|dc.description.abstract||Farmers, researchers, extension officers and policy makers need reliable yield data on banana (Musa spp.) to make informed decisions. There are no standard yield determination methods available for Musa research. Production patterns and types vary distinctly in time and space. This requires special attention when expressing yields as unit mass per unit area per unit time (t ha-1 y-1). Determining banana yield is challenging because: (a) bunch yields are often expressed as fresh weight, but edible dry matter percentage can vary strongly between cultivars and environments; and (b) bunch yields often include the peduncle for which often no uniform cutoff point is used. To obtain realistic and valid yield data and to compare Musa yields with those of other crops, the edible dry matter content needs to be determined and considered. In addition, many cultivars have a highly variable rate of plants reaching flowering and bunch production. Too often, yield calculations are made with the assumption that 100% of the plants produce, whereas in many fields up to 70% of the plants do not produce, especially in lowland plantain fields. Plant densities in farmer fields are often highly variable, which requires careful consideration of (a) the definition of a banana plot; (b) the size of the banana plot; and (c) the number of plants in a banana plot. Often, banana clusters consist of several mats of the same generation, leading to underestimations of the number of mats per unit area. We propose that the definition of a single banana mat is a family of plants with interconnected living corms. Crop cycle duration varies strongly as a function of cultivar, altitude, soil fertility, drought stress, and pest and disease infestation. It can vary from less than one year to over two years, leading to large (>50%) yield differences per unit time between and within cultivars. This paper describes the sources of errors and proposes measures to avoid incorrect or incomplete yield determinations.