Bilingual influences and sources of variability in acceptability judgments: A case study of Chinese

Abstract

The replicability of grammaticality judgments forms the foundation of data quality in linguistic research. Previous work has mostly focused on judgments from ideal “native speakers,” disregarding speakers of different language backgrounds. This study examines whether acceptability judgments in Chinese can be replicated by linguistically diverse native speakers, “monodialectal” and “multidialectal” speakers of Chinese, and then explores how various factors influence such judgments. First, we obtained a representative dataset by randomly sampling 337 minimal pairs from 68 journal articles on Chinese syntax from the past decade. Then, two groups of participants—monolingual Mandarin speakers from Beijing and Mandarin-Cantonese bilinguals from Guangzhou—completed an acceptability rating task (Experiment 1). Two forced-choice experiments (Experiments 2 and 3) were conducted to further examine the unreplicated pairs from Experiment 1. The results of these three experiments showed a convergence rate of 92% between our participants and the syntacticians who authored the examples. Importantly, the language backgrounds of the participants and the authoring syntacticians were not found to play a role in acceptability judgments, whereas sentence length and the language of the journals did. The multilingual status of Cantonese-Mandarin bilinguals has a subtle but limited influence on judgments in Mandarin Chinese. We argue that the reliance on a monolingual “ideal” native speaker for eliciting judgments may have been overemphasized in linguistic research.