The relationship between different levels of pollen production and percentage crossing in alfalfa (Medicago sativa L.) was studied in both hand‐crosses in the greenhouse and bee‐pollinations in the field. White flower color was used as a genetic marker to determine the amount of crossing. Pollen production was rated in 4 classes: 1 = no pollen, 2= trace of pollen, 3 = moderate amount of pollen, and 4 = much pollen dehisced when the flower was tripped.Greenhouse studies indicated that the amounts of sibbed and selfed (nonhybrid) seed were affected by the amount of pollen production in the male‐sterile seed parents. Three‐way random crosses among white‐flowered complete male steriles, white‐flowered sibling partial steriles, and unrelated purple‐flowered fertile plants were used to simulate pollination opportunities within a three‐way alfalfa hybrid. The type of seed produced on class 1 plants was 0% self, 2% sib, and 98% hybrid. The amount of self, sib, and hybrid seed produced on class 2, 3, and 4 plants were 3, 5, and 92%; 12, 27, and 61%; and 18, 40, and 42%, respectively.Field studies were also conducted in which 36 crosses, segregating for varying degrees of male sterility, were assigned a pollen‐production index (PPI) by weighting each pollen‐dehiscence class according to its pollination effectiveness. The weighing factors for classes 1, 2, 3, and 4 were 0.0, 0.1, 0.6, and 1.0, respectively. PPI was negatively correlated with percent crossing for 2 years in a field test. The interaction between years and PPI did not significantly affect, crossing. PPI was significantly correlated with seed yield per plant in 1970, but not in 1971. The equation Ŷ = 98 −0.56x, where Ŷ = predicted percent hybrid seed and x = PPI, was arrived at for the prediction of the amount of pollen dehiscence allowable in the male‐sterile component of an alfalfa hybrid. A hybrid with 95% hybrid seed could not have a PPI of more than 5.4. A hybrid with 75% hybrid seed could not have a PPI of more than 41.1.