Some people believe the assumption that individuals’ speech is indicative of their place of origin. This notion motivated border agencies of several developed nations to use language analysis in the asylum process to distinguish fake asylum seekers from genuine ones. However, several linguists raised concerns over the lack of reliability in the language analysis. This concern motivated a debate over the required qualifications to conduct the analysis.
This study aims to provide empirically-based research contribution to the debate over the supervised native speaker approach versus the expert linguist-only approach in language analysis for the determination of speakers’ origins.
The study comprises two phases: (i) provision of segmental description of four Nigerian English accents (Hausa, Igbo, Kanuri & Yoruba) (ii) accent classification experiments to assess the relative performance of four methods in classifying four Nigerian English accents. A corpus of the four Nigerian English accents was collected from 60 respondents, each accent represented by 15 respondents. The corpus was analysed impressionistically with some acoustic corroboration. The accent classification task involved 118 participants drawn from three human groups—80 Nigerian non-linguists (each L1 group represented by 20 respondents), 25 Nigerian linguists (6 Hausa, 9 Igbo, 5 Kanuri & 5 Yoruba), 13 UK-based phoneticians and an automatic accent recognition system, Y-ACCDIST.
The findings reveal that each of the four methods shows potential in accent recognition. However, overall results indicate that native speakers, regardless of linguistic background, were significantly more accurate in identifying speakers of their accent groups. The findings also reveal that the UK-based phoneticians and Y-ACCDIST were the most accurate in identifying Yoruba-English. Given that Yoruba-English speakers have stereotypes such as [h]-elision and [h]-epenthesis in their speech, it can be speculated that language analysis conducted by non-native speaker linguists can be more reliable if a language variety in question has some stereotypes.
