Stewart, N. (2001). Perceptual categorization. Unpublished doctoral dissertation, University of Warwick, Coventry, England.

The categorization of external stimuli lies at the heart of cognitive science. Existing models of perceptual categorization assume (a) information about the absolute magnitude of a stimulus is used in the categorization decision, and (b) the representation of a stimulus does not change with experience. The three experimental programs presented here challenge these two assumptions. The experiments in Chapter 2 demonstrate that existing models of categorization are unable to predict the classification of items intermediate between two categories. Chapter 3 provides empirical evidence that categorization responses are heavily influenced by the immediately preceding context, consistent with evidence from absolute identification showing people have very poor access to absolute magnitude information. A memory and contrast model is presented where each categorization decision is based on the perceived difference between the current stimulus and immediately preceding stimuli. This model is shown to account for the data from Chapters 2 and 3. Chapter 4 explores the claim that new features may be created on experience with novel stimuli, and that these features serve to alter the representation of stimuli to facilitate new categorization tasks. An alternative account is offered for existing feature creation evidence. However, experimental work re-establishes a feature creation effect. Consideration is given as to how feature creation and memory and contrast accounts of categorization may be integrated, together with extensive suggestions for the development of these ideas.