Preventing insect pests from developing resistance to Bacillus thuringiensis (Bt) toxins produced by transgenic crops is a major challenge for agriculture. Theoretical models suggest that plants containing two dissimilar Bt toxin genes ('pyramided' plants) have the potential to delay resistance more effectively than single-toxin plants used sequentially or in mosaics. To test these predictions, we developed a unique model system consisting of Bt transgenic broccoli plants and the diamondback moth, Plutella xylostella. We conducted a greenhouse study using an artificial population of diamondback moths carrying genes for resistance to the Bt toxins Cry1Ac and Cry1C at frequencies of about 0.10 and 0.20, respectively. After 24 generations of selection, resistance to pyramided two-gene plants was significantly delayed as compared with resistance to single-gene plants deployed in mosaics, and to Cry1Ac toxin when it was the first used in a sequence. These results have important implications for the development and regulation of transgenic insecticidal plants.