Theory to explain how plants defend themselves against herbivorous insects is rich, but can be difficult to test. Biological invasions provide unique opportunities to test and improve upon plant defense theory, as plants experience predictable shifts in insect herbivory after introduction to a new range. Here, we use an invasion to evaluate the power of three cornerstone hypotheses to predict the evolution of defense against herbivorous insects. These three hypotheses represent increasing refinements of classic plant-insect theory regarding optimal defense, and each rests on the same three assumptions: that introduced plant populations escape natural enemies, that insect herbivory reduces plant fitness, and that putative defenses decrease herbivory. These assumptions remain untested in most invasions, which likely contributes to conflicting support for many plant defense hypotheses. We provide evidence that these assumptions are met in common mullein, Verbascum thapsus L. (Scrophulariaceae), which we propose can thus be used as a model system to test plant defense theory. We find that the hypothesis that integrates predictions of within-plant optimal defense theory and the evolutionary dilemma model (i.e., the 'shifting defense allocation' hypothesis) provides strong insights into both invasion and evolution of plant defense. Specifically, we show that introduced populations that escape important specialist herbivores increase the concentration of secondary compounds in high-quality tissue in particular, in this case protecting valuable young leaves from generalist herbivores that dominate in the introduced range. We would not have detected this shift without exploring within-plant defense allocation across native and introduced populations, a task rarely undertaken when assessing evolutionary change in plant defenses. This finding provides broad insight into how native and introduced plant populations alike may respond to shifts in herbivore pressure. We close by highlighting future directions of inquiry using introduced plant populations to develop and test plant defense theory more generally.