Binding location and identity in naturalistic scenes boosts change detection when global configuration is preserved

Our research on visual short-term memory in naturalistic scenes found a boosted change detection for objects when both their spatial location and their semantic identity were jointly manipulated; an effect observed in healthy and pathological populations. However, in such studies, the global scene configuration was disrupted when objects changed location (i.e., empty space in their original position), leaving open a question about its contribution to detection. The current eye-tracking study tackled this issue by manipulating object changes on their identity, location or both features while preserving global scene configuration. Specifically, location changes involved the swap in the position of two objects. Detection was significantly better and faster when both features changed while performances were similar when only one feature changed. Thus, detecting changes for features conjunctions requires less attentional demands, as it can benefit from retrieving only one feature of the two. Additionally, during successful recognitions, the proportion of fixation time and the first-pass gaze duration on the critical object were greater when only its identity changed as compared to changes only in location. Indeed, identity change can only be detected by extracting objects’ semantic information, which requires a higher attentional effort than retrieving its spatial information relative to the scene context. In sum, we observed that objects’ location-identity binding boosts change detection for naturalistic scenes also when their global configuration is preserved. This outcome differs from studies using arrays of artificial objects, whereby feature conjunction results in a greater cost for short-term memory retrieval compared with individual features retrieval.