Two hundred and forty seedlings from each of three diverse populations were measured for coleoptile length. Significant differences between populations were found for mean coleoptile length. Considerable variability was obtained within each population with coefficient of variation values ranging from 32 to 42%. Forty seedlings from each population were placed into one of three selection groups: random (0); shortest 20% (S20); and longest 20% (L20). The plants within each group in each population were allowed to interpollinate. The coleoptile length of 100 seedlings from each group obtained from the interpollination was then measured. This experiment was repeated in each of two successive years. Positive response to selection was found in all populations in both years. The means of the L20 selection group were significantly higher than the means of the 0 group in 6 out of 6 comparisons. The means of the S20 group were significantly lower than the means of the 0 group in only 2 out of 6 comparisons. Selection response values also reflected these results. Therefore, selection for long coleoptile length was considerably more effective and consistent than selection for short coleoptile length. These results suggest that short coleoptile length may be partially dominant to long coleoptile length. No significant correlation between seed weight and coleoptile length was found indicating that selection for coleoptile length on the basis of seed weight would probably be ineffective.