Slide Slam P14 Sandbox Series
Features, flexibility, and fakes: How modifiers change concepts
Abby R. Clements1, Sharon L. Thomspon-Schill1; 1University of Pennsylvania
An important question surrounding concepts is whether or not they have “cores” that define them, separate from any features that they may also probabilistically have. Supporters argue that people tend to feel that most things, especially natural kinds and well-defined categories, have a certain essence that causes their category membership even if they lack key features, although they are usually unable to articulate what precisely that essence might be (Gelman, 2004). Detractors point out that most concepts, particularly artifacts, are impossible to define in any meaningful way, and that if concepts have cores but people never use them to recognize, categorize, or reason about objects, then they are essentially useless (Malt & Johnson, 1992). One difficulty with answering this question is that it is possible that all concepts may have distinctive cores. While it is generally accepted that the core of plants and animals would be built around their genetic composition, artifacts could potentially be defined by many things, such as their function, their material, their appearance, or their intended purpose (Barton & Komatsu, 1989; Bloom, 1998), and different artifacts could possess different cores depending on various idiosyncratic factors. To address this problem, we turned to the class of privative adjectives, which satisfy the constraint, “An [adjective] [noun] is not a [noun]” (Partee, 2007). Many of these adjectives, such as fake, false, counterfeit, imitation, mock, faux, etc., explicitly pick out things that are in some way a representation of the original concept (i.e., a decoy duck is a representation of the duck) and thus have many of the same visual features of the original, but are by definition not true examples of the original concept (so a decoy duck is not a duck). Therefore, privative adjectives could consistently negate the cores of concepts regardless of what those cores might be. In one study, participants were asked whether certain modified concepts did or did not belong to the same category as the bare, original concept. A second study had participants rate the likelihood of sentences that followed the form, “[Concept] has [feature].” The sixty-five concepts were either unmodified or modified by fake; a prototypical adjective; a non-prototypical adjective wherein a typical feature was negated; an orthogonal adjective that made the concept neither more nor less typical; or some combination of the modifiers. The features were either visual or internal. Participants were also asked to produce descriptions for several fake concepts. As predicted, all modifiers increased uncertainty for all conceptual features, replicating prior studies on uncertainty under combination (Connolly et al., 2007). However, fake differentially weakened or negated internal features. While the current study does not directly address the existence of conceptual cores, the results do provide support for them, as they suggest that while the boundaries of conceptual categories can be stretched quite far, internal features are what determine category membership. They also suggest that privative adjectives provide a previously unexplored avenue for determining where conceptual boundaries are and what features are most central to particular concepts.