The primacy of the rising/non-rising dichotomy in American English intonational tunes

In American English, phrase-final pitch trajectories have been described as resulting from a sequence of three tonal elements whose combinations define an inventory of phonologically contrastive nuclear tunes [1]. We investigate the distinctive status of nuclear tunes, testing imitative production of sentences paired with one of 8 nuclear tunes, and testing pairwise perceptual discrimination of the same tunes. Results from group- and individual-level clustering analyses of F0 trajectories of imitated tunes reveal maximally 5 distinct tunes, with the most robust distinctions between two tune classes: rising and non-rising. Converging results are obtained from perceptual discrimination. A further finding is that the phonetic distance between tunes is a good predictor of discrimination accuracy, but accuracy is better than predicted for pairwise discrimination across the rising/non-rising classes, and worse than predicted for tunes grouped together in the rising class. These results suggest a robustness hierarchy of tune distinctions with a primary rising/non-rising distinction. This hierarchy reflects holistic shape distinctions, but does not align with the proposed tripartite composition of tunes.