Abstract
The deployment of attention can be biased by temporal regularities in the environment (Zhao et al., 2013) and representations stored in visual working memory (VWM) (Moorselar et al., 2014). Similarly, regularities in the environment can also facilitate effective storage in VWM, meaning items can be stored more effectively due to chunking (Brady et al., 2009). We investigated how VWM representations of chunked and un-chunked colour pairs impact performance on a visual search task. Participants were presented with VWM displays consisting of three pairs of colours and asked to remember all the colours until presented with a four alternative forced choice for what colour was just presented at the cued location. Colour pairs were selected each trial using a joint-probability matrix, such that each participant was assigned four high-probability pairs to chunk. High-probability pairs were 80 times more likely to be presented than any low-probability counterpart. During the delay period of the VWM task, participants saw a visual search display and were asked to find the coloured diamond with a chip on the top or bottom, among five coloured distractors with chips on the left or right. Across two experiments, evidence of VWM chunking was observed through improvements in VWM accuracy for participants who post-hoc reported being aware of the high-probability colour pairings (F(1, 29) = 12.56, p < 0.001). These chunks, however, did not significantly guide attention during the intervening search task (F(1, 29) = 0.44, p = 0.51). Our experiments demonstrate that under VWM load, chunks exert no significant bias on visual search, despite sharing the qualities of regularity and maintenance in VWM. This provides indirect evidence that the benefits of VWM chunking may be subserved by long-term memory processes.