Abstract
Fixation location during scene viewing can be predicted by salience or task relevance, with the most salient or relevant regions fixated first. To ensure that fixation does not perseverate on the most conspicuous region, the prevailing solution is a mechanism that decreases the likelihood of returning to previously fixated regions, i.e., inhibition-of-return (IOR). Previous work examining the generality of IOR indicates a dependency on viewing task, finding IOR for search tasks but facilitation-of-return (FOR) for memory, evaluation, and free-view tasks. If a mechanism acting as a short-term memory for recently fixated locations inhibits return for some tasks and facilitates return for others, then it might be a component of task-set (i.e., representation of task-relevant stimuli and responses and the corresponding stimulus-response rules) and, therefore, susceptible to influence from a previous task. To investigate this possibility, we examined whether task repetitions and switches influence IOR and scan paths. Participants performed search, memorization, and evaluation tasks in either blocked or mixed order. Analysis of saccadic reaction times in blocked trials to probes appearing either at previously fixated locations or at novel locations replicated evidence of IOR during scene search and FOR during scene memorization and evaluation. This pattern generalized to mixed trials when the task on the previous trial repeated. When the task on the previous trial differed, FOR was still observed during memorization and evaluation but was now also observed during search. Analysis of scan paths largely agreed with recent work in showing above chance return probabilities to previous fixation locations. An important exception, however, was for repeated scene search in mixed task trials. Here, return probabilities were at chance and there was a large spatial bias away from previous fixation locations. We conclude that the visual-saccadic system is flexible and depends on both the current and previous task.
Meeting abstract presented at VSS 2014