Research report Eical evidence for serial sentence processing: a q comparison between non-preferred and ungrammatical continuations

Event-related potentials (ERPs) were used to address a much debated issue in sentence processing research, namely whether one or multiple interpretations are pursued in case of syntactic ambiguities. For example in the syntactically ambiguous fragment ''The man is painting the house and the garage . . . '', 'and' either connects 'the house' and 'the garage', or conjoins two clauses (e.g., ''The man is painting the house and the garage is already finished''). According to serial models, only one syntactic interpretation (the simplest) is pursued first (the first interpretation in the example). If this interpretation is incompatible with subsequently incoming words, the syntactic analysis of the preceding sentence fragment is revised. In contrast, parallel models propose that multiple interpretations of the ambiguity are pursued simultaneously. The two models were tested by comparing ERPs to words that were either ungrammatical, or grammatical but non-preferred continuations of the preceding sentence fragment. In a serial model, these two are not distinguished until after initial revision; in a parallel model, a distinction can be made at an early stage. The results supported a serial model: both with an acceptability judgment and a passive reading task a left lateralized negativity was found for ungrammatical and non-preferred continuations relative to unambiguous grammatical continuations, which indicates that ungrammatical and non-preferred continuations were initially processed in the same way. However, in later time intervals, the ERPs to the ungrammatical continuation showed a posterior positivity (P600), whereas the ERPs to the non-preferred continuation had a more anterior focus, which indicates that they were processed differently.