Recent controversy over the disagreement of population attributable fraction estimates for the obesity-total mortality relation has made the concept of attributable fraction visible in both scientific and popular news. Most of the attention in writings on the attributable fraction has focused on technical matters of estimation and on ensuring a causal relationship between exposure and outcome. Yet some of the most illuminating questions about the attributable fraction have to do with another causal question and how the measure is to be interpreted in light of the answer to this question: What interventions are available to cause the assumed reduction in risk among the exposed and the consequent estimated reduction in disease burden? In this paper, I discuss the limitations to the common interpretations of the attributable fraction and argue that these limitations cannot be overcome merely by better statistical modeling or by use of better data sets. They must be addressed through discussion of specific interventions and the hypothesized causal consequences of such specified interventions.