These results suggest that children’s encoding is guided by the generality of the property they are learning. When asked to identify a target from a set of distractors from the same social category, children in both age groups exhibited superior memory for the targets of historical properties over the targets of morphological properties. Additionally, children remembered morphological properties at a higher rate than historical properties. Given the consistent presentation of properties and the conservative within-subjects nature of the design, these results are especially striking: the only cues to guide children’s encoding were the introductions to each block and the properties themselves. These findings are consistent with the proposal that learning generalizable properties elicits category-level encoding and learning non-generalizable properties elicits individual-level encoding.
However, despite differences in encoding, children did not explicitly generalize morphological properties more than historical properties. This lack of generalization could be an artifact of the memory task, which may have confused children. In the memory task, we asked children to remember the specific individual about whom they had learned the property. Children could have interpreted the demand to remember the specific individual as a cue that the property was specific to that individual. Thus, even if children had originally encoded the properties distinctly (some general, some specific), the memory task may have led to a revised interpretation of all properties as specific. In order to claim that property generalizability affected encoding of targets and properties, we need to ensure that children actually perceived morphological properties as more generalizable than historical properties. Although there is strong support in the existing literature for this association (morphological properties are general whereas historical ones are not, e.g., Gelman, 1988), it would be useful to establish this association for the specific stimuli we used in the experiment. We address this issue in Study 1b, in which we presented children with an independent measure of generalization of morphological and historical properties.