A robust benefit of talker F0 separation is observed in Mandarin cocktail party listening, but the advantage is constrained by the lexical role of F0, with real-word tonal minimal pairs lowering recognition accuracy relative to baseline.
Target word recognition was substantially more accurate when target and masker talkers were of different sexes compared to the same sex.
Real-word tonal minimal pairs interfered with target word recognition accuracy relative to baseline, whereas nonword tonal minimal pairs did not.
F0 serves a dual function in Mandarin that creates a constraint on the talker-separation benefit observed in cocktail party listening.
This research investigated how Mandarin Chinese listeners identify a target speaker's words when another talker is speaking at the same time — the classic 'cocktail party problem.' In Mandarin, the pitch of a voice (fundamental frequency, or F0) plays two roles simultaneously: it helps listeners tell different speakers apart (men typically speak at lower pitches than women), and it also changes the meaning of words, since Mandarin is a tone language where the same syllable spoken with different pitches means completely different things (for example, 'sleep' versus 'dumplings'). The study found a large benefit when the target and masking talkers were of different sexes (85% accuracy) compared to the same sex (48% accuracy), showing that Mandarin listeners, like listeners of non-tonal languages, use pitch differences between speakers to 'tune in' to the right voice. However, the study also found that when the target word had a tonal 'twin' — another real Mandarin word that sounds identical except for its tone — recognition accuracy dropped by about 4% compared to a baseline condition. Importantly, this interference only occurred when the tonal alternate was a real word, not when it was a nonsense word. This suggests that the interference happens at the level of word meaning: when pitch cues point toward a competing real word, listeners are pulled in two directions at once. This research suggests that while tone language listeners are just as skilled as speakers of non-tonal languages at using pitch to separate competing voices in noisy environments, there is a hidden cost: the same pitch information that helps them focus on one speaker can simultaneously activate unintended word meanings. This tension between pitch's two roles — speaker identity and word meaning — is a unique challenge for listeners of tone languages like Mandarin in real-world noisy communication settings.
Wang X, Lee C, Zhang Y, Wiener S. (2026). Processing multi-talker speech in a tone language: Dumplings interfere with sleep at a cocktail party.. JASA express letters. https://doi.org/10.1121/10.0042461