Scottish accents and their value to speech science
Transcription
Scottish accents and their value to speech science
Subtle, Mysterious and Exotic: SCOTTISH ACCENTS AND THEIR VALUE TO SPEECH SCIENCE Professor James M Scobbie With very specific thanks to: Olga Gordeeva, Eleanor Lawson, Marianne Pouplier, Jane Stuart-Smith, Alice Turk, Alan Wrench; all our participants in experiments, fieldwork, outreach and teaching; QMU, Bill Hardcastle, Fiona Gibbon, Nigel Hewlett and the ultras; Kirsty Scobbie for endless support and the idea of doing linguistics in the first place • Spoken language as the object of investigation requires many disciplines and methodologies – social sciences (social and cognitive) – humanities – engineering and physical sciences – physiological and neurological sciences – clinical professions – educational professions • Qualitative and quantitative approaches – Articulatory & Acoustic Phonetics – Phonology Speech Science 2 • • • • • • Accident Insight into a defining human characteristic It’s difficult, satisfying and exciting It’s fun Broad appeal but a small empirical base Always something new to learn, invent or discover • Academia as a profession • Lovely colleagues Why do it? Why did I do it? 3 • An accent of a language is a “sound system” – a system of abstract “phonemes” • around 26 consonants /p t k…/ and 12 vowels /e ɛ …/ • no words have /θ/ (bath with /f/) or /x/ (loch with /k/) – precise spatio-temporal phonetic realisations • coordinated scores of articulatory gestures specified in millimetres and milliseconds • acoustic targets • the precise character of a vowel like /u/ – a choice of which phonemes appear in which words • stair with /e/ or /ɛ/, floor with /e/ or /o/, town with /au/ or /u/ • league with /i/ or /iː/, crisis with /ʌi/ or /aːe/, fur with… What is Scottish English? 4 • Around 7,000 languages and all vary, all change, but – half of these by 0.2% of world’s population – 40% are endangered… up to one language a fortnight dies • Many have characteristics that are rare – Like Scottish English’s rolled /r/? – 20% of languages have a voiced alveolar trill • Some real exotica? (% of languages, % of sounds) – – – – – Short vs. Long Diphthongs (?%) (any diphth = 11%, 10%) Ejective stops (15%, 1.6%, from Acoma to Zulu) Pre-aspirated fricatives (0.2%, 0.2%, Ojibwa) Pharyngealised approx (0.4%, 0.2% Kurdish, Silha) Pharyngealised vowels (1.3%, 3%, Archi etc.) Exotic 5 • We are going to exemplify aspects of Scottish English in some detail – subtleties may leave you mystified – subtle mysteries may intrigue you ☺ • Pronunciation, variation and possible change in word-final consonants – /r/ – /l/ – fricatives /f θ s ʃ/ & /v ð z ʒ/ – stops /p t k/ & /b d g/ • and vowels Mysterious, subtle 6 • Over hundreds of years it’s clear that one phoneme can change into another or disappear – Southern English final /r/ became a vowel – English /kn/ lost the /k/ part • And subtle phonetic variation and change within shorter timescales is normal • How do the abstract “phonological” categories (26 consonants and 12 vowels) relate to the fine-grained “phonetic” details? – Two utterly distinct types of knowledge/phenomena – Two extremes on a continuum A little theoretical question 7 • vowels • stops and fricatives • /r/ and /l/ and more vowels Structure of rest of talk 8 • We use minimal pairs to investigate vowels, and lexical sets – – – – – – heat, niece, greed hate, grade, daze hit, rib, dim hut , hum, bus, het, hem, mess sighed, pie, rise etc., etc. − near, ear, here − bear, bare, fair − fir, birth, dirt − fur, turps, hurt − err, verb, earth SE has ~ 12 vowel phonemes 9 • Small “voicing” effect brute vs. brood /u/ neat vs. need /i/ wrote vs. rode /o/ great vs. grade /e/ • Big “morphological” effect in past tense – brood vs. brewed /u/ – need vs. kneed /i/ voicing effect 80% Relative increase – – – – morphological effect 60% 40% 20% 0% -20% i e ɔ o u • but only for /i/ and /u/ – rode = rowed /o/ – odd = awed /ɔ/ – grade = greyed /e/ Predictable lengthening of /i/ & /u/ 10 Top right – typical SE a. b. c. d. voicing effect 80% Shetlandic-parented Scottish-parented English-parented (male) ditto (female) Relative increase • morphological effect 60% 40% 20% 0% -20% i a. b. c. e o u d. Effect of parental background 11 • Left: side (short [a] part, longer [i] part) • Right: sighed (long [a] part, short [i] part) 5000 0 5000 0 0.663589 0 0 Time (s) 0.7352 Time (s) Predictable short/long /ai/ types 12 • But /ai/ /i/ & /u/ appear unpredictably in certain words – not so rule-governed, especially with word-internal /ai/ – evidence that long vs. short versions of each are emerging • /ai/ in SE can be short (e.g. ice) or long (eyes) – vibes, bribes, Nige, – crisis, psycho, hyphen, rifle, spider, mitre … (next slide) • /i/ in SE can be short (e.g. eat) or long (ease) – league, fatigue, colleague, dweeb, seige, Aristide • /u/ in SE can be short (e.g. oose) or long (ooze) – dude, fugue, cube, stooge, lewd, feud, tube, Froobs • If you have to learn the distribution of long or short duration, then there may be 3 new vowels Emerging vowels 13 Obstr nitrate ibis Pisces Isaac pilot /l/ title idle trifle St Ivel final /r/ mitre spider cypher visor Rottweiler Nasal Brighton taigon hyphen Ivan Simon /ai/ length in trochees by w eak syllable type . % judged "long" 100% 50% 0% obstr l r n voiceless stops voiced voiceless voiced fricatives sonorants postvocalic context Context-specific /ai/ in trochees 14 • The system of abstract categories is complex – vowel quality is dependent on following consonant (trifle vs. rival) or structure (side vs. sighed) – and even non-adjacent consonants (trifle vs. hyphen), an exotic aspect of SE • There may be 3 new vowels, if and when vowel quality breaks free of those environments • We have investigated the categories through intuition and comparison (minimal pairs: hut vs. hat, fur vs. fir), and measurement of acoustic representations using duration and frequency Summary so far 15 Consonant phonemes Labial Lingual td ʧʤ Glottal Stops pb Fricatives fv θð sz Nasals m n ŋ Approx w r l j ʃʒ kg (ʍ x) There are ~26 C phonemes in SE h 16 • place (pre-aspirated /s/) vs. plays (devoiced /z/) Preaspirated fricatives 17 • Ejective stops are pretty common, but there are big idiosyncratic differences as well as social ones realisations of /t/ 70% 60% % of tokens 50% 40% glottal 30% ejective 20% 10% 0% MCF MCM WCF WCM group Ejective stops 18 • New or previously under-reported phonetic variants of stops and fricatives may be quite common in final position – ejective variants of “voiceless” stops /ptk/ – pre-aspirated variants of “voiceless” fricatives /fθsʃ/ • This may be a phonetic change to help maintain the phonemic status quo but it is not clear it is predicted by theories of how sounds “dissociate” in final position for stops Summary: stops and fricatives 19 • It has been observed that final /r/ and final /l/ are weakening and becoming more vowel-like – Acoustic analysis – Articulatory analysis with Ultrasound Tongue Imaging • If and when /r/ and /l/ are “lost” from final position in SE, this will have been because they have “turned into” vowels (cf. RP car) – the number of vowels will increase as /r/ is lost – but phonological merger is also likely • Approximately, but not exactly, a zero-sum game Final /r/ and /l/ and the vowels 20 “the world’s longest tongue”! Annika Irmler from Germany aged 14 … but this is just the tongue tip • most of the tongue is on the inside • it changes shape and location but not volume Vocal tract 22 What ultrasound lets you see • The surface of the tongue • In real time • During speech • Safe 23 • From qualitative “transcription” to quantitative laboratory-based studies with stabilisation UTI: Ultrasound Tongue Imaging 24 Future: corrected high speed data 25 • Since the 1970s coda /r/-“loss” has been reported in working class speech • Stuart-Smith (2003) in a Glasgow corpus including 1415 year old children showed that WC girls have no overtly rhotic consonant for coda /r/ in approximately 90% of cases, boys in about 80% – Middle class children and older adults are rhotic, so the stratified derhoticisation is indicative of change in progress. – /r/ seems to be turning into a vowel right now – Strong impression of pharyngealisation offglide on vowels with monophthongal pharyngealisation on low back ones Losing /r/ in Scotland 26 F3 F2 F3 F2 Derhoticised ear (above) car (below) Rhotic ear (above) car (below) F3 F2 Word-final derhoticisation in ECB08 F3 F2 27 • /r/ follows most vowels as a consonant in its own right in the phonology • We will look at the 9 monophthongs • A typical Edinburgh rhotic speaker – lots of ejectives – loss of the distinction between birth vs. berth – loss of the vowel, not the /r/, in fir, fur, err? – /u/ is fronted and lowered, especially before /r/ Typically rhotic SE 28 • Losses or potential future losses – A phonemic distinction before /r/ for the vowels /ı/ and /ɛr/ • Earth, verb, berth, (err) = third, word, surf, birth, fur • A gain of a vowel – “vocalisation” of /r/ – A new vowel /ɚ/ (or /r/?) in firm, worm, perm i ʉ o iɹ ʉɹ oɹ e ı ɔ eɹ ɚ ɔɹ ɛ a ʌ Rhotic Ed. speaker ɑɹ 29 • /r/ may not be completely absent but is – variable – very weak acoustically (little F3 lowering) – preceded by an audible transition – pharyngeal sounding – More absent-sounding after low back vowels • Some minimal pairs are very hard to distinguish Typical subtle derhoticisation 30 • Losses or potential future losses – the sound of /r/ at ends of words, especially rolled ones – a phonemic distinction or two, maybe • thud/third, hut/hurt, bud/bird, thumb/firm, bust/burst • wad/ward, cod/cord, bomb/ Fawkes/forks, Waugh/war, • Gains in the number of vowels – high and front vowels giving rise to diphthongs – a new contrast of pure vowel qualities giving rise to /ɑː/ • hat /a/ & hut /ʌ/ vs. heart /ɑː/ i ʉ o iə e ı ɔ eə ɛ a ʌ ɛˤ ʉə oʌ ɔˤ ɑː Subtle derhoticising speaker ¡ (ʕ) 31 300 o i e ı F1 (Hz) Dimension 1 ɪ u ɛ ɚ ɔ ɛ ʌ a 1300 3300 F2 (Hz) Dimension 2 Edinburgh south-side 800 32 300 o i e ɛ ı F1 (Hz) Dimension 1 ɛ ɪ ʌ ʌ a 1300 3300 West Lothian F2 (Hz) Dimension 2 ɔ u ɑ 800 33 • Good separation, /u/ is fronted and lowered, Articulation of Ed. vowels 34 • /ı/ is appreciably lower, as expected Articulation of vowels 35 • Tipup (LM17 onset) or tipdown (LM15 onset) Sample ultrasound images of /r/ 36 Tongue blade raising [he] [ɹ] Tongue root retraction [ɹ] [he] [ə] Waterfall time sequence: hair 37 Pre-pausal /r/ may have late (covert?) tip • Low vowels sound derhoticised, acoustically lack F2/F3 approximation, and are near-monophthongs. • Articulatorily a clear rhotic gesture was retained car 38 Let’s try some movies 39 – Red = /¡/ mull (cons) & bulb (vocalised) – Blue = /ı/ film (cons) & pill (vocalised) • Pharyngealised (cons) vs. velarised (vocalised) UTI of laterals 40 • the vowel space revisited Articulation of vowels 41 • Pharyngealisation and velarisation is more extreme in /l/ than in back vowels /l/ compared to /ɔ/ and /o/ 42 • Vocalisation of /r/, /l/ and the vowel system /ı/ /ʌ/ /a/ /ɑ/ V hip hum,hut hat R (firm) fur,hurt far,heart L pill,film Mull,bulb,cult pal • Clear difference between /r/ and /l/ in open and closed syllables • Open syllable /ʌr/ (fur/fir ) and /ar/ are the most likely to lose /r/, then other /ʌ/ Derhoticising speaker 43 • /ai/, /i/ and /u/ might be splitting: long vs. short • /l/ is vocalising but is still different to /o/ • /r/ derhoticisation in West Lothian – acoustic targets are /r/-less – articulatorily, the tip of the tongue gesture may merely be delayed, so /r/ is not lost, not yet – There may be a new long vowel /ɑ/ (far) and short /ʌ/ may able to appear in open syllables (fur) • In rhotic Edinburgh speakers – the tip gesture may be early, making /r/ sound like a vowel (in fur, fir, err) – preaspiration and ejectives are commonplace Summary: variation and change? 44 • Mass observation • Detailed quantitative laboratory-based research – multiple channels • Applications – speech technology – speech and language therapy – second language learning • Distinct languages and traditional dialects are dying out, but accent variation is vibrant and demands closer theoretical attention The future 45 • Accents change – subtle differences in phonetic targets may acquire social distributions • The system largely stays the same in real time – a language maintains its lexical items and contrasts – losses in one area lead to gains elsewhere • Articulatory mechanisms underlying natural speech have rarely been observed before, and new technologies are providing a new point of view – inside in the mouths, and minds, of speakers Conclusions 46