The efficiency of spermatozoa prepared by swim-up or by Percoll centrifugation was assessed in an in-vitro fertilization programme on 71 semen samples of a well-defined quality [total number of type A (WHO criteria) motile spermatozoa]: category I (n = 21) with > 100 x 10(6), II (n = 31) with 15-100 x 10(6), III (n = 11) with 5-15 x 10(6) and IV (n = 8) with < 5 x 10(6) type A motile spermatozoa. Oocytes were inseminated 4 h after oocyte retrieval, alternately with spermatozoa derived from swim-up and Percoll preparation. Both selection procedures resulted in a significantly higher (P < 0.001) percentage motility as compared to fresh semen. For low-quality samples (III and IV), however, swim-up was more effective in selecting highly motile (P = 0.004) and morphologically normal spermatozoa (P < 0.05). For high-quality samples, this difference might have been masked by introducing a swim-up step to remove Percoll particles. Regardless of the initial sperm quality, the mean fertilization rate was significantly higher (P = 0.003) when Percoll-treated spermatozoa were used for insemination (51.3 versus 37.8%). For semen of groups I and II, no difference in fertilization capacity was observed according to the sperm preparation method. Despite the lower percentage motility and normal morphology for the Percoll compared to the swim-up treatment in groups III and IV, fertilizing capacity was significantly (P < 0.001) in favour of this selection method (65.3 versus 26.5% in group III, 47.6 versus 11.6% in group IV).(ABSTRACT TRUNCATED AT 250 WORDS)