Milton Friedman is usually regarded as an instrumentalist on the basis of his infamous claim that economic theories are to be judged by their predictions and not by the realism of their assumptions. This interpretation sits oddly with Friedman's empirical work - e.g., Friedman and Schwartz's monetary history - and his explicit rejection of theories of the business cycle that, while based on accurate correlations, nevertheless do not make economic sense. In this paper, I try to reconcile Friedman's methodological writings with his practices as an empirical economist by, first, taking his roots in Alfred Marshall seriously and, second, by taking the methodological implications of his empirical work seriously. Friedman dislikes the word "cause". Nevertheless, appearances to the contrary notwithstanding, Friedman is best understood as a causal realist - that is, one who understands the object of scientific inquiry as the discovery through empirical investigation of the true causal mechanisms underlying observable phenomena.