A prospective reproductive health surveillance program of semen quality among workers in a chemical facility that until late 1977 had produced 1,2-dibromo-3-chloropropane (96128) (DBCP), was conducted from November 1981 to October 1985. All male employees of the agricultural chemical division of the plant (97 men) were required to participate. Semen was evaluated monthly for the first 3 to 6 months of assignment to the division, with reevaluation after 12 consecutive months. Thirty one of the 97 participants were azoospermic, 21 due to voluntary sterilization, six identified in 1977 as the probable result of DBCP exposure, and four for unknown causes. Of the 66 nonazoospermic participants, 36 were unexposed to DBCP, 12 had uncertain exposure, and 18 had potential exposure of up to 9 years. The nonazoospermic semen was analyzed for sperm concentration, semen volume, total sperm number, percent motile, progressiveness, mean swimming speed, percent normal morphology, and head dimensions. Workers with definite prior DBCP exposure had lower sperm counts (84x10(6) versus 124x10(6)) and lower semen volume (1.6 milliliters (ml) versus 2.15ml) than workers with no prior DBCP exposure, but the differences were not statistically significant. Time between ejaculation and analysis was a statistically significant predictor for all three movement characteristics. Except for an isolated significant increase in sperm concentration with age, no other predictors were significantly associated with any semen outcome variable. The standard deviations of the measured values for an individual were higher than in ejaculates from fertile artificial insemination donors, probably because of less well controlled collection conditions in the workplace setting. The authors conclude that one ejaculate per worker is sufficient for cross/sectional studies, but that even three ejaculates may not provide enough precision for estimating an individual worker's mean.