Children's and Adults' use of Overhypotheses in Explanations About Familiar Kinds

• In Event: Categories and Concepts

Abstract

When a child sees a new animal such as a groundhog, there are many things they could glean from their encounter. They could infer that groundhogs are similar in color, that they are found under porches, or that they eat grass. In these situations, children are faced with a riddle of induction (Hume, 1748/2007): Which generalizations can they make from their experience with a single member of a kind?

One posited solution to this riddle is that children and adults develop overhypotheses (i.e., higher-order generalizations) about features shared among members of a kind (Goodman, 1955/1983; Shipley, 2000). For example, upon learning that rabbits eat carrots, dogs eat bones, and cows eat grass, a child may infer that all animals, including groundhogs, have a characteristic diet. Overhypotheses may, therefore, not only guide inductive inferences, but also provide the basis for children's and adults' concepts of animal kinds. Shipley (2000) proposed that there are five overhypotheses that structure people’s representations of animal kinds (characteristic diet, habitat, mode of locomotion, vocalization, and manner of reproduction). Yet, this proposal has received limited empirical examination.

To explore whether people use overhypotheses, we asked 4- to 6-year-olds (n = 80) to teach an alien puppet about five familiar animal kinds (e.g., dogs) and five familiar artifact kinds (e.g., tables). Adults (n = 80) listed features of the same categories. We manipulated between participants whether animals and artifacts were presented in separate blocks (Compare Condition) or intermixed (No Compare Condition) to examine effects of structural alignment (Christie & Gentner, 2010).

When describing animals, adults mentioned features related to Shipley’s (2000) overhypotheses 32.5% of the time (see Table 1). Children mentioned these features 47.3% of the time. For artifacts, in contrast, participants listed different features: 30.2% of adults’ responses and 46.9% of children’s responses were related to the object’s function or its characteristic location (see Table 2). Though analyses are still ongoing, we will use multinomial regression to determine whether overhypothesis use varies based on age, condition (Compare/No Compare), and item domain (Animal/Artifact). Our preliminary data show that children rely more on overhypotheses than adults, and the youngest children mention these features the most of any age group. The results demonstrate that people (especially young children) use overhypotheses to structure their representations of familiar animal kinds. Further, our studies suggest that because overhypotheses are common in people’s knowledge of animal kinds, it is possible that they may provide a solution to the riddle of induction by limiting the generalizations that children and adults make when they encounter unfamiliar animals.

Authors

• Ella Genevieve Simmons, University of Michigan - Ann Arbor
  Presenting Author

• Kristan Marchak, University of Alberta
  Non-Presenting Author

• Susan A Gelman, University of Michigan - Ann Arbor
  Non-Presenting Author