Some controls control too much.

Investigators of complex phenomena often seek to isolate variables of interest by statistical control. However, they run the risk of studying situations that do not exist in the real world, missing mediational links, and drawing incorrect policy conclusions. In the National Institute of Child Health and Human Development (NICHD) Early Child Care Research Network article in this issue, amount of time in child care was studied controlling for family income and maternal depression. However, use of child care is closely linked with maternal employment, which increases family income and decreases maternal depression, which in turn are linked to children's socioemotional adjustment. Hence, the negative effects of amount of time in child care reported in the article are likely overestimated relative to causal pathways in the lives of real families.