The so-called definitive accent (DA) in Tongan has been analyzed in various ways in the literature: as stress shift from penultimate to final vowel, as simultaneous stress reduction on a penult and stress addition on an ultima, and as addition of a syllable by repetition of the final vowel. This study investigates each of these analyses empirically in order to establish the phonology of DA in Tongan. Our findings support Melenaite Taumoefolau's proposal that definite NPs are formed by repetition of the NP-final vowel, and thus a morphological analysis of DA as reduplicative suffixation. Moreover, our findings substantiate an account of Tongan in which stress is unexceptionally penultimate in a foot, and in which "long vowels" and "diphthongs" are to be considered sequences of two syllables, as suggested by Taumoefolau.
1. INTRODUCTION. (1) Unlike most Polynesian languages, which indicate the difference between definite and indefinite noun phrases (NPs) morphologically, Tongan has frequently been analyzed as using a phonological means of differentiating between the two types of NPs. (2) However, the nature of this phonological marking has been controversial. On one hand, Churchward's 1953 Tongan grammar treats the difference as a stress shift: "Normally the stress falls on the last vowel but one. Examples: mohe, to sleep, mohenga, bed; hae'i, to tear, mahae, torn.... In the following cases, however, it falls on the last vowel: (a) When the last vowel is long. Thus, in kum[bar.a], rat, it falls on the a; in hang[bar.e], to be like, on the [bar.e].... (b) Before an enclitic (3).... (c) In what we shall call the definitive accent" (4). "When a nounal or pronominal group is definite, ... the main stress is shifted from the last vowel but one to the very last vowel ..." (6-7).
On the other hand, Taumoefolau (2002) proposes that "definitive accent is really the repetition of the final vowel" (349). Stress assignment is unexceptional, in this view, with main stress occurring on a penultimate vowel. The relevant operations for a word containing short vowels are schematized in (I).
(I) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
In this paper, primary and secondary stress are indicated following IPA convention; that is, by superscripted and subscripted lines, respectively (e.g., [??]mafa[??]tua, sneeze'). Also, while a long vowel is variously symbolized in the literature as [bar.V], V., or VV, here we use either standard Tongan orthography (usually [bar.V]), or VV as necessary, for clarity of exposition in discussing the location of stress.
The central goal of the current study is to establish the phonology of definitive accent in Tongan by empirically investigating the claims of Churchward, Taumoefolau, and others. Does definitive accent involve rightward stress shift, or the addition of a vowel, or some other mechanism? The paper is organized as follows. Section 2 presents relevant morphological characteristics of Tongan that have to do with the referentiality and definiteness of NPs, and introduces the phenomenon known as definitive accent (hereafter DA). Section 3 reviews previous analyses of DA in Tongan. Section 4 summarizes this study's experimental methods and procedures for analysis. We present results in section 5, and discuss implications for the phonology of Tongan syllable structure and metrical stress in section 6. Concluding remarks are given in section 7.
2. RELEVANT TONGAN MORPHOLOGY
2.1 INDEFINITE ha AND DEFINITE e. Tongan has two articles, ha and e, (4) that are traditionally called the "indefinite article" and "definite article", respectively. These labels, however, are somewhat misleading. To be more precise, the two articles are distinguished from each other in terms of referentiality rather than definiteness. The article ha simply indicates nonreferentiality. In this sense, ha corresponds to English any, rather than the English indefinite article a, in that the latter implies a referential, indefinite, and singular NP. …