Syntactic templates and linking mechanisms: A new approach to grammatical function asymmetries

The comprehension of argument order variations has long posed an interesting challenge for psycholinguistic research. Typically, the processing of varying orders has been described in terms of reconstruction to a base order, i.e. in terms of a positional association between a particular argument and a specific (underlying) position in the phrase structure tree, which guarantees the correct interpretation of that argument. From this perspective, the well-known increases in processing cost for object-initial sentences are engendered by the fact that the object cannot be interpreted in the position in which it is encountered. These assumptions essentially rely on a classical generative perspective on argument interpretation (Chomsky, 1981, 1995, 2000), which posits that interpretation (i.e. thematic role assignment) is accomplished in a base position, while surface argument order is derived via movement. However, they hold even for grammatical theories without movement, e.g. HPSG (Pollard & Sag, 1994), in which interpretation of a fronted object also presupposes an association with a base position. Similar assumptions are also implicit in processing models such as that of Gibson (1998), in which the inverted order results in enhanced prediction and integration cost.