UCLA

This dissertation investigates several aspects of the phonology of Avatime, a Kwa language spoken in the Volta Region of southeastern Ghana. The goal of the dissertation is to use data from original fieldwork conducted by the author to update and supplement the empirical description of Avatime phonology, and to situate several aspects of Avatime phonology in a contemporary theoretical context. The phenomena investigated are (i) the status of the [ATR] contrast in high vowels, (ii) the behavior of the low vowel [a] in [ATR] harmony, and (iii) a tone sandhi process in verbs that is phonologically, morphologically, and lexically conditioned. While these aspects of Avatime phonology have all been investigated in previous work on the language, this dissertation identifies some differences in that are relevant to the description of the language and for phonological theory. Chapter 1 provides an overview of the dissertation, as well as background on the Avatime language and its speakers. Chapter 2 addresses the status of the ATR contrast in high vowels in Avatime. This contrast has long been the subject of debate in the study of Avatime. Some earlier work on the language (Ford 1971) did not mark a contrast between [+/-ATR] vowels in surface forms, but a more recent phonetic study (Maddieson 1995) showed a consistent contrast along the dimension of F1. More recent work on the language (van Putten 2014) has suggested that the contrast is disappearing, especially among younger speakers. This chapter replicates and extends Maddieson’s study, finding that in aggregate, there is still an ATR contrast among high vowels, even for younger speakers. However, this contrast is not produced by all speakers in all contexts. The implications of the loss of contrast in progress for the phonology of Avatime is discussed, specifically the implications for theories of abstractness in phonology (Kiparsky 1973). Chapter 3 develops an analysis of the behavior of the low vowel [a] in ATR harmony in Avatime. This vowel shows an asymmetry in whether it participates in ATR harmony or not: in noun class prefixes and subject agreement prefixes, it harmonizes with the ATR value of the root, re-pairing with the mid vowel [e] in these contexts. In enclitics, however, the low vowel is invariant. Additionally, there is a set of verbal prefixes that have an exceptionally invariant [a]. This chapter argues that these facts can be accounted for in Harmonic Grammar (Smolensky 1986; Legendre et al. 1990; 2006). Specifically, it is shown that the behavior of [a] can be explained as a case of ganging, in which the weights of multiple constraints that disprefer vowel harmony combine to overcome the weight of the constraint that drives harmony. Chapter 4 examines a process of tone sandhi affecting verbs. This process consists of the raising of the tone of verb roots and prefixes in a scalar fashion. Taking Ford’s (1971, 1986) work as a starting point, the chapter presents an overview of this phenomenon in contemporary Avatime. It is shown that the process is conditioned by phonologically, morphological, and lexical information. This type of process, in which multiple sources of non-phonological are required, resembles the cases discussed by Sande (2020) of morphologically-conditioned phonology with two triggers. However, this chapter argues that the tone sandhi process is better analyzed as a case of phonologically-conditioned allomorphy. An analysis is developed using the framework developed by McPherson (2019), in which the lexicon contains generalized frames listing tonal allomorphs for the different classes of verbs. It is shown that this accounts for some previously undocumented aspects of the tone raising process, especially the fact that it may affect words adjacent to the verb itself. Chapter 5 concludes and discusses future directions for research in Avatime phonology both on and beyond the topics addressed in this dissertation.