Abstract
Category learning research typically teaches participants categories, with no consequence for how items on each side of the boundary should be treated. In real life, however, we learn categories that predict further processing goals, which can influence perceptual strategies. Using methods from category learning and perceptual expertise, we studied how people can learn object categories with different perceptual goals within the same feature space. In Experiment 1, participants first learned to categorize a two-dimensional morphspace of complex novel objects into two categories and then learned a different perceptual task, either individuation or local feature judgment, for each category. Later, participants were shown objects from both categories together and had to perform the category-appropriate task for each one. A same-different discrimination test before training and after each training phase measured changes in sensitivity to feature dimensions of the space. After categorization, we found the expected increase in sensitivity along the category-diagnostic dimension. After category-task training there was an increase in sensitivity on both sides of the space for both the category-diagnostic and category-non-diagnostic dimensions although there was evidence that some changes were not global but were applied more locally within the space. In Experiment 2, the same design was used with added emphasis on speeded responses during the category-task training to make categorization-for-perception more automatic. We found the same improvement along the category-diagnostic dimension after categorization and similar patterns of stretching in the feature space in both dimensions after category-task training. We also saw preliminary evidence of greater holistic processing for objects on the side of the space that participants learned to individuate. The results demonstrate the feasibility of studying a more ecological flavor of category learning in the laboratory, to provide a useful model of how continuous perceptual spaces could be carved into distinct domains where objects are perceived differently.
Meeting abstract presented at VSS 2012