The experimental group consisted of men from 81 couples waiting for in vitro fertilization (IVF), about half of whom had sperm dysfunction defined by a negative post-coital test. A diagnostic semen sample was subjected to a hamster oocyte penetration test (HOPT) after stimulation of the acrosome reaction with A23187 ± pentoxifylline and to computerized sperm motility measurements (CASA) as well as conventional semen analysis according to the WHO protocol. Logistic regression was used to identify parameters that predicted the probability of achieving four or more viable embryos at IVF among the 65 couples from whom four or more oocytes were collected. The number of oocytes available and whether the woman had previously been pregnant (ever pregnant) were important factors but once these had been taken into account a number of sperm parameters had additional predictive power. The most useful of these were the percentage sperm static (CASA) or the percent sperm progressively motile (conventional semen analysis) in the Percoll preparation. A model incorporating the number of oocytes collected, ever pregnant and percentage sperm static achieved 85% correct prediction of outcome in the experimental dataset but only 62% correct prediction in an independent set of 280 IVF cycles. The percentage of hamster oocytes penetrated was a significant predictor but had no advantage over simple motility measurements. The results illustrate the difficulty of basing a prognosis for achieving satisfactory fertilization in IVF on the properties of spermatozoa.