This study tests whether spatial dynamics can stabilize metapopulations wit
h a small number of patches and tests the influence of patch arrangement. I
measured persistence of predator and prey protists in replicated microcosm
s with two to four patches. Predators persisted for 85-437 generations (26-
130 d). As predicted by single-species and/or predator-prey metapopulation
models, substantial variation in predator metapopulation persistence was ac
counted for by the amount of patches or habitat, number of dispersal corrid
ors, maximum interpatch distance, and proportion of patches providing colon
ists (which depends on the explicit spatial arrangement of patches). Contra
ry to expectation, persistence was not influenced by loops of patches or pa
tch similarity. Persistence was also shorter in four-patch loops than three
-patch loops, indicating an interaction between patch number and arrangemen
t, which is not predicted by published models. Spatial synchrony of density
fluctuations was central to predator persistence but had complex effects o
n extinction-colonization dynamics, rescue effects, and predator-prey inter
action strength. Levins's model, containing only extinction-colonization dy
namics, predicted patch occupancy for prey but not predators. Predators did
, however, show rescue effects and changes in interaction strength. This wo
rk illustrates the need to combine experimentation with modeling to underst
and the mechanisms of spatial dynamics.