In tone languages, tone is the use of pitch to contrast word meanings; this raises questions of whether, and if so, in what ways and to what extent, the pitches and pitch patterns of tones are preserved when they are sung. This thesis explores the relationship between lexical tones and musical pitch in Chaozhou, a Southern Chinese language characterised by eight tones in isolated speech and even more tonal variations in connected speech. The study employs multiple approaches, including corpus analysis, field research, an observational study, and a production experiment.

The corpus analysis measured the correspondence between tones and melodies in Chaozhou songs. A high degree of tone-melody matching was found: 89% in folk songs and 78% in contemporary songs. It also suggests that singing in Chaozhou follows the rule of tone sandhi (tone alteration in connected speech) observed in Chaozhou speech; tones are more likely to be categorised into high-, mid-, and low-pitch groups, by emphasising the tone-pitch extremes (rather than the ending pitch of tones found in Cantonese songs). The transition from the first note of one melisma to the first note of the next has the most consistent influence on tone-melody matching.

The corpus analysis of notated songs illustrates how tones influence song melody structurally. However, unresolved questions remain regarding how and in what way the tonal patterns are preserved through un-notated musical information during vocal performances. Subsequent fieldwork was conducted to interview Chaozhou singers and record their singing. Analyses of the same song sung by five different singers suggest that the singers may incorporate their spoken language experience into the notes they sing. The result shows that they consistently realised contours of falling tones /53/ and /42/ within individual notes, no matter how they learned to sing the song.

An observational study was performed subsequently to collect a larger dataset, enabling the analysis of ten specific tones sung by 34 Chaozhou singers. Results show that tones were realised to varying degrees when they were sung. Vocal training affected the realisation of falling tones /53/ and /42/ by reducing the extent of the pitch fall; it also affected rising the tone /23/, resulting in a smaller pitch change. However, singers’ prior experience of singing in the language did not greatly affect the realisation of tone. An effect of metrical structure was found only for tones /53/ and /42/. Additionally, there was a tone sandhi effect on the realisation of tones /53/ and /21/. The analysis also hints at the potential effect of melodic context on the pitch change of tones when sung.

A follow-up production experiment examined the effect of melodic context on the realisation of tones /53/ and /35/. Results indicate that tonal contours interact with the melodic context where the tones are sung. The pitch variation in the falling tone /53/ is correlated significantly with the succeeding interval, whereas the pitch variation in the rising tone /35/ is significantly associated with the preceding interval. It was found that increasing the interval sizes leads to greater pitch variations in the sung tones, with each interval increment contributing to a significant difference of nearly half a semitone in pitch variation. The findings also suggest that vocal training affects the realisation of the falling tone /53/ only.

The study concludes that pitch serves a semantic function in Chaozhou songs and habituation to lexical pitch may contribute to the realisation of tones in performances. The concept of authenticity in Chaozhou songs and its connection to cultural identity and boundaries further underscores the significance of tone realisation in Chaozhou singing. Although this study has made significant contributions to the subject of singing in tone languages, more research on the interaction between speech and music remains to be explored.