There has been an increasing emphasis in educational policies, practices and professional development on the capacity of educational leadership to exert a causal impact on student cognitive and non-cognitive outcomes. If the strict criteria of counterfactual causality are adhered to no causal inference could be made about the effects of leadership, given that experimental research designs are inappropriate. Statistical studies of leadership "effects" need to operationalize the concept of leadership and use data from natural settings. They are unable reliably to disentangle associational from counterfactual causal relationships. Given this dilemma, the article argues for using less purist criteria for causal inference but making this methodological approach explicit. Relationships observed in both quantitative and qualitative data can be interpreted only through insights derived from theoretical and conceptual knowledge of the area of inquiry. "Causal social explanation" (as distinct from counterfactual causality) provides a justification for "mixed methods" to build a more secure empirical basis for knowledge about the effects of leadership. (Contains 1 figure and 6 notes.)