Committee: Christopher T. Collins (co-chair), Gillian Gallagher (co-chair), Lisa Davidson, Andy Chebanne (University of Botswana), and Doug H. Whalen (CUNY Graduate Center; Haskins Laboratories)
This dissertation examines the acoustic phonetic properties of consonant-tone interaction in the critically-endangered and little studied Khoisan language Tsua, using speech production data from original field research in Botswana. Tsua exhibits an extreme tonal depressor effect that lowers a post-consonantal, root-initial High tone's production by 50 Hz or more as triggered by voiced obstruents, aspirated obstruents or the glottal fricative /h/, a typologically rare pattern. However, this inquiry finds that 19.3% of the root-initial High tones that are expected to be depressed by the triggering context are produced without depression. It is argued that tonal depression exceptions involving voiced obstruents correspond to historically nasal sonorants or nasalized clicks, consonants which are not depressors in Tsua synchronically or diachronically. Thus, voiced obstruents are depressors if they have not undergone a historic sound change.
To tease apart the acoustic effects by depressor type, F0 contour patterns are evaluated using Smoothing Spline Analysis of Variance (SS ANOVA), a domain-general statistical tool for the rigorous comparison of curves along multiple reference points (Gu 2002, 2013). Three conditions in the implementation of depressed High tones are tested via SS ANOVA on normalized F0 z-scores: (1) when the depressed High tone follows voiced versus aspirated obstruents and is followed by a Mid tone; (2) when the depressed High tone follows voiced versus aspirated obstruents and is followed by a Low tone; and (3) when the depressed High tone is followed by a Mid versus Low tone regardless of depressor type. The effect on Mid tones preceded by depressors is examined as well.
The results indicate that there is no significant difference for a root-initial depressed High tone whether it is produced following a root-initial aspirated obstruent or a root-initial voiced obstruent and when it precedes a Mid tone. There is a significant difference between the F0 onsets and maxima for a root-initial depressed High tone as produced following aspirated obstruents as opposed to voiced obstruents when it precedes a Low tone. There is also a significant difference between a depressed High tone when it precedes a Mid versus Low tone regardless of depressor type. There is no evidence to suggest that Mid tones undergo depression in Tsua. It is argued that the F0 shapes of the tonal melodies are more important for discrimination as opposed to whether they are produced at a significantly higher or lower Hz value.
Tsua tonal depression may be the result of a Low tone insertion rule. However, the purported rule is not completely isomorphic as the Low tone insertion rule for depression in Ikalanga proposed in Hyman and Mathangwane (1998). Moreover, the Tsua depression pattern is not fully accounted for by the [L/voice] feature in Bradshaw's Multiplanar model of consonant-tone interaction. Formally, I account for the Tsua depression pattern via the Low tone insertion rule in (a).
(a) Tsua tonal depression rule
∅ → L / [-sonorant, +slack] ___ H [-H] #
The rule in (a) proposes that [+slack] is the key feature that unifies the depressor types pending further acoustic and articulatory supporting evidence. What is important is that a Low tone insertion account is plausible given modifications to previous proposals.
It is unclear whether the rule in (a) is present in Tsua's phonology or if the current depression pattern is a remnant of Tsua's phonology from long ago. As a first step towards determining whether the rule is synchronically active, a two-part tone Wug Test is proposed for a future experiment. If the rule is found to be productive, this raises the interesting possibility that click replacement may interact with rule application in some way. Such a possibility has not been explored in the Khoisan literature. The findings reported in this dissertation expand our knowledge of tonal phonetics by showing what is possible in a typologically rare tone system and highlight the importance of statistical methods in phonetic fieldwork.