*Stabilizing selection is a key evolutionary mechanism for which there is relatively little experimental evidence. To date, stabilizing selection has never been observed at the whole-genome level. *We tested the effect of selection on genome size in a field experiment using seeds collected in a population of Festuca pallens with a highly variable genome size. Using flow cytometry, we measured the genome size in germinating seedlings and juvenile plants grown with or without high intraspecific competition (908 individuals). Above-ground biomass and leaf number were used as measurements of individual vegetative performance. The possible confounding effect of seed weight was controlled for in a separate experiment. *Growth under high competition had a significant stabilizing effect on genome size. Because no relationship was observed between genome size and vegetative performance, we assume that the elimination of plants with extreme genome sizes was the result of decreased survival as a consequence of some unrecognized stress. *Our results indicate that genome size may be under direct selection. The equal disadvantaging of either large or small genomes indicates that the selection for optimum genome size in species may be fully context dependent. This study demonstrates the power of competition experiments for the detection of weak selection processes.