Abstract
People can only store a limited amount of information in visual working memory. What is the nature of this limit, and does its expression depend on task demands or is it fixed? Here we show evidence that participants have control over both the quantity and fidelity of items stored in working memory. Participants were briefly shown five colored circles and asked to remember the colors over a short delay, and then responded to one randomly probed item by adjusting the probed item’s color using a continuous color wheel. This task affords a measure of the quality and quantity of stored representations (Zhang & Luck, 2008). Consistent with past work we found that participants could store 3.9 objects and that the precision of an object in memory was 22.0° (standard deviation of error over color space). In another condition, we altered the task demands: Participants made a response to all five items and the trial was only considered correct if all responses were within ±90° of the correct answer. In this task, knowing something about all items, even at a poor resolution, is optimal for performance. If strategic manipulation of working memory contents is possible, we expect a tradeoff in performance—participants will store more items but items will be stored less precisely. The results show exactly this; participants stored more items (4.4 vs. 3.9, p<0.005) but at worse precision (26.3 vs. 22.0, p<0.005). This was true even when participants were not cued whether to report one or all items until after stimulus offset, demonstrating that these tradeoffs cannot be attributed to differences during encoding. These results suggest that observers can strategically tradeoff memory quantity and quality, and that capacity limits measured in standard working memory tasks should be taken to reflect a common strategy, rather than fixed limits on working memory.
Meeting abstract presented at VSS 2013