Typologies in equilibrium
Donnerstag, 15. März 2018, 16:15 Uhr bis 18:00 Uhr
In a landmark paper, Maslova (2000) argued that the
synchronous frequencies of a typological variable do not reveal
distributional universals. As there is no guarantee that the underlying
dynamic process has reached equilibrium, observed frequencies may
reflect properties of ancestor languages rather than functional
tendencies. As a remedy, Maslova proposes to estimate the transition
rates between types from diachronic data and to compute the equilibrium
distribution analytically instead.
Probably due to the sparsity of diachronic typological evidence, this
program has not been realized so far. Techniques from the *phylogenetic
comparative method* (cf. Nunn, 2011) in computational biology, however,
paired with the newly available electronic typological data sources,
afford an alternative way to realized Maslova's goal.
Once a typological variable and a collection of languages has been
fixed, the workflow is as follows:
1. Infer a (distribution of) phylogeny(ies) from lexical data.
2. Estimate the transition matrix between the values of the variable.
3. Calculate the equilibrium distribution of this Markov process.
I conducted two case studies: (A) The major word order types, using
WALS, and (B) case marking alignment patterns, using the data from
Bickel et al., 2014. In both studies, I used the ASJP database (Wichmann
et al., 2016) for the first step.
In (B), we found substantial differences between equilibrium and
observed values. The corresponding results for (A), however, suggest
that equilibrium has been reached
Raum: SOD 1 105