Constraints on the acquisition of anaphora in adult English as a Second Language: A developmental model

This dissertation is an attempt to subsume the acquisition of anaphora in Second Language acquisition (L2) under the framework of Universal Grammar (UG), showing that L2 acquisition of anaphora is constrained by the incorporation of two sets of cognitive principles: the learning principles (the Subset Principle, in particular) and the language principles (the notion of c-command, in particular). The general hypothesis tested is that all structural restrictions of pronominal anaphora are developmentally acquired. That is, the hypothesis tested was that learners who rely on the learning principle (i.e., the Subset Principle) as their learning strategy initially would use the restriction based solely on directionality and later would have to retreat from that overgeneralization by taking the structural factor (i.e., c-command) into account. The linguistic properties involved in the study are: (1) the directionality, (2) the notion of command (c-command and d-command).

In the experiment conducted to examine the constraints on the L2 acquisition of anaphora, the general hypothesis, as well as more specific hypotheses, were supported. In the experiment, 107 students from three designated language groups (Japanese, Spanish, and Arabic) were grouped into two language proficiency levels (high and low). Subjects were asked to take a written comprehension test containing 42 token sentences designed to elicit their judgments on pronominal reference.

Results showed (1) that subjects allowed anaphora more frequently when the direction of anaphora is forward than when the direction of anaphora is backward, (2) that subjects based their coreference judgments on the notion of c-command, (3) that subjects' order of development in their acquisition of anaphora in terms of principles used is: Initially, 'directionality only' and later 'the notion of command'.

Several implications for general theories of linguistics were drawn: First, c-command is part of Universal Grammar (UG), and this supports the claim that both L1 and L2 learners have access to Universal Grammar (UG). Second, d-command is not part of Universal Grammar, and this implication indirectly supports the claim that only those principles embedded in 'core grammar' (e.g., c-command) are part of the mental structure of the language learner, and need not be learned from data. Third, forward anaphora was not restricted in all cases, and this indicates that forward anaphora preference is a universal and might be a part of UG.