The factors that contribute to success and failure in introductory programming courses continue to be a topic of lively
debate, with recent conference panels and papers devoted to the subject (e.g. Rountree et al. 2004, Ventura et al., 2004,
Gal-Ezer et al., 2003). Most work in this area has concentrated on the ability of single factors (e.g. gender, math
background, etc.) to predict success, with the exception of Wilson et al. (2001), which used a general linear model to
gauge the effect of combined factors. In Rountree et al. (2002) we presented the results of a survey of our introductory
programming class that considered factors (such as student expectations of success, among other things) in isolation.
In this paper, we reassess the data from that survey by using a decision tree classifier to identify combinations of
factors that interact to predict success or failure more strongly than single, isolated factors.