We investigated the response of maize kernel number to plant density in four hybrids released in Argentina between 1965 and 1993. Assuming kernel number is the main yield component, and using as a framework the relationship between kernel number per plant (KNP) and plant growth rate bracketing silking (PGR_s), we tested the alternative hypotheses that modern hybrids produce more kernels because they have (a) greater PGR_s or (b) more kernels per unit PGR_s than their older counterparts. Three experiments were carried out including a range of plant densities from 3–5 to 15–18 plantsm⁻². PGR_s was calculated from shoot dry matter measured 10 days before and 20 days after silking. Shoot dry matter, grain yield and its components were measured at physiological maturity. Grain yield of the oldest hybrid averaged 7.7tha⁻¹, and increased with year of release at a rate of 173kgha⁻¹ per year. The response of grain yield to plant density was curvilinear. Kernel number per square meter accounted for most of the variation in yield with both year of release and plant density. For both sources of variation, there was a trade-off between kernel number and mass. Both PGR_s and KNP decreased with increasing plant density in all four hybrids. Whereas variation in PGR_s among hybrids was small, the oldest hybrid set 93 (low density) and 113 (high density) kernels per unit PGR_s in comparison to the newest that set 167 and 193. We conclude that more kernels per unit PGR_s, rather than greater PGR_s, accounted for the genetic improvement of yield potential in the hybrids investigated.