Comprehension and production of relative clauses in Cantonese children with and without developmental language disorder (DLD)

Abstract

This thesis studies the comprehension and production of relative clauses (RCs) in Cantonese-speaking children with and without Developmental Language Disorder (DLD). It consists of three studies that examine Cantonese RCs incorporating typological distinct properties of the language and reports empirical findings that bear on the theoretical dichotomy of domain-specific versus domain-general accounts in both typical and atypical language development. Study one presented two novel corpus studies that considered the typological characteristics of Cantonese as an attributive clause language and examined the acquisition of prototypical RCs and their related noun-modifying clause constructions (NMCCs) in a broader conceptual context. It investigated the developmental trajectory and characteristics of NMCCs, including both conventional RC-type NMCCs and gapless NMCCs, in 78 monolingual Cantonese-speaking children's naturalistic speech aged 1;7 – 5;6. Results posed challenges to structurally-oriented accounts that consider structural constraints as the primary determinants affecting acquisition outcomes, but rather, findings were discussed in light of constructivist perspectives (eg. Lieven & Tomasello, 2008) that focus on form-function pairings and conceptualize constructions in an interrelated network. Study two reported the first experimental study that investigated RC comprehension offline and online in Cantonese-speaking children with and without Developmental Language Disorder (DLD). Following a similar design in Frizelle and Fletcher (2014), this study compared children with DLD (n=22) with their age-matched typically-developing peers (AM-TD, n=23) aged between 6;6 - 9;7 and language-matched, younger TD children (YTD, n=21) aged between 4;7 – 7;6. This second study used a referent selection eye-tracking task to test the predictions from domain-specific versus domain-general accounts of Cantonese RCs and DLD children, with regards to three dimensions: (i) two RC types (SRCs vs ORCs); (ii) relativization strategies (CL vs ge3) and (iii) DLD vs TD peers. The developmental pattern cannot be adequately accounted for by domain-specific structural perspectives that conceptualize the nature of processing demands in terms of structural constraints (eg. structural distance (Hawkins, 1999, 2004) or structural intervention (Friedmann, Belletti & Rizzi, 2009)); but are well-predicted by domain-general emergentist-constructivist approaches (eg. O'Grady, 2010, 2011, 2021) that allow multiple factors to jointly determine acquisition outcomes. Study three extended the investigation to a wider range of relativized positions and presented the first empirical study that examined RC production in Cantonese-speaking children with and without DLD. To evaluate the applicability of the noun phrase accessibility hierarchy perspective (NPAH, Keenan & Comrie, 1977) versus domain-general emergentist-constructivist approaches on Cantonese RCs, this study assessed the same three groups of children from study two and used a sentence repetition task to test (i) the relative difficulty between RC types; (ii) the relative difficulty within an RC type; and (iii) the relative difficulty between relativization strategies. The specific pattern of results is not consistent with the predictions based on NPAH, but maps well onto a multifactorial, domain-general account of acquisition that identifies a core role for language-specific properties and learner's experience. These new findings contribute novel naturalistic and experimental data on the developmental patterns of RCs in Cantonese-speaking children with and without DLD. The pattern of results exhibited across the three studies consistently challenge the domain-specific structurally-oriented perspective to RC acquisition, and are best predicted and accounted for by domain-general emergentist-constructivist approaches that are multifactorial and give primacy to the interaction of multiple factors such as learner's experience, language-specific properties in form-function mappings and relationships between constructions, cognition as well as processing.