Abstract
Recognition costs incurred by rotated objects have usually been interpreted as evidence for viewpoint-dependent recognition. However, recent findings from repetition blindness and attentional blink experiments (Harris & Dux, in press; Dux & Harris, VSS 2004) suggest that initial recognition is actually mediated by orientation-invariant representations and that the orientation effects arise during a later stage of processing which enables conscious recognition and report. We detail a two-stage framework of familiar object recognition which accounts for these results and other findings in the object recognition literature. In stage 1, the object's identity is recovered from memory via an orientation-invariant representation. However, before the object can be consolidated, a second attention-demanding stage of processing is required, in which the object's identity and its orientation at a particular moment are integrated, in order to give rise to a conscious percept anchored in space and time. According to the model, orientation effects on recognition arise during the second stage, because of a discrepancy between the spatial orientation of the stimulus and the expected orientation retrieved from memory. This discrepancy has to be resolved before an episodic representation of the object can be consolidated and reported.
Funded by ARC grant DP0211342