Abstract
A prevalent view of working memory is that visual information is actively maintained in the form of perceptually integrated objects. Such a reliance on object-based representations would lead one to expect that after an object is fully encoded into working memory, all features of that object would need to be maintained as a coherent unit. Here, we evaluated the predictions of this object-based account by testing whether observers can prioritize specific features of an object, after that object is fully encoded in working memory. On each trial, participants were presented with two sample gratings that varied in color and orientation (700 ms), followed by a mask array (100 ms) after a 200-ms interval. Both features were equally relevant at this encoding stage. Critically, at 500 ms into the retention period, participants either received a feature cue that indicated with 80% validity which feature dimension was most likely to be subsequently probed, or a neutral cue that provided no information, requiring maintenance of all the encoded features. Following the cue onset, participants maintained the features for a variable duration (2-4s) until they were asked to report either color or orientation of one of the items by adjusting the feature appearance of a probe stimulus. We used a mixture model (Zhang & Luck, 2008) to separately estimate the number and precision of successfully remembered features. Compared with the neutral cue, the feature cue led to improved memory precision for the cued feature, as well as a rapid increase of memory failure for the uncued feature. Our results demonstrate that working memory resources can be redeployed across features of objects during maintenance, conferring enhanced precision and temporal stability to the prioritized features.
Meeting abstract presented at VSS 2015