Generalization of a Semantic-Syntactic Structure

Abstract

Two experiments examined the development of generative responding and the generalization of these responses with young children. In both experiments a model emitted personified sentences attributing human actions or characteristics to a nonhuman noun, and subjects were reinforced for emitting responses that were personified but were different from the model's and thus "generative." In Experiment I each response by the model was followed by a response of the subject to the same noun. In Experiment 2 the model emitted personified responses to five nouns, and the subject was then reinforced for personified responses to five different nouns. Both experiments developed generative responding, although this developed more quickly in Experiment 2. In Experiment 2 responding during "probes" appeared more controlled by the reinforcement contingencies and less by the model's "incorrect" responses than in Experiment 1. Most significantly, in Experiment 2, unlike Experiment 1, generative responding persisted during generalization conditions utilizing novel stimuli with no modeling. The results and procedures were interpreted in terms of the implications of reinforcement and stimulus control variables in the acquisition of complex language behavior.