Dunn et al. (2011) apply computational statistical techniques to crosslinguistic grammatical data taken from WALS and some individual grammatical descriptions, comparing the outcomes to those obtained from more traditional lexical classifications; they claim that the results of their analysis show that (i) “contrary to the generative account of parameter setting [...] the evolution of only a few word order features of languages are strongly correlated” (2011: 9) and (ii) “contrary to the Greenbergian generalizations, [...] most observed functional dependencies between traits are lineage-specific rather than universal tendencies” (2011: 9). They conclude that their “findings support the view that [. . .] cultural evolution is the primary factor that determines linguistic structure” (2011: 9)
However, we show that their experiment and the associated claims are flawed in three respects and therefore fail to satisfactorily approximate their goal.
