Natural selection and phenotypic plasticity can both produce locally differentiated phenotypes, but novel environments or gene combinations can produce plasticity that works in opposition to adaptive change. The red-shouldered soapberry bug (Jadera haematoloma) was locally adapted to feed on the seeds of an introduced and a native host plant in Florida in the 1980s. By 2014, local differentiation was lost and replaced by phenotypically similar populations all adapted to the introduced host, likely as a result of gene flow. Here, I quantify the effects of these two host plants on individual performance, natural selection, and phenotypic plasticity. I find that the seed coat and seedpod of the native host have strong negative effects on juvenile survival and adult reproduction compared to the introduced host. I find support for the hypothesis that the seedpod is driving diversifying natural selection on beak length, which was previously locally adapted between hosts. I also find maladaptive plasticity induced by host plant: bugs develop beak lengths that are mismatched with the seedpod size of the host they are reared on. This plasticity may be the result of gene flow; hybrids in the 1990s showed the same pattern of maladaptive plasticity, and plasticity is stronger in the present in areas with high gene flow. Although ongoing natural selection has produced locally adapted genotypes in soapberry bugs, maladaptive plasticity has masked the phenotypic difference between populations in the field.