The Weekly Newsletter of MIT Linguistics

Linguistics Colloquium 10/24: Larry Hyman

Larry Hyman (University of California, Berkeley)

Title: Tonal and Non-Tonal Intonation in Shekgalagari

Time: Friday, October 24th, 2008, 3:30pm, Room 32-141

There will be a party in honor of Larry, beginning at 6:30pm, at Donca’s place.

The study of intonation in a (fully) tone language presents both a challenge and an opportunity. The challenge is to see how a language which exploits F0 mainly for the purpose of lexical and grammatical contrasts succeeds in encoding the functions often expressed by means of intonation in non-tonal languages. As is well-known, word-level distinctions can be quite rich in tone systems, which contrast up to five pitch levels and a dozen or more tonal contours or clusters, e.g. Wobe (Kru; Liberia) (Bearth & Link 1980, Singler 1984). Word-level tones, in turn, can be subject to considerable manipulation by the postlexical phonology, where juxtaposition, syntactic conditioning, or phonological phrasing can modify the word-level inputs and introduce additional pitch features, e.g. the phrase-final H% boundary tone of Kinande (Bantu; DRC) (Hyman 1990). Particularly when tone systems are complex in these ways, the question is how there can be much room left for intonation to modify or add pitch specifications without obscuring the word-level tonal contrasts. On the one hand, there are specific strategies that different tone languages employ to keep tones and “intonemes” separate. For example, in Mazahua (Otomian; Mexico), there are no lexical tonal contrasts on the last syllable of a word. As Pike (1951:101) puts it, “The pitches of all syllables which do not immediately precede word space are those of the tonemic system. The pitch of any syllable immediately preceding word space is part of the intonemic system.” On the other hand, there are languages where intonation clearly overrides lexical tones. In Coreguaje (Tukanoan; Colombia), for instance, CVCV nouns contrast H-H, H-L, L-H, and L-L tones. However, when such nouns occur in isolation, their tones merge as L-HL with statement intonation and H-L with question intonation (Gralow 1985). If these two strategies can be termed “accommodation” vs. “submission”, a third option is “avoidance”: In many languages with highly developed tones systems there doesn’t seem to be “structured” as opposed to what Ladd (1996) terms “paralinguistic” intonation, e.g. the raising or lowering of pitch associated with excitement, fear, etc. Can a language do without such structured intonation, and if so, what does it put in its place? The strongest limiting cases are probably languages with highly developed tone systems. This constitutes the opportunity side of the above-mentioned challenge: Since they have more reason to resist, tone languages offer a particularly appropriate forum for investigating the essential properties of intonation, e.g. the universal tendency to phonologize Gussenhoven’s (2005) “three biological codes”.

In this paper we take a close look at the intonational properties of Shekgalagari, a Bantu language of the Sotho-Tswana (S.30) subgroup spoken in Botswana. We begin by presenting the tone system, which has two underlying tones /H, Ø/ and a derived phonemic downstep (!H). We then turn to the relatively rich intonation system. We start with what we term “declarative” penultimate length + L tone (PLL): the vowel of a prepause-penultimate syllable is lengthened and a marked L is assigned to its second mora. We then go through the different contexts where PLL typically fails to occur: (1) yes-no and WH-questions; (2) imperatives and hortatives (e.g. ‘may he enter!’); (3) vocatives and exclamatives (e.g. ‘what a fool!’); (4) paused lists (e.g. ‘I ate corn… rice… and beans’); (5) ideophones (e.g. ‘it went splash!’). While other Southern Bantu languages have also been noted to “suspend” penultimate lengthening in questions, Shekgalagari is unusual in having so many other contexts where the vowel remains short. Recall that the declarative not only lengthens the penultimate vowel, but also assigns a L tone. It is therefore striking that all five of the above are speech act types where speakers might be expected to raise their voice, and hence resist the L tone. We claim therefore that non-declarative = unmarked in Shekgalagari: penultimate lengthening (+ L) is not suspended, but rather is not assigned in questions, imperatives, vocatives etc. (Shekgalagari also differs from related languages in not assigning PLL when the prepause word is monosyllabic.)

After illustrating the above speech act types, we discuss their interactions. For example, words which occur in paused lists may optionally lengthen their final vowel in the declarative, but not in questions or imperatives. Also, polysyllabic ideophones obligatorily devoice their final vowel, but not in hortatives (‘may it go splash!’) and questions (‘did it go splash?’). We then discuss two types of intonational overlap. The first, termed “emphasis”, allows PLL to be assigned in all of the above contexts except (1), with varying results, e.g. making a WH-question or imperative seem more like a statement, repeating the question or command in exasperation, or other “emphasis”. Although receiving PLL, the intention of these utterances remains clear, since the speech act is still decipherable from the various redundancies, e.g. a WH word, the lack of a subject in an imperative, PLL on the final word of the paused list, etc. However, if PLL is assigned to a yes-no question, there is no such redundancy, and the result is a statement. In addition, an intonation which we call “urgency” takes the declarative form with PLL and raises the whole register, exaggerating the pitch intervals (e.g. ‘Fire!’, ‘Thief!’). We interpret this as a case of Ladd’s paralinguistic intonation.

After demonstrating that Shekgalagari has a rich and interesting intonational system, we draw two conclusions. The first is that the so-called avoidance strategy needs to be refined: A tone language may resist a tonal implementation of intonation, which poses potential complications, but there are other intonational features that do not necessarily have to be avoided: lengthening, devoicing, glottalization, breathiness, and even nasalization. Perhaps we have been too tonocentric? On the other hand, Shekgalagari confirms the strategy of accommodation: While many Eastern and Southern Bantu languages have penultimate lengthening, it has long been recognized that such lengthening exists only in languages which have lost the historical lexical vowel-length contrast. (Languages which have preserved the contrast would, in turn, tend to avoid penultimate lengthening, which could merge long and short vowels.) Although the effect of the L of PLL is rather noticeable (short H-H alternates with long HL:-L), here too there is no loss of contrast, and similarly for the other intonations. We end with some speculations on the limitations on what features can be used for intonation and on how lexical contrasts accommodate, submit to, or avoid intonational competition.