Effects of talker-to-listener distance on tone

• We model the effects of talker-to-listener distance (TLD) on tone.
• We examine the effects of TLD on tone perception by L1 and L2 listeners.
• Farther TLDs help tone perception by increasing pitch, intensity and duration.
• Two invariant tone features preserve tone information under changes in TLD.

This paper explores the effects of talker-to-listener distance (TLD) on tone and examines its impact on tone perception by both native listeners and second language learners.Speakers naturally adjust vocal effort to talk to people at different distances, which leads to changes not only in intensity, duration and formant frequencies, but also in fundamental frequency (f0). Fundamental frequency is the primary acoustic cue that differentiates Mandarin lexical tones from one another. This study aims to answer the question of whether changes in f0 as a function of TLD affect tone perception by native (L1) and second language (L2) listeners. If so, what are the specific changes that have an impact on tone perception?The production study investigates the acoustic correlates of the effects of TLD on tone, using 7959 monosyllabic Mandarin speech files recorded by three speakers under 11 levels of TLD. The perception study explores the effects of TLD on L1 and L2 tone perception by 2 native listeners and 52 L2 listeners.The effects of TLD on speech production are systematic, and we present models and analyses with sufficient detail to simulate these effects. Intensity, duration, and initial and maximum f0 increase along with TLD, while time-normalized toneshapes remain invariant. The results of the perception study show that native listeners' performance is robust under changes in TLD, while L2 listeners' perception of tone interacts with TLD. One significant finding is that tone 3 recognition improves with TLD.This work investigates the relationship between speech production and perception when acoustic attributes change naturally in response to the demands of speech communication. The results have potential applications in speech synthesis, pronunciation training and second language testing.