Maffongelli, L. 1 , Droege, A. 1 , Schlesewsky, M. 2 & Bornkessel-Schlesewsky, I. 1
1 University of Marburg
2 Johannes Gutenberg-University Mainz
A prominent debate in the recent psycholinguistic literature concerns the claim that the language processing system constructs the most plausible combination of the words it encounters independently of grammatical constraints ("semantic attraction"). This proposal was based on the finding of P600 rather than N400 effects for "reversal anomalies" in English/Dutch (e.g. "The hearty meals were devouring ..."; Kim & Osterhout, 2005). The present ERP study used a novel manipulation in Italian to examine the predictions of semantic attraction. Italian has richer agreement and a more flexible word order than English and subject-verb agreement is the most salient cue to Agent-choice (MacWhinney et al., 1984). We crossed the factors animacy and agreement (literal translations; number counterbalanced across full materials): [a] the artist(SG) paints(SG) ...; [b] the picture(SG) paints(SG) ...; [c] the artist(SG) paint(PL) ...; [d] the picture(SG) paint(PL) ...). At the position of the verb, both non-agreeing condition [c]/[d] engendered an identical N400-P600 response in comparison to control condition [a] (see Haupt et al., 2008, for N400-P600 correlates of word order reanalysis). The agreeing condition with an animacy mismatch [b] showed only a P600. These findings are incompatible with a semantic attraction-based account: if NP1 and the verb were first combined in the most plausible way, the agreement mismatch should be processed differently depending on whether semantic attraction yields an Agent [c] or Theme [d] reading for NP1 (only subjects agree with the verb). This was not the case. We thus conclude that the P600 effect for sentences such as [b] should not be viewed as evidence for a semantic attraction-based conflict, but rather as reflecting ill-formedness categorisation (Bornkessel-Schlesewsky et al., in press), with the absence of an N400 likely due to lexical priming (Lau et al., 2008).