It is widely assumed that resistance to consumers (e.g., predators or patho
gens) comes at a "cost," that is, when the consumer is absent the resistant
organisms are less fit than their susceptible counterparts. It is unclear
what factors determine this cost. We demonstrate that epistasis between gen
es that confer resistance to two different consumers can alter the cost of
resistance. We used as a model system the bacterium Escherichia coli and tw
o different viruses (bacteriophages), T4 and lambda, that prey upon E. coil
. Epistasis tended to reduce the costs of multiple resistance in this syste
m. However, the extent of cost savings and its statistical significance dep
ended on the environment in which fitness was measured, whether the null hy
pothesis for gene interaction was additive or multiplicative, and subtle di
fferences among mutations that conferred the same resistance phenotype.