Abstract
The present study investigated object-based feature encoding in visual short-term memory (VSTM) for two features on the same dimension (two colors or two orientations) when they occur on different parts of the same object. In Experiment 1, mushroom-like objects were constructed. For each mushroom-like object, the cap carried one color and the stem carried a different color; the stem and the cap were either attached (forming a bicolored object), or detached (forming two separate objects). Using the change detection paradigm, in different blocks, participants either monitored the cap color, the stem color, or both the cap and the stem colors. A change in color occurred in 50% of the trials and participants had to detect this change. Performance dropped when both the cap and the stem colors had to be monitored compared to when only the cap or only the stem color had to be monitored. However, this drop in change detection performance did not differ whether caps and stems were attached or detached. In other words, there was no object-based advantage for the bicolored mushroom-like objects in VSTM. In Experiment 2, the same results were obtained for objects with two orientations. In Experiment 3, with the same mushroom-like objects, when participants were asked to monitor the color of the caps and the orientation of the stems, a significant object-based advantage was observed such that change detection performance was better when the cap and the stem were attached than detached. The present results thus impose a major constraint on object-based feature encoding theories by showing that only features from different dimensions can benefit from such object-based encoding in VSTM.