An Analysis of the Evidential Use of German Perception Verbs

This paper presents a frame-theoretic account of the inferential and non-inferential use of German perception verbs like klingen ‘sound’, aussehen ‘look (like)’, and sich anfühlen ‘feel (like)’. We argue that a proper analysis of constructions based on these verbs requires explicit reference to object attributes like SOUND, SIGHT, and TOUCH, to which we refer as ‘dimensions’. It will be shown that a frame analysis, in which object dimensions can easily be represented as frame attributes, is ideally suited for the treatment of perception verbs of this type. Following Barsalou (1992), we define frames as recursive attribute-value structures, which are represented as directed labeled graphs with the arcs corresponding to attributes and the nodes to their values (cf. Petersen 2007). Furthermore, we assume that the knowledge about admissible frames is represented in type hierarchies which restrict the set of appropriate attributes and their values for object classes. Given our frame model, the admissible inferential and noninferential uses of perception verbs are captured as constraints on attributes assigned to the subject referent, the verb and the predicative complement of the verb. Finally, a comparison between the German data and verb-based evidentials in French will show that the inferential use of perception verbs exhibits marked differences cross-linguistically.