Using two studies, this paper argues that the controlled study of how specific variables affect others and the study of the same when embedded in complex educational environments should be based on two sets of assumptions underlying two paradigms, analytic and systemic, that differ as do quantitative and qualitative approaches. (RLC)