We know little about how natural selection on a species is altered when a closely related species consuming similar resources appears in its environment. In a pond experiment with threespine sticklebacks I tested the prediction that divergent natural selection between competitors is frequency-dependent, changing with the distribution of phenotypes in the environment. Differential growth and survival of phenotypes in a target stickleback population were contrasted between two treatments. In one treatment an offshore zooplankton feeder (the limnetic stickleback species) was added to the same pond as the target. In the other treatment I added the benthic stickleback instead, a species adapted to feeding on invertebrates from sediments and inshore vegetation. The target population was ecologically and morphologically intermediate with phenotypic variance artificially inflated by hybridization. Growth rates of phenotypes within the target population differed between treatments as predicted by character displacement. The impact of adding a second species always fell most heavily on those phenotypes in the target population resembling the added species most closely. However, those individuals in the target population that most resembled the added species did not experience reduced survival. Instead, consistent survival differences between populations suggested the presence of an inshore−offshore gradient in mortality risk. These results provide further support for the hypothesis of character displacement in sympatric sticklebacks. They suggest that displacement along the resource gradient also led to divergence in vulnerability to agents of mortality, probably including predation.