Abstract
Stimuli previously associated with reward slow response times (RTs) when presented as irrelevant distractors in subsequent, unrewarded tasks (value driven attentional capture, VDAC). Typical VDAC training requires search for one of two experimentally-determined, colored circles and an orientation judgement of a line inside the color-defined target. Reward follows correct responses, associating high- or low-value with specific colors. Distractors rendered in high-value colors consistently slow RTs in an unrewarded test phase, an outcome that is rarely observed for low-value colors. Might this be due to a failure to adequately learn the reward contingencies during training? 22 observers underwent a modified training phase. On each trial, two objects were presented. Each object was comprised of distinct features: color, shape, and internal line orientation. Participants chose one object and received high, low, or no reward. Only four colors appeared and two were consistently paired (high- or low-value and a no-value match). The task was to maximize earnings by learning which specific feature predicted reward. Training was followed by the standard VDAC test phase. During training, each value stimulus was chosen significantly more often than its no-value match, confirming learning for both high- and low-value colors. However, only high-value colors engendered VDAC during test, as is typical. Using maximum likelihood estimation, individual RT distributions were fit with a three parameter, exponentially modified Gaussian function and the condition means of the resultant distributions were compared, converging with results from model-free analyses. Stimuli associated with low reward consistently fail to generate VDAC. Our results rule out the possibility that this is due to a failure to learn, as participants developed clear preferences for both low- and high-value colors during training. More research is needed to explain how reward learning interacts with other aspects of cognition to produce robust capture effects.