16. 4. / 17:30 /
č. 300 |
1. Distributional Typology: a probabilistic approach to linguistic universals and areal diffusion
Over the past two decades, linguistic typology has been moving increasingly away from its original goal of classifying languages into ideal types that would be constrained by categorical universals. What has been emerging as a new paradigm instead starts from the distribution of structures in the world, asking “what’s where why?” I present here a concrete approach to this question, called ‘Distributional Typology’. The approach starts from causal theories on the forces that affect language change, from processing preferences to the historical contingencies of language contact. The predictions of these theories can then be tested against fine-grained matrices of cross-linguistic diversity, using statistical methods for estimating diachronic trends from synchronic distributions. |
17. 4. / 12.30 /
č. 18 |
2. Government vs. agreement in typological perspective
Case government and verb agreement operate along fundamentally similar mechanisms: both rely on selecting and separating specific argument sets, e.g. subject vs. object sets, and both share the function of signalling dependency relations (Lehmann 1988). Also, both mechanisms are open to a similar range of cross-linguistic variation, e.g. both can be sensitive to referential properties or to co-argument properties. Given all these similarities, one would expect that the worldwide distributions of government and agreement are statistically interlocked. This expectation is not empirically supported. Instead, the distribution of both is chiefly driven by deep-time areal diffusion processes, while the distribution of case is in addition affected by word order conditions (as already proposed by Hawkins). These findings suggest that government and agreement are more independent of each other than is traditionally assumed. Further support for this conclusion form the fact the histories of the two phenomena tend to be subject to radically different factors. |