Prosodic boundary strength and prevocalic T-glottaling in Massachusetts English

Kamil Kaźmierski
Adam Mickiewicz University, Poznań, Poland

Accents 2023 Łódź:: November 30 - December 2, 2023

kamil.kazmierski@amu.edu.pl
Slides available at: kazmierski-accents2023.netlify.app

1 / 31

T-glottaling in American English
2 / 31

T-glottaling in American English
3 / 31

T-glottaling in American English
Unexpected: Competes with flapping4 / 31

Ex. 1: Intervocalic /t/ as a glottal stop

5 / 31

Ex. 2: Intervocalic /t/ as an 'incomplete' glottal stop

6 / 31

T-glottaling; gender by age effect

Kaźmierski, Kamil. 2020. Prevocalic t-glottaling across word boundaries in Midland American English. Laboratory Phonology 11(1), 1-23.

7 / 31

Method :: Data

Speakers from Massachusetts (N = 20)
Interview speech (UMass Amherst lab)
ASR with WebMAUS (LST Radboud University)
aligned with FAVE-Align (Rosenfelder et al. 2014)
Audio & TextGrids available at https://osf.io/mfu5y/

8 / 31

Method :: AnalysisCorpus querying in Labb-CAT  (Fromont & Hay 2012)
1,652 two word sequences with word-final pre-vocalic /t/Contexts with word$_2$ starting in /ən/ and /ɪn/ omitted

Manual annotation of /t/ based on audition + spectrogramsBoundaries hand-corrected

Automatic annotation of prosodic boundaries with wavelet prosody toolkit (Suni et al. 2020)
 Bayesian Mixed-effects Binomial Logistic Regresion Modeling with brms (Bürkner 2017) in R (R Core Team 2023)
9 / 31

1/7: Glottal stop [ ʔ ]

no release burst
identifiable hold stage
irregular vibration before and/or after

19 yo, Northampton

10 / 31

2/7 Incomplete glottal stop [ ◌̰ ]

no release burst
irregular vibration throughout

21 yo, Framingham

11 / 31

3/7 Glottal reinforcement [ ʔ͡t ]

formant movement indicating tongue gesture
release burst
irregular vibration

19 yo, Northampton

12 / 31

4/7: Inaudible release [ t̚ ]

formant movement indicating tongue gesture
no release burst
no irregular vibration

29 yo, Worcester

13 / 31

5/7 Flap [ ɾ ] flap

clear auditory impression
hold stage + release
decrease in intensity

18 yo, West Roxbury

14 / 31

6/7 'regular' release [ t ]

identifiable hold stage
discernible release burst

21 yo, Worcester

15 / 31

7/7 Deletion [ ∅ ]

no discernible trace of the consonant

20 yo, Westwood

16 / 31

Wavelet prosody toolkit

wavelets in action

https://github.com/asuni/wavelet_prosody_toolkit

17 / 31

Results18 / 31

Overall distribution of allophones

19 / 31

ʔ and ◌̰ → ʔ

20 / 31

ʔ and ◌̰ → ʔ

ɾ → ɾ

20 / 31

ʔ and ◌̰ → ʔ

ɾ → ɾ

the rest → 🚮

20 / 31

ʔ and ◌̰ → ʔ

ɾ → ɾ

the rest → 🚮

N = 1,652 → 1,517

20 / 31

Distribution by strength of prosodic boundary

21 / 31

Distribution by strength of prosodic boundary

21 / 31

Duration of $w o r d_{1}$

22 / 31

Individual variation

23 / 31

Modeling results24 / 31

Modeling results :: Average probability of ʔ

Probability of glottaling over flapping, shaded area indicates 95% Credible Interval

25 / 31

Modeling results :: Predictors (fixed effects)

26 / 31

Random forest

27 / 31

Discussion :: Production planning

28 / 31

ConclusionsT-glottaling rate: empirical 32%, fitted: 8%, 95%CrI = [4%, 15%]
Most common realizations: [ɾ] > [ʔ] | [ ◌̰ ] > [ ∅ ] , next to no [t]
More prominent boundary, longer first word, lower predictability  are positively associated with probability of t-glottaling
The less likely the two words are to be planned together, the more likely t-glottaling becomes

29 / 31

Thank you!

wavelets in action

This research was supported by the Polish-American Fulbright Commission

kamil.kazmierski@amu.edu.pl

30 / 31

wavelets :: prosodic vs. syntactic boundaries

31 / 31

No automatically detected prosodic boundary at X: "I don't get it X I hate", but there is a syntactic boundary. Maybe glottaling is a more direct reflection of syntax than of prosody? Or maybe the objective, signal derived prosody is not the right approach?

Help

Keyboard shortcuts

↑, ←, Pg Up, k

Go to previous slide

↓, →, Pg Dn, Space, j

Go to next slide

Home

Go to first slide

End

Go to last slide

Number + Return

Go to specific slide

b / m / f

Toggle blackout / mirrored / fullscreen mode

Clone slideshow

Toggle presenter mode

Restart the presentation timer

?, h

Toggle this help

Tile View: Overview of Slides

Toggle scribble toolbox

Prosodic boundary strength and prevocalic T-glottaling in Massachusetts English

Kamil Kaźmierski
Adam Mickiewicz University, Poznań, Poland

Accents 2023 Łódź:: November 30 - December 2, 2023

kamil.kazmierski@amu.edu.pl
Slides available at: kazmierski-accents2023.netlify.app

1 / 31

T-glottaling in American English
2 / 31

T-glottaling in American English
3 / 31

T-glottaling in American English
Unexpected: Competes with flapping4 / 31

Ex. 1: Intervocalic /t/ as a glottal stop

5 / 31

Ex. 2: Intervocalic /t/ as an 'incomplete' glottal stop

6 / 31

T-glottaling; gender by age effect

Kaźmierski, Kamil. 2020. Prevocalic t-glottaling across word boundaries in Midland American English. Laboratory Phonology 11(1), 1-23.

7 / 31

Method :: Data

Speakers from Massachusetts (N = 20)
Interview speech (UMass Amherst lab)
ASR with WebMAUS (LST Radboud University)
aligned with FAVE-Align (Rosenfelder et al. 2014)
Audio & TextGrids available at https://osf.io/mfu5y/

8 / 31

Method :: AnalysisCorpus querying in Labb-CAT  (Fromont & Hay 2012)
1,652 two word sequences with word-final pre-vocalic /t/Contexts with word$_2$ starting in /ən/ and /ɪn/ omitted

Manual annotation of /t/ based on audition + spectrogramsBoundaries hand-corrected

Automatic annotation of prosodic boundaries with wavelet prosody toolkit (Suni et al. 2020)
 Bayesian Mixed-effects Binomial Logistic Regresion Modeling with brms (Bürkner 2017) in R (R Core Team 2023)
9 / 31

1/7: Glottal stop [ ʔ ]

no release burst
identifiable hold stage
irregular vibration before and/or after

19 yo, Northampton

10 / 31

2/7 Incomplete glottal stop [ ◌̰ ]

no release burst
irregular vibration throughout

21 yo, Framingham

11 / 31

3/7 Glottal reinforcement [ ʔ͡t ]

formant movement indicating tongue gesture
release burst
irregular vibration

19 yo, Northampton

12 / 31

4/7: Inaudible release [ t̚ ]

formant movement indicating tongue gesture
no release burst
no irregular vibration

29 yo, Worcester

13 / 31

5/7 Flap [ ɾ ] flap

clear auditory impression
hold stage + release
decrease in intensity

18 yo, West Roxbury

14 / 31

6/7 'regular' release [ t ]

identifiable hold stage
discernible release burst

21 yo, Worcester

15 / 31

7/7 Deletion [ ∅ ]

no discernible trace of the consonant

20 yo, Westwood

16 / 31

Wavelet prosody toolkit

wavelets in action

https://github.com/asuni/wavelet_prosody_toolkit

17 / 31

Results18 / 31

Overall distribution of allophones

19 / 31

ʔ and ◌̰ → ʔ

20 / 31

ʔ and ◌̰ → ʔ

ɾ → ɾ

20 / 31

ʔ and ◌̰ → ʔ

ɾ → ɾ

the rest → 🚮

20 / 31

ʔ and ◌̰ → ʔ

ɾ → ɾ

the rest → 🚮

N = 1,652 → 1,517

20 / 31

Distribution by strength of prosodic boundary

21 / 31

Distribution by strength of prosodic boundary

21 / 31

Duration of $w o r d_{1}$

22 / 31

Individual variation

23 / 31

Modeling results24 / 31

Modeling results :: Average probability of ʔ

Probability of glottaling over flapping, shaded area indicates 95% Credible Interval

25 / 31

Modeling results :: Predictors (fixed effects)

26 / 31

Random forest

27 / 31

Discussion :: Production planning

28 / 31

ConclusionsT-glottaling rate: empirical 32%, fitted: 8%, 95%CrI = [4%, 15%]
Most common realizations: [ɾ] > [ʔ] | [ ◌̰ ] > [ ∅ ] , next to no [t]
More prominent boundary, longer first word, lower predictability  are positively associated with probability of t-glottaling
The less likely the two words are to be planned together, the more likely t-glottaling becomes

29 / 31

Thank you!

wavelets in action

This research was supported by the Polish-American Fulbright Commission

kamil.kazmierski@amu.edu.pl

30 / 31

wavelets :: prosodic vs. syntactic boundaries

31 / 31

T-glottaling in American English

T-glottaling in American English

T-glottaling in American English

Unexpected: Competes with flapping

Ex. 1: Intervocalic /t/ as a glottal stop

Ex. 2: Intervocalic /t/ as an 'incomplete' glottal stop

T-glottaling; gender by age effect

Method :: Data

Method :: Analysis

1/7: Glottal stop [ ʔ ]

19 yo, Northampton

2/7 Incomplete glottal stop [ ◌̰ ]

21 yo, Framingham

3/7 Glottal reinforcement [ ʔ͡t ]

19 yo, Northampton

4/7: Inaudible release [ t̚ ]

29 yo, Worcester

5/7 Flap [ ɾ ] flap

18 yo, West Roxbury

6/7 'regular' release [ t ]

21 yo, Worcester

7/7 Deletion [ ∅ ]

20 yo, Westwood

Wavelet prosody toolkit

Results

Overall distribution of allophones

Distribution by strength of prosodic boundary

Distribution by strength of prosodic boundary

Duration of word1word1

Individual variation

Modeling results

Modeling results :: Average probability of ʔ

Modeling results :: Predictors (fixed effects)

Random forest

Discussion :: Production planning

Conclusions

Thank you!

wavelets :: prosodic vs. syntactic boundaries

T-glottaling in American English

Help

“Prosodic boundary strength and prevocalic T-glottaling in Massachusetts English”

T-glottaling in American English

T-glottaling in American English

T-glottaling in American English

Unexpected: Competes with flapping

Ex. 1: Intervocalic /t/ as a glottal stop

Ex. 2: Intervocalic /t/ as an 'incomplete' glottal stop

T-glottaling; gender by age effect

Method :: Data

Method :: Analysis

1/7: Glottal stop [ ʔ ]

19 yo, Northampton

2/7 Incomplete glottal stop [ ◌̰ ]

21 yo, Framingham

3/7 Glottal reinforcement [ ʔ͡t ]

19 yo, Northampton

4/7: Inaudible release [ t̚ ]

29 yo, Worcester

5/7 Flap [ ɾ ] flap

18 yo, West Roxbury

6/7 'regular' release [ t ]

21 yo, Worcester

7/7 Deletion [ ∅ ]

20 yo, Westwood

Wavelet prosody toolkit

Results

Overall distribution of allophones

Distribution by strength of prosodic boundary

Distribution by strength of prosodic boundary

Duration of word1word1

Individual variation

Modeling results

Modeling results :: Average probability of ʔ

Modeling results :: Predictors (fixed effects)

Random forest

Discussion :: Production planning

Conclusions

Thank you!

wavelets :: prosodic vs. syntactic boundaries

Duration of $w o r d_{1}$

Duration of $w o r d_{1}$