With random assignment of subjects into experimental groups, we can assume that any observed difference between groups should be due only to the variable we are manipulating.
