ABSTRACT of paper presented at the 5th Conference on Laboratory Phonology, Northwestern University, July 1996, and submitted for publication in the published conference volume. WHAT IS A STARRED TONE? Evidence from Greek Amalia Arvaniti (amalia@zeus.cc.ucy.ac.cy) * D. Robert Ladd (bob@ling.ed.ac.uk) + Ineke Mennen (ineke@ling.ed.ac.uk) + It is generally assumed in autosegmental descriptions of intonation (e.g. Pierrehumbert 1980 on English, Grice 1995 [henceforth G95] on Palermo Italian) that one tone in a pitch accent is somehow central and the other (if present) is peripheral. Yet the intuitively sensible notion of a central or "starred" tone is hard to define in a way that is both phonetically explicit and general enough to cover a range of languages. This paper aims at a language-independent definition consistent with instrumental phonetic evidence. In past work, the "centrality" of a starred tone has been identified with (1) metrical strength (G95) and (2) phonetic centrality, i.e. occurrence of the tonal target within the time span of the accented vowel (G95, Silverman and Pierrehumbert 1990 [henceforth SP90]). Though the connection to metrical strength is interesting, by itself G95's proposal is just a translation from one abstract notion to another and cannot serve as the phonetic basis for a definition. Phonetic centrality is more promising, as it seems to work for American English and Palermo Italian. However, it is not universal: in Glasgow English nuclear accents and Modern Greek prenuclear accents, a L tone is aligned just BEFORE the accented syllable and a H tone is aligned just AFTER the accented syllable, so that during the accented syllable itself the pitch contour rises steadily. Neither tone is phonetically central. How should such cases be accommodated in a general theory of starred tones? In a previous paper (Arvaniti and Ladd 1995 [henceforth AL95]), the designation of the starred tone was based on relative PHONETIC INVARIANCE of alignment and scaling. AL95 showed that L tones in Greek prenuclear accents, though they occur outside the accented syllable, are virtually invariant in their alignment, whereas the alignment of the H tone may vary with on the number of following unstressed syllables. AL95 also demonstrated invariance in the scaling of the L tone: the L does not depend on the number of unstressed syllables between H tones (as it might do if it merely reflected e.g. a "sagging transition" between accents). Given these preliminary findings, AL95 concluded that the Greek prenuclear accent must be L*+H. Intuitively, however, this is unsatisfying. Impressionistically the accents sound high, not low rising like an English L*+H. Moreover, there remains the fact that the L is always aligned outside the syllable, while the H may be aligned within the accented vowel in cases of adjacent accented vowels. In this paper we reexamine AL95's conclusions in the light of more extensive instrumental measurements of Greek prenuclear accents in controlled speech materials. We preserve the assumption that phonetic invariance is the correct basis for analysing a tone as starred, but now suggest that the weight of evidence actually favours the analysis L+H* for the Greek prenuclear accents. We will present instrumental evidence for the following specific conclusions. (1) The temporal invariance of the L tone is replicated: L is aligned on average 10-15 ms before the beginning of the onset consonant of the accented syllable. (2) However, when the number of following unaccented syllables is kept constant, the alignment of the H is also invariant. In words with antepenultimate stress, H occurs about 10-15 ms after the beginning of the vowel of the following unaccented syllable. We expect to present evidence (experiments in progress) of comparable regularities for words with penultimate or final stress, i.e. for very regular temporal adjustments of the sort that often arise from "tonal crowding" (cf. G95, SP90). (3) Moreover, the invariant scaling of L applies only if there is at least one unstressed syllable between adjacent accents. When no unstressed syllable intervenes, L is "undershot", while the scaling of the adjacent H tones is unaffected. Together, findings (2) and (3) suggest that H exhibits as much temporal regularity as L, and that H exhibits more scaling invariance than L. This makes it plausible (contra AL95) to treat H as the starred tone. Further evidence for this view may come from studying the scaling behaviour of tones when pitch range is expanded (cf. Liberman and Pierrehumbert 1984). Specifically, we expect (on the basis of preliminary evidence) to find that prenuclear L and H are both raised when range is expanded, but that the L tone in questions, which is impressionistically more like an English L*, is constant or lowered. For Greek, then, our findings suggest that invariance of alignment and scaling are indeed the best diagnostics of starred tones. It may be possible to extend this as the basis for a cross-linguistically valid definition. Alternatively, it may be that the abstract centrality of a starred tone, like stress, has a range of acoustic correlates that can differ in detail from language to language. In some languages or for some accent types, "phonetic centrality" (alignment within the accented vowel) may be more important than for others. References Arvaniti, A. & Ladd, D. R. 1995. Tonal alignment and the representation of accentual targets. ICPhS 13, vol. 4, 220-223. Grice, M. 1995. The intonation of interrogation in Palermo Italian. Tuebingen, Niemeyer. Liberman, M. & Pierrehumbert, J. 1984. Intonational invariance under changes in pitch range and length. In M. Aronoff and R. Oerhle (eds.), Language sound structure (MIT Press, Cambridge), pp. 157-233. Pierrehumbert, J. 1980. The phonetics and phonology of English intonation. MIT PhD dissertation, published 1988 by IULC. Silverman, K. & Pierrehumbert, J. 1990. The timing of prenuclear high accents in English. In J. Kingston and M. Beckman (eds.), Papers in Laboratory Phonology I (Cambridge, CUP), pp. 71-106. ************************* Addresses: * Dept. of Foreign Lgs and Lit, Univ. of Cyprus, POBox 537, 1678 Nicosia, Cyprus. Phone (+357) 2 368772, ext. 355. + Dept. of Linguistics, Adam Ferguson Building, Univ. of Edinburgh, Edinburgh EH8 9LL, Scotland. Phone (+44) 131 650-3961.