Abstract
Visual working memory (VWM) is a limited resource, but a prolific mechanism for subverting capacity limitations is chunking (Miller, 1959). For example, when colours within a large display are reliably paired, participants show better memory over time, as they chunk the consistent pairings (Brady et al., 2009). However, whether the benefits in VWM reflect the formation of long-term memories of the chunks (Huang & Awh, 2018), or compression of the VWM representation itself (Brady et al., 2009) is debated. Here we investigated if chunks are vulnerable to proactive interference from competing representations in long-term memory. We manipulated previous designs to include colour pairs that are both consistent in orientation (red left, blue right), and colour pairs that swap in orientation (red left, blue right or blue left, red right). A binomial mixed effects model revealed that consistent chunks were remembered better than chunks with swapping orientations over the course of the experiment despite equal frequency of all orientations (β = 0.177, SE = 0.034, z = 5.079, p < 0.001). An additional binomial mixed effects model on error trials investigated if participants were biased to select the target colour’s chunk pair (lure) over other foils. We found a significant interaction of lure selection; when consistent chunks were probed, participants were less likely to choose lures over the course of the experiment, but they were more likely to choose lures for swapping chunks with each passing block (β = -0.122, SE = 0.043, z = -2.836, p < 0.005). Taken together, our study provides evidence that chunks are non-independent representations susceptible to compounding proactive interference from previous exposures. This lends indirect support to the role of long-term memory mechanisms in the process of VWM chunking.