covert articulation of scottish english /r/ now you see and hear it… now you don’t mfm 14 2006...
TRANSCRIPT
Covert articulation of Scottish English /r/ now you see and hear it…
now you don’tMFM 14 2006Manchester
James M Scobbie Speech Science Research Centre, QMUC
Jane Stuart-Smith English Language, Glasgow
Overview
• Plus ça change, plus c'est la même chose When is phonological change phonological?How is fine phonetic detail grammaticalised?What are phonological features?What is a phonological inventory?
• Coda /r/ derhoticisation in Scottish EnglishStudy 1: Auditory and acoustic – socially stratifiedStudy 2: Ultrasound Tongue Imaging – pilot
Coda /r/ in Scottish English• Scottish English is typically described as rhotic
(e.g. Wells, 1982: 10-11) • Coda /r/ is “phonetically” variable
[] - trills are rare and/or stereotypical (Ladefoged and Maddieson, 1996: 236)
[] - alveolar taps are more often noted (e.g. Johnston 1997)
[] [] – approximants – retroflex and post-alveolar - are also common
(e.g. Johnston 1997)
Coda /r/ is changing • Changes to coda /r/ have been reported in
working-class speakers in Edinburgh (e.g. Romaine 1978) and Glasgow (Johnston 1997; Stuart-Smith 2003) to a very weak approximant vowels produced with secondary articulation (e.g.
pharyngealization / uvularization)vowels without any audible secondary articulation,
i.e. similar to vowels in syllables without /r/
Characteristics of /r/
• Differing acoustic properties for approximants
(e.g. Ladefoged and Maddieson 1996)
lowered F3 – retroflex and post-alveolar approximants
high F3 – uvular articulations
• Coda /r/ in Dutch also shows variable ‘deletion’
(Plug and Ogden 2003; Scobbie & Sebregts 2005)
longer vowels differing vowel and consonantal qualitycovert post-alveolar articulations
Study 1: Coda /r/ in Glaswegian
• 12 male working-class informants1m = 10-11 years2m = 12-13 years3m = 14-15 years4m = 40-60 years
• Words selected from larger wordlist
hat ban fan cat
heart barn farm card far car
Study 1: Coda /r/ in Glaswegian
• Impressionistic auditory analysistranscription
• Acoustic analysisduration of vocalic portionvowel quality by formant analysis (midpoint; every 5
pulses up to and including end of vocalic portion)
Auditory results
Older speakers showed most articulated /r/ - [] [] []:
[] 4m1_farm and even []:
[] 4m2_car
Auditory results
Younger speakers showed: weakly approximated [] [] []:
[] 3m1_far pharyngealized/uvularized vowels:
[a] 2m1_card
Auditory results
Younger speakers showed - vowels with no audible ‘colouring’
[] 1m3_car
odd instances of vowels followed by [h] or []
[] 3m3_far
Acoustic analysis - duration
Overall, the vocalic portion of words with /r/ is longer than those without /r/ (p =.0039).
0
50
100
150
200
250
300
350
400
hat cat ban fan heart card barn farm far car
ms
1m1
1m2
1m3
Age group 1
Acoustic analysis - duration
This is regardless of whether an apical /r/ is heard (red dots) or not.There is also some variation.
0
50
100
150
200
250
300
350
400
hat cat ban fan heart card barn farm far car
ms
3m1
3m2
3m3
Age group 3
Acoustic analysis – vowel quality
Midpoint formant values show that words with /r/ are generally more retracted than for words without /r/.
500
600
700
800
900
1000
1100
10001200140016001800F2 (Hz)
F1
(Hz)
a - 1m2 ar - 1m2 a - 1m3 ar - 1m3
Age group 1
Acoustic analysis – vowel quality
Words heard with /r/ (red dots), tend to be even more retracted.
500
550
600
650
700
750
800
850
10001100120013001400150016001700
F2 (Hz)
F1 (H
z)
a - 3m1 ar - 3m1 a - 3m2 ar - 3m2 a - 3m3 ar - 3m3
Age group 3
Acoustic analysis – vowel quality
Sample tracks (3m1 ‘rhotic’) shows slight dip in (high) F3 in most words with /r/.
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
endend-1end-2end-3end-4end-5
heart F1
heart F2
heart F3
card F1
card F2
card F3
barn F1
barn F2
barn F3
farm F1
farm F2
farm F3
far F1
far F2
far F3
car F1
car F2
car F33m1
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
endend-1end-2end-3end-4end-5
cat F1
cat F2
cat F3
hat F1
hat F2
hat F3
3m1
Acoustic analysis – vowel quality
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
endend-1end-2end-3end-4end-5
ban F1
ban F2
ban F3
fan F1
fan F2
fan F3
cat F1
cat F2
cat F3
hat f1
hat f2
hat f3
3m3
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
endend-1end-2end-3end-4end-5
heart F1
heart F2
heart F3
card F1
card F2
card F3
barn F1
barn F2
barn F3
farm F1
farm F2
farm F3
far F1
far F2
far F3
car F1
car F2
car F33m3
Sample tracks (3m3 ‘pharyngealized /r/’) shows high, flat F3.
Phonological Implications• Has /r/ changed phonologically?
How can we tell? If only from neutralisation then “phonology” is thin
• What is changing in speakers’ grammars? Features and phonotactics?
o Place, manner, timing, duration, phonation all affected
Fine-grained phonetic targets?o Articulatory or acoustic? o How is variation encoded?
Why ultrasound?
• Ultrasound Tongue Imaging (UTI)Relatively informalDynamicReal-timeImage of whole mid-sagittal tongue surfaceImpressionistic and objective analyses
• /r/ is characterised byOpen approximationMultiple articulations
Study 2. Pilot 1. Field transcription
• Glasgow Science Centre, QM open days, Edinburgh International Science FestivalLive qualitative analysisNumerous subjects (dozens)All age groups, wide spectrum of social mixHandheld probe plus microphonePossible to record data for re-analysis
• Visual and auditory transcription
Pilot 1. Preliminary results
• Lots of inter-speaker variation• Acoustically derhoticised /r/ is often
Acoustically something else (cf. Study 1)Articulatorily present
oMay involve retracted tongue rootoMay be anterior
– retroflex or bunched (inter & intra-speaker variation)
• Little or no meta-linguistic self-awareness of change or variation in /r/ among ScotsCf. labiodental /r/, vocalised /l/ and others
Study 2. Pilot 2. Lab study
• Laboratory recordingsStill piloting methodHead stabilisationHigher sampling rate to become available
• Subject read from semantic-class wordlist e.g. “eyes, hair, teeth, nose, ear, mouth”
Study 2. Pilot 2. UTI lab subjects
• Control rhotic speaker, female (23) Argyll• UTI shows characteristic retroflex /r/
bar harm
Pa ham
Study 2. Pilot 2. continued
• Derhoticiser, male (22) Edinburgh• Impressionistically
Coda /r/ vary from weak approximants to vocalisation
Onset /r/ is approximant or fricativeMedial /r/ may be tapOnset clusters are tapped, approx, affricated
• Other variables also suggest he is comparable to derhoticisers from Study 1
Pilot 2. Vowel space & inventoryVowel space
200
300
400
500
600
700
800
9001100130015001700190021002300
F2
F1
i
3e
3E
3o
3I
3a3A3
Pilot 2. UTI – derhoticising speaker
• He has acoustic (and articulatory) rhotics
• Approximantsrain
• Taps ferry
Pilot 2. Acoustics – higher V + /r/
• Weakly rhoticised forms shading into derhoticised centring glides & diphthongs
Pilot 2. continued – lower vowels + /r/
• Derhoticisation is more frequent, with relatively monophthongal productions – yet no mergers?
• Weak syllables may sound highly vocalised
Articulatory dynamics with UTI
• Scobbie & Sebregts (2005) at MFM Dutch derhoticisationCovert /r/ reflex
oeasier to see, harder to hearo late, devoiced, weakened, coarticulated
• Scottish pilot speaker also has visible but not so audible anterior lingual constrictions
UTI orientation• A frame of [] from rain
• Tongue surfaceis the clearestfeature – whiteline
• Internalstructures are visible and helpgin transcription
UTI – derhoticising speaker
• Covert anterior rhotic-like post-alveolar tongue movement in derhoticised wordscar, storm, suburb
car towards end of phonation car target 120ms later
covert tip raising
Summary & discussion
• Fairly extreme auditory derhoticisation Listeners hear little rhoticity from speakers like thisProbably can acquire “same” contrasts, lexical sets
• Articulatory evidence of an [] (and an /r/)Anterior gestures are delayed and/or weakPosterior (pharyngeal?) gestures also seen
Targets
• We assume acoustic derhoticisation and covert articulatory targets are required in the grammarAre the targets compatible or incompatible?Speaker-hearer models suggest there is no need to give
either priority… they are in equilibrium
• Various modelsDemands from speech production tend to make speakers
economical with effort and reduce contrastivityPerceptual demands from listeners tend to make speakers
enhance contrasts
• Covert articulation is the opposite • Speakers / hearers have social demands too (Foulkes
& Docherty 2005)
Rough exemplar model
• A shared lexicon is crucialHighly detailed lexical entries (exemplars)Quantity of stored memories causes overlap and
abstraction of commonalitiesAbstraction = formation of
o categorical features (recurrent if functionally-motivated)o gradient tendencies (may also be recurrent)
• Sociophonetic variation is crucialIt stretches and structures phonetic variationLearning and abstraction are not replication of input
Rough exemplar model
• Within a prosodic position, nothing is gained by positing independent labels such as “/r/” in addition to the fine social and phonetic detail plus recognising emergent recurrent categories
(cf. Docherty 1992, Scobbie 2006)
Rough Model
• We create a system mediated by the input
• Our intended output is mediated by our articulation
• Cognitive knowledge has to reflect all three loci
TheSpeaker
HearerTheCommunity
Conclusion
• Derhoticisation is a typical phenomenon of central phonological interest
• To merely describe the linguistic situation in Scottish EnglishWe need more phonetic detail We need more social detail
• To develop theories of the traditional core topics of phonology We need new quantitative evidence of all sorts
THE END
Thanks for listening and watching