Perception and social evaluation of gendered adjective-noun combinations in Mandarin

Abstract

Mandarin lacks grammatical gender, but gender meanings can still be conveyed through lexical stereotypes. This study examines how native Mandarin listeners socially evaluate gendered adjective-noun combinations in speech across various personality dimensions, and whether such evaluations are influenced by voice gender and by the disclosed AI nature of the voice. Seventy-seven native Mandarin speakers listened to sentences containing different gendered adjective–noun combinations, produced in AI-generated female and male voices, and rated the speakers on friendliness, trustworthiness, fluency, openness, and education level. Results showed that gender-congruent combinations were rated more positively than incongruent ones, with the strongest penalty found when female-typical adjectives modified masculine nouns. Female-typical adjectives were also evaluated more favorably in a female voice. Explicit disclosure that the voice were AI-generated did not significantly affect listener ratings, but voices perceived as human were evaluated more positively than those perceived as artificial. These findings provide empirical evidence that, in a language without grammatical gender, social evaluation of gendered language can be shaped by lexical gender stereotypes, paralinguistic voice cues, and perceived humanness.