Realisations of a single high tone in Northern Sotho

This article reports on a production study that investigates the realisation of a single high tone in the verbal constituent in Northern Sotho, a Bantu language spoken in South Africa. The parameters of variation investigated are based on existing descriptive and theoretical literature and relate to numbers of syllables in the verb stem, morphosyntactic constituency and verb-internal morphological boundaries. The results are interpreted as both phonological and phonetic influences on high tone realisation in this language: phonologically, a high-toned object concord causes a peak shift one syllable to the right. Phonetically, the study shows that the F0 peak associated with a high tone is not necessarily reached within the syllable carrying the high tone but only later (peak delay) depending on the segmental make-up of the tone-bearing syllable and its position within the utterance. The segmental make-up of the tone-bearing syllable leads to systematic surface variation in tone realisation. By collecting controlled acoustic data on tone realisation, this study provides a ground for cross-dialectal comparison of Southern Bantu tone. Introduction The current study reinvestigates the phonetic realisation of a single high tone in Northern Sotho, a Southern Bantu language belonging to the Sotho-Tswana group. It analyses acoustic data from four speakers of the same age group1 using stimuli which are controlled for interfering segmental influences. Parameters of variation within the stimuli set are: tense-mood-aspect, length of the verb stem (ranging from disyllabic to quatrosyllabic verb stems), and morphological origin of the single high tone (ranging from subject concord, tense marker, mood marker, object concord to verb stem). These parameters have been reported in the literature to influence the realisation of high tones in Bantu languages in general. The current study does thus not aim at providing an exhaustive investigation into the overall tone system of Northern Sotho. Coherent descriptions are available by Lombard (1976) and Ziervogel et al. (1969), as summarised in Table 1 below. The aim of the present study is rather the investigation into the detailed realisation of one single high tone in the verbal domain, factoring out potential interferences with respect to stimuli, speakers or transcribers. This is done through controlling for segmental and grammatical contexts, using acoustic data, and collecting several repetitions as well as several speakers, which nevertheless form a homogeneous group with respect to geographical origin and age. Trends observable in all speakers are the main focus of our interest, as those are the ones that are most likely characteristic of the variety as a whole. Despite our empirical approach, ambiguities in the interpretation of these data remain, and we try to give reasonable interpretations in those cases. The data collected in the current study are transcribed using the familiar, categorical notation of high and low tones. When comparing the transcriptions resulting from the interpretation of the acoustic data to the existing descriptions of corresponding structures in other Sotho-Tswana varieties (see the next section), it emerges that there is no perfect match to any reported variety. Instead of postulating a new variety with unique tonal features, the study is meant to advocate the methodology used and to encourage dialectal studies into tone that gather data that can be compared across dialects, speakers and transcribers. Zerbian and Barnard 358 However, besides describing the tonal patterns found in the four speakers who participated in our study, the current article also differentiates between the phonological and phonetic factors that interact in the surface realisation of tone in these contexts in Northern Sotho. Whereas the exploration of the phonetic-phonology interface by means of acoustic studies is not a new methodology as such, it is hardly ever applied in Bantu tonology (with the notable exception of Myers, 1998b, 1999, 2003 on Chichewa and Kinyarwanda). In the current study morphosyntactic constituency and length of the verb stem are investigated as the independent variables in high tone realisation based on findings from the theoretical literature. Previous acoustic research has concentrated on high tone alignment in relation to syllable length (Myers, 1999, 2003). In the current study, syllable length is kept constant. The remainder of the article is structured as follows: the next section summarises tone in Sotho-Tswana by reviewing the existing tonal descriptions with respect to the constructions investigated in the current study. The section following then describes the production study that was carried out in order to investigate the phonetic realisation of high tones in the speech of four speakers of Northern Sotho. The results are then presented; and the final section discusses the results both with respect to the phonetics-phonology interface of tone as well as with specific reference to Northern Sotho. Tone in Sotho-Tswana This section lays out the phonological and phonetic characteristics of the Sotho-Tswana tone system that pertain to the alignment of a single, phrase-medial high tone that is surrounded by low tones. The realisation of high tones which are adjacent to other high tones shows different characteristics and is dealt with in a separate study for Northern Sotho (Zerbian & Barnard, unpublished data). Basic characteristics The languages of the Sotho-Tswana group share the basic tonal characteristics with the majority of Bantu languages. They use tone to distinguish grammatical and lexical meaning. Their tone inventory consists of two tones, H and L. The high tone (H) is the active tone in the languages of the Sotho-Tswana group as it participates in tone spread and is subject to positional restrictions, such as the avoidance of two adjacent high tones immediately next to each other (so-called OCP context) and the banning of phrase-final high tones (see Kisseberth & Odden, 2003 for a general discussion of these aspects). Low tones (L) can appear next to each other without inducing any tonal changes and they also appear on phrase-final syllables. The current study focuses on the realisation of high tones because high tones are the ones that are assumed to be specified underlyingly whereas low tones are considered to be either inserted late in the phonological derivation or only implemented phonetically. The view of underspecification of low tones has found phonetic support in Myers’ (1998b) acoustic study of low tones in Chichewa, a Bantu language spoken in Malawi. As high tones are specified in phonological representation they have a phonetic pitch target. Low tones, being underspecified phonologically, lack a phonetic pitch target and can be derived via (sagging) transitions between high tones (Myers, 1998b; but see Xu & Xu, 2005 on English intonation; Chen & Xu, 2006 against sagging transition as a viable articulatory mechanism). Similarly to other Bantu languages, the languages of the Sotho-Tswana group show an asymmetry in the tonal characteristics of their noun and verb system with nouns being more tonal than verbs. However, Bantu languages are agglutinative languages and can thus have a complex, extended verb word in which tones interact with each other morphologically. The realisation of a single high tone within the morphologically complex verb word is the focus of the current study. Determining factors in tone alignment There are various parameters that are known from the literature to influence (high) tone realisation in Bantu languages. These are grammatical aspects such as polarity (negative vs positive) and tense-aspect-mood; pragmatic aspects such as focus and sentence type; prosodic aspects such Southern African Linguistics and Applied Language Studies 2009, 27(4): 357–379 359 as vowel length and size of the tonal domain; and other aspects such as position within a phrase and morphosyntactic constituency. In the current study, only tense-aspect-mood, size of the tonal domain, and morphosyntactic constituency have been varied in a controlled way. All other parameters have been kept constant, as they did not constitute the focus of the current study. Polarity has been kept constant by using only positive target sentences; sentence type by using only positive statements; focus by only using un-focused verb words (Northern Sotho uses non-prosodic means of focusing, see Zerbian, 2006); and position within a phrase by having the target word appear in phrase-medial position via an object or adverb following it. As for vowel length, Northern Sotho does not have phonological vowel length. Only the three parameters relevant for the current study are discussed in the following. Morphosyntactic constituency The verb in Bantu languages shows a rich agglutinative structure, which can be described by the template in (1), where SC and OC refer to subject and object concord markers respectively, and Ext refers to verbal extensions such as causative, applicative or reciprocal. Optional elements are given in brackets. (1) SC – (Tense, Mood, Aspect) – (OC) – verb root – (Ext) Much work (e.g. Barrett-Keach, 1986; Myers, 1987, 1998a) has argued that the Bantu verbal constituent is ‘split’ into two distinct morpho-phonological constituents: the Inflectional stem, comprising subject concord marker (SC) and tense, mood, and aspect prefixes (TMA); and the Macrostem, comprising the object concord marker (OC) and the verb stem with extensions. The proposed structured is shown in (2). (2) Bantu verbal structure (adapted from Myers, 1987; Downing, 1999) The Stem consists minimally of the verb root ending in a final suffix. Derivational suffixes can occur between root and final suffix. The suffix structure is captured in somewhat simplified fashion by ‘extensions’ in (2) as these do not form the focus of attention in the present article. If present, the object concord forms a morpho-phonological constituent with the verb root, termed Macrostem following Hyman & Ngunga (1994). Without going into the morphosyntactic reasoning for adopting such a morphologically split verbal structure (see e.g. Barrett-Keach, 1986; Myers, 1987; Hyman, 1993; Mchombo, 1993; Mutaka, 1994 for evidence from various Bantu languages), the morpho-phonological reasoning for this split-constituent hypothesis is important for the realisation of tone. The stem is commonly subject to prosodic minimality restrictions. It constitutes the domain for vowel harmony, nasal consonant harmony, and restrictions on consonant and vowel distribution (Downing, 1999, 2001). Furthermore, there is tonal evidence that the morphological constituents INFL, Macrostem, and Stem also Verbal constituent