The liaison consonant [z] in French noun phrases has traditionally been assumed to function as a plural marker. The realization of “plural [z]” in N(oun)-A(djective)-combinations is becoming, however, very rare in naturalistic data – except for contexts which allow a proper-name reading. On the one hand, one might think that we are dealing with a recent phenomenon, the beginning of a potential linguistic change in French in the sense of exaptation, reuse of former morphophonological material such as plural markers to signal proper-namehood in the sense of ‘frozen morphology’. If this turns out correct, we expect the productivity of the new synchronic function to increase: New NA-combinations which function as proper names should be realized systematically with liaison, and proper name-marking via liaison should also become possible with other liaison consonants. On the other hand, we may be dealing with a (completed) diachronic process, in that only those NA-combinations which allowed liaison at the relevant point in time may have a liaison consonant in their univerbalized form. That is, new NA-combinations, even though they are used as proper names, do not display a liaison consonant, because liaison is no longer possible. The purpose of this paper was to investigate, based on empirical studies, whether liaison productively marks NA-combinations which function as proper names and distinguishes them from NA-combinations that count as common nouns, or whether we are dealing with a completed diachronic process. In view of the poor productivity observed, we argue that we are dealing with cases of univerbation.