Yield improvement for smallholder farmers in developing countries via crop breeding has been slow, in part because breeders have focused primarily on higher yields rather than stability of yield across environments. We investigated the relationships between (i) grain yield and yield-component traits, (ii) grain yield and yield stability, and (iii) reproductive (R) and vegetative (V) biomass in a field experiment with 18 spring wheat genotypes varying from landraces to recently-released breeding lines in a semiarid region under three different levels of precipitation. Genotypic differences in grain yield varied with the environment. The relationship between grain yield and yield components was also dependent on the environment, making it difficult to identify early-screening parameters for use in breeding programs. A genotype main effects and genotype by environment interaction effect analysis (GGE) showed a negative relationship between mean grain yield and yield stability across environments. There were significant differences among genotypes in the allometric exponent of the R-V relationship (slope of the log R – log V relationship). For the three highest yielding but less stable genotypes, this slope value was > 1, while the five lowest yielding, but more stable genotypes, it was < 1. We hypothesize that the tradeoff between grain yield and yield stability reported here is biologically constrained and cannot be eliminated through breeding. Therefore, a balance between yield stability and high yield should be considered in future wheat breeding programs for smallholder farmers, rather than attempting to improve both.