OBJECTIVE To report the experience in sperm extraction from testicular biopsies (TESE) performed from March 1996 to July 1998 in men with non-obstructive azoospermia (NOA). METHODS Comparisons of age, volume of both testes, serum FSH and testosterone in men, and histology of testicular samples in the first cycles between cycles with spermatozoa found and those without spermatozoa found were performed. Comparisons of fertilization, cleavage and pregnancy rates between cycles with spermatozoa injected and those with spermatids injected were performed. RESULTS Spermatozoa were found in only 12 out of 26 first TESE cycles (46.2%) and other cycles had spermatids (round cells) only. Age of men, history of mumps orchitis/oligozoospermia, volume of both testes and serum FSH/testosterone levels in men were not significantly different between cycles with and without spermatozoa. The fertilization rate was significantly higher in cycles with spermatozoa injected than those with round cell injections (63.3% vs 23.2%, P < 0.0001, Chi-squared test). The pregnancy rate was 14.3% per cycle when spermatozoa were injected. CONCLUSIONS TESE followed by intracytoplasmic sperm injection (ICSI) is an effective treatment in patients with NOA. Less than half of the patients undergoing TESE had spermatozoa recovered. Age of men, volume of both testes and serum FSH/testosterone levels in men were not useful in predicting successful recovery. Compared to using ejaculated and epididymal spermatozoa, fertilization and pregnancy rates were achieved when testicular spermatozoa were used for ICSI.