Statistical Modeling of Writing Processes

Until now, writing has been described in the research literature as the interplay of situations, strategies and phases – with phases being identifiable temporal procedural units with typical dominant writing actions such as ‘formulating’ or ‘source reading’. Phases are recognized as essential for the success of writing processes. At the same time, most scientific approaches to writing base their phase concepts and phase descriptions on introspection, experiments or single case studies. The methodology for a rigorous, objectively verifiable analysis of the structure of writing processes and therefore for an empirically testable explanation of the nature and interplay of phases in writing processes in naturalistic settings has yet to be developed. In this paper we propose that statistical modeling based on progression analysis is an appropriate tool to explore writing phases. In section 1 of the paper we review the most important previous approaches to writing phases. In section 2 we provide the link to current writing research by outlining progression analysis and presenting the data to be analyzed. In section 3 we present a very simple statistical tool for identifying and analyzing writing phases: cusum statistics. In section 4 we move to a more refined procedure based on signal-extraction methods. Section 5 concludes with an outlook on future work in the area of qualitative writing research.