The effects of shadow presence on visual object-recognition

Abstract

Shadows provide valuable cues for many aspects of visual perception. This thesis discusses the definitions of cast and attached shadows, and the different types of shadow borders that exist. Eight experiments investigated whether the presence of shadows affects the speed or accuracy of human object recognition performance. Experiments 1 to 4 investigated the contributions of attached shadows to the recognition of novel objects, using a sequential-matching task modelled on that of Tarr, Kersten, and Biilthoff (1998). Their finding, of faster reaction times associated with the presence of shadows, was not replicated. Reaction times were not affected by the presence or absence of shadows. Across the four experiments, discrimination was either unaffected by shadow presence, or was at its highest when there were no shadows present. In Experiments 5 to 7, the effects of cast shadow presence on object recognition were assessed. Visual cues about the shape of the objects were constrained by manipulating the degree of foreshortening of both the objects, and the shadows cast by the objects. Shadow presence was only of benefit to recognition in highly constrained situations: where the objects were severely foreshortened, while their cast shadows were not. Experiment 8 assessed the affect upon recognition of manipulating shape-from-shading cues independently from shadow-border cues. Shadow presence was only beneficial where shading was negligible. It is suggested that shadow presence may only provide observable benefits to object recognition when other cues to an object's identity, such as bounding contour and shape-from-shading, are minimal. It appears that shadows have the potential to facilitate object recognition, but in most situations their presence will not produce any discriminatory, or reaction time, benefit.