Perceiving and recognising novel objects: detecting configural, switch and shape changes

The representation of the shape, arrangement and relations between object parts is fundamental to many viewpoint invariant or structural description theories of object recognition. View point dependent theories, on the other hand, do not have a parts-focussed approach and instead propose that objects are represented as a collection of 2D images. However, research on simple 2D objects suggests that configural information may be processed differently to shape information. Thus, the aim of the current thesis was a systematic investigation of how visual object property information is extracted, encoded and utilised. This was achieved by examining how the following types of information are utilised in detecting changes to 3D novel objects: (i) the global configuration of the parts of an object, (ii) the arrangement of those parts, and (iii) the identity or shape of an objects parts. Previous studies or visual object processing have typically employed tasks such as matching tasks, visual search and delayed recognition. The change detection paradigm, in which the objective is to detect changes to objects or scenes, has emerged relatively recently. The rationale behind this paradigm is that the types of changes detected reflect the information that is encoded by the visual system. As such, change detection tasks can be used as an effective tool for further investigating the processing and employment of visual object properties. The data reported in this thesis demonstrates that configural information is important for novel object perception and recognition, more so than the shape or relative arrangement of parts. Specifically, changes to the configuration of an objects parts were detected quicker and more accurately than changes to either the shape or arrangement of an objects parts. This pattern of results was obtained regardless of whether the object was 2D or 3D, whether attention was focussed on the location of the change or was distributed, or whether configural information was defined in terms of categorical or coordinate relations. The configural advantage was found using objects or different size, orientation, and complexity, and was found at all levels of object processing (eg, change detection, across object rotation in depth, and recognition). In conclusion, the current research suggests that perception and recognition of novel objects utilises similar types of object property information. Not only is configural information attended before part shape or relative location, but also, configural information is utilised quickly and accurately over a wide range of tasks. This consistency in the pattern of results suggests that configural information plays an important role in the processing of visual information in general. In particular, global configural information may be encoded quickly and used as a framework for further analysis of detail features such as shape.