Divergent resistance traits among plant species should select for different offensive abilities in specialists and generalists. In particular, a specialist may be able to optimize offense on its own host; a generalist’s optimal offense may be broadly useful across host species, but not optimized for variation in resistance within a particular host species. We tested these hypotheses by comparing aspects of performance of a specialist-generalist Leptinotarsa species pair in experiments using wound-induced and un-induced controls of four Solanum species including the major hosts of both herbivore species and a non-host species. Other experiments were restricted to the specialist’s host using plants grown from roots or seeds, which differed in a putative resistance trait. We predicted that (i) the specialist would have a performance advantage over the generalist on its own host on the leaves of both damaged and undamaged plants, but (ii) would lose this advantage on other plants, and that (iii) the performance of the generalist would suffer disproportionately by feeding on wound-induced leaves of the specialist’s host, reflecting a cost for maintaining non-specific offensive strategies. Results provided support for prediction (i) and partial support for prediction (ii). A significant interaction between herbivore species and damage treatment suggested a possible tradeoff between diet breadth and performance on damaged plants. However, the generalist was not disproportionately affected by feeding on wound-induced leaves of the specialist’s host plant, providing no support for prediction (iii).