Constructional grounding: The role of interpretational overlap in lexical and constructional acquisition

Adopting a constructional view of grammar, this dissertation addresses the question of how one conventional linguistic sign---i.e., one lexical unit or grammatical construction---can be based on another. Traditionally, based-on relations between sips are regarded either as arbitrary results of historical change or as properties of a stable system used by adults. This work proposes a third view. Using longitudinal corpus data, it argues that certain sips are related primarily through a dynamic process in early acquisition rather than through static principles of the linguistic or conceptual system instantiated in the minds of adult speakers. In this process, called constructional grounding, a sign that is relatively easy for children to learn (the source construction) serves as the model for another more difficult sign (the target construction), because it occurs in contexts in which it exemplifies important properties of that sign in a way that is especially accessible to children.; Three case studies are presented, each using seven longitudinal American English corpora from the CHILDES archive. In the first study, the source construction is the nonsubject WH-question and the target is a semi-idiomatic construction called the "What's X doing Y?" construction (e.g. What is this scratch doing on the table?). In the second study, the source is the deictic there-construction and the target is the existential there-construction. In the third study, the source is the normal visual sense of the verb see, and the target is the metaphorical mental sense (e.g. Now I see your point.). In all three studies the data show the same tendency in the children's productions: first they produce clear instances of the source construction, then they produce overlap utterances that have the formal and semantic-pragmatic properties of both the source and the target construction, and only after this overlap stage do they produce clear instances of the target construction. This tendency is interpreted as indicating that children use the source constructions, whose meanings are more intersubjectively available in simple adult-child interactions, to "bootstrap" the target construction. The conditions that enable this learning procedure are show to be the natural result of historical semantic change.