We used 22 yr of capture–mark–reencounter (CMR) data collected from 1988 to 2009 on about 12,500 birds at what went from three to five coastal colony sites in Massachusetts, New York, and Connecticut, United States, to examine spatial and temporal variation in breeding dispersal/fidelity rates of adult Roseate Terns (Sterna dougallii). At the start of our study, Roseate Terns nested at only one site (Bird Island) in Buzzards Bay, Massachusetts, but two more sites in this bay (Ram and Penikese Islands) were subsequently recolonized and became incorporated into our CMR metapopulation study. We examined four major hypotheses about factors we thought might influence colony-site fidelity and movement rates in the restructured system. We found some evidence that colony-site fidelity remained higher at long-established sites compared with newer ones and that breeding dispersal was more likely to occur among nearby sites than distant ones. Sustained predation at Falkner Island, Connecticut, did not result in a sustained drop in fidelity rates of breeders. Patterns of breeding dispersal differed substantially at the two restored sites. The fidelity of Roseate Terns at Bird dropped quickly after nearby Ram was recolonized in 1994, and fidelity rates for Ram soon approached those for Bird. After an oil spill in Buzzards Bay in April 2003, hazing (deliberate disturbance) of the terns at Ram prior to the start of egg-laying resulted in lowering of fidelity at this site, a decrease in immigration from Bird, and recolonization of Penikese by Roseate Terns. Annual fidelity rates at Penikese increased somewhat several years after the initial recolonization, but they remained much lower there than at all the other sites throughout the study period. The sustained high annual rates of emigration from Penikese resulted in the eventual failure of the restoration effort there, and in 2013, no Roseate Terns nested at this site.