Daniel McCloy Linguistic Phonetics Laboratory Department ...
1
Figure 1: Formant values, means, and ±σ̂ confidence regions Figure 2: Illustration of “reduction score” technique Figure 3: Reduction scores (±σ̂) grouped by vowel /i/ /u/ /e/ /o/ Hz Hz Hz Hz Hz Hz Hz Hz 2700 2600 2500 2400 2300 2200 2100 2000 1900 Hz 1800 Hz 200 500 400 300 mean of R2 condition defined as “100% reduced” mean of R1 condition defined as “0% reduced” 100 80 60 40 20 0 R1 R2 E2 E3 E1 R1 R2 E2 E3 E1 R1 R2 E2 E3 E1 R1 R2 E2 E3 E1 R1 R2 E2 E3 E1 /i/ /u/ /e/ /o/ /a/ Results Tables 1-4: Bonferroni-corrected pairwise t-tests for /i/, /u/, /e/, and /o/. (All results for /a/ were non-significant and are not reported.) Table 1: Pairwise t-tests for /i/ Table 2: Pairwise t-tests for /u/ Table 3: Pairwise t-tests for /e/ Table 4: Pairwise t-tests for /o/ 2 F 1 F 2 F 1 F 2 F 1 F 2 F 1 F e+kan >0.05 >0.05 eCaN 0.0005 0.0087 >0.05 >0.05 eC+an 0.0430 >0.05 >0.05 >0.05 >0.05 >0.05 eC# 0.0000 0.0000 0.0287 0.0188 >0.05 >0.05 >0.05 >0.05 n a + C e N a C e n a k + e # e 2 F 1 F 2 F 1 F 2 F 1 F 2 F 1 F o+kan 0.0008 >0.05 oCaN 0.0000 0.0002 0.0000 >0.05 oC+an 0.0000 0.0059 0.0000 >0.05 >0.05 >0.05 oC# 0.0000 0.0000 >0.05 0.0000 0.0000 >0.05 0.0000 0.0089 n a + C o N a C o n a k + o # o 2 F 1 F 2 F 1 F 2 F 1 F 2 F 1 F i+kan 0.0247 0.0029 iCaN >0.05 >0.05 >0.05 >0.05 iC+an 0.0000 0.0000 0.0014 0.2553 0.0001 0.0012 iC# 0.0000 0.0000 0.0001 0.0028 0.0000 0.0000 >0.05 >0.05 n a + C i N a C i n a k + i # i 2 F 1 F 2 F 1 F 2 F 1 F 2 F 1 F u+kan 0.0073 >0.05 uCaN 0.0161 0.0064 >0.05 >0.05 uC+an 0.0000 >0.05 >0.05 >0.05 >0.05 >0.05 uC# 0.0000 0.0224 0.0007 >0.05 0.0000 >0.05 0.0173 >0.05 n a + C u N a C u n a k + u # u significant difference in either F1 or F2 significant difference in both F1 and F2 Background ◆ Vowels in closed syllables tend to be shorter than vowels in open syllables (maddieson 1985). ◆ Closed-syllable vowel quality changes (a.k.a. reduction or laxing) are widely attested, e.g., French (dell 1995), Portuguese (redenbarger 2005), ai (apiluck 1996), and Klamath (blevins 1993). ◆ Morpheme boundaries shown to influence closed-syllable reduction; in Ngayu Dajak (iso 639-3: nij, austronesian, borneo), /toso+n/ “breast of ____” shows some vowel reduction, but less than tautomorphemic /toson/ “red squirrel” (brunelle & riehl 2002). Methods Four Indonesian males (age 20-21), Standard Indonesian L1. Fully randomized word list: 75 distractors + 225 tokens per speaker (5 vowels × 5 conditions × 3 words per condition × 3 repetitions). Two reference conditions and three experimental conditions: r1 CV] word (canonical unreduced) r2 CVC] word (canonical reduced) e1 CV+kan] word (circumfixes ber-kan, me(N)-kan, memper-kan) e2 CVC+an] word (circumfixes pe-an, ke-an) e3 CVCaN] word (monomorphemic) eory predicts a single intervocalic C should syllabify into the onset of the following syllable; if so, all three experimental conditions should syllabify as CV.CaN, and V should manifest unreduced. Stress is penultimate unless penult vowel is /ə/, so all experimental conditions have equivalent stress. Segmentation in Praat (boersma & weenink 2011) ◆ Measurements extracted from midpoint of modally voiced portion of vowel. ◆ Modal voicing : = low cycle-to-cycle changes, minimal aperi- odic noise, glottal pulses visible in high frequencies; this en- sures reliable F1 measurements. ◆ Word-final vowels (condition r1), measurement was done at 50ms aſter the onset of modal voicing (if modally voiced por- tion exceeded 100ms) to minimize effect of diphthongization due to word-final lengthening. ◆ Modal-voiced portions shorter than the 25ms analysis window, and any measurements >2 s.d. from the vowel mean were re-measured by hand. Pairwise t-tests (Bonferroni corrected) performed using R (r development core team 2011). /i/ iCaN patterns with i# (unreduced). i+kan falls in the middle, significantly different from the endpoints but not distinct from iCaN. iC+an patterns with iC# (reduced). /u/ u+kan, uCaN and uC+an pattern together, partway between (and significantly different from) the endpoints u# (unreduced) and uC# (reduced). /e/ e+kan patterns with e# (unreduced). eCaN and eC+an pattern with eC# (reduced). /o/ all conditions distinct except oCaN and oC+an. /a/ no significant differences found among any of the reference or experimental conditions. Discussion ◆ Some results support hypothesis, others unexpected; no unified pattern across vowels. ◆ Trends suggest that within-category phonetic variation may index morphological information. ◆ Perceptual salience of differences unknown. ◆ Gradient vowel reduction → gradient syllabification? ◆ Future directions: duration; individual differences in production; perception study. Research Question Is vowel reduction in Indonesian (iso 639-3: ind, austronesian, indonesia) mitigated by morphemic structure? Hypothesis Morpheme boundary will interfere with resyllabification in condition E2 (see below) which will thus have a reduced vowel. Conditions E1 and E3 should pattern with the canonical unre- duced condition (R1). References Apiluck, T. (1996). e mid central vowel [schwa] in ai. Mon-Khmer Studies, 26, 65-78. Blevins, J. (1993). Klamath Laryngeal Phonology. International Journal of American Linguistics, 59, 237-279. Boersma, P., & Weenink, D. (2011). Praat: doing phonetics by computer. Retrieved from http://www.praat.org/ Brunelle, M., & Riehl, A. (2002). Vowel Laxing in Ngaju Dayak. Working Papers of the Cornell Phonetics Laboratory, 14, 1-15. Dell, F. (1995). Consonant clusters and phonological syllables in French. Lingua, 95, 5-26. Maddieson, I. (1985). Phonetic cues to syllabification. In V. A. Fromkin (Ed.), Phonetic Linguistics : Essays in Honor of Peter Ladefoged (pp. 203-221). Orlando: Academic Press. R Development Core Team. (2011). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.R-project.org/ Redenbarger, W. J. (2005). Pre-Nasal Laxing in European Portuguese. Selected Proceedings of the 7th Hispanic Linguistics Symposium. Somerville, Mass.: Cascadilla Press. Vowel laxing in Indonesian as a test case for interaction of morphological and syllabic structure Daniel McCloy 5pSC21 Linguistic Phonetics Laboratory Department of Linguistics University of Washington
Transcript of Daniel McCloy Linguistic Phonetics Laboratory Department ...
Figure 1: Formant values, means, and ±σ̂ con�dence regions
Figure 2: Illustration of “reduction score” technique
Figure 3: Reduction scores (±σ̂) grouped by vowel
/i/ /u/
/e/ /o/
Hz
Hz
Hz
HzHz
Hz
HzHz
2700 2600 2500 2400 2300 2200 2100 2000 1900 Hz1800
Hz200
500
400
300
mean of R2 condition de�ned as “100% reduced”
mean of R1 condition de�ned as “0% reduced”
100
80
60
40
20
0R1 R2 E2 E3 E1R1 R2 E2 E3 E1 R1 R2 E2 E3 E1 R1 R2 E2 E3 E1R1 R2 E2 E3 E1
/i/ /u/ /e/ /o/ /a/
ResultsTables 1-4: Bonferroni-corrected pairwise t-tests for /i/, /u/, /e/, and /o/. (All results for /a/ were non-signi�cant and are not reported.)
Table 1: Pairwise t-tests for /i/
Table 2: Pairwise t-tests for /u/
Table 3: Pairwise t-tests for /e/
Table 4: Pairwise t-tests for /o/
2F1F2F1F2F1F2F1F
e+kan >0.05 >0.05eCaN 0.0005 0.0087 >0.05 >0.05eC+an 0.0430 >0.05 >0.05 >0.05 >0.05 >0.05eC# 0.0000 0.0000 0.0287 0.0188 >0.05 >0.05 >0.05 >0.05
na+CeNaCenak+e#e
2F1F2F1F2F1F2F1F
o+kan 0.0008 >0.05oCaN 0.0000 0.0002 0.0000 >0.05oC+an 0.0000 0.0059 0.0000 >0.05 >0.05 >0.05oC# 0.0000 0.0000 >0.05 0.0000 0.0000 >0.05 0.0000 0.0089
na+CoNaConak+o#o
2F1F2F1F2F1F2F1F
i+kan 0.0247 0.0029iCaN >0.05 >0.05 >0.05 >0.05iC+an 0.0000 0.0000 0.0014 0.2553 0.0001 0.0012iC# 0.0000 0.0000 0.0001 0.0028 0.0000 0.0000 >0.05 >0.05
na+CiNaCinak+i#i
2F1F2F1F2F1F2F1F
u+kan 0.0073 >0.05uCaN 0.0161 0.0064 >0.05 >0.05uC+an 0.0000 >0.05 >0.05 >0.05 >0.05 >0.05uC# 0.0000 0.0224 0.0007 >0.05 0.0000 >0.05 0.0173 >0.05
na+CuNaCunak+u#u
signi�cant di�erence in either F1 or F2signi�cant di�erence in both F1 and F2
Background◆ Vowels in closed syllables tend to be shorter than vowels in open
syllables (maddieson 1985).
◆ Closed-syllable vowel quality changes (a.k.a. reduction or laxing) are widely attested, e.g., French (dell 1995), Portuguese (redenbarger 2005), �ai (apiluck 1996), and Klamath (blevins 1993).
◆ Morpheme boundaries shown to in�uence closed-syllable reduction;in Ngayu Dajak (iso 639-3: nij, austronesian, borneo), /toso+n/ “breast of ____” shows some vowel reduction, but less than tautomorphemic /toson/ “red squirrel” (brunelle & riehl 2002).
MethodsFour Indonesian males (age 20-21), Standard Indonesian L1.
Fully randomized word list: 75 distractors + 225 tokens per speaker(5 vowels × 5 conditions × 3 words per condition × 3 repetitions).
Two reference conditions and three experimental conditions: r1 CV]word (canonical unreduced) r2 CVC]word (canonical reduced)
e1 CV+kan]word (circum�xes ber-kan, me(N)-kan, memper-kan) e2 CVC+an]word (circum�xes pe-an, ke-an) e3 CVCaN]word (monomorphemic)
�eory predicts a single intervocalic C should syllabify into the onset of the following syllable; if so, all three experimental conditions should syllabify as CV.CaN, and V should manifest unreduced.
Stress is penultimate unless penult vowel is /ə/, so all experimental conditions have equivalent stress.
Segmentation in Praat (boersma & weenink 2011)◆ Measurements extracted from midpoint of modally voiced
portion of vowel.◆ Modal voicing := low cycle-to-cycle changes, minimal aperi-
odic noise, glottal pulses visible in high frequencies; this en-sures reliable F1 measurements.
◆ Word-�nal vowels (condition r1), measurement was done at 50ms a�er the onset of modal voicing (if modally voiced por-tion exceeded 100ms) to minimize e�ect of diphthongization due to word-�nal lengthening.
◆ Modal-voiced portions shorter than the 25ms analysis window, and any measurements >2 s.d. from the vowel mean were re-measured by hand.
Pairwise t-tests (Bonferroni corrected) performed using R (r development core team 2011).
/i/ iCaN patterns with i# (unreduced). i+kan falls in the middle, signi�cantly di�erent from the endpoints but not distinct from iCaN.iC+an patterns with iC# (reduced).
/u/ u+kan, uCaN and uC+an pattern together, partway between (and signi�cantly di�erent from) the endpoints u# (unreduced) and uC# (reduced).
/e/ e+kan patterns with e# (unreduced).eCaN and eC+an pattern with eC# (reduced).
/o/ all conditions distinct except oCaN and oC+an.
/a/ no signi�cant di�erences found among any of the reference or experimental conditions.
Discussion◆ Some results support hypothesis, others
unexpected; no uni�ed pattern across vowels.
◆ Trends suggest that within-category phonetic variation may index morphological information.
◆ Perceptual salience of di�erences unknown.
◆ Gradient vowel reduction → gradient syllabi�cation?
◆ Future directions: duration; individual di�erences in production; perception study.
Research QuestionIs vowel reduction in Indonesian (iso 639-3: ind, austronesian, indonesia) mitigated by morphemic structure?
HypothesisMorpheme boundary will interfere with resyllabi�cation in condition E2 (see below) which will thus have a reduced vowel. Conditions E1 and E3 should pattern with the canonical unre-duced condition (R1).
ReferencesApiluck, T. (1996). �e mid central vowel [schwa] in �ai. Mon-Khmer Studies, 26, 65-78.
Blevins, J. (1993). Klamath Laryngeal Phonology. International Journal of American Linguistics, 59, 237-279.
Boersma, P., & Weenink, D. (2011). Praat: doing phonetics by computer. Retrieved from http://www.praat.org/
Brunelle, M., & Riehl, A. (2002). Vowel Laxing in Ngaju Dayak. Working Papers of the Cornell Phonetics Laboratory, 14, 1-15.
Dell, F. (1995). Consonant clusters and phonological syllables in French. Lingua, 95, 5-26.
Maddieson, I. (1985). Phonetic cues to syllabi�cation. In V. A. Fromkin (Ed.), Phonetic Linguistics : Essays in Honor of Peter Ladefoged (pp. 203-221). Orlando: Academic Press.
R Development Core Team. (2011). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.R-project.org/
Redenbarger, W. J. (2005). Pre-Nasal Laxing in European Portuguese. Selected Proceedings of the 7th Hispanic Linguistics Symposium. Somerville, Mass.: Cascadilla Press.
Vowel laxing in Indonesian as a test case for interaction of morphological and syllabic structureDaniel McCloy 5pSC21Linguistic Phonetics Laboratory Department of Linguistics University of Washington
