Measurement of Semantic Knowledge and Its Contribution to Object Recognition Performance

Abstract

How do people differ in their ability to recognize objects, and where do these differences come from? How much of the differences between people in a given domain stem from domain-general abilities, and how much from domain-specific experience? The goals of this thesis are to create measures of non-perceptual semantic knowledge acquired with experience in a domain, and to explore how such knowledge can help us understand individual differences in high-level cognition. We created the Semantic Vanderbilt Expertise Test (SVET), which focuses on one important aspect of semantic knowledge that can be measured across categories: acquisition of relevant nomenclature. We implemented SVETs for eight categories: cars, planes, Transformers, dinosaurs, shoes, birds, leaves, and mushrooms. In Study 1, the SVET was tested and refined with an online sample. The SVET showed good reliability and coverage of the full range of performance. In Study 2A, the SVET was tested with the VET (Vanderbilt Expertise Test, a measure of visual memory) for all categories, measures of domain-general visual and verbal abilities (Gf), and self-reports of category experience in a university sample. The SVET showed good validity as it was more correlated with experience and VET performance for the same category than with the average of other categories. We tested the hypothesis that the only source of domain-specific shared variance between the VET and SVET is experience with a category. This hypothesis was partly supported for six of the eight categories tested. In Study 3, we explore lateralization of visual object recognition performance as a function of semantic knowledge. Overall, this work contributes to the new research area of individual differences in object recognition and provides a carefully tested and refined tool, the SVET, to the psychology community for future use in this field.