Feature As the Basic Storage Unit of Visual Working Memory

B. Wang,Z. Wang
DOI: https://doi.org/10.1167/14.10.38
2014-01-01
Journal of Vision
Abstract:Visual-working-memory (VWM) is crucial for complex cognitive tasks, such as learning and reasoning. Previous studies suggest that VWM stores integrated object, rather than independent features (Luck & Vogel, 1997). Memory accuracy is the same whether the participant needs to maintain one or all features from the same object. Several recent studies, however, have challenged this theory by demonstrating that features from the same object can be stored independently in VWM (Fougnie & Alvarez, 2011). Using change detection task (CDT), six experiments were conducted to resolve this controversy. Experiments 1-3 allocated two colors to either two or six to-be-remembered objects, whose identity was defined by color and shape (and/or location). The object-based theory predicts worse CDT performance when the number of to-be-remember objects was six. Experiments 1-3, however, consistently showed that CDT performance was unaffected by the number of to-be-remembered objects, even when possible perceptual grouping was precluded (Exp. 3). Experiments 4 and 5 further showed that this observation could be generalized to other feature dimensions (spatial frequency and orientation). One might suggest that this observation was obtained because only task-relevant features (e.g., color) were stored in VWM. To rule out this possibility, Experiment 6 was modified as following: a) Changes could happen to all object-defining features (color and orientation); and b) The to-be-remembered objects had two values from one feature dimension, whereas the number of feature values from the other dimension was the same as the number of the to-be-remembered objects. Replicating previous findings, the overall CDT performance declined as the to-be-membered objects increased from two to six. However, this decrease was mainly attributed to changes in the feature dimension with six values. This behavioral dissociation suggests that feature-binding did not happen and strongly support the theory that the units of storage in VWM are features, rather than objects. Meeting abstract presented at VSS 2014
What problem does this paper attempt to address?