The effects of cassava mosaic virus disease (CMD) on yield in fully and partly infected stands of cassava were investigated in field trials in Uganda in 1990-91 and 1991-92. Three cultivars (Ebwanateraka, Bao and Bukalasa 1 l), each at three levels of cutting infection (O%, 50% and 100%) and harvested 510 and 15 months after planting (MAP) were used in a randomised block design with split-split plots and four replicates. Moreover, yield and growth data for individual infected and uninfected plants were considered in relation to the health status of their nearest neighbours. In each experiment, fresh tuberous root yields of plants from 100% infected plots gave sigdicantly lower yields than those from 0% or 50% infected plots at each harvest date and the losses were greatest in cv. Bao. Yields of plants from 0% and 50% plots for each of the three cultivars were not significantly different, 10 and 15 MAP. The loss in yield differed between cultivars and harvest dates. Fresh stem, leaf and root yields and the number of tuberous roots were influenced by the health status of the plants harvested and that of their nearest neighbours. Uninfected plants surrounded by infected ones had more roots and heavier total fresh root, stem and leaf weights than those surrounded by uninfected ones. Overall, 26% and 42% compensation was recorded in 1990-91 and 1991-92, respectively. The effects of CMD on cassava production and of compensation in mixed stands of infected and uninfected plants are discussed, especially in relation to control strategies such as roguing.