In this study we investigated the within- and between-population genetic variation using microsatellite markers and quantitative traits of the shea tree, Vitellaria paradoxa, an important agroforestry tree species of the Sudano–Sahelian region in Africa. Eleven populations were sampled across Mali and in northern Côte d’Ivoire. Leaf size and form and growth traits were measured in a progeny test at the nursery stage. Eight microsatellites were used to assess neutral genetic variation. Low levels of heterozygosity were recorded (1.6–3.0 alleles/locus; HE = 0.25–0.42) and the fixation index (FIS = −0.227–0.186) was not significantly different from zero suggesting that Hardy–Weinberg equilibrium is encountered in all populations sampled. Quantitative traits exhibited a strong genetic variation between populations and between families within populations. The degree of population differentiation of the quantitative traits (QST = 0.055–0.283, QSTmean = 0.189) strongly exceeds that in eight microsatellite loci (FST = −0.011–0.142, FSTmean = 0.047). Global and pairwise FST values were very low and not significantly different from zero suggesting agroforestry practices are amplifying gene flow (Nm = 5.07). The population means for quantitative traits and the rainfall variable were not correlated, showing variation was not linked with this climatic cline. It is suggested that this marked differentiation for quantitative traits, independent of environmental clines and despite a high gene flow, is a result of local adaptation and human selection of shea trees. This process has induced high linkage disequilibrium between underlying loci of polygenic characters.