I studied whether changing a partner, and thus changing the likelihood of human leukocyte antigen (HLA) sharing between mating partners, affects the risk of preterm delivery in the subsequent pregnancy. I identified a total of 128,239 women who had two consecutive births during 1989-1991 through data linkage of the California birth certificates. Paternal date of birth and names on the records of the two consecutive births were compared to determine whether the same father was reported on both records. Three cohorts of women were formed on the basis of the gestational age of their first delivery: <34, 34-36, and >36 weeks. If parental HLA sharing is associated with preterm delivery, the likelihood of HLA sharing was expected to be in a decreasing order from most likely among a <34-week cohort to least likely among a >36-week cohort. Among women in the <34-week cohort, changing partners resulted in a 33% reduction in the risk of early preterm delivery in the subsequent pregnancy compared with those who did not change partners [95% confidence interval (CI), 0.52-0.88]. In contrast, among women in the >36-week cohort, changing partners led to a 16% increase in the risk of early preterm delivery in the subsequent pregnancy (95% CI = 1.04-1.30). Among women in the 34-36-week cohort, changing partners did not affect the risk of preterm delivery (95% CI = 0.78-1.25). These estimates were adjusted for maternal race/ethnicity, age, educational level, prenatal smoking, prenatal care, parity, and interval from birth to conception of the subsequent pregnancy. The findings from this study suggest that the effect of changing paternity depends on the pregnancy outcome with the previous partner and support the hypothesis that parental HLA sharing may be related to preterm delivery.