A Two‐Stage Model of Category Construction

Abstract

The current consensus is that most natural categories are not organized around strict definitions (a list of singly necessary and jointly sufficient features) but rather according to a family resemblance (FR) principle: Objects belong to the same category because they are similar to each other and dissimilar to objects in contrast categories. A number of computational models of category construction have been developed to provide an account of how and why people create FR categories (Anderson, 1990; Fisher, 1987). Surprisingly, however, only a few experiments on category construction or free sorting have been run and they suggest that people do not sort examples by the FR principle. We report several new experiments and a two‐stage model for category construction. This model is contrasted with a variety of other models with respect to their ability to account for when FR sorting will and will not occur. The experiments serve to identify one basis for FR sorting and to support the two‐stage model. The distinctive property of the two‐stage model is that it assumes that people impose more structure than the examples support in the first stage and that the second stage adjusts for this difference between preferred and perceived structure. We speculate that people do not simply assimilate probabilistic structures but rather organize them in terms of discrete structures plus noise.

An algebra of human concept learning

“…This division of the trends in the data to more and less essential elements resonates with evidence that human learners represent induced concepts as a ''discrete structure plus noise,'' i.e. a simple rule-based core plus a more detailed and idiosyncratic periphery (Ahn & Medin, 1992;Medin, Altom, & Murphy, 1984;. Specifically the linear component of the power series is very similar to the network of dependency relations proposed by Sloman et al (1998).…”

“…Recently, many authors have proposed hybrid models that involve both rule-extraction (e.g., prototype formation) and exemplar storage (Ahn & Medin, 1992;Ashby, Alfonso-Reese, Turken, & Waldron, 1998;Erickson & Kruschke, 2002;Medin et al, 1984;Smith & Sloman, 1994). The common premise of these models is that human learners seek to extract common tendencies from examples, but also store individual cases, particularly exceptional ones.…”

The Contribution of Argument Structure Constructions to Sentence Meaning

“…When participants were done with the sorting, with the piles still in front of them, they were asked to briefly describe how they had performed the sorting and whether they had noticed any repeated words across the sentences. The experimenter took notes as accurately as possible and the notes were used to identify participants' basis for sorting (Ahn & Medin, 1992;Medin et al, 1987).…”

“…Participants sorted unidimensionally even when they were explicitly told to pay attention to all of the dimensions of the stimuli. Unidimensional sorting has been found even with large numbers of dimensions (Smith, 1981), ternary values on each dimension (Ahn & Medin, 1992), holistic stimuli, and stimuli for which an obvious multidimensional descriptor was available (Regehr & Brooks, 1995). Lassaline and Murphy (1996) hypothesized that the reason participants do not sort categories according to family resemblance even when the category has a family resemblance structure is that family resemblance sorting is computationally more difficult.…”

The Origin of Exemplar Effects in Rule-Driven Categorization.

“…However, the categoryconstruction literature shows that participants use simple and typically one-dimensional rules to sort items and that the influence of interstimulus similarity on sorting behavior is quite limited (Ahn & Medin, 1992;Regehr & Brooks, 1995). Thus, before Allen and Brooks's (1991) work, it was not clear to what extent exemplar memory would develop and influence categorization when participants can also rely on an explicit rule.…”