Abstract
Although it has been reported that visual short-term memory could hold representations of 3–5 feature-bound objects, Wheeler and Treisman (2002) showed that change detection performance in retaining binding information was significantly worse than the performance in retaining visual feature information. They attributed this binding memory deficit to retrieval processes based on the significant improvement of performance by using a single probe display. However, change detection tasks may be insufficient to evaluate feature-binding because participants can use other types of memory representations. This study examined whether binding memory deficit reflects memory retrieval or maintenance, using tasks to evaluate binding memory more strictly; type identification task and relevant-feature switch detection task. As in typical change detection tasks, sample and test displays were presented with a blank period in between. The stimuli were objects defined by shape and color and during delay period, properties of two objects, either both shape and color, shape alone, or color alone, could switch. To investigate the effect of retrieval, single- and whole-probe trials were randomly mixed within experimental blocks. First experiment with the standard change detection showed significant performance improvement in the single probe condition, replicating Wheeler and Treisman. In the type identification task, participants identified the switch type (shape, color, shape-and-color, and no-switch), and in the relevant-feature switch detection task, participants were asked to detect switches involving shape, and switches involving color in different sessions. Both tasks need discrimination of different switch types, not only detecting the presence of any change. If the binding memory deficit reflects retrieval processes, then these tasks should also show the single-probe advantage. Results showed no single-probe advantage in these two tasks, suggesting that binding memory deficit reflects maintenance, not retrieval. Previous findings using change detection tasks may reflect memory system unrelated to feature binding.
Support from 21st Century COE (D-10 to Kyoto Univ.), and PRESTO from JST.