Changes of Mind in Visual Working Memory Retrieval

Poster Presentation 16.304: Friday, May 15, 2026, 3:45 – 6:00 pm, Banyan Breezeway
Session: Visual Working Memory: Performance, influences

Adam Triabhall1, Ivan Tomic2, Paul Bays1; 1University of Cambridge, 2University of Zagreb

Storage processes in visual working memory (VWM) have been extensively investigated, but retrieval mechanisms are less well understood. Previous work has drawn a parallel between VWM retrieval and perceptual decision making, hypothesizing that both are underpinned by sequential accumulation of evidence towards decision bounds. In perceptual tasks, strong evidence for accumulation-to-bound has come from "changes of mind", dynamic reversals of initial decisions explained by continuing accumulation of evidence after response initiation. To test for the existence of changes of mind in retrieval from VWM, we conducted four experiments using a modified cued-recall design in both online and laboratory settings. On each trial, participants were shown an array of visual objects and subsequently required to report the location of one target object – identified by a cue matching one or more of its non-spatial features – via a rapid movement to its remembered location within a circular response space. In different experiments we manipulated speed pressure, set size, and cue feature dimension(s). We developed a novel method for detecting and quantifying changes of mind based on movement trajectories within a circular response boundary, revealing their consistent presence at low frequencies in all experiments. Spontaneous response updating occurred predominantly between locations of different objects that had been present in the memory array, consistent with continuing accumulation of evidence regarding cue features after response movement initiation. Their prevalence increased with manipulations predicted to decrease memory evidence strength, including set size. Importantly, changes of direction away from correctly identified targets toward incorrect locations were less frequent than the converse, suggesting an adaptive mechanism that supports accurate retrieval. To our knowledge, these findings provide the first evidence for changes of mind in VWM, identifying accumulation-to-bound as a fundamental component of the retrieval mechanism, and establishing a new source of behavioral evidence for investigating working memory.

Acknowledgements: BBSRC; Gonville & Caius College