Domain-initial Strengthening in Taiwanese: A Follow-up Study, Exams of Japanese Language

Download Domain-initial Strengthening in Taiwanese: A Follow-up Study and more Exams Japanese Language in PDF only on Docsity! 1 Domain-initial strengthening in Taiwanese: a follow-up study Wendy Hayashi1, Chai-Shune Hsu2 and Patricia Keating Introduction In a previous study (Hsu & Jun 1998; reported in Keating et al. 1998, this volume) we investigated whether Taiwanese consonants show domain-initial strengthening of consonant articulations of the sort we have found in other languages. Taiwanese is of interest in this regard because its prosodic domains are less well-established than in other languages. Most previous work on the prosody of Taiwanese (and related dialects, e.g. Chen 1987) has concerned the tone sandhi group (the domain in which tone sandhi takes place), but this domain is not strictly layered under the Intonational Phrase, and Hsu & Jun (1996) found no strengthening of consonants at the beginning of the tone sandhi domain. Therefore Hsu & Jun (1998) compared the articulation of the consonant /t/ when initial in only three prosodic domains: Syllable, Word, and Intonational Phrase. The measures of articulation were peak linguopalatal contact, Voice Onset Time, and acoustic closure duration. The result of the experiment was that, while both speakers distinguished all three domains by duration, and neither speaker distinguished them by VOT, only one of the two speakers distinguished all three by contact. The other speaker made no articulatory difference between Word-initial and Intonational-Phrase-initial /t/s. While this result was unexpected, Keating et al. were reluctant to conclude that Taiwanese behaves differently from other languages studied, because the Taiwanese study was the smallest of the set, and the corpus was not designed to be exactly parallel to those of the other studies. However, it would certainly be an interesting result if Taiwanese, a tone language, were in fact different from the other languages in this respect. Therefore a second study was undertaken, designed to be more like those of French and Korean (this volume). To this end, additional prosodic domains were identified for Taiwanese, and an additional consonant was included. Methods Two subjects participated in the study, one female (Speaker 1, the second author) and one male (Speaker 2). These are the same subjects as in the original study. The test domains began with unaspirated /t/ or /n/, followed by /a/, and were also preceded by /a/. The corpus for /t/ consisted of sentences containing real words, and is shown in Table 1. In some cases the target syllable was underlyingly /ta!/; however, syllables closed by /!/ become open syllables within the tone sandhi domain. The corpus for /n/ consisted of reiterant versions of the /t/ corpus, in which all syllables in the model sentences were instead pronounced as /na/. Both speakers were able to produce reiterant speech without difficulty. The domains included in this study, along with the test sentences in the /t/ corpus, are listed in Table 1. The Syllable and Word domains differ from the previous study in that our test 1 Department of Linguistics, UCLA 2 Now at Nuance Communications 2 words are not reduplicated forms of the test syllables. The Intonational Phrase domain in our study is specifically defined as not set off by pause, but only by a large break. The Utterance domain in our study, as in Cho & Keating’s Korean study (1999, this volume; also Keating et al. 1998 and this volume) is defined as an Intonational Phrase set off by pause; in Taiwanese it may also be marked by a small rise in pitch at the end. The other new domain in the present study is a small phrase, consisting of a heavy subject Noun Phrase. This domain is not tonally marked, but is characterized by a break greater than that between words. ----------------------------------------------------------------------------- Table 1. Corpus for /t/. Taiwanese forms are given in IPA plus punctuation. The test word is in boldface and the test syllable is underlined. Tones are marked only for the test syllable and the syllables on either side of it. Domain Taiwanese English meaning Utterance wa u k'ua(-ti+ papa55. ta23ta55 k'ai ia! be lai? I can see Dad. Why isn’t Tata here yet? Intonational Phrase wa k'ua(-ti+ a! papa55, ta23ta55 k'ai ia! be lai? I see it. Dad, why isn’t Tata here yet? Small Phrase hit e la2 e papa55 ta31-ti$31 tsit-tsia katsua! That person’s dad stepped on a cockroach. Word wa ka li k42, papa55 ta31-ti$31 tsit-tsia katsua! Let me tell you, Dad stepped on a cockroach. Syllable wa kina kua((-ti+ ta33ta33 tsim53a k+ i2 kia(. Today I saw Auntie Tata and her child. ----------------------------------------------------------------------------- Speaker 1 read ten repetitions of each test sentence containing /t/ and six repetitions of the reiterant versions with /n/. Speaker 2 read fifteen repetitions of each test sentence containing /t/ and ten repetitions of the reiterant versions with /n/. Electropalatographic and audio signals were recorded with the Kay Elemetrics Palatometer, as in all our previous studies. Peak EPG contact was measured over the whole palate: the percentage of electrodes contacted in each data frame was calculated, and the frame with the most contact was identified for each test consonant. Also, for each token, the number of data frames showing a complete stop occlusive seal was counted. This gives a measure of articulatory duration. Two acoustic measures were also taken. The first was the VOT of /t/, as in previous studies. The second was a set of formant frequency measures: F1 and F2 were measured at the mid-point of the vowel following the test consonant, using LPC with a 25-ms window. Thus we could test whether the vowel of a domain-initial syllable shows any effect of its prosodic position. Analysis of Variance was followed by Fisher’s PLSD post-hoc tests in pairwise comparisons of the different prosodic positions, for each speaker and each consonant.