The study of selection in natural populations is a neglected field. So far only a few investigators have succeeded in measuring lifetime fitness or selection on multiple phenotypic characters. Studies of multiple selection episodes are also rare. Most traits of interest to ecologists and evolutionists are polygenic with complex patterns of genetic coupling (Wright. 1978), but only a few field workers have combined studies of selection and inheritance. Consequently we do not yet have a clear view of multivariate selection in nature or its evolutionary consequences. In this paper we apply statistical measures of selection discussed in the companion paper (Arnold and Wade, 1984) to the analysis of multiple selection episodes. The advantage of the measures we use is that they are directly related to coefficients of selection used in theoretical equations for evolutionary change. We analyze selection exerted during particular segments of the life cycle by dividing total fitness into multiplicative components (e.g., viability, fecundity, fertility). Selection on a phenotypic trait during a particular part of the life cycle can then be measured by the statistical relationship between the corresponding fitness component and the trait. Furthermore, the portions of selection corresponding to particular fitness components or selection episodes will sum to the total selection acting on the trait. This additive property of the selection measures depends on a multiplicative definition of fitness components and has several uses. Our partitioning of selective force has three principal uses. First, it allows a characterization of selection when it is possible to measure portions of fitness but impractical to record total, lifetime fitness. A particularly useful property is that one obtains the same estimate of selection during an episode using complete or incomplete data on lifetime fitness. Second, even when total fitness can be measured, it will often be desirable to measure the forces exerted by separate episodes of selection, since selection may change in magnitude and direction during different stages of life. For example, in a re-analysis of Howard's (1979) data on the fitness of male bullfrogs during a single reproductive season, we found that sexual selection accounted for most of the selection on body size with only minor contributions from two forms of natural selection. Finally, partitioning of selective forces may help identify the actual agents of selection. Thus a demonstration that selection acts during a particular short segment of the life cycle can narrow the field of candidate selective agents. For example, we briefly discuss some examples of strong directional selection arising from severe weather conditions.