Item Response Theories of Semantic Cognition

The boundaries of natural language categories are notoriously vague. It is rarely the case that language users agree upon the extension of a particular category. In this dissertation a formal framework for the study of vagueness in natural language categories is introduced. It contends that inter-individual differences in judgments of category membership are due to the use of distinct membership criteria. These criteria are imposed on a latent scale that organizes potential referents with respect to the target category. While the membership criteria are said to differ from individual to individual, the scale is assumed to be common to all respondents. The position of a respondent's membership criterion along the scale indicates whether she takes a broad or a narrow view on what should be included in the category. The more to the left of the scale the membership criteria are positioned, the more liberal the respondents. Membership criteria that are positioned more to the right of the scale designate more conservative respondents. This is the case because the criteria behave as probabilistic thresholds. Referents that surpass a criterion have a higher probability of being endorsed than of being denounced as category members. The reverse holds for referents that fall short of the criterion. When a referent coincides with a criterion it is as likely to be endorsed as it is to be rejected. The distance between criterion and referent along the scale determines the relative probability of both categorization outcomes. The further away from the criterion a referent is located, the higher its probability of being called a category member (when it surpasses the criterion) and the higher its probability of being called a non-member (when it falls short of the criterion). The vagueness of a particular referent thus also depends on its own position along the latent scale. The further along the scale referents are positioned, the higher the probability that they will be endorsed as category members.The development of the formal framework proceeds in seven chapters. Chapter 1 identifies the proposed account of vagueness with the Rasch model (Rasch, 1960). The Rasch model is part of the family of item response models. It is argued that these models, with their emphasis on differences among both referents and respondents, constitute an appropriate formal framework for the study of vagueness. Chapters 2 and 3 embed the framework in a general theory of categorization, known as the Threshold Theory (Hampton, 2007). This allows for a substantive interpretation of the Rasch model's assumptions and establishes references for its evaluation and development. Both chapters include illustrations of how to apply the Rasch model to answer questions about categorization differences with respect to natural language categories. Chapter 4 demonstrates one of the benefits of having an explicit, formal model at one's disposal. It comes natural to put the model's assumptions to test, allowing one to gain new insights about the phenomenon of study. It appears notto suffice - as the Rasch model does - to assume that only the comparison between criterion and referent position informs the decision to include a referent as a category member or not. The decision might also depend upon the conclusions that were reached for other referents. This warrants the application of more complex models from the item response framework. Chapters 5 and 6 argue for the applicability of the proposed framework beyond the study of extension differences among adult language users. They suggest it might be used to study other aspects of semantic cognition as well. In Chapter 5 it is brought to bear on the question of how children proceed to acquire categories that do not have fixed extensions. In Chapter 6 it is employed to answer the question of how a category's extension and intension relate. The chapters employ item response models that are different from the Rasch model to answer these questions. Chapter 7 concludes with an overview of other directions to take the formal framework in to further our understanding of vagueness in particular, and semantic cognition in general.status: publishe