1 Tone Sandhi Production of YI in Mandarin Chinese Final Presentation Handout Wei-li Yang2 Introduction • The study: acoustic analysis of the alternations of yi in tone before the other lexical tones. • Yi : /ji/ ‘one’ • Hereafter: ji3 The Alternations of the Tone of yi (Chao 1948) Tonal Context Alternations Examples In isolation High jí, ar, sa n ‘one, two, three’ Followed by a falling tone (F) Rising ji kwai taŋ ‘one CL candy’ Elsewhere: followed by a high tone (H), rising tone (R), or falling rising tone (FR) Falling ji tşşşşi pi ‘one CL pen’ ji tʰjao ɕian ‘one CL thread’ ji tɕʰi ‘together’ • No other morpheme has this exact alternation pattern (Chao 1948).4 Introduction • Study on peak alignment variations in Mandarin tones (Xu 1998, 1999, 2001) - F: a relatively early high point in a syllable - R: a relatively late high point - H: high point is neither extremely early nor late - Peak alignment: F<H<R (< later than)5 Motivation • Tone sandhi in Mandarin: (Zee 1980, Kratochvil 1984, Peng 2000, etc.) • Well studied: a FR tone alternates to a R tone when it is followed by another FR • Little research on the specific alternation pattern of ji6 Research Questions • Research questions: 1. Acoustic analysis: does the tone of ji alternate in the way that Table 1 shows? 2. Compared with another high-toned morpheme wu ‘dark, black’, does the tone of ji differ from that of wu in the same tonal context? - wu: sonorant+high vowel - wu has a lexical high tone independent of the following tonal context.7 Hypothesis 1 • The tone of ji alternates in the way that Table 1 shows. - ji in isolation: H - ji + H, ji + R, ji + FR: F - ji + F: R8 Hypothesis 1 • Peak delay in ji: - No difference between ji+H, ji+R, and ji+FR - ji+H, ji+R, or ji+FR < ji in isolation < ji + F9 Hypothesis 2 • ji differs from wu in tone in the same tonal context (except for ji and wu in isolation) - Peak delay: yi + H < wu + H - Peak delay: yi + R < wu + R - Peak delay: yi + FR < wu + FR - Peak delay: yi + F > wu + F10 Method • 5 subjects were recorded, but only the data of 1 subject was utilized. • From Taiwan • Speak Mandarin Chinese as first language • Trilingual: Mandarin/Taiwanese/English11 Materials A B Mandarin Character & Pinyin Romanization IPA Gloss Mandarin Character &Pinyin Romanization IPA Gloss 1 Isolation 一 yī /jí/ ‘one’ 烏 wū /wú/ ‘black’ 2 Followed by H 一淹再淹 yī yiēn tsài yiēn /jí yién tsâi yién/ ‘to be flooded again and again’ 烏煙瘴氣 wū yiēn zhàng qí /wú yién tşâŋ chî/ ‘to be heavily polluted’ 3 Followed by R 一枚 yī méi /jí měɪ/ ‘a CL (e.g. said of ring)’ 烏梅 wū méi /wú měɪ/ ‘dark plums’ 4 Followed by FR 一里 yī lĭ /jí li/ ‘a unit of length equal to half kilometer’ 烏鯉 wū lĭ /wú li/ ‘black carp’ 5 Followed by F 一輛 yī liàng /jí liâŋ/ ‘a CL (said of vehicle)’ 烏亮 wū liàng /wú liâŋ/ ‘glossy black’12 Materials • Carrier sentences: the morphemes before and after the phrases are toneless • Examples: I think the situation that it is flooded again and again is very annoying. • 10 sentences *12 repetitions = 120 total utterances - /wɔ dʒyɛ də jí yién tsâi yién də chǐŋ ɕǐŋ hən tʰaʊ jiên/ I think one flood again flood NOM situation very annoying13 Recording Procedure • Recorded on a digital recorder in a sound-treated booth • Timed Powerpoint presentation • One utterance per slide • The stimuli were given in Mandarin Chinese characters.14 Measurements Procedure • Recorded data was transferred to a computer and analyzed on Praat. • Boundaries: onset and the offset of the target syllable • Measurements: Peak delay from the onset of the target syllable - F0 Maximum and its timepoint in the target syllable - timepoint of the onset of the target syllable • A script was run to measure the peak delay in ji and wu15 Test of the Hypothesis 1 • Significant difference in peak delay in ji: - ji + H, ji + R, or ji + FR < ji in isolation < ji + F The tone of ji alternates in the way that Table 1 shows.16 Test of the Hypothesis 2 • Significant difference in peak delay (except for ji and wu in isolation): - yi + H < wu + H - yi + R < wu + R - yi + FR < wu + FR - yi + F > wu + F  ji differs from wu in tone in the same tonal context17 Results Example: Subject F1 Time (s)0 4.76127010002000300040005000Time (s)0 4.761270500 ji • Broken pitch track • Where is the peak? • Cannot measure the MaxF0 and its timepoint • Cannot measure the peak delay18 Results Subject F2 Time (s)0 2.813360500Time (s)0 2.81336010002000300040005000 ji • Only utilized the data from subject F219 Mean Peak delay in ji by Different Tonal Contexts 0 20 40 60 80 100120Cell Mean for Peak Delayji ji+H ji+R ji+FR ji+F Cell Bar Chart Grouping Variable(s): Tonal Context Error Bars: ± 1 Standard Deviation(s) Inclusion criteria: ji20 One-way Anova: Peak Delay in ji and Tonal Contexts • Significant difference in overall F test (F(4,55)=4.488, p<.05). 4 12828.600 3207.150 4.488 .0033 17.954 .928 55 39299.583 714.538 DF Sum of Squares Mean Square F-Value P-Value Lambda PowerTonal ContextsResidualANOVA Table for Peak Delay Inclusion criteria: ji21 Post Hoc: Peak Delay in ji and Tonal Contexts 41.833 21.870 .0003 S30.583 21.870 .0070 S36.833 21.870 .0014 S31.000 21.870 .0063 S-11.250 21.870 .3071 -5.000 21.870 .6486 -10.833 21.870 .3252 6.250 21.870 .5692 .417 21.870 .9697 -5.833 21.870 .5951 Mean Diff. Crit. Diff. P-Valueji, ji+H ji, ji+R ji, ji+FR ji, ji+F ji+H, ji+R ji+H, ji+FR ji+H, ji+F ji+R, ji+FR ji+R, ji+F ji+FR, ji+F Fisher's PLSD for Peak Delay Effect: Tonal Contexts Significance Level: 5 % Inclusion criteria: ji22 Post Hoc: Peak Delay in ji and Tonal Contexts • Peak delay: ji in isolation is significantly longer than ji in the other tonal contexts • Peak delay: ji+H, ji+R, ji+FR, and ji+F are not significantly different from each other23 Hypothesis 1 • Do the results support hypothesis 1? No. • We only know that there is difference between the tone of ji in isolation and the tones of ji followed by the different lexical tones. • No