The Influence of Shape on Human Categorization

Abstract

An object’s shape is one of the most salient and informative features that it possesses. Indeed, an object’s shape usually defines what it is. Most of the time shape is all that is needed to rapidly assign an object to a category, in turn giving us a wealth of information about the object. This interaction of shape and categorization is the central topic of this dissertation. Study 1 (Tiedemann, Morgenstern, Schmidt & Fleming, 2022) investigates the problem of one-shot learning – the human ability to infer categories from just one example – using a drawing-based generative task. We show that robust and varied perceptual categories can be inferred from viewing just one object. The part structure as well as highly distinctive individual parts of shapes were the main driving force in the creation of these categories. Study 2 (Tiedemann, Schmidt & Fleming, 2022) focusses on the effect of a shape’s part structure on superordinate categorization, specifically the category-boundary between animals and plants. The results demonstrate that subtle differences in curvedness and symmetry of parts, alongside the existence or absence of higher order parts can determine the superordinate category of objects. Together, theses studies show that human perceptual categories, both known and novel, are heavily influenced and delineated by specific shape features, chief among them an object’s part structure. Crucially, the specifics of part structure are a key ingredient in the human ability to learn categories quickly, with few examples.