Leaves display an enormous array of sizes and shapes. Although these attributes appear to have evolved primarily in response to abiotic conditions in the plant’s habitat, the importance of insect herbivores as additional selective agents is still poorly understood. A necessary requirem ent for leaf size and shape to evolve in response to attack by insects is that insects must respond to and/or be affected by, leaf morphology. We tested leaf-shape preferences in adult flea beetles (
Phyllotreta
spp.) feeding on the highly variable rosette leaves of
Capsella bursa-pastoris
. Contrary to theoretical expectation (Brown & Lawton 1991), leaves with deeply lobed margins were more intensely damaged, both in field-collected and experimental plants. In two ancillary experiments with
Capsella
, we found that
Spodoptera
caterpillars showed no preferences for leaf shape, but that adult vine weevils (
Otiorhynchus sulcatus
) did, preferring (as predicted), undivided over divided leaves. We conclude that Brown & Law ton’s (1991) hypothesis is at best weakly supported by laboratory data for vine weevils, refuted by laboratory data for
Spodoptera
, and consistently refuted by both laboratory and field data for flea beetles. Although the experiment tried to reduce confounding variables to a minimum, interpretation was complicated by correlations between leaf shape and other developmental parameters of the plants, and highlights the difficulty of disentangling leaf-shape effects from other confounding factors.