The driving factors of compressed semantic representations in visual working memory

Visual working memory (VWM) has a notably limited capacity of 3-4 items. Recent studies have suggested that this capacity can be larger for semantically meaningful information. However, whether the shared semantics across objects has additional benefits on VWM capacity remains controversial. While some studies find that same-category objects facilitate VWM performance, suggesting efficient coding, or compression in VWM, others cast doubt on the compression hypothesis. To contribute to resolving this controversy, we will test a series of factors that differ across studies that support compression and those that do not. These factors are the availability of WM resources and the difficulty of extracting semantic congruency, as determined by memory load of semantically congruent vs. incongruent items, respectively (Experiment 1), the expectations regarding the benefits of summary statistics, as determined by the probe type being single-probe or whole-display report (Experiment 2), the contribution of perceptual vs. conceptual similarity, as determined by shared semantics being at the basic-level vs. high-level (Experiment 3), and the duration of the encoding display (Experiment 4). We will present participants with a change detection task where participants will be given a probe and will be asked if the probe items match the memory items. Larger d’ for congruent objects as opposed to incongruent ones will be used as an index of semantic compression. By exploring the effects of conceptual knowledge on VWM storage, this study will develop a better understanding of VWM organization.