Abstract
Aim: Can we hold more familiar items than novel items in visual working memory (VWM)? Previous studies suggest not: familiar polygons are not remembered more efficiently than novel ones and upright letters are not remembered better than inverted letters. In this study we report an exception to this rule when faces of different familiarity were tested. Methods: Our study was revised from Buttle & Raymond (2003) who tested perception of briefly presented and masked faces. To study VWM rather than perception, we presented faces for a long duration (500ms or several seconds) and inserted a long retention interval (1000ms) in a change detection task. VWM for three types of faces was tested: superfamiliar faces of celebrities or one's friends, moderately familiar faces that had been viewed 160 times before, and novel faces. Subjects held 4 faces in VWM while performing an articulatory suppression task. Results: VWM for superfamiliar upright faces (celebrities or friends) was significantly better than that for novel upright faces, but this enhancement was not observed for recently trained upright faces and was eliminated by the inversion of faces. These results extend the superfamiliarity effect observed by Buttle & Raymond (2003) from a perceptual task to a VWM task. Conclusion: Our study suggests VWM can be influenced by long-term visual memory, but only if the set of items that must be distinguished in VWM are subtly different and that subjects have achieved superfamiliarity with the stimuli.
This study was supported by NSF 0345525 and Harvard College Research Program