids and dot-configurations were different on many dimensions and a question for future work is what are the important stimulus differences that result in more or less learning transfer?