Abstract: The push-pull system may offer a promising crop diversification strategy for small-holder farmers because it is associated with maize stemborer and Striga suppression, and can improve soil fertility and provide feed for livestock. However, the system has often been assessed as a package and the contribution of each component is not clear. In addition, it is unclear how the system performs when crop components are changed to better meet household needs, and how it functions in different landscape settings. Here we evaluate the potential of the push-pull system to suppress maize stemborer Busseola fusca (Fuller) infestations in three landscapes in the Rift Valley region of Ethiopia along a gradient of landscape complexity. Within each landscape, experimental plots were established on four representative smallholder farms. At each farm we used a split-plot factorial design with main plots surrounded by Napier grass or not, and subplots consisting of sole maize, a maize-bean intercrop, and a maize-Desmodium intercrop. In the simple landscape, which was dominated by maize crops, all plots had high stemborer infestation levels, irrespective of within-field crop diversity. However, in the intermediate landscape, subplots with sole maize had higher stemborer infestation levels compared to maize-bean or maize-Desmodium. In the complex landscape infestation levels were low in all treatments. However, these effects did not lead to significant differences in maize grain and stover yields among treatments. Our findings highlight the importance of the landscape surrounding for the performance of the push-pull system.