Do preschoolers make behavioral predictions based on a ‘teleological’ framework?
Csibra & Gergely (1998) proposed that infants take a "teleological stance" when interpreting goal-directed actions; they are sensitive to actions, goals, and physical constraints of an action event, contingent on a principle of rationality. The present experiments tested the plausibility that such a model underlies action prediction that develops into children taking an "intentional stance" (Dennett, 1987), the over-attribution of intentions as a predictive strategy. 3 ½ -4-year-olds were shown animations based on those shown to infants by Gergely et al. (1995). In Experiment 1, children viewed animations of a ball moving in a rational (over rectangle) or non-rational (over nothing) manner, in route to an unambiguous goal (to a triangle) or ambiguous goal (away from a triangle). The objects were then moved, the rectangle taken away, and the child was asked to predict what the ball would do in this new situation. In Experiment 2, the rectangle was included in the test scene in such a way that the child had to choose between the triangle or the rectangle. In Experiment 3, children were given the same animations as Experiment 2, but were given intentional and animacy cues by the Experimenter. There were three dependent measures: Verbal responses, Forced-choice responses (3 locations to choose), and Action Trajectory production (child moves the ball via a touch screen). It was predicted that if children were using the same teleological representation proposed by Csibra & Gergely (1998), then children would verbally reference actions and goals, use intentional language, choose a location consistent with the goal previously viewed, and produce rational (straight-line) trajectories to their predicted location. Moreover, a robust representation was predicted to have high levels of agreement across the three dependent measures. Results indicated nonsystematic patterns of results and little evidence consistent with the predictions of a teleological model. Although children did not show appear to take an "intentional stance" in their language in any of these experiments, there was limited evidence suggesting that those who did attribute intention had higher agreement across measures than those who did not.