The neurobiology of object constancy
`Object constancy' is the name given to the brain's ability to overcome the myriad environmental obstacles to visual perception and produce a stable, consistent internal representation of object shape. Changes in object orientation represent one such confound. It can be inferred from the time taken to recognise misoriented objects that we encode specific object views based on our experience of those objects and their typical orientations ('viewpoint-dependent recognition'). Such studies also suggest that we may recognise certain objects in a manner that is not dependent on their orientation ('viewpoint-invariant recognition'). Further studies indicate that the time to resolve two angularly disparate shapes (`mental rotation') increases as a function of their angular disparity. It is hypothesised, based on these findings, that viewpoint-dependent recognition and mental rotation share a common mechanism for transforming the global stimulus percept into alignment, but that viewpointinvariant recognition is achieved by some other, non-transformational means. This thesis presents studies that examine the cortical correlates of viewpoint-dependent and viewpointinvariant object recognition using novel objects to eliminate the confounding effects of prior experience. It also presents a study that directly compares the cortical correlates of mental rotation, viewpoint-dependent and viewpoint-invariant recognition. Further comparison of these object constancy processes is then made using electrophysiological markers of visuospatial transformation. The findings of these studies indicate that viewpoint-dependent recognition and mental rotation recruit a bilateral parietal-premotor network for the manipulation of global stimulus percepts, hypothesised to be the same mechanism as that used for physical object manipulation and prehension. Viewpoint-invariant recognition does not appear to recruit such a mechanism, and this process appears to be less expensive in terms of cognitive resources than transformational object constancy mechanisms. Thus, implementation of a viewpoint-invariant mechanism to recognise misoriented objects is preferable, but may not be possible where stimulus features are few or ambiguous. In recognising misoriented objects, viewpoint-dependent and viewpoint-invariant mechanisms initially proceed in parallel, but successful recognition of object invariant features may be sufficient to terminate the viewpoint-dependent mechanism.