Designing a sociolinguistic research project.pdf

sociolinguistica

1. Tuesday, 22
nd
October, 2013
Michael Ramsammy
[email protected]

2. `
Starting point for any research project: ◦identify a problem, e.g.
xa hitherto unreported or unexplained observation
xa hitherto untested theoretical prediction
xa conflict in existing literature
`
Define an appropriate research question: ◦narrow and specific
`
Decide upon an appropriate methodology
that enables you to test hypotheses about
your research question.

3. `
Watt (1999) reports that dialect levelling in
Tyneside has resulted in younger speakers
avoiding traditional, local pronunciations of
FACE
and
GOAT
in favour of General Northern or RP-like
productions.
◦Older speakers use [ɪə, ʊə]
◦Younger speakers use [eː, oː] or [eɪ, oʊ].
`
This study was carried out more than 10 years
ago. ◦Under a situation of ongoing dialect levelling, we would
predictthat younger generations of speakers would
continue to abandon traditional dialect forms in favour
of supralocal or standardised forms.
Hitherto untested theoretical prediction!
Now formulate a narrowand
specificresearch question.

4. `
“Variation in
FACE
and
GOAT
vowels in Tyneside
English”
`
“Intergenerational variation in the realisation
of
FACE
and
GOAT
in Tyneside English: an
acoustic study of dialect levelling in progress”
◦What evidence can analysis of realisations of
FACE
and
GOAT
in two age groups (18-25 and 40-50)
reveal about processes of dialect levelling reported
in previous sociophonetic studies on Tyneside?

5. `
Hypothesis 1: current data mirror the results reported in Watt
(1999) ◦suggests that dialect levelling has not progressed significantly in the last
10 years.
`
Hypothesis 2: very high use of [eː, oː] and/or [eɪ, oʊ] variants by
young speakers, tokens of [ɪə, ʊə] are infrequent (and may not
occur at all); [ɪə, ʊə] more common amongst older speakers, use
of [eː, eɪ, oː, oʊ] less frequent. ◦suggests change has continued along the same trajectory over the last 10
years.
`
Hypothesis 3: change is asymmetrical — i.e.
FACE
may show
different rates of levelling from
GOAT
.
◦suggests that dialect levelling has continued, but younger speakers may
have come to evaluate change in
FACE
differently from change in
GOAT
.
Now, what methodology would be most appropriate for testing
these hypotheses?

6. `
To collect data appropriate for addressing
research questions, consider the following when
planning the research proposal. ◦Who are the informants going to be?
◦How will you gain access to informants?
◦How much data will you need?
◦What social factors should be investigated?
◦What data collection strategies will you use?
◦How will you analyse the data once collected?
◦What are the limitations of a given methodology?
◦What ethical issues need to be brought into
consideration?
◦What other problems might arise?

7. data-collection:
qualitativeand quantitative
techniques

8. `
The staple of sociolinguistics: ◦widely used to study many different types of variables.
◦provide recordings of spontaneous speech that can be
analysed in a variety of ways.
`
Typically conducted 1-to-1 or 1-to-2: ◦1-to-1 appropriate for non-vulnerable adults
◦1-to-2 more appropriate for vulnerable informants
xe.g. young informants like school pupils
`
Interviewer asks interviewee(s) a series of
questions: ◦semi-structured interview procedure framed around an
interview schedule
.

9. `
Designed to allow informants to speak freely on a
wide range of topics. ◦The purpose of this is to collect recordings of
spontaneous speech recorded under more-or-less
informal conditions.
`
The task of the interviewer is: ◦to help the informant to relax
xcrucial for collecting useable data
◦to engage with the informant on a personal level
xestablish a rapport: show interest in the informant’s
responses, be polite and honest.
xit’s not a job interview: at best, interviews become casual
chats between two people.

10. `
See appendix of ◦Tagliamonte, S. (2006).
Analysing sociolinguistic variation
.
Cambridge: CUP.
Get informant talking,
check s/he is suitable for
the study.
Some people love
talking about their
families and growing
up: can be a great ice-
breaker towards the
beginning of the
interview.
Others don’t like to say
too much.

11. `
Allow the informant to speak freely. ◦Deviations are fine, just use the schedule to bring up new topics
of conversation as necessary.
◦Fine to spend a long time on one particular section and do other
quickly (or miss them out).
`
Sometimes recommended to end with a free question
(Dörnyei 2007: 138): ◦‘Is there anything else that you would like to add?’
◦‘What should I have asked you about that I didn’t think to ask?’
`
Other tasks can be incorporated into the interview too: ◦picture description tasks
◦story re-telling tasks
◦reading tasks

12. `
Shifts the focus away from the informant’s
experiences and opinions.
`
Some informants are great at this, others
have more trouble!

16. `
Why might we want to use read speech in an
interview? ◦Stylistic effects: compare more informal, spontaneous
speech with more formal, structured speech.
◦Increase token count: depending upon the variable
under investigation, there’s no guarantee you’ll get
enough tokens to be able to conduct a full analysis.
◦Co-occurrence of variables: it can be very difficult (and
sometime impossible) to test for co-occurrence effects
using spontaneous speech. xe.g. what if use of a given variant is dependent upon — or at
least strongly influenced by — use of a variant of another
variable?
x/i/-lowering in Mancunian English (i.e. pronouncing
unstressed word-final -
Y
as [ɛ]) is a working-class feature.
xIt seems to have some relationship with /t/-glottaling.

17. `
Always run them after the main interview, never
before.
`
Data are often better if word lists are embedded in
carrier phrases: (adhere to guidelines from e.g.
experimental phonetics).
`
Use distractors: ◦the number of distractors should be at least equal (and
preferably greater) than the number of target sentences.
x
They had a good party.
x
They sent a large parcel.
x
They sent a long letter.
x
They bought a large ladle.
`
Fully randomise all stimuli.

18. `
http://www.projet-pac.net/
Sample word
lists and
reading text
available here

19. `
Group interviews designed to elicit qualitative responses from
participants: ◦‘collective brainstorming […] participants thinking together, inspiring and
challenging each other’ (Dörnyei 2007: 114).
`
Usual to repeat focus groups with different individuals. ◦But important to participants to have similar enough backgrounds that
they all meet sampling criteria.
`
Interviewer takes more of a back seat: ◦the core purpose is to facilitate interaction between participants.
`
Focus groups are often used in combination with other methods
as a secondary method. ◦Pilot studies:
xhelp researchers to get an idea of the workability of a particular method, or
of the potential validity of working hypotheses before running a full-scale
study).
◦Follow-up studies:
xin cases where a particular result proves difficult to interpret and the
researcher needs further insights/input.

20. `
Researcher based in a particular community of interest: ◦observes and analyses key cultural practices in the community and how
these define patterns of linguistic behaviour.
`
Suitable for long-term research: ◦usually a period of 6 months to a year is considered necessary.
◦typically takes time to find a clear focus for a study:
x‘the exact focus of the research will evolve contextually and ‘emerge’
in situ
only after some fieldwork has been done’ (Dörnyei 2007: 131).
`
e.g. Palo Alto High School (Eckert 2008: 457ff., see also Eckert
2000) ◦Observed and interviewed high school students.
◦Defined two categories of student:
x
new wavers
x‘dressed almost exclusively in black with distinctive pegged pants (pants that
narrowed at the ankles)’
x
preppy group
x‘wore pastel colors and straight-leg designer blue jeans’
◦Ethnographic study of the local, cultural practices in the community of
practice can inform analyses of variation in ways that are not available
from interviewing groups of individuals.

21. data-collection:
qualitative and quantitative
techniques

22. `
Very common in studies aiming at uncovering opinions,
attitudes, beliefs towards linguistic variables: ◦very versatile, can be customised to many different types of
research question.
◦allow for quick and easy collection of data.
`
Many different formats used: ◦pen-and-paper surveys
◦online surveys
◦surveys administered as part of a focus group
`
Survey design and respondent sampling are crucial: ◦questions that are misleading, biased or difficult to understand
may skew results.
◦results may also be skewed if the sampled population is not
balanced/representative of the target group.

23. `
As with everything, questionnaires need a clearly
defined introduction and end.
`
The main questionnaire should be varied. ◦Use a variety of different question types:
xotherwise participants get very bored
xmethodologically sound to test for the same thing in a
number of different ways.
`
Where possible, questions should be fully
randomised (follow guidelines from e.g.
experimental psychology): ◦randomise questions.
◦randomise stimuli, response scales, multiple choice
options etc.

24. `
Open ended questions: ◦e.g. How are you finding this lecture today? Please give as many details as possible.
____________________________________________________________________________
`
Likert scales: ◦strongly agree–agree-neither agree nor disagree-disagree-strongly disagree
`
Semantic differential scales: ◦ratings based on semantically contrasting terms
xe.g. How are you finding this lecture?
useful/informative……………………………………… pointless/dull
`
Numerical rating scales: ◦ratings based on a (clearly defined!!!) numerical scale:
xe.g. How are you finding this lecture?
Very Not at
all
654321
useful
informative
pointless
dull

25. `
True/false questions: ◦Self explanatory, but sometimes yield surprising results!
`
Multiple choice questions ◦e.g. How do you feel this morning? (Tick as many as applicable, or none.)
`
Rank order questions ◦e.g. Rank the following celebrities according to how great you think their contribution
to humanity has been (where 1 =
no discernible contribution
and 5 =
immeasurably
great contribution
):
†happy†hungry
†tired†creative
†drunk†confident
†enthusiastic†irritable

26. `
Designed to test respondents’ reactions to certain voices.
`
We wanted to know what attitudes to certain dialect
features were: ◦TH-fronting (neutralisation of /θ/~/f/ contrast)
◦ING-variation (producing gerundive –
ing
as [ɪŋ] or [ɪn] (or [ɪŋɡ])
◦T-glottalisation (use of [ʔ] for /t/)
`
3 surveys created for 3 major UK cities: ◦London
◦Manchester
◦Edinburgh
`
http://fluidsurveys.com/surveys/mr-K9b/copy-lon-1/

27. Introductory
page gives
brief
information
about who is
running the
survey, who we
consider to be
eligible
participants
(we won’t pay
you if you
don’t meet
these criteria)
and a data
security
statement.

28. Practice question briefly familiarises participants with the task.
They can decide at this point whether they want to continue!

29. Semantic
differential
question
Numerical
rating scale

30. Multiple choice
question
Open ended
question
Multiple choice
question
Open ended
question

31. Include different
tasks to elicit the
same information:
i.e. where is this
speaker from?

32. Supplementary questions about respondents’ speech habits at end
of main questionnaire

33. Collect demographic/personal information at the very end

34. Repeat data security statement
if collecting personal details.
Collect demographic/personal information at the very end
Often, people will only
consider taking part in this
sort of survey if there is a
guaranteed incentive.

35. Variationist studies:
Auditory coding
vs
acoustic/instrumental analysis

36. `
Survey responses can be submitted for statistical testing
with relatively little manipulation.
`
Collate data and code for relevant variables, e.g. ◦Respondent factors:
xage, sex, region, class
◦Question/stimulus factors:
xFor our 3 cities studies: speaker, speaker sex, speaker age, speaker’s
region
Relatively simple coding allows us to address a variety of key research
questions:
e.g. Do male respondents react differently to the occurrence of non-standard
features when they are produced by male speakers and female speakers?
Do working-class respondents have different attitudes to non-standard features
depending upon whether the speaker is working class or middle class.

37. The coding procedure will depend upon the type
of questionnaire, and the types of research
hypotheses that you’re aiming to address.
Relatively simple coding allows us to address a variety of key research
questions:
e.g. Do male respondents react differently to the occurrence of non-standard
features when they are produced by male speakers and female speakers?
Do working-class respondents have different attitudes to non-standard features
depending upon whether the speaker is working class or middle class.

38. `
Interview data often form the basis of
variationist studies
.
`
Variationist studies have a different focus from attitude
studies. ◦What is the scope of variation of a particular variable?
◦What is the influence of social factors on that variation?
◦How do these factors interact with linguistic factors?
`
Answering these sorts of questions requires more complex
analyses of audio recordings.
`
Auditory coding of interview data is a very commonly used
method in sociolinguistics. ◦Essentially, data processing is a transcription task:
xlisten to audio recordings and code the corpus for different realisations
of a particular variable based upon what you hear.

39. `
Study on TH-fronting in adolescent speech
comparing London with Edinburgh.
`
Interview data supplemented with read speech
data: ◦collected in two schools, pupils interviewed in pairs.
◦21 participants from Edinburgh (8 males, 13 females)
and 24 from London (12 males, 12 females)
`
Occurrences of /θ/ throughout the corpus of
interviews were auditorily categorised as: ◦[θ, f, h, ð, ʔ, ∅]
◦carried out by two paid transcribers.

40. `
Each realisation of /θ/ was then coded for linguistic and social factors:
This coding system
permitted us to
address numerous
research questions.

41. `
Study on
happ
Y
and lett
ER
lowering in Mancunian English.
`
20 Mancunians participated in interviews: ◦working-class and middle-class informants were recruited by
social networking.
◦data from a reading task form the basis of the analysis since
tokens of -
Y
and –
ER
in spontaneous speech were too few to
support statistical analysis.

42. `
Formant frequency measurements were extracted
from each token of -
Y
and –
ER
.
◦Normalised F1and F2 measurements were used to test
contextual variation in realisation of vowels both in
height (F1) and backness (F2).
Measurements
taken at vowel
midpoint

43. `
Formant frequency measurements were extracted
from each token of -
Y
and –
ER
.
◦Normalised F1and F2 measurements were used to test
contextual variation in realisation of vowels both in
height (F1) and backness (F2).
Realisations were coded for phonological factors
(e.g. word-position, phonotactic context)
and social factors (speaker sex, social class). Significant effects of phonological and social factors were
observed:
•Word-final prepausal context favours lowering of both
vowels.
•All WC speakers exhibit the –
Y
lowering pattern, but only
a subset use –
ER
lowering.
•Neither pattern significant for MC speakers.

44. `
Recent sociophoneticwork has moved beyond
using auditory and acoustic analysis: ◦begun to incorporate the use of experimental phonetic
techniques into studying variation.
`
Ultrasound study of /ɹ/-variation in Scottish
English: ◦Lawson, E., J. Stuart-Smith & J. Scobbie. (2008).
Articulatory insights into language variation and change:
preliminary findings from an ultrasound study of
derhoticization in Scottish English.
University of
Pennsylvania Working Papers in Linguistics
14/2
(selected papers from NWAV 36).

45. Lawson et al. (2008:103)
Allows for a
‘microscopic’,
frame-by-frame
analysis of variants
of (ɹ).
This level of detail
is simply
impossible to
achieve either
through auditory or
acoustic analysis.

46. email: [email protected] office: 1.13 DSB
office hour: Tuesdays 2pm-3pm
(or email for an appointment)