This paper reassesses the genetic classification customarily associated with what we refer to as “unspoken languages,” that is, language forms that are, or were at some earlier stage, fulfilling exclusively nonspoken functions (e.g. Medieval and Modern Hebrew, Medieval Latin). We argue that if claims of genetic affiliation are to have any empirical linguistic content, and not simply sociopolitical significance, then genetic descendance from an earlier linguistic form must require the existence of uninterrupted normal transmission. Since the conditions of transmission attested in the case of “unspoken languages” arguably fail to qualify as such, we conclude that these cannot possibly be genetic descendants of the older spoken languages whose names they usually bear. We present our own hypothesis as to the nature and genesis of “unspoken languages,” with particular emphasis on the process of relexification of the user's native language with lexical items of the “ancestor” language. As a revealing test case, we discuss Modern Hebrew, claiming that its apparent affinity to Biblical and Mishnaic Hebrew is indeed only a reflection of relexification, not of descendance. Finally, we point out some general methodological implications for genetic and historical linguistics.