The costs of freedom: an ERP – study of non-canonical sentences Mike Matzke a , Heinke Mai b , Wido Nager c , Jascha Ru ¨sseler b , Thomas Mu ¨nte b, * a Department of Neurology II, University of Magdeburg, Magdeburg, Germany b Department of Neuropsychology, University of Magdeburg, Universitatsplatz 2, Geba ¨ude 24, 39106 Magdeburg, Germany c Department of Neurology, Medical School Hannover, Hannover, Germany Accepted 5 March 2002 Abstract Objectives: The present investigation explored the electrophysiological correlates of working memory during sentence comprehension. Methods: Event-related brain potentials (ERP) were recorded from 29 channels in 22 subjects, while they read German sentences having subject-first (canonical) or object-first (non-canonical) word orders. Results: Three different ERP effects were observed: a negativity (maximum at Fc5) differentiating unambiguous object-first and subject- first sentences, interpreted as reflecting the demands of the object-first sentences on working memory; a second negativity (maximum at F7) to the subject noun-phrase in object-first sentences, interpreted as indicating retrieval of verbal material. Finally, a parietal positivity was found for ambiguous sentences that turned out to have a non-canonical word order, which was interpreted as indicating revision and reevaluation processes. Conclusions: The present data underscore the different roles of working memory in comprehension. q 2002 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Working memory; Language processing; Event-related potential; Word order; Non-canonical sentences 1. Introduction The past few years have seen enormous advances in the conceptualization of the role of working memory in language processing (Just and Carpenter, 1992; Just et al., 1996; Caplan and Waters, 1999). One of the factors that influences the demand on working memory is syntactic complexity of a sentence. In these sentences, it can be diffi- cult to compute the syntactic structure and to assign thematic roles (‘who did what to whom’). This assignment, which can be thought of as matching structural and lexical information, may require the availability of a sequence of words in verbal working memory. Empirical studies show that different languages draw on different sources of infor- mation for role assignment. MacWhinney et al. (1984), for example, have shown that English subjects rely primarily on word order for role assignment, while Italian subjects used agreement information as the most important cue, and German subjects apparently employed agreement and animacy information. Japanese uses yet another kind of information, the introduction of particles, to determine a constituent’s role. In the present investigation, we are interested in the use of working memory during sentence processing in German. Because of its relatively rich morphology, German allows deviations from the standard subject–verb–object (SVO) word order. This is exemplified by the following sentences: (1) Der begabte Sa ¨nger entdeckte den talentierten Gitar- risten The gifted singer (Masc. Nom.) discovered the talented guitar player (Masc. Acc.) . (2) Den begabten Sa ¨nger entdeckte der talentierte Gitarr- ist. ‘The gifted singer (Masc. Acc.) discovered the talented guitar player (Masc. Nom.) .’ meaning: The talented guitar player discovered the gifted singer. (3) Die begabte Sa ¨ngerin entdeckte den talentierten Gitar- risten. The gifted singer (Fem. Nom.?Acc.?) discovered the talented guitar player (Masc. Acc.) . (4) Die begabte Sa ¨ngerin entdeckte der talentierte Gitarr- ist. ‘The gifted singer (Fem. Nom.?Acc.?) discovered the talented Clinical Neurophysiology 113 (2002) 844–852 1388-2457/02/$ - see front matter q 2002 Elsevier Science Ireland Ltd. All rights reserved. PII: S1388-2457(02)00059-7 www.elsevier.com/locate/clinph CLINPH 2001576 * Corresponding author. Tel.: 149-391-6718-469; fax: 149-391-671- 1947. E-mail address: thomas.muente@medizin.uni-magdeburg.de (T. Mu ¨nte).