Distributional universals and the rate of type shifts: towards a dynamic approach to “probability sampling”

However, this can be done by means of a relatively simple and feasible technique that may significantly increase the linguistic value of statistical typological observations. The basic idea is very simple: what we need is a representative sample of pairs of closely related languages (with a roughly similar time of divergence). This gives us an additional statistic, the frequency h of divergent pairs – an estimate of divergence rate, which can significantly restrict the set of possible diachronic scenarios (hence, of plausible linguistic interpretations of typological tendencies). Methodologically, the proposed technique remains within the realm of (synchronic) typology, in the sense that only present feature values are sampled; no reconstructions are needed.