Evaluating the role of self-description in demarcating accents

This study tests if speakers’ self-description of accent, using both pre-selected labels and free-classification, is predictive of their linguistic productions. Based on the production of diphthongs in a sample of 186 speakers from Southeast England, we identify three clusters of speakers, representing Standard Southern British English, Estuary English and Multicultural London English. We explore participants’ self-description of accent guided by pre-selected accent labels, such as 'London' or 'Queen's English', and their unguided, self-description of accent in their own terms. The former corresponded to the clusters of vowel features in 62% of cases. Unguided descriptions revealed some patterns, but typically speakers favoured geographical terms such as 'British' or 'London'. We conclude that self-description is a relatively poor predictor of accent, especially in a complex sociolinguistic context.