Rolling.Classify is a recently developed tool for studying collaboration (Eder, Rybicki, and Kestemont 2016; Eder 2016) that builds on earlier work that tested successive overlapping sections of texts (van Dalen-Oskam and van Zundert 2007, Burrows 2010, Hoover 2012).

The power and ease of use of Rolling.Classify (and its related Rolling.Delta) have led to several studies based on various kinds of texts.. Rigorous testing of this new method on problems with known solutions seems especially important because its results vary greatly with the choice of classification method other parameters. I will begin with simulated collaborations comprising text sections of varied lengths assembled to model different kinds of collaboration. I will then test collaborations with known contributions by the authors, and finally some in which no clear evidence of the nature of the collaboration exists.