The Application of Universal Hierarchies during Sentence Processing: Evidence for Incremental, Interactive Thematic Processing

Ina Bornkessel¹, Matthias Schlesewsky², and Angela D. Friederici¹
bornke@cns.mpg.de, schlesel@rz.uni-potsdam.de, angelafr@cns.mpg.de
1 Max-Planck-Institute of Cognitive Neuroscience, Leipzig
2 Department of Linguistics, University of Potsdam

In this paper, we provide experimental evidence in support of the hypothesis that thematic information is processed independently of syntactic information but interacts with the latter in a post-initial processing stage. We do so by examining a novel type of thematic ambiguity, namely an ambiguity not with respect to the individual thematic role label borne by an argument, but rather with respect to the hierarchical thematic ordering between the arguments of a sentence. The relevant contrast is illustrated in (1).

(1)

a.

... dass der Gaertner dem Jaeger vertraut.
that the-NOM gardener the-DAT hunter trusts

b.

... dass der Gaertner dem Jaeger auffaellt.
that the-NOM gardener the-DAT hunter notices

Whereas the subject-experiencer verb in (1a) calls for a thematic structure in which the nominative argument (the Experiencer) is higher than the dative argument (the Theme), relations are reversed in the structure projected by the object-experiencer verb in (1b) (i.e. here, the dative-marked Experiencer is thematically higher than the nominative-marked Theme). Under the assumption that hierarchical thematic relations are assigned incrementally, a reanalysis of this ordering should be necessary when an object-experiencer verb (1b) is encountered clause-finally.

In a first experiment using event-related brain potentials (ERPs) we observed a parietal positivity between 300 and 600 ms post-onset of the verb in structures such as (1b) in comparison to structures such as (1a). This effect was independent of the word order of the clause, i.e. it was entirely dependent on the thematic manipulation at the position of the verb, and must therefore be interpreted as a thematic reanalysis effect (i.e. a reanalysis effect that is independent of syntactic structure).

Having shown that thematic reanalysis exists independently of syntactic reanalysis, we conducted a second (speeded grammaticality judgement) experiment, in which we directly compared thematic reanalysis and structural reanalysis (in the form of subject-object ambiguities disambiguated towards an OS order). This experiment revealed that the garden path effect for ambiguous object-initial sentences is no longer visible when the disambiguation is effected by an object-experiencer verb, thus showing that thematic information interacts with syntactic information (number agreement) during disambiguation and the subsequent reassignment of grammatical functions.

We discuss our results in the context of Friederici's (1999) neurocognitive model of language comprehension and situate thematic processing and its interaction with syntactic processes in the second phase of comprehension, i.e. following initial phrase-structure assignment and preceding processes of syntactic reanalysis and repair.