typological and articulatory perspectives on context...

Mapping from phonetic biases to phonological patternsArticulatory studies

Typological and articulatory perspectives oncontext e�ects

Je� Mielke

North Carolina State University

May 29, 2014

Mielke Typological/articulatory/context 1/37


From sound changes to phonological patternsFrom phonetic biases to sound changes

Overview

How do speech production and perception map onto the typologyof synchronic sound patterns?

I Phonological patterns database: What frequent synchronicpatterns call for an explanation via sound change?

I Questioning the role of social identity in the early stages ofsound change.

I Articulatory studies of covert innovations that can actuate andadvance without speaker awareness.




P-base (Mielke, 2008; Brohan and Mielke, 2014)

I Database of 4560 phonologicalpatterns in 537 languages

I Phonological segments associatedto feature vectors.

I Patterns labeled according to input,output, context,assimilatory/non-assimilatorychange, etc.

I Web interface:http://phon.chass.ncsu.edu/

pbase-app/

basic_consonants

p t̪ t tl ʈ c k k͡pq ʔ

b d̪ d ɖ ɟ ɡ ɡ͡bɢ

pʼt̪ʼ

tʼcʼ

kʼqʼ

ɓɗ̪

ɗʄ

ɠ

ɠ͡ɓ

ʛ

pɸ

pf t̪θ tθ ts tʂtçtɬ tʃtɕʈʂ

cçqχ

bv d̪ð d̪zd̪ð dz dʐdʝ dʒ dʑɖʐ ɟʝ

ɸ f θ s ɬ ʃ ɕʂç x χħ hβ

v ð z ɮ ʒ ʑ ʐʝ ɣ

ʁ ʕ ɦ

m n̪ n ɳ ɲ ŋ ŋ͡mɴ

v̌r̪

rʀ

vɾ ɾð

ɾ̪ ɾ

ɺɽ

ʋ ɹ

ɻ j ɥɰ wl̪ l ɭ

ʎ


http://phon.chass.ncsu.edu/pbase-app/


4560phonological rules

100%

other30.5%

epenthesis7.1%

deletion17%lenition

13.8%

palatal-ization3.2%

otherassimilation

24.8%

finaldevoicing

1.1%

gliding2.5%

other30.5%

Vepen2.3%

glideepenthesis

2%glottal

epenthesis1.6%

Cepen1.2%

Vdeletion

6.2%

glidedeletion

2%

glottaldeletion

2.3%

Cdeletion

6.4%

lenition13.8%

palatal-ization3.2% other

assimilation18.5%regressive

nasal placeassim4.5%

Vnasalization

1.8%final

devoicing1.1%

gliding2.5%



Context-sensitive phonological patterns

I Recurrent phonologicalpatterns re�ecting recurrent(context-sensitive) soundchanges.

I What is missing?

I QUESTION: What doyou think are the relativerates of occurrence ofcontext-free andcontext-sensitive soundchange?

other30.5%

Vepen2.3%

glideepenthesis

2%glottal

epenthesis1.6%

Cepen1.2%

Vdeletion

6.2%

glidedeletion

2%

glottaldeletion

2.3%

Cdeletion

6.4%

lenition13.8%

palatal-ization3.2% other

assimilation18.5%regressive

nasal placeassim4.5%

Vnasalization

1.8%final

devoicing1.1%

gliding2.5%




Well-established bias factors (Garrett and Johnson, 2013)

Bias factors Representative sound changesProduction/perception biasMotor planning Consonant harmony; anticipatory

displacementAerodynamic constraints Rhotacism, other fricative-to-glide

shifts; �nal devoicingGestural mechanics Palatalization; umlaut; VN > V;

vowel coalescence

Systemic biasAuditory enhancement Interdental fricative labialization;

back vowel rounding




Sound change actuation and social factors. . .

I Janda and Joseph (2003):Sound change is initiated byphonetic factors, but spreadprimarily by social factors.

I But how?

I Garrett and Johnson (2013)model: �We will furthersuggest that social factorsinteract with bias variationin ways that lead to soundchange.� (+ �gures →)

–––– -Alan-Yu-c-drv Allan-Yu (Typeset by SPi) of September , :

OUP UNCORRECTED PROOF – REVISES, //, SPi

. Phonetic bias in sound change

Phonetic distribution

Phonetic output (with a bias factor)

Nu

mbe

r of

inst

ance

s

–4 –2 0 2 4 6

050

150

250

Population distribution

Social group identity

Freq

uen

cy

–2 0 2 4 6 8 10

050

100

150

–2 0 2 4 6–2

02

46

8

Bivariate random selection

Phonetic output

Soci

al id

enti

ty

–2 0 2 4 6

–20

24

68

Cycle number 0

Phonetic output

Soci

al id

enti

ty

–2 0 2 4 6

–20

24

68

Cycle number 25

Phonetic output

Soci

al id

enti

ty

–2 0 2 4 6

–20

24

68

Cycle number 50

Phonetic output

Soci

al id

enti

ty

Figure . Simulation of a gradient phonetic bias. The starting phonetic and social identitydistributions are shown in the histograms. The results of a bivariate random selection fromthese distributions is shown in the top right panel. Social group differences are indicated onthe vertical axis, which measures an arbitrary ‘social identity’ parameter. Phonetic output isshown on the horizontal axis, where a value of zero indicates a voiced fricative production, anda value of four indicates a voiced approximant production.The bottom panels show the gradualphonetic drift, from iteration to iteration of the simulation, as the phonetic target includesapproximated variants for one social group, and persistent phonetic instability for the othergroup who do not allow the inclusion of approximated variants to influence the target.

so that the phonetic distribution has a longer tail in one direction than it does in theother. The speech community in this simulation was also characterized by a bimodalsocial stratification with fifty per cent of exemplars produced by one social group andfifty per cent by another group of talkers. Each dot in the top right graph represents anexemplar in the sociophonetic space defined by phonetic output and social identity.At the start of the simulation there is no correlation between the phonetic and socialvalues; the bias factor is equally likely to affect the speech of each population group.The bottom row of graphs shows how this phonetic system evolved over the course offifty iterations of simulated imitation.As seen in Figure ., the phonetic output of the two simulated groups of speakers

diverges. One group (centered around social identity index value ) maintained thestarting phonetic realization—a situation of persistent phonetic instability, where anaerodynamic bias factor influences about ten per cent of all /z/ productions, butthis bias factor does not induce phonetic drift. The other group (centered aroundsocial identity index value ) shows gradual phonetic drift, so that by the end of the




Sound change actuation and social factors. . .

I Janda and Joseph (2003):Sound change is initiated byphonetic factors, but spreadprimarily by social factors.

I But how?

I Garrett and Johnson (2013)model: �We will furthersuggest that social factorsinteract with bias variationin ways that lead to soundchange.� (+ �gures →)

–––– -Alan-Yu-c-drv Allan-Yu (Typeset by SPi) of September , :

OUP UNCORRECTED PROOF – REVISES, //, SPi

. Phonetic bias in sound change

Phonetic distribution

Phonetic output (with a bias factor)

Nu

mbe

r of

inst

ance

s

–4 –2 0 2 4 6

050

150

250

Population distribution

Social group identity

Freq

uen

cy–2 0 2 4 6 8 10

050

100

150

–2 0 2 4 6

–20

24

68

Bivariate random selection

Phonetic output

Soci

al id

enti

ty–2 0 2 4 6

–20

24

68

Cycle number 0

Phonetic output

Soci

al id

enti

ty

–2 0 2 4 6

–20

24

68

Cycle number 25

Phonetic output

Soci

al id

enti

ty

–2 0 2 4 6–2

02

46

8

Cycle number 50

Phonetic output

Soci

al id

enti

ty

Figure . Simulation of a gradient phonetic bias. The starting phonetic and social identitydistributions are shown in the histograms. The results of a bivariate random selection fromthese distributions is shown in the top right panel. Social group differences are indicated onthe vertical axis, which measures an arbitrary ‘social identity’ parameter. Phonetic output isshown on the horizontal axis, where a value of zero indicates a voiced fricative production, anda value of four indicates a voiced approximant production.The bottom panels show the gradualphonetic drift, from iteration to iteration of the simulation, as the phonetic target includesapproximated variants for one social group, and persistent phonetic instability for the othergroup who do not allow the inclusion of approximated variants to influence the target.

so that the phonetic distribution has a longer tail in one direction than it does in theother. The speech community in this simulation was also characterized by a bimodalsocial stratification with fifty per cent of exemplars produced by one social group andfifty per cent by another group of talkers. Each dot in the top right graph represents anexemplar in the sociophonetic space defined by phonetic output and social identity.At the start of the simulation there is no correlation between the phonetic and socialvalues; the bias factor is equally likely to affect the speech of each population group.The bottom row of graphs shows how this phonetic system evolved over the course offifty iterations of simulated imitation.As seen in Figure ., the phonetic output of the two simulated groups of speakers

diverges. One group (centered around social identity index value ) maintained thestarting phonetic realization—a situation of persistent phonetic instability, where anaerodynamic bias factor influences about ten per cent of all /z/ productions, butthis bias factor does not induce phonetic drift. The other group (centered aroundsocial identity index value ) shows gradual phonetic drift, so that by the end of the




Typical patterns for new sound changes (Labov, 2001)

I Age strati�cation(more advanced amongyounger people)

I The curvilinear pattern(more advanced amongthe middle social classes)

I Absence of style shifting(because the new varianthas little social indexicalvalue, and may not evenbe perceived).

Labov Driving forces Page 9

Figure 2. Age stratification of the fronting of /aw/ in the Philadelphia Neighborhood study [N=112]. Vertical axis shows expected values of F2 for each age range, calculated by adding age regression coefficients to the constant [Source: Labov 2001, Ch.5].

1500

1600

1700

1800

1900

2000

2100

Under 20 20-29 30-39 40-49 50-59

Age

Social class distributions drawn from the same regression analysis are

shown in Figure 3, with the characteristic curvilinear pattern. The upper and middle working classes are in the lead, while the lower working class is significantly behind. The middle and upper classes show progressively lower values.

Figure 3. Social stratification of the fronting of /aw/ in the

Philadelphia Neighborhood Study [N=112]. Expected values of F2 calculated by adding social class regression coefficient to the constant. [Sourrce: Labov 2001, Ch. 5].

Age strati�cation in Philadelphia /aw/fronting (Labov, 2001)




Typical patterns for new sound changes (Labov, 2001)

I Age strati�cation(more advanced amongyounger people)

I The curvilinear pattern(more advanced amongthe middle social classes)

I Absence of style shifting(because the new varianthas little social indexicalvalue, and may not evenbe perceived).

Labov Driving forces Page 10

1550

1600

1650

1700

1750

1800

1850

1900

1950

2000

2050

Lower

working

class

Middle

working

class

Upper

working

class

Lower

middle

class

Upper

middle

class

Upper

class

This curvilinear pattern in Philadelphia was located as a test of the

hypothesis that linguistic change stems from a group located in the center of the socioeconomic hierarchy, a relationship found in New York City (Labov 1966), Norwich (Trudgill 1974), Panama City (Cedergren 1973) and Cairo (Haeri 1996). In smaller communities, it has been found that sound changes can serve as symbols of local identity (Labov 1963, Hazen 2000, Holmquist 1988, Frazer 1983). In close studies of social networks in the high schools of Detroit suburbs, Eckert has found differentiation of stages of the Northern Cities Shift by the adolescent realizations of social class (1999). The most recent stages of the Northern Cities Shift are significantly associated with membership in the polar group of“Burn-outs,” youth who show a generalized resistance to institutional norms and adult-sponsored activities.

More detailed studies of the social characteristics of the leaders of linguistic change show that they are female members of the highest status local group, upwardly mobile, with dense network connections within the local neighborhood, but an even wider variety of social contacts beyond the local area (Labov 2001, Ch. 11). Moreover, these leaders of change show a history of nonconformity in other respects, and show a life-long history of resistance to repressive or unjust authority (Labov 2001, Ch. 12). The same configuration appeared in Haeri’s study of palatalization in progress in Cairo Arabic (Haeri 1996).

The curvilinear pattern in Philadelphia/aw/ fronting (Labov, 2001)




A sound change trajectory

nominalvariation

exaggeration stratifiedfeature

indexicalfeature

nominalvariation

ka kaki

tʃi




A sound change trajectory

nominalvariation

exaggeration stratifiedfeature

indexicalfeature

nominalvariation

ka kaki

tʃi

I QUESTION: What proportion of potential soundchanges maintain their trajectory at each stage?




Innovations that do not seem innovative

I Language users overwhelmingly use forms that are already inuse.

I CLAIM: The typology of sound change is the typologyof ways to do something you don't realize you aren'talready doing.

I Predicts another form of underphonologization(Moreton, 2008)

I Innovative forms may have phonetic, structural, orcontact-based sources. . .




Innovators and early adopters (Milroy and Milroy, 1985, 381-2)

�[B]y making a further distinction between INNOVATORS of alinguistic change and the EARLY ADOPTERS, we have suggesteda principled reason for the di�culty experienced in observing theintroduction of an innovation into a community. This may be seenas the earliest stage of a linguistic change � at least from the pointof view of the community which is adopting it.�




Internal sources of innovation

◦ ◦

I innovative variants overtly present in other people's speech(spread of an innovative variant)




External sources of innovation

◦ ◦

I output of a mechanical processI covert reanalysis (e.g., Blevins' CHANCE)I an interpretation of aggregate data about the languageI an output present in another phonological contexts (analogy)

or in another speech style



Covert motivationsVowel studies

Articulatory di�erences: overview

I Persian /t/ deletion:deletion and gestural overlap

I English /s/ retraction:degree of motivation for coarticulation is covert

I French rhotic vowels:shift from front/rounded to bunched, covert shift to retro�ex

I English /æ/ tensing:managing multiple context e�ects




Falahati (2013): /t/ deletion in Persian

197.7 197.8 197.9 198.0

1510

50

-5t101 vaxt besheh

time (s)

dist

ance

(m

m)

t

b

C1 (col 15)

C2 (col 28)

/Xt#b/ → [Xb]

tongue dorsumtongue blade

516.5 516.6 516.7 516.8 516.9 517.0 517.1

1510

50

-5

t331 saxt joon

time (s)

dist

ance

(m

m)

x t

C1 (col 10)

C2 (col 28)

/Xt#j/ → [Xj]

tongue blade

tongue dorsum




Falahati (2013): /t/ by perception and production

S6

S7

S2

S1

S3

S5

S4

% of tokens

subj

ect

0.0 0.2 0.4 0.6 0.8 1.0

/t/ percept/gesture

[t]/full) ∅/full ∅/partial) ∅/none




Falahati (2013): /t/ by perception and production

S6

S7

S2

S1

S3

S5

S4

% of tokens

subj

ect

0.0 0.2 0.4 0.6 0.8 1.0

/t/ percept/gesture

[t]/full) ∅/full ∅/partial) ∅/none

I Mixture of deleted and overlapped /t/s.I Di�cult to match actual rate of deletion without knowing the

rate of overlapped /t/sI Frequency matching results in incrementation of deletion rate.




Baker et al. (2011): [s]-[ô] articulatory di�erence

5 10 15 20 25

0.3

0.4

0.5

0.6

0.7

0.8

Articulatory Difference ([s]t/st[r])

Ret

ract

ion

Rat

io

RetractorsNon−retractors

[StôIN]

[stôIN]

I Articulatory bias towardretraction is variableamong non-retractors.

I Hidden parameter makescompensation forcoarticulation moredi�cult.

I Acoustic target matchingwithout compensation forindividual bias couldresult in (forwards orbackwards)incrementation.




Rhotic vowels in Canadian French

p n ø

pneu

Time (s)0.8192 1.169

0

6000

Fre

quen

cy (

Hz)

Time (s)3.985 4.3610

6000

Fre

quen

cy (

Hz)

p n ø

pneu

b œ K

beurre

Time (s)1.596 2.2690

6000

Fre

quen

cy (

Hz)

Time (s)3.318 3.8870

6000

Fre

quen

cy (

Hz)

b œ K

beurre

Time (s)0.08746 0.6123

0

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

Time (s)2.657 3.1610

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

←F3

p n ø

pneu

Time (s)0.8192 1.169

0

6000

Fre

quen

cy (

Hz)

Time (s)3.985 4.3610

6000

Fre

quen

cy (

Hz)

p n ø

pneu

b œ K

beurre

Time (s)1.596 2.2690

6000

Fre

quen

cy (

Hz)

Time (s)3.318 3.8870

6000

Fre

quen

cy (

Hz)

b œ K

beurre

Time (s)0.08746 0.6123

0

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

Time (s)2.657 3.1610

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

←F3

p n ø

pneu

Time (s)0.8192 1.169

0

6000

Fre

quen

cy (

Hz)

Time (s)3.985 4.3610

6000

Fre

quen

cy (

Hz)

p n ø

pneu

b œ K

beurre

Time (s)1.596 2.2690

6000

Fre

quen

cy (

Hz)

Time (s)3.318 3.8870

6000

Fre

quen

cy (

Hz)

b œ K

beurre

Time (s)0.08746 0.6123

0

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

Time (s)2.657 3.1610

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

p n ø

pneu

Time (s)0.8192 1.169

0

6000

Fre

quen

cy (

Hz)

Time (s)3.985 4.3610

6000

Fre

quen

cy (

Hz)

p n ø

pneu

b œ K

beurre

Time (s)1.596 2.2690

6000

Fre

quen

cy (

Hz)

Time (s)3.318 3.8870

6000

Fre

quen

cy (

Hz)

b œ K

beurre

Time (s)0.08746 0.6123

0

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

Time (s)2.657 3.1610

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

rhotic non-rhotic

I Perceptually and acoustically similar to English [ô]

I Not obviously borrowed from English




Rhotic vowels in Canadian Frenchp n ø

pneu

Time (s)0.8192 1.169

0

6000

Fre

quen

cy (

Hz)

Time (s)3.985 4.3610

6000

Fre

quen

cy (

Hz)

p n ø

pneu

b œ K

beurre

Time (s)1.596 2.2690

6000

Fre

quen

cy (

Hz)

Time (s)3.318 3.8870

6000

Fre

quen

cy (

Hz)

b œ K

beurre

Time (s)0.08746 0.6123

0

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

Time (s)2.657 3.1610

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

←F3

p n ø

pneu

Time (s)0.8192 1.169

0

6000

Fre

quen

cy (

Hz)

Time (s)3.985 4.3610

6000

Fre

quen

cy (

Hz)

p n ø

pneu

b œ K

beurre

Time (s)1.596 2.2690

6000

Fre

quen

cy (

Hz)

Time (s)3.318 3.8870

6000

Fre

quen

cy (

Hz)

b œ K

beurre

Time (s)0.08746 0.6123

0

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

Time (s)2.657 3.1610

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

←F3

p n ø

pneu

Time (s)0.8192 1.169

0

6000

Fre

quen

cy (

Hz)

Time (s)3.985 4.3610

6000

Fre

quen

cy (

Hz)

p n ø

pneu

b œ K

beurre

Time (s)1.596 2.2690

6000

Fre

quen

cy (

Hz)

Time (s)3.318 3.8870

6000

Fre

quen

cy (

Hz)

b œ K

beurre

Time (s)0.08746 0.6123

0

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

Time (s)2.657 3.1610

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

p n ø

pneu

Time (s)0.8192 1.169

0

6000

Fre

quen

cy (

Hz)

Time (s)3.985 4.3610

6000

Fre

quen

cy (

Hz)

p n ø

pneu

b œ K

beurre

Time (s)1.596 2.2690

6000

Fre

quen

cy (

Hz)

Time (s)3.318 3.8870

6000

Fre

quen

cy (

Hz)

b œ K

beurre

Time (s)0.08746 0.6123

0

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

Time (s)2.657 3.1610

6000

Fre

quen

cy (

Hz)

p a K f œ˜

parfum

rhotic non-rhotic

I Perceptually and acoustically similar to English [ô]

I Not obviously borrowed from English




Rhotic vowels perception study (Lamontagne and Mielke, 2013)

I Anecdotally, rhotic vs. nonrhotic is non-salient.

I Francophones are less sensitive to rhoticity of /ø/ and of /œ̃/compared to anglophones from North Carolina.

I Rhoticity is more salient to francophones who do not producerhotic vowels.

I Rhotic /ø/ vs. English [@~] is more distinct than rhotic /ø/vs. non-rhotic /ø/.




Rhotic vowels corpus study (Mielke, 2013)

I Based on two existing corpora (Poplack, 1989; Poplack andBourdages, 2010): Corpus du français parlé à Ottawa-Hull

(adults from Ottawa and Gatineau, recorded in 1982, andCorpus du français de l'Outaouais au nouveau millénaire

(students and teachers recorded in the last decade)

I 21 hours from 75 speakers analyzed using a French forcedaligner (Milne, 2013).




F3 by birth year (speaker means)

2000

2250

2500

2750

1900 1925 1950 1975birth year

norm

aliz

ed F

3 phone

ø

œ̃

œ




F1, F2, and F3 by birth year

1000

2000

3000

1900 1925 1950 1975birth year

norm

aliz

ed fo

rman

t

formant

F3

F2

F1

phone

ø

œ̃

œ




F1, F2, and F3 by birth year

1000

2000

3000

1900 1925 1950 1975birth year

norm

aliz

ed fo

rman

tformant

F3

F2

F1

phone

ø

œ̃

œ

I Backing/rounding shifted to rhoticity after 1960 withoutspeaker awareness.

I The most rhotic speakers are least aware of the di�erence.




Two English [ô] variants (Delattre and Freeman, 1968)

Bunched Retro�ex




Ultrasound data collection methods (Ottawa)

Ultrasound imaging

I Terason T3000 + Ultraspeech (Hueber et al., 2007)

I Mid-sagittal ultrasound and face video captured at 30 fps.

I Palatron system for head movement correction (Mielke et al.,2005; Baker, 2005)

Participants

I 23 francophones from Quebec, Ontario, and New Brunswick

I 16 females, 7 males, aged 18-38 (born 1973-1993)

Stimuli

I 30-35 target words with /ø œ̃ œ/ and 77-108 �llers

I 3 repetitions in carrier phrases:2 × �Je dis .� 1 × �Je dis encore.�




Sample ultrasound images for /ø/ in pneu

Bunched Retro�ex Non-rhotic

.avi ø œ̃ œ .mwv ø œ̃ œ .avi ø œ̃ œ .mwv ø œ̃ œ .avi ø œ̃ œ .mwv ø œ̃ œ




Quantifying concavity: parfum: bunched rhotic

300 350 400 450

-300

-280

-260

-240

-220

-200

-180

frr28

word[, 1]

-wor

d[, 2

]

0.343

−0.475




Quantifying concavity: parfum: non-rhotic

350 400 450 500

-300

-280

-260

-240

-220

-200

-180

frr10

word[, 1]

-wor

d[, 2

]

0.064

−0.559




Quantifying concavity: parfum: retro�ex rhotic

300 350 400 450

-300

-250

-200

-150

frr4

word[, 1]

-wor

d[, 2

]

0.315

0.297




Rhoticity by F3 and concavity: /ø/

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35

0.6

0.7

0.8

0.9

1.0

/ø/ rhoticity

concavity (area)

norm

aliz

ed F

3

10

11

12

13

14

15

1617

18

2021

22

24

25

26

28

3

4

5

6

7

8

9

Change #1 (rhoticity):

I Covert and gradual

I Linear concavity/rhoticity relationship.

Change #2 (retro�exion):

I Covert and abrupt

I No intermediate levelsof retro�exion.

I Bunching andretro�exion areinterchangeable forextreme rhoticity.




Rhoticity by F3 and tongue blade angle: /ø/

-0.8 -0.6 -0.4 -0.2 0.0 0.2

0.6

0.7

0.8

0.9

1.0

/ø/ rhoticity

tongue blade angle

norm

aliz

ed F

3

10

11

12

13

14

15

1617

18

2021

22

24

25

26

28

3

4

5

6

7

8

9

Change #1 (rhoticity):

I Covert and gradual

I Linear concavity/rhoticity relationship.

Change #2 (retro�exion):

I Covert and abrupt

I No intermediate levelsof retro�exion.

I Bunching andretro�exion areinterchangeable forextreme rhoticity.




Rhotic vowels summary

Bunched

8 9 10 11 12 13 14

-9-8

-7-6

-5

X

Y

frr3/22/F/Aylmer QC

ɛøœ̃

I similar to a mid front vowel

I gradual exaggeration of lowF3 or enhancement of low F2

Retro�ex

8 9 10 11 12 13 14

-9-8

-7-6

-5-4

X

Y

frr4/24/M/OTTAWA ON

ɛøœ̃

I no gradual path from a midfront vowel

I diagnostic of new target?




Managing invariance across contexts

Guenther et al. (1999): articulatory variability reduces

acoustic variability in English /ô/.

I /ô/'s low F3 can be achieved by increasing thepalatal constriction length, by increasing thefront cavity length, and by decreasing theconstriction area.

I Speakers exploit trading relations between them.

I /ô/ tongue postures resemble those required foradjacent sounds, and speakers use remainingdegrees of freedom to achieve /ô/'s acoustictarget.




/ae/ tensing contexts

I Tensing before /m n/I Lingual articulations o�set the acoustic e�ects of nasalization

in English and enhance it in French, with idiosyncraticarticulatory strategies (Carignan et al., 2011; Carignan, pear).

I De Decker and Nycz (2012) found interspeaker di�erences inlingual articulations for pre-nasal /æ/ in New Jersey.

I Tensing before /g N/I Variable anticipation of formant transitions associated with

velars

I Tensing before /f T s/

I etc.


/ae/ in North American English (Labov et al., 2005)



Ultrasound data collection methods (Raleigh and Ottawa)

Ultrasound imaging

I Terason T3000 + Ultraspeech (Hueber et al., 2007)

I Mid-sagittal ultrasound video captured at 60 fps.

I Articulate Instruments ultrasound headset

Participants

I 20 English speakers from the U.S. and Canada + 1 from theU.K.

Stimuli

I 120 monosyllabic words, randomized and repeated 3 timeswithout carrier phrase

Collaboration with Chris Carignan & Robin Dodsworth (NCSU).




Ultrasound PCA (Hueber et al., 2007; Carignan and Mielke, ms)

(�gure from Gick et al. 2013)

0 10 20 30 40 50 60

4030

2010

0

nov01.pc1

0 10 20 30 40 50 60

4030

2010

0

nov04.pc1

0 10 20 30 40 50 60

5040

3020

100

nov06.pc1


Vowel space (diagonals)

2000 1800 1600 1400 1200 1000

800

700

600

500

400

300

nov20 normalized and selected F1 and F2

normalized F2 frequency (Hz)

norm

aliz

ed F

1 fre

quen

cy (H

z) IY1

IY0 IH1IH1L

EY1EY1L

EH1

EH1L

AE1 AE1L

ER1ER0 ER1L

AA1

AA1RAW1 AO1

AO1R

AO1L

OW1

OW1L

UW1UW1L



/æ/ tensing before /n/

-0.5 0.0 0.5 1.0

010

0020

00

nov11 c("F1n_frequency", "F2n_frequency")

time (normalized)

NDG

-0.5 0.0 0.5 1.0

-10

12

nov11 art.Z2Z1

time (normalized)

I pre-/m n/ tensing for 19 of20 North Americans(Wilmington, NC example)

I smallest di�erence:Fargo, ND

I no peak: Newfoundland

I no di�erence: UK





-0.5 0.0 0.5 1.0

010

0020

00


time (normalized)

MNNG

-0.5 0.0 0.5 1.0

-10

12

nov11 art.Z2Z1

time (normalized)









-0.5 0.0 0.5 1.0

010

0025

00


time (normalized)

NDG

-0.5 0.0 0.5 1.0

-10

12

nov12 art.Z2Z1

time (normalized)









-0.5 0.0 0.5 1.0

010

0020

00


time (normalized)

NDG

-0.5 0.0 0.5 1.0

-10

12

3

nov21 art.Z2Z1

time (normalized)









-0.5 0.0 0.5 1.0

010

0020

00


time (normalized)

NDG

-0.5 0.0 0.5 1.0

-20

12

3

nov13 art.Z2Z1

time (normalized)








/æ/ tensing: /g/ > /d/

-0.5 0.0 0.5 1.0

010

0020

00


time (normalized)

NDG

-0.5 0.0 0.5 1.0

-10

12

3

nov06 art.Z2Z1

time (normalized)

I /g/ > /d/ by end of vowelfor all speakers (velar pinch)

I from 2nd half of vowel formid-Atlantic, Bu�alo, all butone Southern speaker(Harrisburg, NC example)

I from 1st half of vowel for allNorthwest, Northern(−Bu�alo & +Virginia)(Olympia, WA example)

I entire vowel for all Ontariospeakers (Barrie example)





-0.5 0.0 0.5 1.0

010

0020

00


time (normalized)

NDG

-0.5 0.0 0.5 1.0

02

46

8

nov04 art.Z2Z1

time (normalized)









-0.5 0.0 0.5 1.0

010

0020

00


time (normalized)

NDG

-0.5 0.0 0.5 1.0

-20

12

34

nov20 art.Z2Z1

time (normalized)








/æ/ tensing: /N/ > /g/

-0.5 0.0 0.5 1.0

010

0020

00


time (normalized)

NNGG

-0.5 0.0 0.5 1.0

-10

12

3

nov06 art.Z2Z1

time (normalized)

I 16/20 North Americans: nopre-/N/ peak, but tenserthan pre-/g/(Harrisburg, NC example)

I pre-/N/ > pre-/n/(Fargo, ND)

I pre-/N/ = pre-/g/(Barrie, ON)





-0.5 0.0 0.5 1.0

010

0025

00


time (normalized)

NNGG

-0.5 0.0 0.5 1.0

-10

12

nov12 art.Z2Z1

time (normalized)








-0.5 0.0 0.5 1.0

010

0020

00


time (normalized)

NNGG

-0.5 0.0 0.5 1.0

-20

12

34

nov20 art.Z2Z1

time (normalized)







/æ/ tensing across contexts

I Managing invariance across contexts is challenging

I Tensing before /m n/ indicates distinct vowel target(like /f T s/ in Philadelphia)

I Tensing before voiced velars appears to be di�erent degrees ofanticipation of velar constriction




From phonetic biases to phonological patterns. . .

I Typologically frequent types of phonological patterns matchexpected production/perception e�ects.

I Losses in the sound change→phonological pattern mapping:I QUESTION: What are the relative rates of occurrence of

context-free and context-sensitive sound change?I QUESTION: What proportion of potential sound changes

maintain their trajectory at each stage?

I Non-obvious innovations: Sound change is predicted to favorparameters that can change and progress without beingnoticed, e.g.:

I ambiguity between intentional and non-intentional sourcesI inter-speaker di�erences in magnitude/nature of covert e�ectsI multiple phonetic motivations for similar e�ects


References

Thanks

I Support: This work has been supported by the University of Ottawa andSSHRC grants #410-2007-0735, �Measuring the Phonetic Similarity of SpeechSounds� and #410-2010-0552, �Data Mining Sound Patterns�, and CFI grant#15834 �Sound Patterns Laboratory/Laboratoire des structures sonores�, andthe NCSU Department of English and CHASS.

I Collaborators: Anthony Brohan (MIT), Christopher Carignan (NCSU),Robin Dodsworth (NCSU), Je�rey Lamontagne (Ottawa→McGill)

I Thanks also: Diana Archangeli, Adam Baker, Will Dalton, Nathalie Dion, RezaFalahati, Brendan Henry, Laura Kastronic, William Labov, Lyra Magloughlin,Kimberley Miller, Peter Milne, Elliott Moreton, Shana Poplack, Megan Risdal,Joseph Roy, Alan Yu

I P-base: http://phon.chass.ncsu.edu/pbase-app/



References

Baker, A. (2005). Palatoglossatron 1.0. University of Arizona, Tucson, Arizona.

Baker, A., Archangeli, D., and Mielke, J. (2011). Variability in English s-retractionsuggests a solution to the actuation problem. Language Variation and Change,23(3):347�374.

Blevins, J. (2004). Evolutionary Phonology. Cambridge University Press, Cambridge.

Brohan, A. and Mielke, J. (2014). Frequent segmental alternations in P-base 3.Submitted for Hyman/Plank phonological typology volume.

Carignan, C. (to appear). An acoustic and articulatory examination of the �oral� in�nasal�: The oral articulations of French nasal vowels are not arbitrary. Journal ofPhonetics.

Carignan, C. and Mielke, J. (ms). A PCA-based technique for automatically extractingarticulatory signals from ultrasound video. North Carolina State University ms.

Carignan, C., Shosted, R., Shih, C., and Rong, P. (2011). Compensatory articulationin american english nasalized vowels. Journal of Phonetics, 39:668�682.

De Decker, P. M. and Nycz, J. R. (2012). Are tense [æ]s really tense? the mappingbetween articulation and acoustics. Lingua, 122:810�821.

Delattre, P. and Freeman, D. C. (1968). A dialect study of American r's by x-raymotion picture. Linguistics, 44:29�68.

Falahati, R. (2013). Cluster simpli�cation in Persian. PhD thesis, University ofOttawa.

Garrett, A. and Johnson, K. (2013). Phonetic bias in sound change. In Yu, A. C. L.,editor, Origins of sound change: Approaches to phonologization. Oxford UniversityPress, Oxford.


References

Gick, B., Wilson, I., and Derrick, D. (2013). Articulatory Phonetics. Wiley-Blackwell,Malden, MA.

Guenther, F., Espy-Wilson, C., Boyce, S., Matthies, M., Zandipour, M., and Perkell, J.(1999). Articulatory tradeo�s reduce acoustic variability during American English/r/ production. Journal of the Acoustical Society of America, 105 (5):2854�2865.

Hueber, T., Aversano, G., Chollet, G., Denby, B., Dreyfus, G., Oussar, Y., Roussel, P.,and Stone, M. (2007). Eigentongue feature extraction for an ultrasound-basedsilent speech interface. In IEEE International Conference on Acoustics, Speech and

Signal Processing, pages 1245�1248, Honolulu, HI. Cascadilla Press.

Janda, R. D. and Joseph, B. D. (2003). Reconsidering the canons of sound-change:Towards a big bang theory. In Historical Linguistics 2001. Selected papers from the

15th International Conference on Historical Linguistics, Melbourne, 13�17 August

2001., pages 205�219.

Labov, W. (2001). Principles of Linguistic Change: Social Factors. Blackwell, Oxford.

Labov, W., Ash, S., and Boberg, C. (2005). The atlas of North American English:

Phonetics, phonology and sound change. De Gruyter Mouton.

Lamontagne, J. and Mielke, J. (2013). Perception of Canadian French rhotic vowels.Poster presented at Acoustics 2013, Montreal.

Mielke, J. (2008). The Emergence of Distinctive Features. Oxford University Press,Oxford.

Mielke, J. (2013). Ultrasound and corpus study of a change from below: Vowelrhoticity in Canadian French. In Penn Working Papers in Linguistics 19.2: Papers

from NWAV 41.


References

Mielke, J., Baker, A., Archangeli, D., and Racy, S. (2005). Palatron: a technique foraligning ultrasound images of the tongue and palate. In Jackson, S. and Siddiqi,D., editors, Coyote Papers vol. 14.

Milne, P. M. (2013). The interaction between speech styles and phonology. PhDthesis, University of Ottawa.

Milroy, J. and Milroy, L. (1985). Linguistic change, social network and speakerinnovation. Journal of Linguistics, 21:339�384.

Moreton, E. (2008). Analytic bias and phonological typology. Phonology,25(1):83�127.

Poplack, S. (1989). The care and handling of a mega-corpus. In Fasold, R. andSchi�rin, D., editors, Language Change and Variation. John Benjamins,Amsterdam.

Poplack, S. and Bourdages, J. S. (2010). Normes et variation en français : Rivitalitéentre l'école, la communauté et l'idéologie. Research project funded by SSHRC (#410-2005-2108 [2005-2008]).


typological and articulatory perspectives on context...

Documents