1. Postvocalic /r/ and Class Mobility in Edinburgh
Lauren Hall-Lew (U of Edinburgh) & Victoria Dickson (U of Oxford)

2. Postvocalic /r/ in Edinburgh English
Literature review, part 1:
Lawson, E., Scobbie, J. M., & Stuart-Smith, J. (2011). The social stratification of tongue shape for postvocalic /r/ in Scottish English. Journal of Sociolinguistics, 15(2),

3. Postvocalic /r/ in Edinburgh English
typical Middle Class alveolar or retroflex /r/
typical Working Class derhoticised /r/

4. Lawson et al. (2011:256), Methods
“…a socially-stratified UTI corpus ‘ECB08’, collected in the eastern Central Belt of Scotland in 2008 (Scobbie, Stuart-Smith and Lawson 2008)
…Adolescents from an Edinburgh fee-paying school and from a state-sector school in an economically deprived area of Livingston made up broadly distinguished socio-economic groups of middle- and working-class speakers respectively.
Four male and four female volunteers from the school in Livingston and three male and four female volunteers from the school in Edinburgh…
The data presented here is taken from the word-list section of the corpus.”

5. Lawson et al. (2011), Methods

6. Lawson et al. (2011:265), Results
“Our initial auditory identification of a socially-stratified continuum of weaker (WC) to stronger (MC) auditory variants of postvocalic /r/….”

7. Lawson et al. (2011:265), Results
“…was backed up by the ultrasound investigation.
Quantitative analysis of tongue configurations also showed a continuum from WC boys tending to produce /r/ mainly with canonical tip up and front up tongue configurations to MC girls always using bunched tongue configurations.”

8. Postvocalic /r/ in Edinburgh English
Today’s talk:
Dickson, V. & Hall-Lew, L. (under revision). Class, Gender and Rhoticity: The Social Stratification of Postvocalic /r/ in Edinburgh Speech. Journal of English Linguistics.

9. Research Question What about class mobility?
What if you are born Working Class, but become Middle Class by the time you retire?
What kind of /r/ do you produce?
Note: This question necessitates and unusual focus on an older speaker sample

10. Comparison of Studies Lawson, Scobbie & Stuart-Smith:
auditory & acoustic analysis of wordlists
male & female teenagers from Edinburgh (MC) & Livingston (WC)
Dickson & Hall-Lew:
auditory analysis of spontaneous speech
male & female adults ages from Edinburgh, fee-paying (MC) & state (WC) schools, +
third social class category introduced

11. Three Social Class CategoriesWC
left full-time education at age 16 or younger
had done blue-collar work
parents did blue-collar work
E(stablished) MC
university graduates who attended private schools
had done professional or higher managerial work
parents did professional or higher managerial work
N(ew) MC
some were first in their family to go to university
some had been awarded a scholarship for fee-paying school

12. Participants

13. Postvocalic /r/ & Upward Mobility
Literature Review:
Labov, William. 1966b. The Effect of Social Mobility on Linguistic Behavior. Sociological Inquiry. 36(2)

14. Labov (1966b): Types of Mobility

15. Labov (1966b:197)
The working class “U” shows a surprisingly high set of (r) scores, almost equal to that of the lower middle class “U”.
In actual fact, these persons are members of an upper stratum of the working class, having higher occupational skills than most of the “S” group.

16. Labov (1966b:197)
The lower middle class “U’s” show the archetypal pattern of hypercorrection in the use of (r).
As compared to the smaller lower middle class “S” group, the “U” group shows a much wider range of (r) usage shifting from near zero in casual speech to 72 in the most formal contextual style.

17. Labov (1966b:197)
Both the “U’s” among the working class and those in the lower middle class show this hypercorrect pattern, going beyond the upper middle class standard in their more formal speech.
Therefore, we can infer that the shift to the hypercorrect pattern is more characteristic of upward mobility than of membership in any particular socioeconomic group.”

18. Labov (1966b)

19. Comparison of Studies Labov 1966b: Dickson & Hall-Lew:
binary coding of rhoticity
non-rhoticity as older, WC variant
mobility type (3 levels) crossed by class (4 levels)
Dickson & Hall-Lew:
more than 2 possible rhotic variants
derhoticisation as newer, WC variant
but non-rhoticity as an older, (U)MC variant
mobility type (2 levels) only within 1 class (WC)

20. Data Collection Six 1-hour sessions, Nov 2013 to Jan 2014.
Same-sex same-SEC groups of 2-3 speakers.
Sessions led by 1st author; F, EMC Edinburgh.
Talk prompted by a written list of topics (childhood, education, family, work and life in Edinburgh).
Interpersonal dynamics similar across groups; most had met previously or had mutual friends.

21. Dataset N = 5212 Excluded contexts: /r/ followed by a vowel
/r/ followed by a word-initial /h/ that is deleted
/r/ followed by a word-initial /r/
vowel + /r/ voicing duration < 30ms

22. Coding Decisions
Lawson et al.’s 7-point continuum reduced to an auditorily reliable 4-point continuum:
0=no/de- 1=approx 2=schwar 3=tap/trill
We then converted the data for a binary analysis, comparing a relatively sparse set of non-rhotic/derhotic tokens (N=931) to rhotic ones (N=4230), excluding taps/trills (N=51).

23. Descriptive Results F M
del. / derhotic
approximant
schwar
tap / trill

24. Modeling Dependent variable Linguistic Factors:
auditory coding, treated here as a binary factor
0 = non-/derhotic, 1=rhotic
Linguistic Factors:
syllable stress, word final vs. word internal,
preceding vowel, phrase final vs. phrase internal
manner of the following segment, lexical frequency (BNC spoken)
Social Factors: social class & gender
Random Intercepts: word & participant

25. Modeling
All linguistic factors were either eliminated or models would not converge with them.
Best-fit model estimates:

26. Modeling NMC speakers produce the most rhoticity.
WC men produce the least amount of rhoticity, but also the highest proportion of taps/trills.
The biggest gender difference is among the WC speakers.
Since EMC speakers, especially women, have the 2nd-highest rates of non-rhoticity, we think that the ‘0’ category conflates two differently indexed variants: RP non-rhotic & WC derhotic.

27. Discussion
Is the NMC pattern surprising? Should we expect them to sound more like WC speakers?
No: class attainment seems to be a stronger predictor than the social class of one’s parents.
teens’ orientations to local social structures are more predictive than parents’ class (e.g., Eckert 1989; 2000)
teens’ class aspirations are predictive prior to their entrance into the workforce (e.g., Wagner 2012)
adults’ occupation, alone, is an effective predictor of class-based variation (e.g., Macaulay 1977; Horvath 1985)
class might be better defined according to consumption than production (e.g., Mallinson 2007)

28. Discussion
Do NMC speakers show higher rates of rhoticity than EMC speakers because of hypercorrection?
No: rather than ‘linguistic insecurity’, consider language ideologies (e.g. Yaeger-Dror 1992; Milroy 1999; Preston 2013)
E.g., as exemplified in American geek girl speech (Bucholtz 2008); ‘superstandard English’ contrasts ideologically with both standard and non-standard varieties.
In urban Scotland, superstandard speech has been long associated with Morningside- & Kelvinside Englishes (Johnston 1985); both have been described as having exceptionally high rates of rhoticity.
Our claim: stylisation, not hypercorrection.

29. Discussion
What about the derhotic, tapped, and trilled variants among WC men?
EMC & NMC speakers use the variants situated in the middle of the rhotic continuum.
WC speakers, especially men, index their class identity through divergence from these central MC variants.
This divergence occurs in both directions along the continuum, resulting in the use of both strongly rhotic variants (taps and trills) and the least rhotic variants (vocalised, derhotic).
Our claim: stylisation towards a (masculine) Scottish Working Class style

30. Discussion
Do NMC speakers show higher rates of rhoticity than MC speakers because of hypercorrection?
What about the derhotic, tapped, and trilled variants among WC men?
A reviewer suggests that we can’t talk about style because we aren’t analysing intraspeaker variation.
We are planning on being clearer in the revisions that we’re not talking about (and not able to talk about) that kind of style; rather, we’re talking about stylisation.
Lit Review for derhoticisation indexing working class masculine Scottish? E.g., Lawson (Robert), 2011

31. Help, please! How should we model variation in R?
Coded on a 7-point scale
Distribution across points highly unbalanced
Coder confidence fuzzy between points 1&2, 3&4, 6&7
Modeled as a continuous factor
Models then converged and linguistic factors then gained significance. However, how accurate is that?

32. Help, please!
How should we distinguish non-rhotic & derhotic, given probably differences in indexical meaning?
Should we code for the preceding vowel quality, and do a mini-analysis only on those tokens?
If so, is one vowel (e.g., /a/) sufficient?
Note that the corpus isn’t phone-segmented (yet).
If we don’t do this, does it undermine our analysis? (The reviewers didn’t ask for it!)

33. Thank you!

34. Modeling (/r/ as continuous)