Abstract
We use visual working memory to temporarily store visual information about our environment. However, our environment is mostly visually static and as such, memory can often be ‘offloaded’ onto the environment. As a result, there is a trade-off between choosing to internally store information or to externally sample it. In the present study, we explored how this trade-off changes as reliability of access to the environment changes, by submitting participants to a copying task, in which they copied a layout of stimuli on the left side of a computer screen to the right side of the screen. The example layout intermittently disappeared throughout a trial, the timing of which was varied across conditions. As the example layout disappeared for greater amounts of time, participants sampled it less often (and thus memorised more items at once) than in the baseline condition, in which the example was always visible. We then designed and ran a computational cognitive model. This model explored different combinations of strategies regarding the number of stimuli it attempted to remember with each gaze toward the example layout, and regarding how often a stimulus was rehearsed in memory after its first encoding. We then compared human data and model data on three outcome variables: (1) the number of crossings from the right side of the screen to the example layout on the left side; (2) the completion time of trials; and (3) the number of fixations per second. We present a model which approximates human data fairly accurately, and show that incorporating specific memory strategies caused models to fit more strongly than when they were disregarded. Our findings reveal a clear shift in the usage of visual working memory when reliability of visual access changes, and validate the storage/sampling trade-off in environments with varying reliability of visual access.