Incremental interpretation and discourse complexity

We will present two self-paced reading studies that provide new evidence on the incrementality of semantic interpretation, in particular with regard to the notion of discourse complexity as introduced by Gibson's distance-based dependency locality theory (DLT; cf. Gibson, 1998, 2000). More specifically, we will focus on the contribution of referential processing on sentence complexity. The experiments compared the processing of simple definite DPs like der Lehrer ('the teacher') and complex DPs containing a possessive NP like Peters Lehrer ('Peter's teacher'). While simple DPs introduce only a single discourse referent, complex DPs introduce two discourse referents and some relation between them. This additional processing effort is reflected by increased reading times. Introduction In prior work (Bader and Häussler 2005) we investigated simple and complex DPs in experiments focussing on the processing of number ambiguous nouns in German. A number ambiguous noun like Lehrer is usually disambiguated by a determiner (der Lehrer 'the teacher' vs. die Lehrer 'the teachers'), but remains ambiguous when specified by a possessive proper name (Peters Lehrer 'Peter's teacher(s)'). We found that number ambiguous nouns cause garden-path effects on the clause-final disambiguating verb. Reading times on the verb correlate substantially with number preferences determined by a corpus analysis. The more often a noun occurs in the singular, the more often it is analyzed as a singular noun during first-pass parsing, leading to difficulties when encountering a verb specified for plural. Furthermore, reading times at the noun were longer in ambiguous DPs than in unambiguous DPs. As an alternative or in addition to ambiguity, the increased reading times might be attributed to complexity. While unambiguous DPs introduce only a single discourse referent, ambiguous (= possessive) DPs introduce two discourse referents and some relation between them. This might cause additional processing effort which is reflected in reading times. Since complexity and ambiguity were confounded in these experiments we conducted two further experiments separating both factors by using two types of unambiguous material. Both experiments used a word-by-word non-cumulative self-paced reading procedure. ITRW on Experimental Linguistics Athens, Greece August 28-30, 2006 ISCA Archive http://www.isca-speech.org/archive