Different procedures were investigated for the dilution of human cryopreserved semen and the preparation of an enriched population of motile spermatozoa for assisted reproduction. The dilution of a 0.25 ml straw of cryopreserved human semen by addition of 2.0 ml Ham's F-10 buffer in one step caused a large decrease in the proportion of motile spermatozoa. This was due to osmotic stress because many of the diluted spermatozoa exhibited swollen tails. To a large extent the damage could be avoided by adding the buffer in 0.10-ml aliquots at 30-s intervals. Spermatozoa obtained after such dilution of cryopreserved human semen were subjected to the swim-up procedure, to centrifugation on two-step gradients of Nycodenz or Percoll, or to filtration through glass fibre paper and compared with respect to yield, motility parameters and penetrating ability in the hamster egg test. The swim-up procedure yielded spermatozoa with excellent motility but only 12% of the available motile spermatozoa were recovered. On both Nycodenz and Percoll gradients, > 40% of the available motile spermatozoa were recovered and the average velocity of the spermatozoa was not significantly less than for the swim-up technique. When A23187 was used to promote acrosome reactions in the hamster egg test, Percoll-prepared spermatozoa achieved an average of 8.6 decondensed sperm heads/egg compared to 1.9 for Nycodenz and 1.3 for the swim-up procedure. The yield from glass fibre paper filtration was only 12% and the velocity of the spermatozoa and their performance in the hamster egg test was significantly poorer than in all the other methods. We conclude that cryopreserved semen must be diluted with great care and that centrifugation on two-step Percoll gradients is an effective method to separate fertile spermatozoa from this semen for use in clinically assisted reproduction. © 1992 Oxford University Press.