Crosses between certain genotypes of common bean result in dwarfing of F1 plants and lethal dwarfing in a proportion of the F2 population. This is under the control of the semi-dominant alleles, DL1 and DL2 at two complementary loci which are expressed in the root and shoot respectively. The various DL genotypes can be simulated by grafting. The graft combination DL1DL1dl2dl2/dl1dl1DL2DL2 was found to have a significantly higher root dry matter fraction than either parent. Lethally dwarfed plants (DL1DL1DL2DL2) and the analogous lethal graft combination (dl1dl1DL2DL2/DL1DL1dl2dl2) exhibit failure of root growth and have very low root fractions. Hybrids or graft combinations with failed roots ceased growth and accumulated large amounts of starch throughout their hypocotyls. In sterile culture, both lethal dwarfs and lethal graft combinations were able to grow roots if sucrose was added to the growth medium. This indicates that a failure of sucrose translocation to the roots is probably responsible for failed root growth. Data from screening the DL genotypes of 49 cultivars could be fully explained using the DL system hypothesis, and grafting proved to be efficient for identifying DL genotype. The DL system might be of fundamental importance in root–shoot partitioning. Current evidence favours the hypothesis that failure of root growth is the outcome of excessively high sink strength of shoots compared to roots, which might arise from signalling incompatibilities between the genotypes.