Item

Abstract

A central goal of linguistics is to explain the constraints on linguistic diversity. The landscape of linguistic diversity is vast, its boundaries invisible from a distance and its details obscure from up close. So-called linguistic universals stand as beacons as we try to map this unruly geography. Word order universals ― as described by Greenberg and elaborated by Dryer ― are the paradigm case of a clear structual constraint on the variation of language. In some schools of linguistics the word order universals have been taken as direct evidence of innate parameters, abstract cognitive structures shaping the ontogenetic development and the evolution of language. In other schools of linguistics, the dependencies between aspects of word order are understood as statistical tendencies reflecting universal systems biases. In this paper I argue that the light cast by these beacons might be deceiving us, and report a reanalysis of well-studied word order universals using phylogenetic methods. These tools have been pioneered in biology and evolutionary anthropology to solve "Galton's problem", the statistical interdependence of observations coming from genealogically related taxa (be they cultures, languages, species). A statistical, model based approach to variation in word order types in four language families shows that functional dependencies between word order traits vary greatly by lineage. Instead of being tightly constrained by cognitive factors specific to language, the patterns in word order diversity are the product of cultural evolution constrained by the current state of the system, so that future states are canalized within attraction basins specific to the particular linguistic lineage.