Input cues affecting the organization of grammatical knowledge in early acquisition.

摘要

The fundamental problem of grammatical acquisition is converting a set of unanalyzed sentences into knowledge of a set of patterns that can accurately describe the grammatical patterns of a language. One proposal to allow this problem to become tractable is that learners use low-level input cues to help them organize input in a way that makes patterns more easily discoverable. Through three sets of experiments using artificial language learning methods, this work tests whether patterns and cues that signal those patterns emerge simultaneously in development, the most likely scenario under which cues could help learners. Chapter Two describes experiments investigating the utility of intonation variability as a cue to grammatical phrase structure and finds that intonation helps adult learners to acquire patterns better early in acquisition, even when the intonation cue is taken away at test. Chapter Three describes a set of experiments that manipulates function words---critical cues to grammatical structure---in a variety of ways. First, function words are manipulated so that they either do or do not contain phonetic cues to category. Phonetic cues to category did not help---and actually hindered---learning of the artificial language. Second, the presence and absence of function words is manipulated during exposure and at test. More function words in the input helped learners acquire patterns better, and more function words at test generally---though not always---hindered learners' performance. Chapter Four includes a set of experiments that manipulate variability and repetition, two aspects of input that can help learners discover patterns but are also inversely proportional to one another. Adult learners appear to learn better from more variable input, while 12-month-old infants learned better when the input contained more repeated sentences. Chapter Five extends a previous framework that explains findings in early speech perception and word learning research within a single mechanism. The extension proposed in Chapter Five broadens the scope of explanation to include early grammatical development.