Strict criteria on phonological categoryhood coupled with strict privativity of representation inevitably lead to a conclusion that sonorants must not contain a prime responsible for voicing. Assuming that this prime is also not supplied to sonorants in the course of phonological derivation, this class of segments, contrary to observed patterns, should be inactive with respect to voicing phenomena. Presonorant sandhi voicing in Cracow-Poznań Polish is used to show how such apparent patterns can be dealt with without compromising the above theoretical assumptions. This however has consequences which bear on almost every aspect of laryngeal phonology. Some of them include: arbitrariness of the relation between phonology and phonetics, emergent nature of laryngeal categories, minimization of the role of phonological computation, re-evaluation of typical analytical criteria for deciding on phonological representation of laryngeal distinctions, which are used in phonological practice, as well as a possibility that the prime [voice], or its theoretical counterpart in various models, is not present in some ‘voice’ languages.