Visual working memory precision is under voluntary control

Visual working memory allows us to guide behavior based on recently seen information. Most research builds on the assumption that people always try to store visual information as precisely as possible, regardless of what the upcoming action requires. In everyday life, however, different (memory-guided) actions demand different levels of precision. Here we asked whether people flexibly adjust how deeply they encode, how much effort they put in maintaining visual information based on the task demands. Across three experiments participants memorized up to four colors, while we manipulated the required precision for a delayed report to be considered correct. We found that when required precision was higher, reports became slower and (in some cases) more precise. Crucially, we found that higher precision demands led to (1) stronger pupil constriction during encoding, indicating deeper encoding, and (2) larger pupil dilation during maintenance, indicating greater cognitive effort. Moreover, this adaptive increase in maintenance effort occurred not only when precision demands were known before encoding (Experiments 1 and 2), but even when they were revealed only after encoding (Experiment 3). Thus, the maintenance of visual information can be flexibly tuned in response to changing task demands. Yet, this flexibility had limitations: with increasing precision requirements, participants increased effort even when it no longer improved performance. Together, these findings show that visual working memory is not a passive store that is always maximally loaded. Instead, how visual information is encoded, maintained, and used is tuned to the demands of upcoming behavior.