early adult lifespan Richard Rhodes University of York - PowerPoint PPT Presentation

Changes in the voice across the early adult lifespan Richard Rhodes University of York richard.rhodes@york.ac.uk

2 What? • Longitudinal acoustic study • Formants and fundamental frequency • 8 speakers • Real-time over 3 decades • 7 year intervals • Ages 21, 28, 35, 42, 49

3 Why? • Need for real-time (spontaneous) speech research • Applications ▫ Speaker comparison cases with long delay  e.g. Yorkshire Ripper Hoaxer (27 years, R v Humble 2005) ▫ Building reference populations for LRs and ASR ▫ Foils for voice parades ▫ Any application of speech science where aging is apparent

4 Why does it matter? • Should be aware if speech features change significantly through early adulthood • Important to be able to estimate direction and magnitudes of change • Assumption in linguistics that language is set and unchanging by adulthood ▫ Currently being challenged

5 ‘7 Up’ Dataset • Recorded every 7 years, from ages 7 to 49 (currently) • Spontaneous speech in an interview setting • Short-term non-contemporaneous data • (Licensed by ITN for Granada) • Michael Apted (1964)

6 ‘7 Up’ Dataset • Average sample length 5 minutes • Vowel token Ns range from 4-30

7 ‘7 Up’ Subjects • 8 subjects (6 male, 2 female) • Geographically – 2 highly mobile, 2 somewhat mobile, 4 stable • Range of accents/regions

8 Tests • f0 • F1, F2, F3 of 9 monophthongs • Future study: ▫ Diphthongs ▫ Voice quality ▫ Temporal features ▫ Consonantal features

9 Predictions: physiology • Reduction in f0 • 10% (Hollien & Shipp, 1956) or 14Hz (DeCoster and Debruyne, 2000) • Sensitive to smoking (Verdonck-de-Leeuw & Mahieu, 2004) • Less marked for females (Linville, 2001)

10 Predictions: physiology • Reduction in formant frequencies ▫ Endres et al. (1971), Linville (2001 etc.), Reubold et al. (2010) • Contraction/reduction of the vowel space ▫ Ratstatter and Jacques (1990), Ratstatter et al. (1997)

11 Predictions: physiology F2 • 21 -> 49 years old we would expect that... 49 • Vowel space is F1 ▫ contracted 21 ▫ displaced

12 Predictions: sociophonetics • Speakers adjust for mainstream accent changes ▫ Queen’s English - Harrington et al. (2000 etc.) • Cases involving geographical mobility (Neil) ▫ Salient between-accent differences reflected ▫ Compare phonology of each region/accent

13 Results • Generalised results (all speakers) ▫ f0 ▫ F1, F2, F3 ▫ Vowel space • Specific case study examples ▫ Illustrate predictions about physiology ▫ Illustrate sociophonetic predictions

14 Results: f0: general patterns • Males: minor decreases in f0 ▫ ( average 3% between 21 and 49) • Females: exhibited decreased f0 ▫ ( 8% and 23% between 21 and 49) ▫ 23% decrease was a habitual smoker

15 Results: f0: by speaker 250 230 210 Andrew 190 Bruce 170 Neil 150 Hz Nick 130 Symon 110 Tony 90 Lynn 70 Suzy 50 1 2 3 4 5 49 21 28 35 42

16 Results: F1: general patterns • Average F1 decrease : 8.5% • All speakers • Close front vowels reduced more than open vowels

17 Results: F1: by speaker Mean percentage F1 decrease between 21-49 0 % 5 10 15 20 25

18 Results: F1: by vowel Mean percentage F1 decrease between 21-49 -10 ʊ o a: ʌ a e ı i: u: -5 0 5 % 10 15 20 25

19 Direction p ≤ 0.01 p ≤ 0.05 Results: F1: significance Decrease *** * Increase *** * Mixed *** * F1 Andrew Bruce Lynn Neil Nick Suzy Symon Tony ʊ \ \ \ ** n \ * \ o \ * ** *** n * *** n a: \ * \ * ** *** * * ʌ \ * n ** n n n n a n n n n n ** n n e * *** * *** n n * * I * * *** *** *** *** *** * i: *** *** \ *** n *** *** n u: \ * ** ** n n ** \

20 Results: F2: general patterns • Average F2 decrease: 3.7% • 6/8 speakers • Close front vowels reduced less than other vowels

21 Results: F2: by speaker Mean percentage F2 decrease between 21-49 -4 -2 0 2 % 4 6 8 10

22 Results: F2: by vowel Mean percentage F2 decrease between 21-49 -4 ʊ o a: ʌ a e ı i: u: -2 0 2 % 4 6 8 10 • /u:/ fronting in English ▫ (Hawkins and Midgley, 2005)

23 Direction p ≤ 0.01 p ≤ 0.05 Results: F2: significance Decrease *** * Increase *** * Mixed *** * F2 Andrew Bruce Lynn Neil Nick Suzy Symon Tony ʊ \ \ \ n n \ n \ o \ n * n n n n n a: \ n \ n * * *** n ʌ \ * * * *** n n *** a n * *** * *** n * * e n n n * *** n * n I ** ** n *** n n *** ** i: n * \ n * * n n u: \ n n n n n n \

24 Results: F3: general patterns • Average F3 decrease: 2.2% • 6/8 speakers • Far less consistent and marked than F1 and F2

25 Results: F3: by speaker

26 Results: F3: by vowel Mean percentage F3 decrease between 21-49 -6 -4 ʊ o a: ʌ a e ı i: u: -2 0 % 2 4 6 8 10

27 Direction p ≤ 0.01 p ≤ 0.05 Results: F3: significance Decrease *** * Increase *** * Mixed *** * F3 A B L Ne Ni Su Sy T ʊ \ \ \ n \ n \ n \ o \ n n n *** n n * a: \ n \ * n ** n n ʌ \ * n * n *** ** * a ** *** n *** n ** *** * e n * n *** n n * n I ** ** n *** n *** *** ** i: n n \ n n n n n u: \ n n * n n * \

28 Results: vowel space area: by speaker VSA percentage decrease between 21-49 -50 -40 -30 -20 Andrew Bruce Lynn Neil Nick Suzy Symon Tony -10 0 % 10 20 30 40 50 60

29 Case example: Bruce (non-mobile) Hz Bruce • non-mobile • SE England • upper middle Mean F1 class

30 Case example: Bruce (non-mobile) Hz Mean F1 Mean F2

31 Case example: Bruce (non-mobile) Hz Mean F1 Mean F2 Mean F3

32 Case example: Bruce (non-mobile)

33 Case example: Bruce (non-mobile)

34 Case example: Bruce (non-mobile)

35 Case example: Bruce (non-mobile)

36 Case example: Bruce (non-mobile) Area Vowel space area

37 Case example: Neil (mobile)

38 Case example: Neil (mobile) • /a/ very likely to show F2 reduction • Liverpool /a/ is relatively back (low F2) ▫ Ferragne & Pellegrino (2010) • -> SSBE /a/ = increase in F2

39 Case example: Suzy (RP)

40 Case example: Suzy (RP) • Expect /a/ to show decrease in F1 • Following mainstream pattern in RP for /a/ to lower ▫ Hawkins and Midgley (2005) • Increased F1

41 Summary • Vowel formant frequencies are reduced across most speakers ▫ F1 – 8.5% ▫ F2 – 3.7% ▫ F3 – 2.2% • Some vowels are more/less robust to changes • Fundamental frequency exhibits some reduction ▫ More marked in female speakers & smoker

42 Implications • Be informed about likely changes as a result of long- term delays in casework • Be aware of age-correlated features when building reference populations • Be aware that language and speech is flexible in adulthood, especially in cases of mobility

43 Thank you for your time and suggestions richard.rhodes@york.ac.uk

44 Formant Summary by speaker Mean percentage F1-3 decrease between 21-49 18% 16% 14% 12% 10% F1 8% 6% F2 4% F3 2% 0% -2% -4%

45 Formant Summary by vowel Mean percentage F1-3 decrease between 21-49 Mean percentage F1-3 decrease between 21-49 18% 25% 16% 20% 14% 12% 15% 10% F1 F1 8% 10% 6% F2 F2 5% 4% F3 F3 2% 0% 0% -2% -5% ʊ o a: ʌ a e ı i: u: -4% -10%