Abstract
It is known that subjects can exert voluntary control over what is encoded into working memory. Does implicit knowledge also influence what is encoded into this online workspace? To examine this question, we measured subjects' ability to detect changes in an array of colored squares, following a brief delay period that required the items to be maintained in working memory. Unbeknownst to subjects, one quadrant of the display (dominant quadrant) was more likely than the others (nondominant quadrants) to contain the changed item. Change detection accuracy was significantly higher in the dominant quadrant, suggesting that items from this quadrant were more likely to be encoded into working memory. Across four separate experiments, subjects were not significantly better than chance (6/52) at identifying the dominant quadrant. Moreover, those subjects that did identify the correct quadrant did not show a larger bias towards the dominant quadrant. This bias effect was not due to a reduction in the response threshold for indicating changes in the dominant quadrant (Experiment 2) or a speed-accuracy trade-off in the nondominant quadrants. In another experiment, we examined whether mnemonic resolution was also improved in the dominant quadrant (Experiment 3). We used a cross-category/within-category change detection paradigm that enables separate measures of the number and the resolution of the representations in working memory. This experiment also showed that a larger number of items were encoded from the dominant quadrant, but the resolution of these representations was unaffected. Thus, implicit knowledge influences which items are encoded into working memory, but not the clarity with which those items are represented. These results suggest that the encoding of items into working memory is influenced by implicit knowledge of which locations are likely to contain useful information.