Recent research has shown that speech perception can easily be influenced by the overall accent of the speaker. This paper investigates whether such accentinduced bias also occurs in speech transcription by professional and linguistically trained coders and to what extent such a bias may affect linguistic analyses. The paper compares the transcriptions of the Bären vowel in Standard German with the acoustic values of these vowels, as well as sociolinguistic analyses based on both of these. The results of the two analyses turn out to be considerably different. Further examination shows that the coders only partly relied on the acoustic values. The residual does not consist of random errors, but correlates with the degree of accentedness of the speakers. It is concluded that 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 were quite different from the acoustic reality.