Upload
dinhthuy
View
222
Download
0
Embed Size (px)
Citation preview
1/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
The Phonology and Phonetics of Ambiguity:A Case Study of /v/
Christina Bjorndahl
Cornell University
January 13, 2012
2/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Motivation: Russian /v/
Patterns as an obstruent• Undergoes final devoicing• Undergoes regressive voicing assimilation
Patterns as a sonorant• Does not trigger regressive voicing assimilation
“. . . the Standard Russian V . . . occupies an obviously intermediateposition between the obstruents and the sonorants”
Jakobson (1978)
3/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Inventories containing /v/
Frequency of /v/
• PBase: 200/548 (36.5%)• UPSID: 95/451 (21.1%)
Frequency of /v/ & /f/
• PBase: 167/548 (30.5%)• UPSID: 77/451 (17.1%)
Frequency of /v/ & /f/no /w, B
fl, V/
• PBase: 39/548 (7.1%)• UPSID: 23/451 (5.1%)
Figure: Mielke (2011), website
4/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Languages with /v, f/ no /w, B, V/ (PBase)
Indo-European (20)• Slavic (6): Bulgarian, Czech, Russian, Serbo-Croatian, Slovene,
Ukranian• Germanic (4): Afrikaans, Danish, Swedish, Yiddish• Indo-Aryan (3): Dhivehi, Burgenland-Romani, Sepecides-Romani• Indo-Iranian (2): Persian, Northern Talysh• Italic (2): Italian, Miogliola Ligurian• Armenian (1): Armenian• Baltic (1): Latvian• Greek
Austronesian (6)• Kwamera, Pileni, Rotuman, Sinaugoro, Tuvaluan, Wolio
5/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Languages with /v, f/ no /w, B, V/ (continued)
• Altaic (Turkic): Iranian Azari, Turkish• Niger-Congo: Bantu• Creoles: Berbice Dutch Creole, Kristang• Afro-Asiatic (Semitic): Hebrew• Arawakan: Resigaro• Austro-Asiatic: Khmer• Oto-Manguean: Lealao Chinanteco• Sino-Tibetan: Bisu• Uralic (Finno-Ugric): Hungarian
6/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Overview: Cross-linguistic patterning of /v/
ObstruentPolish*, Greek
SonorantSerbian, Icelandic
AmbiguousRussian, Hungarian
Factors• Structure of the inventory• Static patterning (distribution)• Dynamic patterning (phonological processes)• Phonetics
6/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Overview: Cross-linguistic patterning of /v/
ObstruentPolish*, Greek
SonorantSerbian, Icelandic
AmbiguousRussian, Hungarian
Factors• Structure of the inventory• Static patterning (distribution)• Dynamic patterning (phonological processes)• Phonetics
7/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Greek: Inventory
Labial Interdental Alveolar VelarStop p (m)b t (n)d k (N)gFricative f v T D s z x G
Affricates >ts
>dZ
Nasal m nLateral lRhotic r
8/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Greek: Word-initial Clusters
p b t d k g f v T D x G s z r l m np pt (ps) pr pl pnb br blt ts tr tmd dz drk kt (ks) kr kl kng gr glf ft fT fx fr flv vD vG vr vlT Tr Tl TnD Drx xt xT xr xlG GD Gr Gl Gns sp st sk sf sx smz zv zGrlm mnn
9/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Greek: Phonological Processes
Final DevoicingN/A
Voicing AssimilationLittle evidence of voicing assimilation as an active process; wordscan only end in vowels, [s, n]
(1) a. /tis Dino/ > [tiz Dino], I give herb. /tis varvaras/ > [tiz varvaras], Barbara’sc. /tous barbaDes/ > [touz barbaDes], the uncles, acc.d. /tis mamas/ > [tiz mamas], the mother’s
(2) [evGlotos] ‘eloquent’ ∼ [efstaTia] ‘steadiness’ (same prefix)
10/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Serbian: Inventory
Labial Alveolar Palatal VelarStop p b t d k gFricative f s z S„ Z„ x
Affricates >ts«
>tS«
>dZ«
>tS„
>dZ„
Nasal m n ñLateral l LRhotic rApproximant v j
11/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Serbian: Word-initial Clustersp b t d f s z x m n l r v j
p pl pr (pj)b bl br (bj)t tl tr tv (tj)d dl dr dv (dj)k kl kr kvg gl gr gvf fl fr (fj)s sp st sf sx sm sn sl sr sv (sj)z zb zd zm zn zl zr zv (zj)x xl xr xvm ml mr (mj)n (nj)l (lj)rv vl vr (vj)j
12/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Serbian: Phonological Processes
Final DevoicingN/A
Voicing Assimilation
(3) a. /s-paziti/ > [spaziti], observeb. /s-gaziti/ > [zgaziti], tramplec. /s-loziti/ > [sloziti], put togetherd. /s-variti/ > [svariti], digest
(4) [ovca], sheep
13/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Russian: Inventory
Labial Dental Palato-Alveolar VelarStop p b t d k g
pj bj tj dj (kj) (gj)Affricates ts tSj
Fricative fj vj s z S Z xfj vj sj zj (xj)
Nasal m nmj nj
Lateral llj
Rhotic rrj
Approximant v, vj j
14/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Russian: Word-initial clusters
p b t d k g f v s z S Z x m n l r jp ps pS pn pl pr pjb bl br bjt tk tv tl tr tjd dv pn dl dr djk kp kt kv ks km kn kl krg gb gd gv gz gn gl grf ft fk fs fl fr fjv vd vz vZ vm vn vl vr vjs sp st sf sv sx sm sn sl sr sjz zb zd zg zv zm zn zl zr zjSZx xv xl xrm Sm ml mr mjn njl ljrj
15/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Russian: Phonological Processes
Final Devoicing
(5) a. [sled-a] ∼ [slet], track (gen./nom.sg)b. [mil] ∼ *[mil
˚], dear
c. [prav-a] ∼ [praf], right (fem./masc.)
Voicing Assimilation
(6) a. /v ruke/ > [v ruke], in one’s handb. /v gorode/ > [v gorode], in the cityc. /v supe/ > [f supe], in the soup
16/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Russian: Phonological Processes (cont)
Voicing Assimilation (cont)
(7) a. /ot-pustitj/ > [otpustitj], releaseb. /ot-brositj/ > [odbrositj], throw asidec. /ot-nesti/ > [otnesti], carry awayd. /ot-vesti/ > [otvesti], lead away
17/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Summary: Phonological Patterning of /v/
Obstruent /v/
Greek
Sonorant /v/
Serbian
Ambiguous /v/
Russian
Unknown /v/
Yagaria• CV syllable structure• No voicing assimilation• No nasal harmony
No evidence either way as to how/v/ is categorized with respect tosonorancy.
17/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Summary: Phonological Patterning of /v/
Obstruent /v/
Greek
Sonorant /v/
Serbian
Ambiguous /v/
Russian
Unknown /v/
Yagaria• CV syllable structure• No voicing assimilation• No nasal harmony
No evidence either way as to how/v/ is categorized with respect tosonorancy.
18/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Phonetic Study: Motivation
Why do a phonetic study?
• Because it hasn’t been done. . .• Given the diversity of patterning, it’s hard to understand howGreek /v/ and Serbian /v/ are tokens of the same type;perhaps understanding the phonetics of [v] in these languagescan shed some light.
• Recent explanations of Russian /v/ rely on the “phonetics” of[v] being somehow “intermediate”, but this hasn’t beenquantified. (Padgett (2002); Kavitskaya (1999)).
• Phonetic studies that have been done have only focussed on asingle language (Lulich (2004) for Russian; Kiss and Bárkányi(2006) for Hungarian).
19/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
How many [v]’s exist?
To what extent does the three-way patterning of /v/ correlate withthe phonetic realization of [v]?
3Relationship between phonological patterning and phonetics is 1-1.
1Relationship highly abstract; explanation for patterning purelygrammatical.
2• {Greek, Serbian} vs. {Russian}• {Serbian, Russian} vs. {Greek}• {Greek, Russian} vs. {Serbian}
20/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Phonetic Study
• Greek, Serbian, Russian• 7 native speakers of each language• Cornell University or University of Toronto• SD722 digital recorder; 44100 Hz, 16-bit• Hand-segmented in Praat• High pass filtered at 1000 Hz• Resampled to 22050 Hz & analysed in Matlab
21/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Stimuli
• [f, v, T, D, x, G, s, z, m]
• C1VC2V(C)• Stressed in both syllables• Flanking vowels: /a, o/
• 5 randomized word lists
Greek[eGrapsa ____tris fores]
Serbian[kaZe jetsa ____opet]
Russian[sveta skazala ____opet]
22/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Measures & Normalization
Measures• Spectral moments
• Centroid• Skewness
• (Relative duration)
Windowing
• Hann window, 23.2ms• windowed every 10ms• 512-point FFT• spectral measures middle 3windows averaged
OutliersEliminated outliers with z-score > 2.5; mean calculated across alltokens of segment within language.
23/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Spectral Measures: Centroid and Skewness
Figure: Spectrogram of [sa], Russian
24/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Spectral Measures: Centroid and SkewnessExample Spectrum: [s]
25/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Spectral Measures: Centroid and SkewnessExample Spectrum: [a]
26/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Spectral Measures: Centroid and SkewnessExample Spectrum: [f]
27/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Normalization & Predictions
NormalizationFor each measure and for each speaker s,
• µ[f ],s = mean value for utterances of [f], averaged acrosswords and repetitions of that speaker
• For each measurement xi of speaker s, the relative measure xiis xi − µ[f ],s
Thus, xi provides a measure of the relative difference in thespectral measures of [v, s, z, m] to [f], for each speaker.
PredictionThe more obstruent the articulation of [v], the smaller thedifference in centroid and skewness from µ[f ].
28/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Relative Centroidword initial, stressed syllable
29/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Relative Skewnessword initial, stressed syllable
30/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Summary and (Tentative) Conclusions
These data point to 2 kinds of [v], and group Greek and Russiantogether, to the exclusion of Serbian [v].
• Measures for [v] in Greek and Russian are similar to eachother, and are close to [f].
• Measures for [v] in Serbian are more similar to sonorant [m],and differ greatly from [f].
Conclusions?• Serbian /v/ is really [V] – transcriptions are wrong!
Unlessthere exists an even “more sonorant” /v/ and [v]...
• Russian /v/ is phonetically an obstruent... Or is it?Languages with contrasting labial approximants were excluded– maybe that’s what it takes to get a true “obstruent” [v]
30/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Summary and (Tentative) Conclusions
These data point to 2 kinds of [v], and group Greek and Russiantogether, to the exclusion of Serbian [v].
• Measures for [v] in Greek and Russian are similar to eachother, and are close to [f].
• Measures for [v] in Serbian are more similar to sonorant [m],and differ greatly from [f].
Conclusions?• Serbian /v/ is really [V] – transcriptions are wrong! Unlessthere exists an even “more sonorant” /v/ and [v]...
• Russian /v/ is phonetically an obstruent...
Or is it?Languages with contrasting labial approximants were excluded– maybe that’s what it takes to get a true “obstruent” [v]
30/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Summary and (Tentative) Conclusions
These data point to 2 kinds of [v], and group Greek and Russiantogether, to the exclusion of Serbian [v].
• Measures for [v] in Greek and Russian are similar to eachother, and are close to [f].
• Measures for [v] in Serbian are more similar to sonorant [m],and differ greatly from [f].
Conclusions?• Serbian /v/ is really [V] – transcriptions are wrong! Unlessthere exists an even “more sonorant” /v/ and [v]...
• Russian /v/ is phonetically an obstruent... Or is it?Languages with contrasting labial approximants were excluded– maybe that’s what it takes to get a true “obstruent” [v]
31/31
Introduction Distribution Phonological Patterning Phonetic Study Conclusion
Thank you!