Aarhus University Seal

Marjoleine Sloos: Beware, Coder Bias!

Recent research showed that speech perception is easily influenced by other factors like social indexation (age, gender, level of education, profession) and also the overall accent of the speaker. Such indexation serves in-group/out-group distinction and probably also facilitates speech perception.

Marjoleine Sloos

Info about event

Time

Tuesday 1 April 2014,  at 11:00 - 12:30

Location

Aarhus University-DK, Nobelparken, building 1483-3, IMC meeting room (312)

Organizer

Interacting Minds Centre

 However, it has been shown that misperception or misinformation about the social class of the speaker leads to bias in perception and subsequent judgements about the speech. Linguistics and its applied fields crucially rely on the perception of the linguistic researcher. If linguists are as susceptible as naïve listeners to accent-induced bias, the linguistics analysis is potentially affected. My research focuses on accent-induced bias in speech transcription by professional and linguistically trained coders and the central question is to which extent such a bias may affect the linguistic analysis. 

The case study is a sociolinguistic analysis of transcriptions of the Bären vowel in Standard German compared to the same sociolinguistic analysis of the acoustic values of these vowels. The results of the two analyses turn out to be considerably different. Further examination into the correlation between the codings on the one hand and the acoustic values and degree of accent of the speakers on the other hand shows that the coders only partly relied on the acoustic values. The codings also correlated with the speakers' degree of accent. This accent-induced coder bias led the coders to transcribe the codings according to their expectations about the pronunciation in the local dialect—expectations that turned out to be quite different from the acoustic properties. 

We will discuss these findings and current follow-up research on this topic.