Predictive spatial working memory content guides visual search

Abstract

In visual search tasks, repeating spatial contexts that are predictive of target location facilitate detection (Chun & Jiang, 1998). In past studies, the predicted spatial configurations appeared concurrently with target information. Here we examined whether repeatedly presented working memory (WM) arrays could also serve as contextual cues. Participants performed visual search while maintaining a WM array presented at the beginning of each trial. In the learning phase, each WM array was paired with a specific target location. The paired displays were repeated throughout the entire learning session. In the test phase, half of the pairings remained constant (old condition); the other half switched to new, unpaired locations (new condition). If participants learn the associations between the representations maintained in WM and target locations in the search displays, then search performance should be better in the old condition than in the new condition. In Experiment 1, four colour patches were used as WM stimuli and paired with specific target locations in the subsequent visual search display. These predictive colour WM representations in the old condition did not improve search performance. In Experiment 2, four locations were used as WM stimuli and paired with specific target locations in the visual search. These predictive spatial WM representations improved search performance. Chance performance in a subsequent recognition test revealed that the learned associations were implicit. Experiment 3 demonstrated that spatial WM displays did not serve as contextual cues when the spatial representations were not held in WM. These results indicate that “contextual cueing” can occur even when the predictive context is not concurrent with the target. Implicit contextual cueing of attention by WM was specific to spatial WM in the present study.

Keywords:

• Attention
• Working memory
• Visual search
• Contextual learning

Acknowledgements

This work was supported by Ministry of Science and Technology of the Republic of Korea 21st Century Frontier Research Program Brain Research Centre Grant (M103KV010018-07K2201-01810). MMC was supported in part by NIH NEI EY014193. We thank Keetaek Kham, Jongsoo Baek, and Do-Joon Yi for their giving helpful comments on the experiments and Yael Yungster for her assistance in editing the manuscript.

Notes

¹We thank an anonymous reviewer for raising this alternative.

²We thank James R. Brockmole for raising this point of view.