EEG-Untersuchungen zur Informationsstruktur und Verarbeitung koreferenter Nominalphrasen

The present study examined the processing of antecedent-anaphor-pairs in the context of different discourse structures. The main intention was to investigate the influence of context on the analysis of co-referent noun phrases during visual presentation. In three experiments a set of semantically related nouns, namely hyponyms and hyperonyms were presented to participants. The influence of context on the processing of focussed elements as well as the processing of anaphoric elements was examined. A further aspect was the examination of word order manipulations in the German middle field between non pronominal accusative and nominative arguments. Context is especially important with regard to the resolution of anaphoric relations, i.e. in establishing and maintaining reference throughout the discourse. Typically, the resolution of anaphoric expressions during on-line comprehension is established using personal pronouns. However, there are also other means of establishing anaphoric relations, for example by the use of hyperonyms or even identical words. In the present study, we examined the time course of the establishment of hyperonymic anaphoric relations by comparing hyperonymic anaphors (e. g.: “fish” to the antecedent “carp”) to repeated noun phrases (e. g.: “carp” – “carp”). The repetition or reactivation of an argument has been shown to lead to a reduction of the N400 component in terms of event-related brain potential (ERP) measures. Depending on the semantic relatedness and the distance between prime and target the reduction of the component varies and is largest for consecutive word repetition. The results of the present study indicate that the reduction of the N400 component is influenced by an additional factor, namely the strength of processing, which is apparently related to context structure. Results of experiment 1 showed, that semantic as well as spatial distance between antecedent and anaphor did not lead to the expected reduction of the N400 component. Instead the strength of both effects was identical. Only the latency for hyperonymic anaphors increased by approximately 70 ms. The delay might show the time necessary to reactivate the concept of the previously presented hyponym. The observed effects can be interpreted as an evidence that processing and discourse integration of different non pronominal anaphors rely on very similar mechanisms in the context of complex sentences. Apparently the semantic difference causes a temporal shift of the successful integration but not an increased processing demand as compared to the integration of identical anaphors. The structure of the context seems crucial for that. On changed discourse conditions, as tested in experiment 3, this effect was not present. For the examination of the processing of scrambled structures, subjects read target sentences with such manipulations in its embedded clause (e. g.: „…, dass der Junge den Fisch …“ = „… that the boy [nominative] the fish [accusative] …” vs. the scrambled counterpart: „…, dass den Fisch der Junge …“ = „… that the fish [accusative] the boy [nominative] …”). A specific event related potential, the so called scrambling negativity, can be usually expected for non pronominal accusative arguments in the initial position of the German middle field compared to initial nominative arguments. In the present study, the effect was not observed for all initial accusatives. This happened probably due to the word by word presentation of all target sentences leading to local ambiguities of some articles. In this initial position of the German middle field they could now also be interpreted as pronouns. A pronoun would be base-generated in this so called “Wackernagelposition”. This interpretation is supported by the fact, that the scrambling effect was observed on nouns following articles with no observable scrambling effect in experiment 3. This would mean that due to the ambiguity the effect was shifted to the element solving the ambiguity. Behavioural data of experiment 3 revealed another interesting finding. As expected participants judged sentences with an initial accusative determiner less acceptable than sentences with an initial nominative in the middle field of target sentences. However conditions with an indefinite accusative determiner were judged more acceptable than those with a definite accusative article. This might be caused by an unexpected exchange of thematic roles.