Evaluating Competition-based Models of Word Order

The ordering of constituents in semi-free word order languages has attracted considerable attention in theoretical linguistics. Three types of models have been proposed to explain word order preferences, based on (a) weighted constraints, (b) Optimality Theory (c) syntactic weight. All three models use grammatical competition to explain the interaction of word order constraints. They rely on intuitive judgments or corpus studies, but have not been evaluated against experimental data. This is the purpose of the present paper. We report the results of a magnitude estimation experiment investigating word order in German, focusing on the interaction of verb position, case marking, pronominalization, and information structure. The experimental data are compatible with models (a) and (b), indicating that relativized (ranked or weighted) constraints are essential in explaining word order preferences. Model (c), on the other hand, is not compatible with the data.