Tone contour identification from the perspective of production

Abstract Contour shape identification is a critical yet challenging task in the study of tone systems. A common issue is inconsistent transcriptions of a single tone as different contour shapes across alternative data sources of the same dialect. We attribute this difficulty to the significant distortion of target contours in their phonetic realizations. To address this problem, the current study establishes a mapping between pre-defined target contours and their realized production forms. Realization data of pre-defined target contours are obtained through an imitation task in a controlled production experiment. The scope of the experiment is restricted to level, rising, and falling pitch contours. The analysis focuses on descriptive parameters derived from the pitch lines resulting from our experiment. The results reveal that while range and slope parameters can exhibit ambiguity in distinguishing level from non-level contours, a set of parameters related to the steadiness of rising and falling pitch yields a clear distinction between level and non-level (rising/falling) contours. The effectiveness of these steadiness parameters is further validated through a case study of four tones in two Chinese dialects (Xiangtan and Jian’ou), demonstrating their utility in distinguishing the target contours and improving tonal analysis.