Abstract
Capture of visual attention by working memory representations has been a central topic in cognitive science since 2000. However, researchers still reached little consensus on several questions, including (1) the role of verbal encoding, (2) the potential strategic component in attention capture, (3) the effect of memory prioritization, (4) the maximum number of prioritized representations that can capture attention at the same time and (5) the role of mnemonic fidelity. Here, we use meta-analysis to comprehensively summarize the previous literature and try to provide insights on the questions. A total of 79 experiments (total sample size N = 1964) is included in the final meta-analysis. The search latency difference between the condition in which the search distractor is memory-matching and the condition in which the distractor is neutral has been extracted from each of the included experiments as the raw effects. We first observed a significant attention capture effect (d = 0.732, p < .001) for memory set size of one and two with high between-study heterogeneity (I2 = 77.8%) and a moderate publication bias (Rosenthal’s N = 3670). Based on subgroup analyses and multivariate meta-regressions, we found that deprioritized items can still capture attention, which is inconsistent with the hypothesized role of WM prioritization as the boundary condition of WM-driven attention capture. There is no effect of suppressing verbal encoding. However, the proportion of distractors in the search array and the validity of the memory cues can all independently predict the effect size of attention capture, which supports the strategic control of attention capture. Furthermore, the proportion of search distractors and the memory cue validity interact with memory item repetition. Based on the results, we propose that visual attention capture is a process driven by multiple interacting components.