Yao Yao (The Hong Kong Polytechnic University) will give an online colloquium. Here is the Zoom link.
Separating listener-based and talker-oriented influences in prediction: A study of gender effects on sentence prediction in Mandarin Chinese
When listening to a talker, we are constantly predicting the upcoming word the talker will produce. There are reasons to believe that sentence prediction can be jointly influenced by the listener’s own production system (e.g., Pickering and Garrod “prediction-by-simulation” account) and the listener’s model of the talker (i.e., communication accommodation theory). In this study, we test the extent to which sentence prediction is modulated by listener- and talker-oriented factors, using a series of cloze tests with Mandarin Chinese participants. The crucial factors we vary are the gender (sex) of the listener and the talker. During the cloze test, the participant (listener) is asked to guess the final word of a sentence a talker says about their profession, hobbies, styles, etc. – topics that are widely associated with gender-related stereotypes in human societies. We first conducted the cloze test in the baseline condition, where talker identity was not revealed and participants only saw the sentence frame on the screen (e.g., My favorite color is ____). In this condition, participants’ cloze responses were significantly influenced by their own gender: female participants produced more feminine-oriented word predictions than male participants. We then tested a set of gender-cued conditions, where talker gender was revealed either by a personal name (e.g., John says, my favorite color is ___) or a gendered voice (e.g., a male-sounding voice saying My favorite color is). Overall, participants did adjust their responses to align with the cued gender of the talker, with more feminine/masculine responses for cued female/male talkers. We also observed interactions between talker gender and experimental design on the genderedness of the cloze responses: the effect of talker gender was stronger in the name-cued conditions than the voice-cued ones and stronger when talker gender was mixed instead of blocked in a testing session. Our results provide evidence for the co-presence of listener-based and talker-oriented influences in the sentence prediction task. We discuss implications of current findings for understanding the integration of linguistic and social information in language comprehension.
