Abstract
The classification of unwritten languages, evidence from archaeology, and parallel research in population genetics have revived the interest in genealogical linguistics and expanded its methods. The development can only be welcomed, provided, of course, the potential of each method is properly recognized. The use of the lexical-resemblance criterion for uncovering patterns of kinship among languages provides a possibility for exploring areas which the comparative method could not reach. But patterns of kinship are one thing, ancestral forms another. This paper will argue that, without recognizing the principle of a developmental process in linguistics and without having access to the corresponding empirical data, reconstruction efforts are doomed to produce insufficient and misleading results.