Abstract
Recent studies have revealed that a spatial bias emerges toward a location where a target appears with high probability. However, it is not clear whether the spatial bias can distinguishably emerge when different context contains different probabilistic information. In the current study, participants searched for a target among multiple distractors. The context (i.e., the color of stimuli) predicted which quadrant is more likely to contain a target. For example, a black target appeared more frequently in one quadrant of the search display, and a white target appeared more frequently in another quadrant. If probability learning is context-specific, a spatial bias to different quadrants would emerge depending on a given context. In contrast, an equal spatial bias to two quadrants would emerge regardless of contexts, if probability learning is context-independent. In Experiment 1, when a search display contained only one context (either black or white stimuli), attention was equally biased to both target-frequent quadrants regardless of the context, showing context-independent spatial bias. This result was not due to lack of time to process context information (Experiment 2). However, context information was not critical in Experiments 1 and 2, because each search display only contained a single context. In Experiment 3, a search display contained both black and white stimuli and participants were pre-cued which color context they should use. Results showed that a significant context-specific spatial bias emerged as context became task-relevant. Context-specific spatial bias could have been the result of increased task difficulty as the set size doubled. However, context-specific spatial bias was still not found when search became difficult by modulating target-distractor similarity, suggesting that task relevance, not task difficulty, influences context-specific spatial bias. These results demonstrate that statistical knowledge can be distinguishably learned for different contexts, when the contexts are relevant to the task.
Meeting abstract presented at VSS 2018