Theory on plasticity driving speciation, as applied to insect–plant interactions (the oscillation hypothesis), predicts more species in clades with higher diversity of host use, all else being equal. Previous support comes mainly from specialized herbivores such as butterflies, and plasticity theory suggests that there may be an upper host range limit where host diversity no longer promotes diversification. The tussock moths (Erebidae: Lymantriinae) are known for extreme levels of polyphagy. We demonstrate that this system is also very different from butterflies in terms of phylogenetic signal for polyphagy and for use of specific host orders. Yet we found support for the generality of the oscillation hypothesis, in that clades with higher diversity of host use were found to contain more species. These clades also consistently contained the most polyphagous single species. Comparing host use in Lymantriinae with related taxa shows that the taxon indeed stands out in terms of the frequency of polyphagous species. Comparative evidence suggests that this is most probably due to its nonfeeding adults, with polyphagy being part of a resulting life history syndrome. Our results indicate that even high levels of plasticity can drive diversification, at least when the levels oscillate over time. Predictions from the oscillation hypothesis of diversification in plant-feeding insects were tested in a lineage containing many extreme generalists, the tussock moths (Lymantriinae). Taxa with higher diversity of host use were found to contain more species, and this was true also for taxa containing the most polyphagous single species. These observations support the hypothesis that plasticity in host use promotes diversification.