Ill-defined causal questions present serious problems for observational studies—problems that are largely unappreciated. This paper extends the usual counterfactual framework to consider causal questions about compound treatments for which there ...