The acquisition and processing of relative clauses : experimental evidence from Mandarin

Abstract

This thesis studies the acquisition and processing of relative clauses (RCs) in Mandarin-speaking children. The first study examined the production of a wide range of Mandarin relativized positions including Subject (S), Agent (A), Patient (P), Indirect Object (IO), Oblique (OBL) and Genitive (GEN). One hundred and thirteen Mandarin monolingual children aged 3;0 to 5;0 were tested by a sentence repletion task adapted from Diessel & Tomasello (2005). Children showed similar patterns across age groups. The difficulty ranking is S>A=OBL>P>GEN ('> easier'), and IO (double object datives) > IO (prepositional datives). Developmental predictions based on Noun Phrase Accessibility Hierarchy (NPAH, Keenan & Comrie, 1977) cannot adequately account for this specific pattern of results. The second study revisited the issue of subject-object processing asymmetry in Mandarin RC acquisition and processing. This is the first study using an online method to study this issue in child Mandarin, and the first study comparing two types of subject and object relative clauses: I) RC de DCL N (DCL-RC): relatives with the head nouns specified with a demonstrative (D) and a classifier (CL), and II) RC de N (DE-RC): relatives with bare head nouns. Thirty-six four-year-old Mandarin monolingual children were recruited. Children's eye movements were coded when they heard the test sentences and chose a referent that the sentence describes (Brandt, Kidd, Lieven, and Tomasello, 2009; Chan, Yang, Chang & Kidd, 2018; Rahmany, Marefat & Kidd, 2014). Online results revealed different asymmetry patterns for the two types of RCs. For DCL-RCs, children showed an object advantage, whereas for DE-RCs, the same children showed a subject advantage. This differential pattern of results is not predicted by theories that make general predictions of subject or object advantage within a language (e.g. structural distance hypothesis (Lin & Bever, 2006); Dependency Locality Theory (Gibson, 2000)), but maps well onto the distributional properties/frequencies in the input. The third study investigated the subject-object processing asymmetry in Mandarin and extended it to a bilingual context. Forty-six Kam-Mandarin bilingual children aged 5;11-10;3 were recruited in a Kam village in Mainland China, and were tested on the comprehension of head-final subject and object RCs in both Kam (L1) and Mandarin (L2) using a picture-pointing task. As expected, children found object RCs more difficult than subject RCs in both Kam and Mandarin, but they found object head-final RCs significantly more difficult in Kam than in Mandarin even though Kam is their L1 and the stronger language (for the younger group). This specific pattern of results cannot be adequately accounted for by structural perspectives to RC acquisition/processing (e.g. structural distance hypothesis by Lin and Bever, 2006 and structural intervention accounts by Friedmann, Belletti and Rizzi, 2009) but can be predicted by approaches that consider how relationships between constructions impact acquisition/processing outcomes (Rowland, Noble & Chan, 2014): Kam, but not Mandarin, has competing head-final and head-initial RC constructions. The current studies are significant in many ways both empirically and theoretically. Empirically, they bring in novel online data comparing two different RC types on the issue of subject-object asymmetry in young Mandarin-speaking children; novel comprehensive developmental data from a wide range of relativized positions that go beyond the frequently studied subject and object RCs in child L1 Mandarin; and novel developmental data from Kam-Mandarin bilingual children that allow us to reflect on the nature of difficulty in comprehending object RCs in Kam versus Mandarin. Theoretically, the developmental patterns exhibited are not predicted by several theoretical perspectives (e.g. Dependency Locality Theory by Gibson, 2000; and perspectives based on the Noun Phrase Accessibility Hierarchy, Keenan & Comrie, 1977). Rather, they support approaches that emphasize a close relationship between acquisition/processing and similarity to other structures and language specific distributional properties/frequencies of the input (Diessel & Tomasello, 2005; Chen & Shirai, 2015).