Cosmopolitan pests can consist of geographic populations that differ in their current host ranges or in their ability to colonize a novel host. We compared the responses of cowpea-adapted seed-beetle populations (Callosobruchus maculatus [F.]) from Africa, North America, and South America to four novel legumes: chickpea, lentil, mung bean, and pea. We also qualitatively compared these results to those obtained earlier for an Asian population. For each host, we measured larval survival to adult emergence, and used both no-choice and choice tests to estimate host acceptance. The pattern of larval survival was similar among populations: high or moderately high survival on cowpea, mung bean, and chickpea, intermediate survival on pea, and very low survival on lentil. One exception was unusually high survival of African larvae on pea, and there was modest variation among populations for survival on lentil. The African population was also an outlier with respect to host acceptance; under no-choice conditions, African females showed a much greater propensity to accept the two least preferred hosts, chickpea and lentil. However, greater acceptance of these hosts by African females was not evident in choice tests. Inferences about populations differences in host acceptance can thus strongly depend on experimental protocol. Future selection experiments can be used to determine whether the observed population differences in initial performance will affect the probability of producing self-sustaining populations on a marginal crop host.