Low soil nitrogen (N) and drought impede maize production in the small-scale farming sector in Zambia; and adoption of new cultivars with improved tolerance might enhance production. This study: a) assessed farmer preferences for maize cultivars; b) determined genotype x environment interaction effects among popular maize cultivars under contrasting soil fertility levels and; c) investigated landraces for tolerance to low N and drought using S1 selection. The study was carried out in Zambia from 2004-07. Farmer preference influencing the adoption of maize cultivars was investigated using both formal and informal surveys in Luangwa, Chibombo and Lufwanyama rural districts representing the three agro-ecological regions of Zambia. Focus group discussions and personal interviews were used to collect data on issues that affected maize production in these areas. It has been found that although farmers perceived landraces to be low yielding, they believed that they were superior to improved cultivars for: tolerance to drought; tolerance to low soil fertility; grain palatability; grain storability; and poundability. The need for food security, their inability to apply fertiliser, and their need for drought tolerant cultivars significantly (p ≤ 0.05) influenced farmers in adopting cultivars. The farmers would readily adopt cultivars that address these concerns. The predominant use of certain landraces (76%) reflected their superiority in meeting some of these needs. The performance of nine popular cultivars (three for each of hybrids, OPVs and landraces) under contrasting levels of soil fertility, across six environments (ENVs) in the three agro-eological regions, was evaluated. An ENV was defined as season x location combination. The fertilizer treatments were full fertilization, basal dressing, top dressing and nil fertilization. The cultivars exhibited significant non-crossover type of genotype x fertilisation interaction effects at three ENVs, while the genotype x fertilisation interaction effects, were non-significant at the other three ENVs. The cultivars exhibited dynamic stability by increasing grain yield (GY) when fertilization was increased. Landraces yielded higher than all open pollinated varieties and were generally higher yielding than two hybrids. Based on average rank for GY, the five highest yielding cultivars were MRI724, Gankata, MM603, Kazungula and Pandawe. Superiority of landraces revealed their genetic potential for GY under low soil fertility and they should be used as germplasm in developing cultivars targeting such environments. Ninety-six local landraces were selfed to generate S1 lines (2004/05 season) which were crossed to a tester (2005/06 season). Testcrosses were evaluated under optimal, low N, and drought conditions (2006/07 season). Data on GY, anthesis-silking interval, number of ears per plant, leaf senescence, leaf rolling, tassel size and grain texture were recorded in all the trials during the study period. Testcrosses, their S1 parents and landraces that were superior under low N, drought, optimal conditions and across environments were identified; these should be used to develop varieties targeted to a particular environment. Selection for tolerance to drought also selected for tolerance to low N. Selection for low N tolerance also selected for GY under drought and optimal conditions. Therefore, in selecting for tolerance to abiotic stresses, use of optimal and managed stress environments was effective. The following landraces were superior at 10% selection intensity: LR38, LR84 and LR86 (optimal, low N and drought conditions); LR11, LR35 and LR76 (low N and drought conditions); LR12 (optimal and drought conditions); LR40 and LR93 (low N conditions only); LR79 (drought conditions only) and; LR74 and LR85 (optimal conditions only). These landraces should be used as source germplasm targeting respective environments. Significant (p ≤ 0.05) positive general combining ability effects for GY under both low N and drought conditions were found implying that additive gene action conditioned GY under the abiotic stresses. The heritability for GY under low N (0.38), and drought (0.17) conditions, was low suggesting that selection based on GY alone was not effective. The genetic correlation for GY between optimal, and either low N (rG=0.458), or drought (rG = 0.03) environments, was low (rG < 0.5) suggesting that indirect selection would not be effective either. Therefore, use of secondary traits for selection is discussed. The study established that most farmers depended on local landraces for seed and would adopt low input improved varieties that yield higher than the landraces. Some landraces were found superior to some improved cultivars under contrasting fertilisation regimes. The study also found that landraces had genetic variation for tolerance to low N and drought. Landraces, S1 lines and testcrosses superior under low N, drought, optimal conditions and across environments were selected and they should be used to develop cultivars targeting respective environments. Policy implications of these results are discussed. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2007.