LOT summer school Ultrasound, phonetics, phonology: Articulation for Beginners! James M Scobbie CASL Research Centre With special thanks to collaborators Jane Stuart-Smith & Eleanor Lawson Joanne Cleland & Zoe Roxburgh Natasha Zharkova, Laura Black, Steve Cowen Reenu Punnoose, Koen Sebreghts Sonja Schaeffler & Ineke Mennen Conny Heyde Alan Wrench (aka Articulate Instruments Ltd) for AAA software and UTI hardware Various funding – thank you to ESRC, EPSRC, QMU June 2013 Ultrafest! November 2013 Edinburgh • Scottish English 1. Rhotic tongue shape 2. Derhoticisation among WC speakers 3. Vowel system generally • • Is it time for some nitty gritty stuff? Scottish English again 4. Fronted /u/ • Extensions, if time – – Northern Irish /u/ and diphthongs Dutch /r/ Sociophonetics / Lg var & change • Retroflexion vs. bunching for /r/ is claimed to make little or no acoustic difference in US English (Boyce & Espy-Wilson 1997, Zhou et al 2008, Guenther et al 1999) – Sustained phonations • There is no social variation in the appearance of the variant shapes (Mielke et al, 2010, Twist et al 2007) – USA, among rhotic speakers • We find strong and consistent MC (bunch) vs. WC (retroflex) difference in Scottish English 1. Tongue shape for /r/ LM16 “par” TIP UP (retroflex) EF6 “far” FRONT BUNCHED LF1 “purr” FRONT UP EM3 “purr” MID BUNCHED avi • Lawson, Scobbie and Stuart-Smith (2011) • Overlay (n=9-12), each frame a speaker Social variation – MC more bunched • contra Mielke, Baker and Archangeli for US Eng Social variation – WC more tip-up • 2 raters, 49% identical rating and 90% agreement to adjacent categories within a 5 point rating scale, then the 10% redone. Results present average • From light to dark: results • 2x2 Chi2 shows main effects of – class p<0.001 (MC bunching, except for LM15) – gender p<0.01 (female bunching, in WC) results • Should stats be done on a speaker basis? Individual tokens and speakers • MC: Female (left) and Male (right) palates /o/ Mean /r/ with 1 s.d. • WC: Female (left) and Male (right) Mean /r/ (rotated & translated to /o/) • I couldn’t find the chart! • 7/8 WC tip up + 1 rather hyper-triller • 8/8 MC bunched Confirmation – glasgow 2011 • Pre-pausal WC /r/ looks “more retroflex” than MC /r/ • • What about non-prepausal contexts? When the tongue tip raises, we lose image – are these characterisations really what we think they are? • We come back later to – – the 4-way impressionistic categorical analysis WC pre-pausal /r/ tends to be late in its postalveolar constriction as well Summary – tongue shape for /r/ • Vernacular Scottish English is variably derhotic – breaking / diphthongisation before overt rhotic consonant – weakening acoustic rhoticity (loss of trill & high F3 in approx) – social and age-grading provide apparent-time evidence – high % r-loss in contemporary Glasgow (StuartSmith 2007) vs. literature provides evidence of realtime change 2. Derhoticisation and covert /r/ Scottish coda /r/ is weakening in WC speakers Romaine (1979); Speitel and Johnston (1983); Stuart-Smith (2007) • MC auditorily strong, postvocalic /r/ variant (traditionally labelled as an alveolar or retroflex approximant far purr poor • WC auditorily weak, “derhoticised /r/”, including pharyngealised vowels and plain vowels with no /r/ apparent far purr poor Auditory variation in Scottish coda /r/ F3 F2 F3 F2 Derhoticised ear (above) car (below) Rhotic ear (above) car (below) F3 F2 F3 F2 Word-final derhoticisation in ECB08 “…by it’s thought, his passenger. Now, the incident happened at the town’s Hole Farm Road. I went there today and found one woman, young mother, Denise Ponsonby, who claims to have witnessed everything “I heard some screaming – ehm – and I turned to see two men running in the middle of the road – ehm – more* or less - - it, - I heard the guys screaming help – ehm When I’ve turned round – ehm - I seen one man chasing the other – ehm – and then I seen a knife.” *mair /mer/ Typical derhoticised coda /r/ (radio) WC speakers in Edinburgh and Glasgow begin to use derhoticised variants.. .. MC and UC Scots begin to use non-rhotic forms: Morningside Kelvinside. 1700 Derhoticisation first evident in S.E. England, unstressed syllables.. 1800 1900 M&K accents are stigmatised. Non-rhotic forms fall out of favour with MC speakers. 2000 Continuing social and geographical spread of non-rhotic variants in England. Anglo vs. vernacular Scottish “r-loss”: non-rhoticity vs. derhoticisation • 2 raters, 49% identical rating and 90% agreement to adjacent categories within a 5 point rating scale, then the 10% redone. Results present average • From light to dark: results • On the 9-point compromise scale • 2-way ANOVA for class and scale showed no interaction and 2 main effects • Social class p<0.001 • Gender p<0.05 On the 9-point compromise scale • Stuart-Smith reports derhoticised rimes often sound “pharyngealised” for consistent or variable speakers – /ir/ [iə]… (fronter higher vowels, a centralising diphthong) – /ɔr/ [ɔˁ]… (lower backer vowels, a pharyngeal offglide) – /ar/ [ɑ]… (even rhotic speakers have allophonic [ɑɹ] ) • /r/ has multiple gestures (pharyngeal + post-alveolar) where the latter is more “consonantal” (Sproat & Fujimura 1993) & the anterior gesture could show – weakening – temporal delay • Gradual gestural change with complex acoustics • Recall that WC both derhoticise, and are retroflex Articulatory hypotheses • In onset, pharyngeal and anterior gestures are more simultaneous • In coda, dissociation occurs, to varied degrees Tongue blade/tip raising [ɹ] [he] Tongue root retraction [ɹ] [he] [ə] Waterfall time sequence: hair Spontaneous speech, raw video • • Scobbie & Stuart-Smith (2008) find that articulatorily a strong rhotic gesture may be retained, delayed beyond strong source energy into (near-)silence Seen also in Dutch (Scobbie and Sebregts, 2011) car nb bunched /r/ Covert gesture in a derhoticising speaker /r/ maximum constriction after voicing offset before A two-way between – groups ANOVA showed no interaction effects for gender and social class. A main effects model showed a large significant effect for social class F=65.945 p<0.001 η2= 0.328, n=128 Mean proportional lag in CVr## words in ECB08 corpus • Bunched shape is achieved earlier during the rime (Lawson et al) • Tip raising associated with delay and covert /r/ Timing of articlatory blade raising max, relative to the acoustic duration of word 150% 100% 50% tip up front up Timing front bunched dorsal bunched • hut vs. hurt in phrase final position – 2 derhoticising speakers (m & f) – M (LM17) hut vs. hurt – F (LF1) hut vs. hurt2 • Tip up vs. tip down in initial position • Glottal stop vs. glottally reinforced version of /t/ • [folder] Movies (single citation words) • /r/ before a voiceless stop like /p/ or glottal /t/ (or before silence) can be acoustically masked – Environments likely to lead to loss of /r/ • Less likely before voiced lingual consonants due to coarticulation? • Vowels before /r/ in word final pre-pausal position can appear to occur in open syllables for the first time (FUR) – /ʌ/ new phonotactics – /ar/ > /ɑ/ new phoneme – New role for duration in system /ʌ/ vs. /ɑ/ Phonological implications • WC speakers have been observed to have – weak rhoticity during /r/ – breaking & pharyngealisation before /r/ • Ultrasound has shown that – Even tokens without much audible rhoticity at all have a visible /r/ articulation in pre-pausal location, in the silence at the end of the word – WC speakers have more “tip-up” /r/ than MC speakers – MC speakers appear to have a vocalic or syllablic /r/ in some words, like American English /ɚ/ /r/ & derhoticisation summary • Functional explanations emphasise lexicality • Speakers aim to – maximise perceptibility of lexical/grammatical info – minimise effort – also to vary prosodically for information structures • and – express a social identity – vary for social and interpersonal purposes • and – use structured input – deal with novel input Why make “inaudible” gestures? • Acquisition, use, change are socially variable at phonetic and phonological levels Representation Hearing & perception (input vs. intake) acoustic listener oriented output articulation speaker oriented output The speaker-hearer triangle • Are covert articulations long-lasting (i.e. learnable) or a phenomenon found at a point in time when there is an inter-generational loss of /r/? • What do speakers do, when asked to “mimic” a derhotic /r/? – Copy the tongue shape – Copy the timing (late & perhaps covert) – Fail to hear that the derhoticised /r/ is even there at all? • Pilot study by Lawson, with a de-rhoticising model speaker and a derhoticising mimicker. How do covert articulations spread? LM17 “hurt” provided an audio stimulus Brief, delayed tongue-tip raising (derhoticised) LM 17 “hut” Simple tip raising and sometimes none at all folder Original covert contrast in LM17 LM17 “hurt” provided an audio stimulus Brief, delayed tongue-tip raising (derhoticised) Pilot 1’s mimicked version of LM17’s stimulus of “hurt” No tip raising (rless) – makes it rather like “hut?” Mimicry of LM17 Pilot 1’s mimicked version of “hurt” audio stimulus. Pilot 1’s mimicked version of “hut” audio stimulus. •No covert rhotic curl in mimicked HURT. •The durations of mimicked HUT and HURT were almost identical •Both had glottal stop replacement of /t/ Merged hut & hurt in mimicry? • P1 unable to mimic LM17’s hut / hurt contrast • Yet P1 himself has derhotic / covert contrast • In mimicry he appears to be show categorical misperception • He also found connected speech models hard P1’s baseline “hurt” P1’s baseline “hut” • 9 monophthongs in labial & /h/ environments – beam fame hip hem map hum awe hope boom – /ieıɛaɔoʉ/ • 9 (in practice 8) vowels before /r/ – beer bare (fir) herb far fur for bore poor – All vowels take /r/ except /ɪ/ fir (merges with fur) • Issues with UTI and available real lexical items – – – – High/mid vowels are breaking i.e. diphthongal Low /a/ already has categorical allophony Low /ʌ/ does not appear in open syllables Low /ɔ/ has too few minimal pairs (awe vs. or) • We focus on /ʌ/ & /ʌr/ , and on /a/ & /ar/ – HUT, HURT, HAT, HEART 3. Vowel materials for ECB08 open n=41 ʌ a/ɑ p t d m pup pub hub hut butt bud bum mum hum purr fir fur burp (verb herb) hurt Burt bird (perm) firm ma pa baa map - hat - pam palm barb heart part hard harm farm arm par bar harp far parp weak b suburb hammered Two tokens of materials above plus single tokens from: Warm-up liquids: ram, rum, lumber, lamb, cull, Mull, hulk, pill, cult, film, bulb Cool-down vowels: hem, beer, bear, beam, boom, hope, hip, for, awe, poor, fame, bore, hubbub, with extra cool-down materials for MC participants: sure, pure, bare. ECB08 materials: single words • ECB08 corpus shows two (connected?) socially indexed patterns – Variation in tongue-shape – Delayed / weakened post-alveolar constriction • Covert or acoustically weak contrast in WC speakers but no mergers or new vowels yet • Merger of the 3 checked vowels is more common before /r/ in MC speakers than WC speakers, who nevertheless merge fir and fur • Perhaps the bunched shape of MC /r/ is nonaccidentally associated with – Aggressive coarticulation over a preceding central vowel – Early /r/ initiation – Strong rhoticity • Leading to higher likelihood of merger • And the occurrence of a new vowel, rhoticised schwa (“schwar”) or syllabic /r/, whichever seems theoretically less upsetting Checked vowels before /r/ • Mergers acoustics Just the vowels With /r/ too LM16 LM17 LM18 LF1 LF2 LF3 LF4 No recording EM3 EM4 EM5 MC males WC females WC males LM15 EM4 data less reliable - probe shift but still tip down MC females EF3 EF4 EF5 EF6 • MC “V” early in the rime is almost identical to /r/ in average spline-to-spline distance • Not just merger • /r/ vocalisation • Recall that WC speakers derhoticise • 2 “opposite” lenitions of consonantal /r/ Summary and Conclusions • Lex sets BIRD WORD + HERD merged (8/11) – Earth, verb, berth, (err) = firm, word, surf, birth, fur – Monophthong could be rhotic vowel /ɚ/ - it lacks segmentable vowel + transition + rhotic portion • No /a/ split (Pam/palm) – contra Aitken 1979 • /ʉ/ is central and not very high i ʉ o iɹ ʉɹ oɹ e ı ɔ eɹ ɚ ɔɹ ɛ a ʌ verb firm (fur) ɑɹ far Rhotic (MC) speaker • More vowels (and environments) with weak /r/ – No merger of /ɛr/ + /ʌr/ (8/8) -& not [ɚ], but [ʌˁ] (_##) – /a/ “split” (hat/heart) [a] vs. [ɑ] for most derhoticised – /ʌr/ can be very short [ʌˁ] (sir, blur) – /ʌr/ vs. /ar/ (_##) still contrast (car) • Future /ʌ/ merger? (hut/hurt/heart, bud/bird) i ʉ o iə ʉə oʌ e ı ɔ eə ɛ a ʌ ɛˤ herb ɔˤ ʌʕ ɑ far Derhoticising (WC) speaker fur, fir • MC pattern • /r/ remains rhotic but can be more “vowel like”! • /ɛr/ is now merging with /r/ & /ɪr/ • New monophthongal vowel /ɚ/ is descriptive, not causal • /r/ & /ɪr/ had merged in all speakers… without it? – //, /ɚ/, /a/ and /ar/ all remain distinct (& /ar/ >> /r/?) i ʉ o e ı ɔ MC ɛ a(ɑ) ʌ Derhoticisation and the pL inventory • If /r/ is vocalising / derhoticising in WC…? – /r/ // (& /ar/ /ɑ/) in open syllables – a new phoneme and new phonotactic distribution? • /a/ (BRA, PALM) vs. /ɑ/ (FAR, FARM) • // (FIR, FUR) not currently in open syllables – /r/ and /ar/ may merge in some closed syllables? – // may merge with /r/ and /ar/ in closed syll? i ʉ o i ʉ o e ı ɔ MC e ı ɔ WC ɛ a(ɑ) ɛ aɑ ʌ ʌ Derhoticisation and the pL inventory • MC speakers are more phonetically rhotic – Strong rhotic quality to /r/ – New rhotic vowel /ɚ/ instead of V+/r/? – Used in BIRD, WORD, HEARD, leading to merger • WC speakers – Plenty of opportunity to guess what might happen next… “loss” of /r/ leading to new vowels? – Unclear if derhotic BIRD, WORD, HEARD merged • More speculations – /o/ is the new high back corner vowel – /u/ is fronted… but is it lowered phonologically? Summary /r/ and vowels