Acquisition of variation in child language learners

Children exhibit a remarkable capacity to acquire whatever language (or languages) they are exposed to without any formal linguistic instruction. When one thinks of learning a language, one often thinks of learning new vocabulary words and grammatical rules. However, another important aspect of language acquisition that is particularly under-studied is how children come to acquire variation in natural language.

Mastering a language to fluency requires learning more complex and subtle dependencies in the linguistic and social environment. For example, whether an English speaker produces "went shopping" versus "wen' shoppin'" may be determined by many factors, some of which are internal to the utterance (e.g., nearby sounds or the expression's syntactic position in the larger grammatical context) whereas others are external (e.g., social context and speaker identity).

In the context of language learning, children must learn and master the use of these variable forms, applying them in the same contexts and at the same rates as other speakers of their dialect. The proposed research addresses a paradox in how children acquire such variation. In established language communities, children acquire such variation--mastering the relevant articulatory, grammatical and social conditioning factors from a very young age.

On the other hand, children in emerging languages communities will frequently regularize variable forms, removing variation to create a more regular or learnable linguistic system.

The experiments comprising this research address why, when confronted with variability in their primary linguistic data, children sometimes learn variation and its conditioning factors, and other times regularize it. Using recent findings, this project examines the hypothesis that particular properties of the language input may lead children to match variation in one case but regularize in another.

In particular, (1) whether the variation is conditioned on a grammatical or social context, (2) the fluency of the language model, and (3) whether the input is provided by a community of speakers. The project employs methods from the Artificial Language Learning paradigm--an innovative method that enables researchers to conduct language learning experiments with children in a single 30-minute session.

The experiments investigate whether the hypothesized properties of the input lead children to adjust their learning behavior, acquiring variation under some circumstances and regularizing variation under others. The proposed work serve as a foundation upon which further, more nuanced hypotheses about the acquisition of variation in children can be built, as well as provide an innovative experimental approach with which researchers can empirically investigate other important (but still only theoretically explored) questions in child language acquisition

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.