Sensitive periods, during which experiences have a large impact on phenotypic development, are most common early in ontogeny, yet they also occur during later ontogenetic stages, including adolescence. At present, however, we know little about why natural selection favors sensitive periods for some traits early in ontogeny and for others later in ontogeny. This article synthesizes recent mathematical models and empirical studies that explore sensitive periods beyond early ontogeny. Across formal models, we observe two general patterns. First, sensitive periods emerge beyond early ontogeny when an organism’s uncertainty about the environment-phenotype fit increases at later developmental stages. Second, sensitive periods also emerge beyond early ontogeny when cues at later stages reduce this uncertainty more than earlier cues do. In the empirical literature, we observe that traits showing sensitive periods beyond early ontogeny tend to be social traits, particularly among mammals. Connecting theory to data, we hypothesize that mammals have evolved to expect highly reliable information from peers in adolescence to reduce uncertainty about the current and future social environment (e.g. social dominance, mate value). Finally, we highlight current gaps in our understanding, describe how different ways of quantifying sensitive periods influenced observed patterns, and suggest future directions for strengthening bridges between empirical and theoretical studies of sensitive periods. Ultimately, we hope our synthesis will contribute towards an integrative science of sensitive periods across the biological and the social sciences.