The role of brain size as a cost of learning remains enigmatic; the nature and timing of such costs is particularly uncertain. On one hand, comparative studies suggest that congenitally large brains promote better learning and memory. In that case, brain size exacts a global cost that accrues even if learning does not take place; on the other hand, some developmental studies suggest that brains grow with experience, indicating a cost that is induced when learning occurs. The issue of how costs are incurred is an important one, because global costs are expected to constrain the evolution of learning more than would induced costs. We tested whether brain size represented a global and/or an induced cost of learning in the cabbage white butterfly, Pieris rapae. We assayed the ability of full sibling families to learn to locate either green hosts, for which butterflies have an innate search bias, or red hosts, which are more difficult to learn to locate. Naïve butterflies were sacrificed at emergence and congenital brain volume estimated as a measure of global costs; experienced butterflies were sacrificed after learning and change in brain volume estimated as a measure of induced costs. Only for the mushroom body, a brain region involved in learning and memory in other insects, was volume at emergence related to learning or host-finding. Butterfly families that emerged with relatively larger mushroom bodies showed a greater tendency to improve their ability to find red hosts across the two days of host-search. The volume of most brain regions increased with time in a manner suggesting host experience itself was important: first, total number of landings during host-search was positively related to mushroom body calyx volume, and, second, experience with the red host was positively related to mushroom body lobe volume. At the family level, the relative volume of the mushroom body calyx and antennal lobes following learning was positively related to overall success in finding red hosts. Overall, our results suggest that within species, brain size might act as a small global cost of learning, but that environment-specific changes in brain size might reduce the overall costs of neural tissue in the evolution of learning.