Randomness of fertilization was studied in an open-pollinated population of maize (Zea mays L.) through allozyme assays of seedlings from open-pollinated seeds produced on both tasseled and detasseled plants. Mixed-mating-model estimates of the amount of outcrossing (t) were not significantly different from t = 1.00 for four enzyme loci ( Adh1, Idh2, Got1 and Acp1), indicating that fertilizations were at random in the population. However, for loci Prx1 and Est4 , estimates of t were significantly smaller than unity—0.80 and 0.70 for tasseled plants and 0.81 and 0.80 for detasseled plants. The excesses of homogametic fertilizations detected on the detasseled plants could not have been due to selffertilization, s = 1 - t, because the detasseled plants shed no pollen. Analyses of allelic frequencies in the pollen that produced seed on the detasseled plants established that different maternal plants sampled genetically different populations of pollen from the outcross pollen pool. It was suggested that the causes of the differential sampling were temporal variation in the pollen pool, and/or gametophytic selection, correlated with marker-locus genotype. Two-, three- and four-locus interactions among the marker loci were often statistically significant, indicating that the factors responsible for the nonrandom gametic unions observed in the maize population studied were complexly interactive.