In developing Faroese text-to-speech synthesis, the choice of phonological norms has proved problematic. In particular, the literature on Faroese phonology offered contradictory accounts of the realization of word-medial postvocalic stops. Spontaneous speech data from 4 speakers of the Tórshavn dialect were recorded, and their stop production was analyzed. It emerged that there were detailed differences between the 4 speakers in the stop production patterns. Since there seems to be no single "correct" phonological representation of these word-medial stops in the Tórshavn dialect of Faroese, our approach has been eclectic, in that we have chosen those phonological norms for the synthesis which we have judged optimal for the user.