Intonation effects on Cantonese lexical tones in speaking and singing

Abstract

Cantonese lexical tones are preserved in both speaking and singing (which can be regarded as a special kind of speech style). Through three main acoustic experiments, the present thesis is a systematic study of the intonation effects on Cantonese lexical tones in speaking and singing. In Experiment I, three pairs of speech styles were compared: opera speaking versus opera singing, opera speaking versus normal speaking, and normal speaking versus normal singing. A perception test further examined the conclusions of the acoustic experiments from the perspective of listeners. The results of Experiment I showed that speaking and singing speech styles are different mainly in pitch slope. Speaking speech styles are associated with a declining intonation, while singing speech styles tend to be level or even show a slightly ascending tendency. The experimental data also indicate that a higher pitch register is usually employed when a speech type is considered more conspicuous, such as opera speaking (compared with normal speaking) and normal singing (compared with normal speaking). Experiment II focused on the comparison between speaking and singing by layman speakers. Twelve native speakers of Cantonese were asked to read aloud and sing a set of lyrics taken from Cantonese songs. The data analysis was conducted in the framework of a matrix of the two dimensions of pitch configuration and pitch domain. At the syllable level and the utterance-body portion, speaking has a more pitch slope declination, a lower pitch register, and a wider pitch span, while singing has a more ascending pitch slope, a higher pitch register, and a narrower pitch span. Pitch slope at both the syllable and the utterance levels exhibits a downdrift pattern for speaking and an updrift pattern for singing. At the utterance-final portion, speaking shows a strong utterance-final declination effect while singing does not exhibit an obvious utterance-final effect. Experiment III was conducted in the same framework as Experiment II, but with focus on the internal differences of normal speaking. Four sentence types were compared: declarative without a sentence-final particle (SFP, henceforth), declarative-derived question without an SFP, declarative with an SFP, and declarative-derived question with an SFP. The results showed that there is a declination effect in the domains of the syllable level and the utterance-body portion. For utterances without an SFP, the utterance-final intonation is very different: in a declarative, the final syllable has a significantly lower pitch register and a slightly more declining pitch slope; in a question, the final syllable keeps the pitch register unchanged but has a sharply rising contour. For utterances with an SFP, the SFP carries the burden of utterance-final intonation. The utterance-final intonation can be regarded as a kind of segmentless SFP with floating pitch contour to be superimposed to the final syllable. The present study enriches the knowledge of intonation typology and provides further insight into the relationship between lexical tones and intonation. In particular, the updrift intonation of singing fills the vacuum of non-declination in intonation types. The matrix of pitch configuration and pitch domain proposed here is a new model for experimental studies on intonation in tonal languages.