Most herbivorous arthropods are host specialists and the question is which mechanisms drive the evolution of such specialization. The theory of antagonistic pleiotropy suggests that there is a trade-off between adaptation of herbivores to a novel host and their native host. The mutation accumulation hypothesis proposes that herbivores on a novel host lose their adaptation to the native host through the accumulation of mutations with negligible effects on performance on the novel host. Experimental evidence for either of the two hypotheses is scarce. We compared the fitness of two sympatric moth strains from an introduced host and a native host. The strain from the novel host did not perform better on this host than the strain from the native host. The strain from the novel host performed less well on the native host than did the strain from the native host. Hence, selection on the novel host did not result in noticeable gain in performance, but adaptation to the native host was lost. These results are more readily explained by the mutation-accumulation hypothesis than by the trade-off hypothesis.