Experiments designed to reveal variation among individual parasites in preference for different host species may generate misleading results. Apparent variation in the order of preference among host species can be generated solely from variation in the strength of discriminations made within host species. We illustrate this with a study of oviposition preference in the butterfly Melitaea cinxia. All butterflies were tested on the same six individual plants, three Plantago lanceolata (P) and three Veronica spicata (V). Some insects repeatedly preferred all individual P over all individual V or vice versa. We designated these as “pure” species ranks. Other insects repeatedly produced “mixed” ranks, preferring some individual V over some P, and some individual P over some V. We show how a “mixed” rank butterfly could differ from a “pure” rank insect by discriminating either more within plant species and/or less between them. Therefore, discrimination within host species can mask or confound discrimination among species. We discuss implications for the design of preference experiments.