The theory of kin selection is central to the understanding of social evolution. Recent theoretical work suggests a limitation on the action of kin selection in structured populations. The first such limit involves a specialized population structure, termed "budding viscosity," common to social insects, in which new groups are formed by fission with no individual dispersal. An argument based on the separation of kin selection dynamics into among and within group components suggests that kin selection cannot operate in this structure. However, stochastic computer simulations show that random variation among daughter groups can supply the needed among-groups variation and allow kin selection to proceed. The second limit on kin selection involves simple population viscosity, in which individuals disperse limited distances and so are related to their neighbors. Altruism toward neighbors, favored by kin selection, is opposed by local competition. Computer simulation confirms this limitation and shows that some form of specific kin recognition is required to favor the evolution of altruism by kin selection. All applications of kin selection require a measure of genetic relatedness; a computer program for calculating this statistic from genetic data on natural populations is described.