|Dispersion Theory formalizes the structuralist notion of systemic contrast within a constraint-based phonological framework (DT; FLEMMING, 1995, 2002; NI CHIOSAIN & PADGETT, 2001; PADGETT, 2003a,b,c). Bradley & Delforge (in press) propose a DT analysis of sibilant voicing patterns throughout the history of Spanish, from the loss of medieval voiced sibilants through their reemergence in several contemporary dialects. Phonetic effects in sibilant voicing are adequately explained by a distinction between obstruents that are phonologically specified for [voice] and targetless, neutral obstruents that undergo gradient voicing by phonetic interpolation (ERNESTUS, 2003, STERIADE, 1997, 1999). It is possible to incorporate a non-contrastive phonetic category because in DT, systemic constraints govern the well-formedness of phonological contrasts. The present study focuses in greater detail on sibilant voicing in the Spanish of highland Ecuador and takes into account some additional observations by Robinson (1979) that have not been addressed in the DT approach thus far. First, regional variation in the voicing of prefix-final /s/ is shown to depend on whether prefixes are incorporated in the lexical phonology, where devoicing is favored, or in the postlexical phonology, where voicing serves to distinguish morpheme-final sibilants from morpheme-initial ones in phrasal intervocalic contexts. Second, native speaker intuitions regarding the lack of resyllabification of morpheme-final prevocalic [z] are actually predicted by a theory which acknowledges the role of systemic contrast in the postlexical phonology.