Filetype PDF | Posted on 24 Sep 2022 | 3 years ago
Authentication
digital resources
107x Filetype PDF File size 0.20 MB Source: blogs.umass.edu
Voiceless aspirated coronal VOT values
0.03
0.02 Language
English Ivy Hauser
density0.01 Hindi
Within-category variation in production of Hindi and English stops UMass Amherst
0.00 ASA 2018 Minneapolis
−50 −25 0 25 50
centered VOT
Voiceless aspirated velar VOT values
0.03
Introduction Results: Voice Onset Time (VOT) Results: Voicing
0.02 Language
English
→ Question: How does phonological contrast affect within-category ▸ Dispersion Theory (DT) predicts less variation in Hindi because of more contrasts. ▸ The Hindi stop system exploits the voicing dimension in addition to the VOT dimension
density0.01 Hindi
phonetic variation? All graphing and analysis done in R (R Core Team, 2013; Wickham, 2009). (Kagaya et al., 1975; Lisker and Abramson, 1964; Dixit, 1985).
▸ Dispersion Theory: Systems with more phonological contrasts are expected to exhibit 0.00 ▸ Hindi speakers are expected to be more consistent in production of voicing.
Expected results from DT prediction
less within-category variation than systems with fewer contrasts (Lindblom, 1986). −50 −25 0 25 50 Voicing in Hindi voiced stops
centered VOT
– Hypothesis: Variation predicted by number of phonological contrasts in a system. Expected VOT values (DT) All Hindi voiced stops Speaker with least voicing Speaker with most voicing
▸ But phonological contrasts are not unidimensional in phonetic space 100 100 100
0.04
– Difficulties with extending dispersion theoretic principles to consonants 0.03 Language 75 75 75
– What are the relevant phonetic dimensions? What counts as a “system”? 0.02 English oiced stops oiced stops oiced stops
– Proposal: There is more variation along phonetic dimensions which are not as density Hindi 50 50 50
crucial to the perception of phonological contrast. 0.01
0.00 25 25 25
−50 −25 0 25 50 percent of all v percent of all v percent of all v
Case study: Hindi and English stop consonants centered VOT 0 0 0
none part full none part full none part full
▸ Hindi has four contrasting stops at each place of articulation; English has two. degree of prevoicing degree of prevoicing degree of prevoicing
Experimental results
h h Voicing in English voiced stops
Hindi velar stops: g g k k Voiceless aspirated coronal VOT values
h 0.03 All English voiced stops Speaker with least voicing Speaker with most voicing
English velar stops: g k
100 100 100
▸ Dispersion Theory Prediction: Hindi has more contrasts. Variation in Hindi 0.02 Language
English 75 75 75
expected to be more constrained across all relevant dimensions. density0.01 Hindi
h h oiced stops oiced stops oiced stops
→ Hindi [k ] should vary less than English [k ] in voice onset time (VOT). 0.00 50 50 50
▸ Phonetic dimensions alternative: Hindi speakers will only constrain variation along −50 −25 0 25 50 25 25 25
phonetic dimensions which distinguish more phonological contrasts than English. centered VOT percent of all v percent of all v percent of all v
→ Hindi /k/ and /g/ should vary less than English /g/ in voicing. Voiceless aspirated velar VOT values 0 none part full 0 none part full 0 none part full
– Why? Because of how the phonetic dimensions are used in each language. 0.03 degree of prevoicing degree of prevoicing degree of prevoicing
Phonetic dimensions in Hindi and English stops 0.02 Language ▸ Main result: English stops /b d g/ vary more in voicing than Hindi voiced stops.
voiced unvoiced English – Consistent with phonetic dimensions prediction.
density0.01 Hindi → Variation is both within and between speakers.
aspiration VOT
h h 0.00
Hindi system: g ↔ g k ↔ k −50 −25 0 25 50 Discussion and conclusion
h
English system: ∅ ↮ ∅ g ↔ k centered VOT What if VOT variation is present but experiment failed to detect it?
▸ No difference expected in VOT because VOT space of contrasts is the same. Expected VOT values (DT) ▸ If the variation in question in only random variation not structured variation, a finer
0.04
▸ Difference expected in voicing because Hindi has voicing contrasts (Kagaya et al., 1975). Why are Hindi speakers just as variable as English speakers? grained analysis is required.
0.03 Language
(Levene’s Test for homogeneity of variance not significant.) ▸ The Hindi speakers were all multilingual (different L2s) and living in a dominant English
0.02 English
▸ Hindi does not differ from English in VOT – both have a single VOT contrast. speaking environment, could potentially affect VOT variation?
The Experiment density0.01 Hindi
▸ Unvoiced lag time or after closure time (Mikuteit and Reetz, 2007) only distinguishes ▸ It may not be possible to observe the effect in a lab setting.
0.00
▸ Production task. Native speakers of each langauge read CVC words where first one of the contrasts.
consonant was a stop followed by [i a u]. −50 −25 0 25 50 Results consistent with Dispersion Theory with two modifications:
▸ VOT dimension realizes the same contrast in both languages, so no difference expected.
▸ 14 Hindi speakers and 9 English speakers recorded; 7 of each analyzed after exclusions. centered VOT ▸ Prevoicing is a separate dimension from lag time and does not form a continuous scale
▸ Phonological systems exploit multiple phonetic dimensions. with VOT (Mikuteit and Reetz, 2007).
▸ Carrier phrases: Say X again; Dobara X doharao. VOTis only one of the relevant dimensions here. ▸ The “system” of phonological contrasts is defined over multiple phonetic dimensions.
http://blogs.umass.edu/ihauser/ @hauserivy
no reviews yet
Please Login to review.
