Given that seeds fertilized by slowly growing pollen are of low quality genetically, we theoretically reanalyzed the hypothesis that plants selectively abort ovules fertilized later to enhance the mean quality of resulting seeds. We assumed that both superior and inferior pollen exist, the superior pollen growing faster to fertilize ovules, resulting in seeds of higher quality than those of ovules fertilized by inferior pollen. We developed two models to determine the conditions under which selective abortion is favored. In the first model, ovules in one flower are fertilized by pollen grains that arrive at different times, with each visit bringing both fast- and slow-growing pollen. In the second model, ovules in two flowers are fertilized by all pollen grains that arrive at the same time. In the first model, we found that selective abortion based on the order of fertilization is never advantageous irrespective of the duration of the time lag between the two visits. Rather, random abortion is possibly favored. In the second model, although selective abortion based on the order of fertilization can be advantageous, the parameter region favoring it is rather restricted. This is because over production can be advantageous only if the quantity of the superior pollenis not limited in one flower but is limited in the other flower. In addition, the degree of overproduction was very low, implying that the merit of overproduction (increase in the number of superior seeds) is low compared to the cost of overproducing ovules. These results suggest that selective abortion of ovules based on the order of fertilization is not as advantageous as previously considered. (C) 2009 Elsevier Ltd. All rights reserved.