Environmental risks of Bt maize, particularly pollen drift from Bt maize, were assessed for nontarget lepidopteran larvae in maize field margins. In our experimental approach, we carried out 3-year field trials on 6 ha total. Three treatments were used in a randomized block design with eight replications resulting in 24 plots: (i) near-isogenic control variety without insecticide (control), (ii) near-isogenic control variety with chemical insecticide (Baytroid) and (iii) Bt maize expressing the recombinant toxin. We established a weed strip (20 × 1 m) in every plot consisting of a Chenopodium album (goosefoot)/Sinapis alba (mustard) mixture. In these strips we measured diversity and abundance of lepidopteran larvae during maize bloom and pollen shed. C. album hosted five species but all in very low densities; therefore data were not suitable for statistical analysis. S. alba hosted nine species in total. Most abundant were Plutella xylostella and Pieris rapae. For these species no differences were detected between the Bt treatment and the control, but the chemical insecticide treatment reduced larval abundance significantly. Conclusions regarding experimental methodology and results are discussed in regard to environmental risk assessment and monitoring of genetically modified organisms.