We compared the growth, phenology and leaf demography of partly defoliated, connected shoots with that of partly defoliated, severed shoots in four old-field perennials (Solidago canadensis, S. altissima, S. gigantea, Aster lanceolatus) with differing genet architectures (rhizome systems), in a common garden and in the field. Our main hypothesis was that defoliation would have fewer negative effects on shoot performance if shoots were connected than if their rhizomes were severed. Since degree of clonal integration is related to differences in genet architecture, our second hypothesis was that the effects of defoliation would be less pronounced in more integrated than in less integrated clones. Removing about 50% of the total leaf area from shoots had different effects depending on plant species, shoot density, and in particular whether rhizome connections between shoots were left intact or severed. In agreement with our prediction, experimentally isolated shoots in the field or in high density clumps in the garden suffered the most from defoliation, while shoots with intact connections or in low density clumps suffered the least. Our second prediction was neither confirmed nor falsified in the present study. Solidago altissima showed overcompensation in response to simulated herbivory in the common garden, i.e. defoliated shoots grew faster and were larger at harvest than their non-defoliated neighbours.