In this paper I critically evaluate Reisman and Forber’s (Philos Sci 72(5):1113–1123, 2005) arguments that drift and natural selection are population- level causes of evolution based on what they call the manipulation condition. Although I agree that this condition is an important step for identifying causes for evolutionary change, it is insufficient. Following Woodward, I argue that the invariance of a relationship is another crucial parameter to take into consideration for causal explanations. Starting from Reisman and Forber’s example on drift and after having briefly presented the criterion of invariance, I show that once both the manipulation condition and the criterion of invariance are taken into account, drift, in this example, should better be understood as an individual-level rather than a populationlevel cause. Later, I concede that it is legitimate to interpret natural selection and drift as population-level causes when they rely on genuinely indeterministic events and some cases of frequency-dependent selection.