This paper reports on the result of an experiment that was designed to test the different predictions of two phonological analyses of German fricative voicing, one that actively bans voicing in coda position, and one that preferentially preserves voicing in presonorant position. In the experiment, native speakers pronounced words such as gruslig 'spooky' and fasrig 'fibrous', crucial cases where an underlying voiced fricative occurs before a sonorant consonant, and where [zl] and [zr] are not possible onsets. The authors argue that only the positional faithfulness analysis is consistent with their experimental findings, which include considerable variation; coda devoicing cannot account for the full range of data attested.