A multilocational evaluation trial comprising 18 Musa
genotypes was carried out from 1991 to 1994 in three representative locations in the Humid Forest (Onne, Nigeria and M'Balmayo, Cameroon) and the Forest-Savanna (Ibadan, Nigeria) zones of the West and Central Africa. The main objective was to evaluate the performance of improved Musa
germplasm under different agro-ecological conditions, thereby assissing the genotype-by-environment interaction (GxE) for specific traits and yield stability. Hybrids, their parental genotypes together with plantain and banana landraces, were included in the trail for comparison purposes. Differences between the genotypes and environments were significant for all traits. Number of days to fruit filling was the only trait not affected by location effects. Due to a large number of fruits and high fruit weight, the improved germplasm showed heavier bunches than their parents, even though they had fewer hands. The GxE affected all traits, except fruit circumference. Genotype-by-location effects were significant for bunch weight, number of h ands, number of fruits and fruit weight. Most of the traits were not affected by genotype-by-cycle interaction at Onne and at Ibadan. These observations suggest that multilocational trails may be more efficient than signle site trials over several years. Stability analysis of bunch weight and yield potential, based on the phenotypic coefficient of variation, allowed the identification of high and stable yielding genotypes, e.g. the black sigatoka resistant hybrids TMPx 1658-4 and TMPx2796-5.