The neural constraint hypothesis is one of the central ideas for the understanding of insect–plant interaction but there are still knowledge gaps in the data for foraging behavior and the performance of herbivores, and particularly florivores.We used a floriphilic katydid, Phaneroptera brevis (Serville, 1838) (Orthoptera:Tettigoniidae) and a naturalized weed, Bidens pilosa L. (Asteraceae) in caged experiments in an insectary to answer these questions: 1) How does the foraging performance of the floriphilic katydid vary when exposed to a choice in the number of capitula and types of florets of B. pilosa? 2) Does the foraging performance of the katydid, when exposed to multiple choices, improve with time, and are between-individual differences in foraging performance consistent? We observed that having more choices in the floret types and number of capitula is generally associated with a reduced foraging performance of the katydids. Floret types and number of capitula, however, did not have an interactive effect on foraging performance. We also found that the differences in foraging performance in response to choice tend to be consistent between katydids but each katydid became more efficient and decisive over time.That learning and experience can improve the foraging performance of the katydid has provided us with some insights as to how a continuum of efficient and inefficient katydids can be maintained in a population.