Some language universals are historical accidents

In this short paper, I elaborate on previous work by Givón (1971) and Aristar (1991)
to argue that a substantial part of the well-known word-order correlations is best
explained by grammaticalisation processes. Functional-adaptive accounts in terms
of processing or learning constraints are currently weakly substantiated, and they
suffer from the fact that they do not adequately control for language-internal inher-
itance patterns. More generally, historical relatedness between different types of
phrases constitutes an important confound in typological research, one that needs
to be taken seriously before word-order correlations are motivated by anything
other than the diachronic patterns that link the word order pairs in question.