Object Long-Term Value and Novelty Create Incentive Salience Maps that Bias Eye Movements
33.502, Sunday, 18-May, 8:30 am - 12:30 pm, Pavilion
Ali Ghazizadeh1, Okihide Hikosaka1; 1Laboratory of Sensorimotor Research, National Eye Institute, National Institutes of Health
Quick detection of and orientation to objects with incentive salience are key for animal survival. Here we report that long-term reward history and novelty of objects strongly bias free viewing in the absence of any task instructions. Two adult male rhesus monkeys were trained with over 100 similar-looking fractal objects with which they had no prior experience. To create reward history, fractals were randomly associated with large or small reward in an object-directed saccade task over multiple days (good vs. bad objects). Long-term familiarity was established by free viewing or passive viewing of another set of fractals over multiple days with no reward association (neutral familiar objects). Test sessions consisted of trials in which animals freely viewed four random fractals in the absence of stimulus or behavioral contingent reward. In value block, fractals were randomly chosen from good or bad objects. In novelty block, they were randomly chosen from neutral familiar or completely novel objects. Upon display onset, majority of first evoked saccades were toward good or novel objects (p<.001). This indicates that objects long-term value and novelty are detected even at the periphery and evoke quick orienting responses (RT~200ms). Subjects viewed good or novel objects longer than bad or familiar objects (p<.001). Furthermore the rate of exploratory mini-saccades (<2.5 deg) within good or novel objects was significantly higher than within bad or familiar objects (p<.001). This shows that not only good and novel objects are salient (draw initial attention) but that they have incentive quality (are liked). Unlike good objects however, the novelty-driven incentive salience decreased over trials as objects became more familiar. Together these results show that the history of observer-object interactions creates incentive salience maps that guide saccades regardless of physical features and without any explicit targets.