Influence of prosodic information on the processing of split particles: ERP evidence from spoken German.

Spoken language comprehension involves the use of different sources of linguistic information such as prosodic, syntactic, lexical, and semantic information. The question, however, of ''when'' and ''how'' these sources of information are exploited by the language processing system still remains unanswered. In the present study, we used event-related brain potentials (ERPs) to investigate the interaction between prosodic, syntactic, and lexical information during the processing of spoken German sentences. The sentence structure was manipulated by positioning a split particle at the end of the sentences after the occurrence of inflected verb whose lexical entry does not contain a split particle (e.g., *Sie alarmierte den Detektiv an [*She alerted at the detective]) [According to linguistic convention, incorrect sentences are marked by an asterisk.]. The prosodic contour of the verb stems was manipulated such that it marked either the presence of a split particle at a later position in the sentence or not. Participants performed an off-line probe-detection task. ERP data indicate that prosodic information of German-inflected verb stems is consulted on-line by the language processing system (''parser'') in order to ''predict'' the presence of a split particle at a later position in the sentence. An N400 effect was observed for the processing of split particles following verb stems which do not take a particle. However, this effect was only observed when the prosody of the verb stem did signal the presence of a split particle. We argue that the N400 component reflects the high costs associated with the lexical search that the language processing system has to perform when confronted with nonexisting words such as these resulting from the combination of the split particle and the verb stem in the present study. Furthermore, as a general reflection of prosodic processes, a Closure Positive Shift (CPS) was found at intonational phrase boundaries. In sum, the present findings provide strong evidence that prosodic information is a good ''predictor'' of upcoming information during the auditory processing of German sentences.