Contributions of Familiarity and Chunking to Visual Working Memory Capacity

Visual working memory (VWM) is responsible for the temporary storage of visual information required for perception and cognition. The capacity of VWM is surprisingly limited to three or four items. Despite decades of research, the nature of the capacity limit is still unclear, in part due to uncertainty about the main factors contributing to this limit. We approached this issue by exploring two instances in which memory performance is enhanced. Firstly, while controlling stimulus complexity and similarity, familiarity produced significant increases in both encoding rate and capacity. However, familiarity gained from training observers to simply recognise the stimuli did not produce any benefits for change detection. Secondly, the inclusion of statistical regularities in the displays produced significantly improved recall. However, only subjects with explicit awareness of the statistical regularities showed improvement, whereas unaware subjects showed no change in their recall performance. We extended this result by observing whether contralateral delay activity (CDA), a neural marker of the number of item-based representations held in VWM, reduces with explicit chunking. Although recall performance was significantly better, the CDA did not appear to index equivalent number of chunks, suggesting that online representations do not change with the use of explicit chunking. Instead, the behavioural benefit appears to rely on retrieval of a long-term memory representation (LTM) when recall is tested. These results indicate a major influence of LTM in guiding VWM performance. Behavioural data collected at the end of the trial, such as change detection or probed recall, appear inadequate for fully examining the nature of VWM. An embedded-process framework, in which activated LTM representations can fluidly shift into the focus of attention, is useful in interpreting these results and understanding the cognitive processes involved in memory.