Corpus Data in Experimental Linguistics

Abstract

This contribution outlines a novel way of experimental item construction which aims at integrating as much realistic data as possible in experimental studies on phenomena that seem to ban extensive recourse to natural data. As a case in point, we discuss syntactic base positions. In Modified Stimulus Composition, semantically annotated corpus data are extracted and systematically modified so as to fit an indicative test surround. With that, we arrive at near-natural experimental items covering natural lexical variability, thus preventing bias and artificiality. As the systematic derivation is captured in an annotation scheme, the method in addition facilitates further analyses of surface-related phenomena.