In theory, predator-prey pairs with extinction-prone local populations can persist through metapopulation dynamics, wherein local populations fluctuate asynchronously, occasionally providing dispersers that prevent permanent extinction in all patches. A few studies have shown that spatial structure can extend predator-prey persistence. However, no studies have unequivocally demonstrated the asynchrony among patches, low dispersal rates, and rescue effects that prove metapopulation dynamics extend persistence. We used a protist predator-prey pair to show that spatial subdivision lengthens persistence through metapopulation dynamics. The pair comprised the predaceous ciliate, Didinium nasutum, feeding on the bacterivorous ciliate, Colpidium cf. striatum. A replicated experiment assessed how habitat subdivision affects persistence. Undivided habitats were of four volumes: 30, 180, 270, and 750 mt. Subdivided microcosms, or ''arrays,'' were groups of nine or 25 linked 30-mL bottles (270 or 750 mt total volume). In arrays, predators and prey persisted for 130 d (602 prey and 437 predator generations), at which point the experiment ended. Predators went extinct in undivided microcosms of equivalent volumes within a mean of only 70 d. Predators persisted for a mean of just 19 d in isolated 30-mL bottles (equivalent to isolated patches of arrays). In a separate experiment, prey were driven extinct in four of 15 isolated 30-mL bottles, and persistence times of predators were broadly similar. We documented the following hallmarks of metapopulation dynamics: (1) asynchronous fluctuations in different subpopulations; (2) frequent local prey extinctions and recolonizations; (3) persistence of protists in arrays, despite extinction of isolated local populations; and (4) rescue effects in predator populations. Other experiments measured dispersal rates and the effects on local dynamics of immigrant predators and prey, and initial predator : prey ratios. Only a small fraction of protists dispersed within a generation, consistent with metapopulation dynamics. Immigration of predators increased the frequency of local extinctions of prey, and immigration of prey increased the persistence of both predators and prey. Higher initial predator:prey ratios decreased the persistence of prey in undivided volumes. Although the pair persisted regionally in arrays, data indicated that local extinctions of prey were common. In array patches, predator:prey ratios were higher and predator-prey cycles were shorter than in undivided volumes. Dispersal made local dynamics more prone to extinction, yet promoted regional persistence because the risk of extinction of distant subpopulations became independent.