Ugandan smallholder farmers produce the nation's major food crop using numerous banana varieties with distinctive attributes, while coping with important biotic constraints and imperfect markets. This empirical context motivates a trait-based model of the agricultural household that establishes the economic association between household preferences for specific variety attributes (yield, disease and pest resistance, and taste), among other exogenous factors, and variety demand, or the extent of cultivation. Six variety demands are estimated in reduced form, each in terms of both plant counts ("absolute" or levels demand) and plant shares ("relative" demand). Two salient findings emerge from the analysis: 1) the determinants of both absolute and relative demands are variety-specific and cannot be generalized across groups of cultivars; and 2) the determinants of absolute and relative demand are not the same in sign or significance. These findings raise questions about commonly used econometric specifications in the adoption literature. Grouping varieties together masks individual differences, and differences may be important for predicting the adoption of new technologies such as genetically transformed, endemic or local varieties. The development of methods to estimate a complete variety demand system might permit resolution of cross-variety relationships. The purpose of this research is to contribute information of use in identifying suitable local host varieties for the insertion of resistance traits through genetic transformation, and the factors affecting their potential adoption.