In experiment 1, two different structures of German subordinate clauses were tested allowing for the variation of the two factors 'Prosody' and 'Syntax' at two positions, namely penultimate vs. final position of a sentence. The prosodic manipulation of the penultimate word in syntactically correct sentences evoked an N400 corresponding in time with the unexpected terminal word (occurring after the speech stream already had been closed prosodically). We interpret this effect as enhanced lexical integration costs resulting from the preceding prosodic closure. The comparison with syntactically incorrect sentences (created by an illegal word order before the critical word) revealed this "prosody driven N400" to be dependent on the sentence's correctness. When the prosodic manipulation was realized at the sentence final word (prosodically marked as if the sentence would continue, although no further element occurred), we found a right-anterior negativity (RAN), regardless of whether the sentence was syntactically correct or not. This suggests that the RAN might be considered as a pure prosodic effect caused by unconfirmed expectations. A late positivity (P600) was observed for syntactically incorrect sentences (realized though the omission of the sentence final verb), as well as for the pure prosodic manipulation at the sentence final position. The combined data indicate a clear prosody and syntax interaction modulating the amplitude of the P600.
In experiment 2, a conjunct clause was added to all constructions used in the first experiment to test whether the observed effects would occur not only at the end of a sentence, but also at the end of a clause. We similarly observed the RAN and the interactive P600 pattern. However, the "prosody driven N400" was missing at the end of the clause. This might indicate that only sentence final prosody immediately influences lexical integration processes, but not clause phrase final prosody.