Agroforestry (AF) systems have been the focus of numerous research and development projects in southern Africa, yet their adoption rate generally remains low. Employing on-farm, participatory research techniques in southern Malawi, we compared the suitability of three AF-based systems that relay crop the dominant staple, maize (Zea mays), with the perennial legumes Sesbania sesban, Tephrosia vogelii, and Cajanus cajan (pigeonpea). Our secondary objective was to compare two methodologies employed to investigate AF adoption: farming systems based ex ante adoption potential and ex post adoption analysis. Nineteen percent of farmers preferred S. sesban, 26% T. vogelii, and 55% pigeonpea. Between 2001 and 2003, S. sesban adoption ranged from 3 to 6%, T. vogelii from 16 to 20%, and pigeonpea from 76 to 100%. Pigeonpea and T. vogelii were primarily preferred and adopted for their immediate livelihood benefits—a secondary food source in the case of pigeonpea and a fish poison in the case of T. vogelii. Though S. sesban was the most promising in terms of biophysical impacts, many farmers found it labor intensive and its lack of immediate livelihood benefits was a deterrent to adoption. With food insecurity a pervasive hardship in the region, farmers will likely continue to focus on satisfying immediate livelihood needs before prioritizing longer-term soil-quality improvement techniques. Both ex ante adoption potential and ex post adoption analysis contributed distinct and valuable data, and relying on either exclusively would have limited our understanding of the AF systems.