A corpus based analysis of the gendered use of Cantonese sentence final particles

The purpose of this study is to characterize the gendered usage of sentence final particles (SFP) by native Cantonese speakers. Cantonese (a Chinese language mostly spoken in Hong Kong, Macau and the Guandong province) has an unusually rich inventory of around 40 SFP (e.g. Kwok, 1984 or Matthews & Yip, 2011). These elements are described as conveying a variety of meanings dealing with pragmatic parameters like the epistemic status of the speakers, the questions currently under discussion, the mood of the speakers etc. Stereotypes about features of the language that are typically feminine/ masculine frequently make reference to SFP, describing them as markers of the alleged “emotional and polite” feminine register compared to the more succinct and direct masculine register (e.g. Erbaugh, 1985; Chan, 2000). Surprisingly though, there has been no large scale attempt to quantify the gendered differences in the use of Cantonese SFP. The closest work in that domain is that of Chan (2002) who focuses on the analysis of scripted dialogues from a TV show, and the use of a selection of SFP chosen on the basis of the aforementioned stereotypes.
To address the matter in a more systematic way, we analyzed the use of SFP in two different corpora of distinct formalities. One is the conversation part of the HKCanCorp (Luke & Wong, 2015) which contains recordings of conversations between close ones (friends, family colleagues), the other are the Hansard transcripts of the Hong Kong Legislative Council (LegCo) which we annotated by manually adding all occurrences of SFP. Both corpora indications about the speakers’ gender which allows the study of the variation between masculine and feminine speakers.
To assess the over/under use of SFP for each gender, we relied on a manually established list of 43 SFP, based on inventories in the existing literature. We then measured the rate of use of each SFP per utterance per speaker. To evaluate the significance of the gender bias of each SFP we relied on (i) the visual inspection of the Pearson residuals of a fit of the contingency table of the frequencies of each SFP by gender and (ii) the fit of GLMM with optimized random structure to measure how well the individual frequencies of usage predict the gender of the speaker. The results show that overall, women use more SFP than men in conversations, confirming the stereotype about them, although the effect is rather weak (x²=77.006;df=1;p <0:001;φ=0.079) and that while SFP are used more in conversations than in the formal setting of LegCo, a limited set of SFP is still used in that context. Despite the overall overuse of SFP by women, a closer examination of the SFP shows that some have a significant feminine bias (aa3, laa1, lo1, me1, wo3) and others are masculine oriented (ge3, laa3, ze1). We propose to analyze these differences in terms of attributed epistemic authority (Fricker, 2007).
The masculine SFP all involve a form of authority of the speaker: they introduce a new discourse topic (laa3), push (ge3) or refute (ze1) content with a minimal call to addressee. The feminine SFP on the other hand involve the search for common ground (lo1), the softening of statements with an evidential value (aa3) or the indication of a low belief in a non at-issue way (me1, wo3), suggesting that the speaker is not attributed enough authority to simply push their statement forward or refute one. Copyright © 2017 22nd International Conference on the Yue Dialects.