© 2017 John Wiley & Sons Ltd Aim: We tested the effects of season and migratory status (residents-versus-seasonal migrants) on island biogeography of bird assemblages through partitioning beta diversity into richness and turnover components and community nestedness metrics. We predicted that total beta diversity, the richness component of beta diversity and community nestedness will be lower for bird assemblages in winter than in summer, and lowest of all for winter visitors. These predictions were derived from published ideas about resource availability, movement and habitat choice in birds in different seasons. Location: Thousand Island Lake, China. Methods: Bird species were sampled using line transects on 36 islands during five breeding and winter seasons (2009−2014). Birds were grouped into assemblages of winter residents, winter visitors and summer residents. Associations between beta diversity partitioning, island area, isolation and habitat richness were tested using partial Mantel correlations. We complemented these tests with measures of nestedness and null model approaches. Results: Contrary to expectation, beta diversity, nestedness and difference of beta diversity or its components from null models were higher for winter residents than either summer resident or winter visitor assemblages. As predicted, winter visitors showed little association with habitat richness, and beta diversity was rarely different from null communities. Summer residents had the highest correlations of beta diversity components with habitat richness, but showed the lowest level of total beta diversity, a low richness component and were anti-nested (less nested than random). Main conclusions: Substantial differences were found in the biogeography of winter-versus-summer residents, and seasonal visitor (migratory)-versus-resident bird assemblages, which match expectations derived from bird biology and population ecology. Summer residents highlighted the role of habitat-related niche differences, whereas winter residents showed area-related selective extinction. By contrast, winter visitors appeared to be more randomly distributed.