Abstract
Binding of non-spatial features in visual working memory (VWM), and the role of location in binding, remain unknown. Empirical evidence for and against feature binding in VWM has been based on null results. Object file theory postulates that spatiotemporal location is critical in maintenance of feature binding in VWM, but no evidence has been provided for the binding of feature conjunction to location. The current study combined the redundancy gain paradigm and the object reviewing paradigm to investigate the relationship between color-shape binding and location. A set of features was presented in a two-object memory display, followed by a single object probe. Participants judged if the probe contained any features of the memory display, regardless of location. The probe contained two, one, or zero features in the memory display (redundant trials, single-feature trials, or new trials, respectively). Within the redundant trials, the features were either grouped or separated. To investigate the role of location, single-feature and redundant-grouped trials were further classified into location-shared and location-unshared trials The race model analysis examined whether RT advantage of redundant over single-feature trials is due to probability summation, and revealed object-based feature co-activation regardless of location sharing between probe and memory objects, suggesting location-unbound binding of color and shape. Location-sharing benefits were obtained for both redundant and single-feature conditions, indicating maintenance of object files independent of non-spatial feature binding. Amplitude modulations in N350 ERP component reflecting the feature co-activation had the peak in a frontal region, whereas those reflecting the location-sharing benefit had the peak in a posterior region, suggesting different neural mechanisms for the maintenance of color-shape binding and object files. In a location-irrelevant task, color-shape binding is maintained without binding to location, inconsistent with object file theory. These novel findings suggest a parallel maintenance of type and token VWM representations.
Meeting abstract presented at VSS 2015