Aspects of the phonetics and phonology of Southern Sotho /a/

Abstract

Up until now it has been taken for granted that Southern Sotho only has one /a/ phoneme, without any allophonic variation. Barnard and Wissing (2008) indicated the possibility of an opposite situation. In this study we explored this suggestion in detail. We concluded that /a/ is indeed characterized by at least two distinct allophones. Such allophones are ascribed to a process of vowel-to-vowel coarticulation, a specific type of vowel anticipatory assimilation. According to this process, the articulatory features of vowel height and vowel backness of /a/ in the negative formative ha are changed to positions that are closer to that of the mid-high ‘e’ or high ‘i’ of the following syllables, specifically in constructions ha (ke (tlo)) and ha (di), and ha (ba), all followed by a verb. These articulatory features translate to the corresponding acoustic features F2 and F1 respectively. Compared to ‘a’ of ha (ba), taken as the base-line vowel, the differences with respect to both F1 and F2 (and also F2—F1) are statistically significant.

The current results do not support the theoretical view, as propagated by others such as Manuel (1990) with respect to the correlation between the density of a vowel system and the presence or absence of allophonic variation. Manuel presents her results of an investigation of the vowel ‘a’ in Sotho, Shona and Ndebele to be supportive of this theory, which predicts a maximum distribution of vowels in the acoustic vowel space in order to minimize perceptual confusion between vowel categories. We offer some possible explanations for her findings. Her use of unnatural experimental language materials within a carrier phrase instead of real language could possibly force the readers to pronounce this vowel more in isolation, that is, being directed by the orthography rather than in a natural way. It also seems rather unlikely that the Southern Sotho vowels ‘e’ and ‘o’ are rather far apart from ‘a’ in the acoustic vowel space to be a threat to allophonic expansion in the case of /a/, as predicted by this theoretical approach. Our results of an elaborate investigation based on natural Southern Sotho materials clearly refute this.

Barnard and Wissing (2008) offer information concerning allophonic variation of the ‘e’ and ‘o’ vowels of Southern Sotho that may be of great assistance in the development of Human Language Technologies such as automatic speech recognition systems or text-to-speech systems. The present results regarding similar allophones of ‘a’ add up to such a body of knowledge in the case of Southern Sotho.