According to the geographic mosaic theory of coevolution, selection intensity in interactions varies across a landscape, forming a selection mosaic; interaction traits match at coevolutionary hotspots where selection is reciprocal and mismatch at coldspots where reciprocity is not a factor. Chemical traits play an important role in the interaction between wild parsnip (Pastinaca sativa) and the parsnip webworm (Depressaria pastinacella). Furanocoumarins, produced as plant defenses, are detoxified by the webworms by cytochrome P450 monooxygenases; significant additive genetic variation exists for both furanocoumarin production in the plant and detoxification in the insect, making these traits available for selection. To test the hypothesis that differences in selection intensity affect the distribution of coevolutionary hotspots and coldspots in this interaction, we examined 20 populations of webworms and wild parsnips in Illinois and Wisconsin that varied in size, extent of infestation, proximity to woods (and potential vertebrate predators), and proximity to a chemically distinct alternate host plant, Heracleum lanatum (cow parsnip). Twelve of 20 populations displayed phenotype matching between plant defense and insect detoxification profiles. Of the eight mismatched populations, a logistic regression model related matching probability to two predictors: the presence of the alternate host and average content of xanthotoxin (one of the five furanocoumarins produced by P. sativa). The odds of mismatching were significantly increased by the presence of the alternate host (odds ratio = 15.4) and by increased xanthotoxin content (odds ratio = 6.053). Parsnips growing near cow parsnip displayed chemical phenotypes that were chemically intermediate between cow parsnip and parsnips growing in isolation. Rapid phenotype matching in this system is likely due in part to differential mortality every season; larvae transferred to a plant 30 m or more from the plant on which they developed tended to experience increased mortality over larvae transferred to another umbel on the same plant on which they had developed, and plant populations that mismatched in 2001 displayed a change in chemical phenotype distribution from the previous year. Trait mixing through gene flow is also a likely factor in determining mismatch frequency. Populations from which webworms were eradicated the previous year were all recolonized; in three of seven of these populations, infestation rates exceeded 90%. Our findings, consistent with the geographic mosaic theory, suggest that the presence of a chemically distinct alternate host plant can affect selection intensity in such a way as to reduce the likelihood of reciprocity in the coevolutionary interaction between wild parsnip and the parsnip webworm.