Amazonian bird species often have patchy spatial distributions, and previous work has attributed this pattern to habitat specialization and dispersal limitation; however, we know comparatively little about the origins and maintenance of the isolated populations that constitute a patchy distribution. In this study, we ask whether patchy populations are interconnected by dispersal. We formulated 2 alternative hypotheses: (1) patchy populations are relicts of ancient connectivity or dispersal; and (2) patchy populations are centers of local abundance embedded in a matrix of contemporary dispersal or diffuse intervening populations. We confronted these hypotheses with circumstantial evidence derived from a unique suite of noteworthy bird records and geological observations from northeastern Peru. We found support for both hypotheses in different species, and sometimes within single species at different spatial scales. Phenotypically differentiated populations in relictual habitat patches provide strong support for the first hypothesis, whereas populations in recently created, ephemeral habitat patches provide strong support for the second. Colonizations of anthropogenic habitats are further examples of the second process, and they indicate ongoing changes in the porous connectivity of upper Amazonia. These results highlight the need for a conservation approach that accounts for metapopulation dynamics across patchy species ranges.