Thirty fourth-grade students participated in an extended intervention previously successful in fostering skills of scientific investigation and inference, notably control of variables (COV). The intervention was similarly successful for a majority of students in the present study, enabling them to isolate the three causal and two noncausal variables operating in a multivariable system. However, when asked to predict outcomes of various constellations of variable levels, they tended not to take into account the effects of all of the causal variables they had identified. Moreover, they did not adhere to a consistency principle, i.e., that a factor that produces an effect can be expected to produce the same effect in the future, given similar conditions. These findings suggest that COV is not the only challenge students experience in reasoning about multiple variables. Elementary-school students' mental models of multivariable causality appear to deviate from a normative, scientific model, even after they have mastered that aspect of scientific method having to do with the design of controlled experiments to isolate effects of individual variables. The challenges, beyond COV, that appear to be involved in making prediction judgments involving multiple variables warrant attention in the design of curricula to foster development of scientific thinking skills. (Contains 3 tables and 5 figures.)