Department of Linguistics and English Language, Lancaster University, LA1 4YT, United Kingdom
Tel: +44 (0) 1524 594577 Fax: +44 (0) 1524 843085


LVLT talk: 'Sociolinguistics, salience and statistics: outlining some new directions in language attitude research' Kevin Watson & Lynn Clark

Date & Time: 10 March 2011, 12.00-1.00
Venue: County South C89
Who can attend: Anyone

The LVLT talk this week will be given by

Kevin Watson & Lynn Clark


Sociolinguistics, salience and statistics: outlining some new directions in language attitude research



There have been a number of methodological advancements in recent work eliciting and measuring language attitudes. For instance, changes to the questionnaire design (e.g. from Likert scale/semantic differential scales to magnitude continua, see Redinger 2010) have allowed for more detailed and reliable empirical coding of results, and changes to the stimuli (e.g. 'cut and paste' methods in pseudo-matched guise experiments, see Campbell-Kibler 2006; 2008) have provided a better understanding of the complexity of social meaning, from the perspective of the listener. One area of methodology which is still in its infancy concerns the dynamic nature of response elicitation (cf Labov 2008). Evaluative responses (and the thought processes which produce them) are dynamic events, constantly changing through time as the signal unfolds, yet the tools we usually use to capture these reactions allow us only to see a snap shot of this process by recording reactions at a single point in time. Continuous audience response data collection tools have been successfully applied to studies collecting dynamic real-time reactions to performing art (Stevens et al 2009), music (Schubert 2001) and dance (Stevens et al 2009). This paper outlines a new methodology which attempts to employ similar techniques to capture listeners' real-time reactions to variation in a speech signal.

We report on a web-based pilot study which assessed 20 listeners' reactions to 5 regional varieties of English in the UK. We used bespoke audience response software which incorporated Redinger's (2010) magnitude continuum and measured listeners' evaluative responses to the basic dimensions of solidarity and status at one second intervals. Listeners reacted to these 5 regional varieties in largely predictable ways (cf. Coupland & Bishop 2009) which confirms that listeners are capable of relaying attitude reactions in real time. The novelty in this approach, however, is that it can also be used to examine the points in time at which major shifts in response have taken place. This can be done empirically by applying the statistical techniques of Change Point Analysis (using the changepoint package designed for the R environment, Killick 2010).

A preliminary analysis of change points in these data suggests that there are significant shifts in evaluative reaction towards all 5 regional varieties on both dimensions (status and solidarity). There are also some significant similarities in the points at which reactions take place. For instance, 30% of listeners reacted very dramatically only 4-5 seconds in to the Liverpool English audio stimuli (extreme early shifters). 1 second prior to this there is a clear instance of h-dropping in the audio stimulus. As this pilot experiment was conducted on listeners from Scotland (where h-dropping is not typically found), it is plausible that this stark change in the data represents a reaction not only to Liverpool English in general but to h-dropping in particular. In other words, we suggest that these shifts can be correlated with the occurrence of particular linguistic features and so used as a way of uncovering which linguistic features are salient for different listeners in a given variety. To finish, we highlight some limitations with the design of this pilot study and suggest a number of ways in which the technology could be developed in future work.


LVLT workshop on using RBrul (Lynn Clark)

Date & Time: 16 December 2010, 9.30 - 11.30am
Venue: County South C89 (Meeting Room 7)
Who can attend: Internal

As part of the LVLT research group, Lynn Clark is running a workshop about using RBrul. Rbrul (Johnson 2009) runs on the R environment and allows the user to model variable data using logistic regression (including random effects and continuous predictors) with minimal previous knowledge of R, a command-based open-source statistical software. Using real language data, this workshop will take the user through an analysis of variation with Rbrul. By the end of the workshop, participants will be familiar with Rbrul and be able to employ it with their own variable data.

Click here to download a copy of the handout for the workshop

Click here to access the department store data (ds)

Click here to access the t-to-r data (t-to-r_archive_archiveliv_rbrulworkshop)

Please bring your laptop to the workshop

You must have pre-registered for this workshop by getting in touch with Lynn Clark to confirm your place.


LVLT meeting: Kevin Watson & Lynn Clark '/t/s and /r/s and t-to-r in Northern British English'

Date & Time: 28 October 2010, 12.00-1.00
Venue: County South C89 (Meeting Room 7)
Who can attend: Anyone

Meeting of the Language Variation and Linguistic Theory Research Group

Kevin Watson & Lynn Clark will give a talk called

/t/s and /r/s and t-to-r in Northern British English

County South C89 (Meeting Room 7)

All welcome