Blurb:

Clements argued that languages have a tendency to sequence phonemes such that there is maximum difference between the onset and nucleus of syllables. This phenomenon has been labeled the Sonority Sequencing Principle (SSP).













This chapter describes a study which investigated the SSP in Zulu, a Southern Bantu language spoken in South Africa. ‘Ukwabelana’, an open source corpus of Zulu words developed by the University of Bristol contains over 15 000 items (Spiegler, van der Spuy, & Flach, 2010), provided a large pool of data.













A more precise characterization of ‘tendency’ was needed before analysis could proceed. I selected concepts from the discipline of statistics to arrive at a clearer statement of this construct viz. mode as a measure of central tendency and normal distribution as a description of ranking patterns among syllable shapes (after Hinton, 2004). The definition of the sonority sequencing tendency which informed this study was thus: when considering sonority transition patterns across a large number of syllables in a language, the mode configuration (i.e. most popular configuration) will be the ideal (following Clements, 1992, maximum sonority difference between onset and nucleus). All non-ideal configurations should be ranked below the mode; as configurations move further and further away from the mode, fewer and fewer syllables should display these configurations i.e. configurations will be normally distributed from the mode.













For the purposes of this study, I used the sonority rankings provided in Ball, Müller, & Rutter (2010) supplemented by information from a variety of other works (e.g. Cruttenden, 1992). Based on this hierarchy, it was possible to calculate sonority transition scores for syllables. This score was operationalized as the difference in sonority ranking between the onset and the nucleus of a syllable (since the corpus to be analyzed in this study is a Zulu one, and Zulu is a typical Bantu language that does not permit CVC syllables, the analysis was confined to onsets and nuclei).













Transition score distributions according to the predictions of the SSP were compared to transition score distributions actually present in the data. Neither the mode or anti-mode configuration predicted by the accepted conceptualization of sonority matched the mode or anti-mode configurations present in the data. The Spearman’s rho and consequent R2 statistic calculated to determine the association between the predicted and actual arrays was consistent with a very low correlation (R2=0.05).













The tendency of Zulu to sequence syllables such that maximal sonority differences exist between onsets and nuclei would appear to be a very minor one. If some universal pattern of syllable formation does indeed exist, this analysis seems to suggest syllable shape distributions are the product of an organizing principle other than sonority sequencing. Claims that rigid sonority patterning tendencies are ‘hardwired’ into the brain (Christman, 1992; Sussman, 1984) seem less defensible when such sonority trends are absent from natural language data. Similarly, the argument that acoustic peaks and valleys at the borders between syllables help listeners to comprehend the speech signal (Ohala, 1990) seems to deserve less credence, at least in respect of Zulu speakers. Other mechanisms (semantic and syntactic forms of utterances, etc.) should be considered before an accurate picture of how speech sounds are decoded can emerge.