Comprehension of canonical and non-canonical structures within and across the verbal and nominal syntax domains in agrammatism Philip Rausch * , Frank Burchert, Ria De Bleser Potsdam University, Department of Neurolinguistics, Building 24, Karl-Liebknecht-Straße 24-25, 14476 Golm, Germany Introduction This study systematically examines German agrammatic speakers’ comprehension of thematic relationships established canonically and non-canonically within and across the verbal and nominal domains. The results only partially support predictions made by existing theories about the agrammatic comprehension deficit, such as the Trace Deletion Hypothesis (TDH; Grodzinsky, 2006). One major research topic in syntactic theory has been the characteriza- tion of the parallelisms and differences between the syntactic structure pro- jected by verbs and the one projected by nouns (‘CPs’/complementizer phrases and ‘DPs’/determiner phrases, respectively). Theoretical consider- ations suggest that the syntactic structure projected by nouns and verbs and the syntactic operations within each domain are parallel to a consid- erable degree (the CP/DP parallelism; Abney, 1987). The basic paradigm tested here goes back to Chomsky’s (1970) observation that certain syntac- tic operations usually found in sentences seem to take place inside nominal structures as well, such as the derivation of a kind of ‘passive’ construc- tion: (1) a. [ DP the enemy’s Agent destruction N of the city Patient ] b. [ DP the city’s Patient destruction N by the enemy Agent ] These examples illustrate that the agent/patient sequence can be canonical or non-canonical within the nominal DP-domain (as in verbal actives and passives) and Lindauer (1998) argues that the corresponding non-canonical constructions in German are movement-derived. Unlike verbal passives, however, they lack any passive morphology triggering the operation. By testing such constructions and their verbal counterparts, we assess predictions made by existing theories of the agrammatic compre- hension deficit, such as the TDH (Grodzinsky, 2006), which states that comprehension problems reflect the deletion of traces left behind by moved arguments and the interaction of this trace deletion with a specific default strategy. Previous studies investigating the agrammatic syntactic comprehension deficit have focused on constructions involving thematic relations estab- lished in canonical or non-canonical order either within the verbal domain alone (as in verbal active and passive constructions) or across the two domains (as in subject and object relative clauses, where a CP is embedded within a DP). However, no study has been conducted on agrammatic speakers’ interpretation of thematic relationships within DPs only. Con- sidering the theoretical parallelisms and differences between verbal and nominal constructions, our study gives insight into a number of crucial questions about the nature of the agrammatic comprehension deficit. These include the influence of nominal vs. verbal features, the role of mor- phology, the locality of syntactic (movement) operations and the scope of the underlying syntactic impairment. Materials and methods We designed a sentence comprehension test involving semantically reversible canonical and non-canonical structures within a single domain (verbal and nominal ‘actives’ and ‘passives’) and across domains (subject and object relative clauses) in German. By using transitive verbs with a corresponding event nominalization ending in the derivational suffix— ung, we were able to hold thematic relationships constant across domains. Table 1 illustrates the seven experimental conditions (n = 22/condition). In contrast to English, in the German DP genitives may appear either before the head noun (condition 6) or after it (condition 7). However, only genitive proper names like ‘Peter’s’ (and not full genitive DPs like ‘the boy’s’) may appear in the prenominal position in German. Therefore, the agent and patient of each experimental stimulus sentence was either a boy or a man introduced to the participants by a proper name (‘Peter’ and ‘Mister Miller’) before testing. Using proper names thus enabled us to test for comprehension of constructions involving both genitive posi- tions. Each stimulus sentence contained one proper name and one full- DP argument. The relative clauses and the DP-structures were embedded in direct object position of a question for the correct picture. To be able to assess any detrimental influence of the nominalization process itself, we included two lexical distractor pictures (showing a different action with either the correct or reversed thematic relationship) in addition to the tar- get and the role-reversal foil. Participants thus had to choose the correct picture from an array of four. Participants So far, we have tested 6 participants with agrammatism and an equal number of control participants. doi:10.1016/j.bandl.2007.07.055 * Corresponding author. E-mail address: philip.rausch@gmail.com (P. Rausch). www.elsevier.com/locate/b&l Brain and Language 103 (2007) 8–249