Binding identity and orientation in object recognition

Abstract

We tested whether an object’s orientation is inherently bound to its identity in a holistic view-based representation at the early stages of visual identification, or whether identity and orientation are represented separately. Observers saw brief and masked stimulus sequences containing two rotated objects. They had to detect if a previously cued object was present in the sequence and report its orientation. In Experiments 1 and 2, the objects were presented sequentially in the same spatial location for 70 ms each, whereas in Experiments 3 and 4 they were presented simultaneously in different spatial locations for 70 ms and 140 ms, respectively. Across all experiments, observers reported the correct orientation for approximately 70% of the positively identified objects, and were at chance in reporting the orientation when they had not recognized the object. This finding suggests that orientation information is accessed after an object has been identified. In addition, when the two objects were presented sequentially in the same spatial location, orientation errors were not random—observers tended to report the orientation of the alternative object in the sequence, indicating misbindings between the identities and orientations of objects that share spatial location. This susceptibility to binding errors was not observed when the objects were in different spatial locations. These results suggest that identity and orientation may be prone to misbinding, and that spatial location may serve to protect their joint integrity.