THE ACQUISITION OF WH-QUESTIONS IN TASHLHIYT BERBER: NOVEL BEHAVIORAL AND CORPUS STUDIES

Abstract

The central quest in this dissertation is how a child would acquire abstract structure in a language that masks the overt expressions of said structure. More specifically, the interest here is in studying the acquisition of wh-questions in a language with a grammar that shows the following properties: (1) the wh-word is always followed by an obligatory complementizer that is rendered inaudible by the phonology in most instances and (2) it does not differentiate between subject and object Ā-extraction in word order but through verbal and nominal morphology only.The case study explored here is that of the Afro-Asiatic language Tashlhiyt Berber (henceforth TB)(ISO 639-3/shi). In this language, the morphology is the predominant source of information about the position of a gap in Ā-dependencies. The verbal morphology that cues to the subject extraction is the anti-agreement marker on the verb and the nominal morphologies that cues to the extraction of the object is the construct state on the embedded NP. We test the children's knowledge with a behavioral study and make hypotheses about the learning problem through a longitudinal corpus of child-directed Tashlhiyt Berber that involves two sisters from ages 2 years to 4 years and 4 months to 1 year. The comprehension study involved children between ages 2 and 4 and it was designed to test knowledge of reversible action verbs in subject and object wh-questions and relative clauses. We find that children's knowledge of wh-questions and relative clauses in the language is equivalent, meaning that these construction types are not predictive of performance. By ages 3 and 4, TB-acquiring kids show high comprehension of subject extraction which featured the anti-agreement morphology. The behavioral study result that construction type (WH vs. RC) is non-predictive of performance comes despite some of the surface asymmetries between the two constructions and an asymmetry in terms of their input distribution as measured in the corpus. Although the input appears to be informative in terms of the distribution and quality of the data on anti-agreement, the input turns out to be uninformative about the obligatory complementizer and the nominal morphology necessary to infer the grammar of TB Ā-construction. This would be problematic unless the child recruits informative data for well-represented constructions to learn about the less represented constructions in the input, which would require innate knowledge of the abstract identity of Ā-constructions. This is the first work in the Berber developmental literature that combines facts about the morphosyntax with a phonological conspiracy present in the language to formulate a developmental puzzle. Beyond this, the contribution of this dissertation is twofold. Firstly, a significant portion of the findings in it builds on the construction of the first-of-its-kind corpus of caregiver-child interactions in Tashlhiyt Berber. This corpus was collected for the purposes of this dissertation and is meant to be contributed, fully transcribed and media-linked, to the Child Language Data Exchange System (CHILDES) (MacWhinney 2000) repository. Secondly, this contribution diversifies the growing work on the study of V1 languages with anti-agreement phenomena and the idiosyncratic language-specific Construct State from a developmental standpoint.