This article reports on an acoustic study of voicing in obstruents followed by a sonorant across a word boundary in two dialects of Dutch: East- and West-Flemish. In both varieties only gradient phonetic voicing was typically found in word-nal stops when a sonorant followed in the next word. In addition, West-Flemish showed optional categorical voicing in word-nal pre-sonorant fricatives. The voicing of fricatives is argued to be phonological, as it extends beyond the scope of automated coarticulation, and as the data pattern to form a distinct phonetic voicing target. However, the phonetic results do not support the hy-pothesis that West-Flemish sonorants are laryngeally specied, and thus able to spread voicing to neighbouring fricatives. Instead, fricative voicing is proposed to be an optional positional realisation in West-Flemish. Although the process cannot be directly motivated by reference to the phonological specications of the segments surrounding its target, it makes sense in terms of perceptual fac-tors leading to diachronic reanalysis. The West-Flemish positional variation may arise when partially voiced fricatives are perceived and subsequently reanalysed as categorically voiced by listeners, as proposed by Jansen (2004). It is further argued that fricatives are more likely than stops to be reinterpreted as voiced, as additional acoustic cues prevent voiced percepts in passively voiced stops.