As the evolutionary importance of plant tolerance of herbivory is increasingly appreciated, more and more studies are not just measuring a plant's tolerance, but are comparing tolerance among plant genotypes, populations, species, and environments. Here, we suggest that caution must be taken in such comparative studies in the choice of measurement scales (and data transformations) for damage levels and plant performance. We demonstrate with a simple scenario of two plant groups of equal tolerance how the choice of scales can lead one to infer that the first group is more tolerant, the second group is more tolerant, or the two groups are equally tolerant—using the identical dataset. We conclude that to make reliable, logically consistent inferences when comparing tolerances among groups of plants, damage and performance should both be on an additive scale or both on a multiplicative scale.