Causative alternation errors in child language acquisition

This dissertation presents a theory of the acquisition and unlearning of the knowledge underlying children's causative alternation errors, with reduced assumptions about innate grammatical knowledge.; The central claim is that children's knowledge of argument realization regularities takes the form of construction paradigms, which explicitly link verbs and constructions in possible pairs. These paradigms are acquired starting from initial; verb-specific argument realization patterns.; Errors arise because children's construction paradigms contain hypothetical verb-con-struction links. The addition of these links is motivated by shared causative structures among the lexical semantic representations of verbs and constructions. Children prefer to acquire verbs with causative lexical semantic representations when they have the chance. The Intrusive Results Hypothesis is that this leads them to misacquire a class of verbs that adults use to talk about causative events despite having noncausative lexical semantic representations for them. The hypothesis is supported by a corpus study of adult verb usage.; However, analysis of child errors drawn from corpus data falsifies the Strict Causativity Hypothesis, that errors arise because children add hypothetical paradigmatic links between verbs with causative meanings, correct or not, and the causative alternation constructions: verb causativity is a significant factor in inchoativization errors (The blocks knocked down), but not in causativization errors (Isak fell the blocks over).; Construction paradigms accommodate these findings easily: the shared causative representations of the Transitive and Periphrastic Causative constructions motivate the addition of the former to the paradigms of all the verbs that occur in the latter. But there is no periphrastic rephrasing for inchoatives, so the structural commonalities motivating hypothetical links in this case hold only among verb meanings.; Children recover from errors because they have access to negative evidence, produced by a hypothesis-testing language acquisition mechanism. This takes as input a parsed sentence, and uses its meaning to generate a child form. Comparison between the parsed and child form yields positive evidence if the forms match, and negative evidence otherwise. The hypothesis that mismatches are driven by children's sensitivity to pragmatic differences between the causative alternation constructions is supported by corpus study results obtained here and existing experimental results.