This study examines the ba and bei constructions in Mandarin using data from the Tong corpus, a new multimedia longitudinal child language corpus. A unified aspectual account of the two constructions is proposed: both require telic predicates, and should thus correlate with the perfective rather than imperfective aspect for learners. Analysis of corpus and diary data reveals that Tong is generally sensitive to the telic requirement when he begins to use the two constructions around 2;0. His ba and bei sentences occur far more frequently in perfective than imperfective aspect, consistent with the Aspect Hypothesis. However, while the majority of the child's bei passives are perfective, the majority of his ba sentences are produced without overt aspect markers, among which most are irrealis (imperatives, modal sentences, etc.). The difference mirrors the pattern in adult input. These findings are corroborated by additional child Mandarin corpora. The acquisition of the aspectual properties of the ba and bei constructions is influenced by inherent knowledge, input and language-specific features of Mandarin, consistent with the multi-factorial account proposed by earlier studies.

Aims and objectives: Heritage languages spoken by speakers in overseas communities can diverge significantly from the language spoken in the home country. Recent investigations have suggested that some grammatical structures or features are more vulnerable than others. This paper investigates the role of cross-linguistic influence, incomplete acquisition and attrition in heritage Cantonese in contact with English, focusing on the grammar of the pretransitive zoeng-construction in displacement contexts. Methodology: An elicited oral production task modelled on the fruit cart experiment was used to elicit displacement instructions in Cantonese. Fourteen heritage speakers and thirteen émigré speakers participated. All had acquired Cantonese as their first language but experienced a shift of language dominance to English due to immigration and education. Seventeen native speakers of Cantonese in Hong Kong served as the baseline. Data and analysis: The utterances were manually transcribed and coded. Production and error rates were calculated. Statistical results revealed quantitative differences among the three groups of Cantonese speakers. The baseline speakers preferred the zoeng-construction in displacement contexts, whereas the heritage and émigré speakers made greater use of canonical and topicalization structures. Nevertheless, the zoeng-sentences produced by the heritage and émigré speakers were all grammatical and felicitous. Findings: The basic structure of the zoeng-construction is kept intact in less than half of the heritage and émigré speakers’ Cantonese grammar. The zoeng-construction is thus vulnerable to intergenerational language change induced by language contact and individual differences, which is partially attributable to cross-linguistic influence from English. Originality: This is the first experimental study to investigate the grammar of heritage Cantonese. Significance: The study provides new empirical evidence of structural vulnerability and variability of heritage grammar and sheds light on the role of incomplete acquisition, cross-linguistic influence and attrition in such vulnerability.

This article reports an empirical study investigating L2 acquisition of the Mandarin Chinese shì … de cleft construction by adult English-speaking learners within the framework of the Feature Reassembly Hypothesis (Lardiere, 2009). A Sentence Completion task, an interpretation task, two Acceptability Judgement tasks, and a felicity ranking task were administered to learners with intermediate and advanced Chinese proficiency ( n = 76). The results reveal an initial mapping between the target Chinese structure and the English it-cleft construction. The relevant tense, telicity and discourse features are added in an uneven feature-by-feature manner in the subsequent feature reassembly. It is proposed that feature reassembly tasks involving cross-domain operations (e.g. from prosody to syntax) are more complicated and more difficult to accomplish than those taking place within the same linguistic domain.