Studies of tree-to-tree variation in fruit traits are a pre-requisite for cultivar development. Fruits were collected from each of 63 marula ͑Sclerocarya birrea͒ trees in Bushbuckridge, South Africa and from 55 trees from the North Central Region of Namibia. The South African trees were in farmers' fields, communal land and natural woodland, at three sites: Acornhoek road, Allandale/Green Valley and Andover/Wits Rural Facility. The Namibian trees were all from farmers' fields in three areas: North east, North west and West. The fruits were partitioned into skin and flesh/juice to examine the extent of the variation found in different components of marula fruits from different trees. Namibian fruits were significantly larger than those from South Africa ͑26.7 vs 20.1g͒, due to their greater pulp mass ͑22.2 vs 16.2 g͒, especially the flesh/juice component. In South African fruits, those from farmers' fields were significantly larger in all components ͑Fruit mass ϭ 23.6 vs 19.3 and 18.0g in natural woodland and communal land respectively͒. In Namibia, mean fruit mass did not differ significantly across sites ͑25.5 Ϫ 27.0g͒. However, within each sample there was highly significant and continuous variation between trees in the pulp ͑S Africa ϭ 7.5 Ϫ 31.3g; Namibia ϭ 8.3 Ϫ 36.0g͒ and flesh/juice mass ͑S Africa ϭ 2.2 Ϫ 7.6g; Namibia ϭ 3.8 Ϫ 22.6g͒, indicating the potential for selection of trees producing superior products. The fruits of the Namibian trees were compared with the fruits from one superior tree ͑'Namibian Wonder'͒ with a mean fruit mass of 69.9g. The percentage frequency distribution of fruit mass from trees in farmers' fields in South Africa was skewed, while being bimodal in North east and North west populations from Namibia, suggesting that at these sites farmers are engaged in domestication through truncated selection of the best mother trees. It is concluded that there are trees in on-farm populations that have great potential to be propagated vegetatively as selected cultivars.

As part of a wider study characterizing tree-to-tree variation in fruit traits as a pre-requisite for cultivar development, fruits were collected from each of 63 marula ͑Sclerocarya birrea͒ trees in Bushbuckridge, South Africa and from 55 trees from the North Central Region of Namibia. The nuts were removed from the fruit flesh, and the kernels extracted, counted and weighed individually to determine the patterns of dry matter partitioning among the nut components ͑shell and kernel͒ of different trees. Mean nut, shell and kernel mass were not significantly different between the two countries. Between sites in South Africa there were highly significant differences in mean nut mass, shell mass, kernel mass and kernel number. In Namibia, there were highly significant differences between geographic areas in mean shell mass, kernel mass and kernel number, but not in nut mass. These differences had considerable impacts on shell:kernel ratios ͑8.0 Ϫ 15.4͒. In South Africa, mean kernel mass was significantly greater in fruits from farmers' fields ͑0.42 g͒ than from communal land ͑0.30 g͒ or natural woodland ͑0.32 g͒. Within all sites, in both South Africa and Namibia, there was highly significant and continuous variation between individual trees in nut mass ͑South Africa ϭ 2.3 Ϫ 7.