An increasing number of studies comparing legislatures relies on analyses of roll call votes. These analyses are used to infer characteristics of the way in which the legislature works and how their members vote. These inferences are problematic, however, if not all votes in parliament are recorded or the recorded votes are systematically distinct from the rest of the votes. Neglecting the way in which roll call votes are triggered or decided may result in selection bias. In this paper I discuss these problems of selection bias regarding various rules employed in legislatures which may lead to roll call votes. I then present evidence for these selection biases from a unique source of electronically recorded votes, namely all votes decided on between 1995 and 2003 in the Swiss lower house. A careful analysis illustrates to what extent commonly used figures on party discipline are biased because of the selective nature of roll call votes. I also propose a simple way how to correct for these biases, which performs adequately well.