What to Do when You are Unhappy with Language Areas but You do not Want to Quit

Abstract

AbstractThe article deals with the concept of "linguistic area" and its problems.Section 1. deals with the history of the concept, as originally formulated by Trubetzkoy in 1923, showing how logical problems were present since the very beginning.In Section 2. a few possible preconditions for a scientifically sound definition of "area" are proposed and discussed, and their application is exemplified in Section 3. with the case of the languages of Northern Eritrea.It is suggested that the concept of "area" is in principle more interesting in linguistics than in social sciences due to the availability in linguistics of two other unrelated and powerful tools: genetic classification and typology. It is in the light of these other tools that language areas will have to be judged – as what lies beyond the range of both genetic and typological linguistics. It is suggested that ideally, in order for a linguistic area to be "proven": its members will have to be as genetically diverse as possible; and it will not be possible to account for the area-defining features on the basis of typological tendencies and regularities. Moreover, language areas described in terms of similar traits should not be overlapping, and attention will have to be paid in keeping language-external facts (such as historical contact and cultural similarities) at bay and not let them guide our search for language areas.It is further argued that, since one cannot expect the members of the area to be maximally different genetically (i.e., totally unrelated), nor the outcome of contact to be maximally irregular typologically (i.e, typologically impossible), real-world areas cannot plausibly meet these strong requirements in toto. The most one can expect is a gradual implementation of these conditions: language areas will have to be defined negatively as a space-based clustering of features which can not be accounted for genetically or typologically, and membership in them will have to be recognized as continuous rather than discrete.