Abstract
Experience routinely associates natural images with value. For instance, we may associate views of restaurants with bad or good dining experiences. Do such associations affect visual representations? To answer this question, we alternately trained participants to associate visual scenes with high or low reward, and had them categorize the same images as their brains were scanned using fMRI. On every trial of a choice task, they saw an indoor and outdoor scene, and were asked to choose between them. Choices of one image led to a higher monetary gain than choices of the other. These pairs were repeated across blocks, so that learning could lead to higher reward. Between choice task blocks, participants completed a categorization task. They saw the same images, plus novel images, randomly interleaved, and categorized them as ‘indoors’ or ‘outdoors.’ Analysis focused on categorization, when reward association was not relevant to the task. Categorization-phase results showed priming (shorter reaction times compared to novel images) for the first appearance of high-reward, but not low-reward images. By the third repetition, there were no differences in priming between high- and low-reward scenes. In fMRI analyses, we functionally localized the parahippocampal place area (PPA). The PPA showed lower activation (repetition attenuation, a measure of learning) for the first appearance of high-reward images in the categorization phase compared to novel images. In contrast, the low-reward images did not show significant repetition attenuation on the first categorization-phase appearance. Thus, basic visual representations are modified by reward associations, even when the reward is not relevant to the task at hand. Further analyses will focus on the effect of the number of appearances and strength of reward associations on visual representation.
Supported by NIH grant EY014193 to MMC.