Studying connectivity with the help of computer-readable corpora

This paper discusses methodological aspects of the use of electronic language corpora for the study of connectivity. We demonstrate how a corpus-based approach was used to investigate functional characteristics of coordinating elements in sentence- or utterance-initial position across different languages (English, German, Old Swedish and Turkish), across different modalities (written and spoken) and across the diachronic dimension (historic and modern languages). Our focus is on the difficulties we encountered in this study when attempting to transfer corpus-based methods developed for the analysis of corpora of modern, written language to the analysis of corpora of historic or spoken language. We suggest an abstract corpus-linguistic workflow and discuss where and how this workflow differs according to the corpus type, and how well its individual steps are supported by current corpus technology.