Up and Down the Ontological Tree: categories and semantic generalisation

Abstract

Children's understanding of ontological categories has been assessed in two ways: by gauging sensitivity to the sentence anomalies which result when terms are paired with the wrong category of predicate and by gauging the degree to which children's generalisation of a new word to a new referent is constrained by ontological category. The first technique suggests that young children's ontological knowledge is very primitive and the second that it is well‐developed. A problem is that because the latter technique employs picture‐matching, we cannot be sure that the matching is by category and not by perceptual features. We used a semantic generalisation procedure to look systematically at within‐category generalisation between the ages of 2\ years and adulthood from the category of ‘event’ down to that of ‘human’. Our procedure precluded within‐category matching on the basis of perceptual features by using target pictures that showed two objects interacting so that the target object had to be picked out by the experimenter's description (e.g. “what grows in gardens...”). Experiment 1 showed that within‐category generalisation for substance, artefact, plant and animal was possible in school‐age children. Experiment 2 showed that this result held irrespective of the distance apart on the tree of the objects in the target picture. In two further experiments, a different technique was used: single object targets plus multiple comparison pictures. In Experiment 3 there was no within‐category match and in Experiment 4 there was one match. Although the reasons for the pre‐schoolers’ inability to generalise within ontological categories is open to debate, due to the conservative nature of our tests, we can conclude that children above 5 years of age often constrain their interpretations of a novel noun in terms of the ontological category to which the initial referent belongs.