SLIDE 1 Cognitive Mechanism and Locus of Tone Cognitive Mechanism and Locus of Tone Sandhi during Mandarin Chinese Spoken Word Production
Jenn-Yeu Chen1, Train-Min Chen2, & You-Yu Dai3
1 National Taiwan Normal University, Taiwan ROC 師大 2 National Cheng Kung University, Taiwan ROC 成大
g g y
3 National Chung Cheng University, Taiwan ROC 中正
The Third International Symposium on Tonal Aspects of Languages
SLIDE 2
Ackno knowledgement ledgement Ackno knowledgement ledgement
This research was supported by National Science Council This research was supported by National Science Council,
Taiwan, as well as by the Chiang Ching Kuo Foundation. It was part of an undergraduate research project carried out by was pa t o a u e g a uate esea c p oject ca e out y Mr. You-Yu Dai and supervised by the first author.
SLIDE 3 Thir Third T d Tone ne Sandhi Sandhi in Mandarin Chinese in Mandarin Chinese Thir Third T d Tone ne Sandhi Sandhi in Mandarin Chinese in Mandarin Chinese
Mandarin Chinese is a tone language which includes four basic
Mandarin Chinese is a tone language which includes four basic tones (Tone 1 to Tone 4) and a neutral one (Tone 5).
The T3 sandhi is a pronunciation rule of Mandarin Chinese, which
p , states that whenever two T3’s are in succession within the same word or the same prosodic unit, the first one must change into a tone that is indistinguishable from T2 (Xu, 1991).
For example, xi3zao3 (洗澡, “take a bath”) is pronounced as [xi2zao3].
d h bl f h f h
Here, [xi2] is indistinguishable from the pronunciation of a character
carrying a second tone such as 息 (xi2, “rest”).
SLIDE 4
Cat Categorical or Gradient gorical or Gradient Cat Categorical or Gradient gorical or Gradient
Categorical view: the sandhiT3 is categorically T2
Categorical view: the sandhiT3 is categorically T2.
Gradient view: the sandhiT3 retains its phonemic identity, but its
articulation is adjusted so that it sounds similar to T2. This is akin j to a flap in pronouncing an English word latter, which sounds more like a /d/ than a /t/.
Based on the findings in the literature of no differences in f0 and
slope between the sandhiT3 and the T2, Myers and Tsay (2003) d h h dh l d b d suggested that the T3 sandhi, as is employed by Taiwan Mandarin speakers, is categorical rather than gradient.
Th
t d th t B iji M d i k t d d t d
They noted that Beijing Mandarin speakers tended to produce a
lower and flatter sandhiT3 than a T2, and accordingly suggested that the T3 sandhi in Beijing Mandarin is more gradient than that the T3 sandhi in Beijing Mandarin is more gradient than categorical.
SLIDE 5 The T3 Sandhi The T3 Sandhi in the Cont in the Context of a W t of a Word Pr Production Model
We addressed the issue by exploiting the feedforward discrete-
stage model of word production proposed by Levelt and colleagues (Levelt, Roelof, & Meyer, 1999), as well as an experimental paradigm initially designed to test the predictions of th d l (M 1990) the model (Meyer, 1990).
SLIDE 6 Th The L Levelt-R
eyer M r Mode
e e e e e t
eye eye
Word production consists of a stage of word form encoding and a
Word production consists of a stage of word form encoding and a subsequent stage of actual articulation.
Word form encoding consists further of phonological encoding
g p g g and phonetic encoding, which eventually results in the retrieval of syllable-sized abstract motor programs that drive the neuromuscular system for articulation.
Placed in this model, tone sandhi could occur during word form
d f f h f dh b d d encoding or after it. If the former, tone sandhi may be considered as categorical. If the latter, it may be viewed as gradient.
SLIDE 7 The LRM Theory
{horse}-{s} W phonological code retrieval
ω
W O R D
( /h/1, /o/2, /r/3, /s/4 ) 1 ( /I/1, /z/2 ) 2 ω σ’ σ
O R syllabification (prosodification)
/h / / /
R M
phonetic encoding ( h i d i l)
syllabary /ho’r/1-/sIz/2
N C O D (phonetic code retrieval)
[hors] [hor] [sIz]
[ho’r]- [sIz] D I N G articulation “horses” ARTICULATION horses
SLIDE 8 The Im The Implicit Primin licit Priming Pa Paradi digm p g p g g
Learning Phase: Learning Phase: hermit ~ loner economy ~ local flower ~ lotus e g se e g se e
eco o y
Test Phase: Test Phase: hermit loner
screen shows participant says
Homogeneous context set 1 2 3 set 1 2 3 Heterogeneous 4 beacon m ajor loner 5 beadle m aker local context 5 beadle m aker local 6 beaker m aple lotus
SLIDE 9 Im Impli licit Primin Priming Ef Effect ect and i its ts A Accou ccount p c t g ect ect a d ts ccou ccou t
RTs in the homogeneous context are faster than RTs in the
g heterogeneous context. The difference has been referred to as the implicit priming effect of the shared phonological unit (the h ll bl ) segment, the syllable, etc.).
According to the computer implementation of the LRM model, a
i ti h i ll th d ti t suspension-resumption mechanism allows the production system to prepare the word in advance for whatever is known, suspend the operation, and resume it when further information becomes t e ope at o , a esu e t w e u t e
available during a production trial. The advanced preparation includes retrieving the phonological contents of a word, setting up the metrical frame, assembling the segments into a phonological word, retrieving the phonetic syllables from the mental syllabary, d ibl ti ti t f th ti l t t Th and possibly activating part of the articulatory gestures. The more that can be prepared in advance, the larger the effect.
SLIDE 10
Im Implicit Priming of T plicit Priming of Tonal Syllable nal Syllable Im Implicit Priming of T plicit Priming of Tonal Syllable nal Syllable
homogeneous context heterogeneous context context context
qing1 -liang2 (“清涼＂ cool) qing1-liang2 (“清涼＂ cool) qing1 -sheng1 (“輕生＂ suicide) xi2-su2 (“習俗＂ custom) qing1 ting2 fei3 bang4 qing1 -ting2 (“蜻蜓＂ dragonfly) fei3-bang4 (“誹謗＂ slander) qing1 -jiao1 ke4-ben3 q g j (“青椒＂ green pepper) (“課本＂ textbook)
tonal syllable (IP = 46ms) y
SLIDE 11 Im Implicit Priming of A plicit Priming of Atonal Syllable
Im Implicit Priming of A plicit Priming of Atonal Syllable
homogeneous context heterogeneous context context context
fei1-dang4 (“飛彈＂ missile) fei1-dang4 (“飛彈＂ missile) fei2-zhao4 (“肥皂＂ soap) ke2-tan2 (“咳痰＂ expectoration) f i3 ji4 i3 4 fei3-ji4 (“斐濟＂ Fiji) xi3-yue4 (“喜悅＂ joyful) fei4-bu4 qing4-zhu4 (“肺部＂ lung) q g (“慶祝＂ celebrate)
atonal syllable (IP = 12ms) y ( )
SLIDE 12 Rationale of the Present Study Rationale of the Present Study Rationale of the Present Study Rationale of the Present Study
To manipulate context such that target words become
To manipulate context such that target words become homogeneous with respect to the first tonal syllables either before
- r after tone sandhi has applied.
Pre-sandhi heterogeneous / post-sandhi homogeneous:
T3T3, T3T3, T2T3, T2T3 / T2T3, T2T3, T2T3, T2T3
Pre-sandhi homogeneous / post-sandhi heterogeneous:
T3T3, T3T3, T3T2, T3T2 / T2T3, T2T3, T3T2, T3T2 , , , , , ,
SLIDE 13 Post
Sandhi hi Homoge
eous
Sa d
eous
Homogeneous Set 1 2 3 4 1 起舞 /qi3-wu3/ [qi2-wu3] 演講 /yan3-jiang3/ [yan2-jiang3] 魚簍 /yu2-lou3/ [yu2-lou3] 執法 /zhi2-fa3/ [zhi2-fa3] erogeneous 2 乞討 /qi3-tao3/ [qi2-tao3] 眼角 /yan3-jiao3/ [yan2-jiao3] 愚蠢 /yu2-chun3/ [yu2-chun3] 直尺 /zhi2-chi3/ [zhi2-chi3] Hete 3 奇險 /qi2-xian3/ [qi2-xian3] 延緩 /yan2-huan3/ [yan2-huan3] 雨水 /yu3-shui3/ [yu2-shui3] 止血 /zhi3-xie3/ [zhi2-xie3] 祈禱 沿海 語法 紙板 4 祈禱 /qi2-dao3/ [qi2-dao3] 沿海 /yan2-hai3/ [yan2-hai3] 語法 /yu3-fa3/ [yu2-fa3] 紙板 /zhi3-ban3/ [zhi2-ban3]
If dhi d i d f di d ld b
- If tone sandhi occurs during word form encoding, target words would become
homogeneous with respect to the first tonal syllable, and a large implicit priming effect would be expected.
- If tone sandhi occurs after word form encoding, target words would be homogeneous with
respect to the first atonal syllable, and a small effect should be the result.
SLIDE 14 Pr Pre-Sa Sandhi hi Homoge
eous e Sa Sa d
eous
Homogeneous Set 1 2 3 4 1 導演 /dao3-yan3/ [dao2-yan3] 理想 /li3-xiang3/ [li2-xiang3] 古人 /gu3-ren2/ [gu3-ren2] 檢查 /jian3-cha2/ [jian3-cha2] rogeneous 2 搗鬼 /dao3-gui3/ [dao2-gui3] 裏海 /li3-hai3/ [li2-hai3] 骨牌 /gu3-pai2/ [gu3-pai2] 減肥 /jian3-fei2/ [jian3-fei2] Heter 3 倒楣 /dao3-mei2/ [dao3-mei2] 李白 /li3-bai2/ [li3-bai2] 鼓手 /gu3-shou3/ [gu2-shou3] 剪綵 /jian3-cai3/ [jian2-cai3] 4 島國 /dao3-guo2/ [dao3-guo2] 禮儀 /li3-yi2/ [li3-yi2] 谷底 /gu3-di3/ [gu2-di3] 簡短 /jian3-duan3/ [jian2-duan3]
Predictions are reversed:
- If tone sandhi occurs during word form encoding, a small implicit priming effect would be
expected. p
- If tone sandhi occurs after word form encoding, a large effect should be the result.
SLIDE 15
Results of Experiment 1 and 2 sults of Experiment 1 and 2 Results of Experiment 1 and 2 sults of Experiment 1 and 2
Experiment 1 Post-Sandhi Homogeneity: 42 ms effect Experiment 1 Post-Sandhi Homogeneity: 42 ms effect Experiment 2 Pre-Sandhi Homogeneity: 49 ms effect C
t di t d b th di ti
Contradicted both predictions. What happened?
SLIDE 16
An Explanation and Hypo An Explanation and Hypothesis thesis An Explanation and Hypo An Explanation and Hypothesis thesis
(Xu, 1997)
The phonetic forms of T2 and T3 are quite similar. The sizable implicit
priming effect observed in Experiment 1 & 2 could be the result of p g p T2 and T3 sharing the initial portions of their tonal contours.
SLIDE 17
Tests of the Hypo sts of the Hypothesis thesis Tests of the Hypo sts of the Hypothesis thesis
Two control experiments where no tone sandhi was involved Two control experiments where no tone sandhi was involved. Experiment 3: similar tones (T2 and T3), expect an effect
similar to that of Experiment 1 and 2 similar to that of Experiment 1 and 2.
Experiment 4: dissimilar tones (T1 and T3), expect a
reduced effect reduced effect.
SLIDE 18 Experiment 3: T2 and T3 Experiment 3: T2 and T3 no no sandhi sandhi Experiment 3: T2 and T3 Experiment 3: T2 and T3, no no sandhi sandhi
Homogeneous Set 1 2 3 4 執法 奇險 羽毛 掩埋 eneous 1 執法 /zhi2-fa3/ 奇險 /qi2-xian3/ 羽毛 /yu3-mao2/ 掩埋 /yan3-mai2/ 2 直尺 / hi2 hi3/ 祈禱 / i2 d 3/ 語詞 / 3 i2/ 眼球 / 3 i 2/ Heteroge 2 /zhi2-chi3/ /qi2-dao3/ /yu3-ci2/ /yan3-qiu2/ 3 紙錢 /zhi3-qian2/ 乞食 /qi3-shi2/ 魚簍 /yu2-lou3/ 沿海 /yan2-hai3/ q q y y 4 指頭 /zhi3-tou2/ 啟程 /qi3-cheng2/ 愚蠢 /yu2-chun3/ 延緩 /yan2-huan3/
51 ms effect
SLIDE 19 Experiment 4: T1 and T3 Experiment 4: T1 and T3 no no sandhi sandhi Experiment 4: T1 and T3 Experiment 4: T1 and T3, no no sandhi sandhi
Homogeneous Set 1 2 3 4 織女 悽慘 羽毛 眼球 eneous 1 織女 /zhi1-nv3/ 悽慘 /qi1-can3/ 羽毛 /yu3-mao2/ 眼球 /yan3-qiu2/ 2 知己 / hi1 ji3/ 七堵 / i1 d 3/ 語言 / 3 2/ 演員 / 3 2/ Heteroge 2 /zhi1-ji3/ /qi1-du3/ /yu3-yan2/ /yan3-uan2/ 3 指頭 /zhi3-tou2/ 乞食 /qi3-shi2/ 瘀血 /yu1-xie3/ 淹水 /yan1-shui3/ q y y 4 紙牌 /zhi3-pai2/ 啟程 /qi3-cheng2/ 迂腐 /yu1-fu3/ 煙草 /yan1-cao3/
27 ms effect
SLIDE 20 Summar Summary of Findings
Summar Summary of Findings
A full implicit priming effect of the tonal syllable was A full implicit priming effect of the tonal syllable was
- bserved regardless of whether the response words involved
tone sandhi or whether they were rendered homogeneous by to e sa
e e o oge eous y tone sandhi, as long as the tones were similar (T2 and T3).
When the initial syllables of the response words involved
When the initial syllables of the response words involved dissimilar tones (T1 and T3), the implicit priming effect was reduced.
SLIDE 21 Clues fr Clues from resear
- m research in hand and arm
ch in hand and arm mo movements ments
Studying curved hand movements, Hoffman (2010) showed that
when the goal switched after movement onset, subjects could g j adjust their movements smoothly to the new goal. Hoffman suggested that this was accomplished by ‘switching a single control d h h d l h h l command, without the need to re-plan or re-optimize the whole movement or superimpose movements’ (p. 73).
Fl h
d H i (1991) d t t d th t ‘ i d t
Flash and Henis (1991) demonstrated that ‘aimed arm movements
can be elicited in quick succession, without appreciable delays in responding to the target displacement, beyond the normal responding to the target displacement, beyond the normal reaction time’ (p. 220).
SLIDE 22
An An Ar Articulat ticulatory Account of T ccount of Tone ne Sandhi Sandhi An An Ar Articulat ticulatory Account of T ccount of Tone ne Sandhi Sandhi
The results of Experiments 1 to 3 can be explained by assuming
The results of Experiments 1 to 3 can be explained by assuming that the production system prepares the first syllables of the target words in the homogeneous condition all the way after the shared g y phonetic syllable has been retrieved from the mental syllabary. The retrieved phonetic syllable is likely the one with T3. The i i h i i suspension-resumption mechanism assumes operation as soon as the system knows which target word to produce. Upon that time, the system is able to shift the articulatory gesture from that of T3 the system is able to shift the articulatory gesture from that of T3 to that of T2 if tone sandhi requires it or if the target (with an initial T2) requires it.
This articulatory account is more consistent with the gradient
view of tone sandhi.
SLIDE 23 Conclusion Conclusion Conclusion Conclusion
The third tone sandhi in Mandarin Chinese is an articulatory The third tone sandhi in Mandarin Chinese is an articulatory
- peration that is gradient rather than categorical.
SLIDE 24
Th k f tt ti Thank you for your attention. Comments are welcome and appreciated. pp
