Abstract: Richerson, Boyd, and Henrich argue that certain skills exhibit such a high degree of complexity that they likely cannot be invented alone. Such complex skills, they claim, call for an evolutionary explanation, via cumulative, stepwise changes over time, transmitted via social learning, or “cumulative cultural evolution.” Assuming that complexity means the number of parts in a whole or the density of interactions among the parts, it remains unclear whether we are to judge a skill “complex” on grounds of the complexity of its performance, the complexity of its products, or the complexity of its developmental process. In this paper, I argue that the complexity in performance and products, while mostly used in the existing literature, do not simultaneously motivate cumulative explanation and unify many targets of such explanations. As a positive alternative, I argue that complexity in development is a better criterion for filtering skills that require cumulative explanation.