The acquisition of sociolinguistic evaluations among Polish-born adolescents learning English: evidence from perception

Abstract

In order to achieve full native-like competence in a second language, speakers must also acquire sociolinguistic awareness in that language. This paper reports the results of a study investigating the acquisition of sociolinguistic awareness among immigrant Polish adolescents learning English in the UK. This paper asks whether Polish-born adolescents living in the UK can identify different varieties of British English as well as their native-speaker peer group can and whether they share similar evaluations of these varieties of English as their native-speaker peer group. The results of a variety recognition survey suggest that Polish-born adolescents now living in the UK are not yet able to identify different varieties of English. However, the vast majority of evaluations carried out by Polish- and UK-born adolescents were not statistically different. Furthermore, we see clear evidence of the acquisition of the muted evaluations typically associated with the two varieties of English that are most positively and negatively evaluated among the UK-born adolescents: received pronunciation and Birmingham English. We suggest that our study provides a snapshot of the initial stages of the acquisition of attitudes towards variation in a second language.

Keywords:

• immigration
• acquisition
• sociolinguistic awareness
• Polish

Acknowledgements

This research was funded by the Economic and Social Research Council (RES-000-22-3244) and conducted in collaboration with Miriam Meyerhoff. We are extremely grateful to Robert McKenzie and Remco Knooihuizen for providing extensive comments on an earlier draft of this paper and to the audience of the 40th Poznan Linguistics Meeting and the English Language Research Group at Edinburgh University where parts of this work were presented. Thanks also go to Satori Soden who collected the data in London. The high schools we worked in remain anonymous, but we owe a debt of great magnitude to the staff and students in Edinburgh and London who worked with us on this project. In particular, our thanks go to the migrant teenagers, for some of whom the interview and reading tasks were a major effort. We admire and respect the individuals in these schools for the work they are doing. We alone are responsible for any failings in this paper.

Notes

1. The Workers Registration Scheme can be used as a measure of the number of migrant citizens coming to work in the UK, but it provides no indication of their length of stay, and because the Scheme excludes the self-employed, it is likely that these figures hugely underestimate the actual number of migrant workers in the UK.

2. Following Poplack and Tagliamonte (2001, p. 93), our approach assumes that the constraints operating on linguistic variation (uncovered in an analysis of variation through multiple regression) represent the variable grammar of the speech community, and therefore, differences in the constraints and ranking of constraints among UK-born adolescents and Polish-born adolescents living in the UK can be interpreted as ‘diagnostics of fundamentally different underlying grammars’ (Meyerhoff, 2009, p. 303).

3. The VGT is a variant of the MGT which was originally developed by Lambert et al. (1960). The main difference between these two methods of data collection is that a number of different speakers (with different linguistic characteristics) provide the stimulus for a VGT, whereas with an MGT, the same speaker produces a number of different linguistic ‘guises’.

4. Redinger and Llamas (2009) discuss in detail the advantages of employing magnitude estimation techniques in attitude research.

5. Thanks to one of the reviewers of this paper for suggesting this point.

6. Eigenvalues are a measure of the amount of variance accounted for by each of the components extracted in PCA. The Keiser criterion is the most commonly used method for establishing which components are relevant in PCA. The rule of thumb is that factors should be extracted with eigenvalues greater than or equal to 1 (i.e. 10% of the variance). Components with an eigenvalue smaller than this are not contributing more to the model than a single variable and therefore are meaningless.

7. One reviewer questioned the use of ANOVA rather than a simple t-test. While it would have been possible to analyse these data using multiple t-tests, this increases the probability of committing at least one Type 1 error. The ANOVA procedure performs fewer hypothesis tests and therefore reduces the likelihood of experiment-wise error.