Abstract
Protecting information in working memory (WM) from potential distraction is essential for efficiently bridging between perception and subsequent behavior. During natural behavior, we are likely to encounter many forms of distraction after an item is encoded. Compared to the lab, we can also choose when to move away from relevant items (i.e., stop encoding) and when to look back to objects of interest rather than relying on their memory representation. In such cases, reliance on WM is much lower than predicted based on WM capacity measures in typical laboratory tasks. Here, we investigated whether the amount of distraction during the period between encoding information and using it contributes to this under-reliance on WM content. We tracked head, hand, and eye movements in virtual reality during an adapted object-copying task where participants copied a model display by selecting realistic objects from a resource pool and placing them into a workspace. We derived an implicit measure of the tradeoff between reliance on WM and gathering information from the external world during natural behavior. Critically, in a “low distraction” condition, we reduced the distractibility of objects in the resource pool by increasing their transparency to investigate whether and how natural WM encoding and usage changed as a function of distractibility. Reducing distractibility reduced the time participants spent searching the object they had encoded, errors, and the time required to copy the objects. Limiting distractibility also increased reliance on WM. Nevertheless, even in the less distracting condition, WM usage remained low (~2 features). When distractibility was high, participants spent more time encoding each object—reflecting a change in encoding strategy in response to anticipated environmental distraction. This ability to regulate the encoding process and the subsequent usage of perceptual vs. memory information showcases the impressive flexibility of adaptive behavior when trading-off perception and memory.