Abstract
Recent investigations have shown that externally-adjudicated rewards can modulate selection processes both within and between trials. In particular, when participants recently received a reward or a penalty (on the previous trial) or can expect to receive a reward or a penalty in the current trial (based on learned reward-feature contingencies), rewards can strongly guide attention, biasing selection mechanisms towards highly-rewarded information and away from penalty-inducing information. These effects are observed after participants have had extensive experience with the task and with the reward-feature contingencies (how much each feature is typically worth). Here, we investigated whether reward-based effects on attention can be induced on a trial-by-trial basis (i.e., without a consistent association between a level of reward and a specific visual feature). We used the Distractor Previewing Effect, an inter-trial bias of selective attention that is observed in oddball-search tasks: participants are slower to select an oddball target when its defining feature was shared by all distractors on a preceding target-absent trial, and are faster when distractors share a feature with the distractors on a preceding target-absent trial. Previously, we have shown that learned rewards strongly modulate the DPE. When penalties are associated with the color of distractors on a target-absent trial, the ensuing DPE is exaggerated, whereas when high levels of rewards are associated with the color of distractors on the target-absent trial, the ensuing DPE is reversed, showing now an attentional preference to select normally inhibited information. Furthermore, we observed strong within-trial biases such that items defined by rewarded features were preferentially selected, and items defined by penalized features were efficiently rejected. Our current results show that these reward-induced modulations of attention are totally absent when reward levels are randomly assigned to features on trial-by-trial basis. Under conditions of reward-uncertainty, attention ignores rewards, presumably because previous rewarding experiences fail to predict future rewards.
National Science Foundation grant to AL, award number BCS 07-46586 CAR.