1. Adapting to a low-quality plant may require modification of an insect's digestive physiology, oviposition behaviour, or other host-use traits. If colonising a marginal host entails a cost, a decay in adaptation would be expected after selection is relaxed, i.e. if populations on a novel host are reverted to their high-quality ancestral host. 2. Replicate lines of the seed beetle Callosobruchus maculatus (F.) rapidly adapted to lentil seeds; larval survival rose from approximately 1 to ≥ 90%, and oviposition on lentil increased more than two-fold. This study compared egg-laying behaviour in lines that either remained on lentil or were reverted to the ancestral host, mung bean, for 22–62 generations. 3. Consistent with the trade-off hypothesis, females from two reverted sublines showed decreased oviposition on lentil (estimated as lifetime fecundity), but host acceptance in a third subline was unchanged. In a short-term assay, acceptance of lentil by newly emerged females was lower in each reverted subline than in the corresponding non-reverted one. Because effective population sizes (determined from genome resequencing) were large throughout the experiment, this decline in host acceptance is unlikely to be explained solely by genetic drift. 4. Variation among replicates in the magnitude of the reversion effect was also observed in a previous study of larval survival. However, the pattern of variation for survival was not congruent with the pattern of variation for host acceptance in this study. Thus, genes mediating improved performance on lentil appear to be largely independent of those responsible for increased oviposition.