Human spermatids from ejaculate and testicular tissue have been utilized for evaluating human fertilization by intracytoplasmic sperm injection (ICSI) and, where possible, compared with spermatozoa utilizing sibling oocytes. Round and elongated spermatids obtained from ejaculates were either prepared through Percoll gradients or isolated and washed individually using subzonal insemination needles (SUZI; 10-14 microm internal diameter). Seminiferous tubules obtained after biopsy were placed into HEPES-buffered Earle's medium and dissected using 21-gauge needles. Spermatogenic cells and spermatozoa were isolated and washed individually using SUZI needles. Spermatozoa were subsequently injected into the ooplasm using 5 microm (internal diameter) ICSI needles, whereas 8-9 microm (internal diameter) needles were used for spermatid injection. Only metaphase II oocytes (n = 207) were injected: 64 with round spermatids, 92 with elongated spermatids and 51 with spermatozoa; the fertilization rate was 30, 24 and 67% respectively. There was a significant (P < 0.001) increase in the fertilization rate using spermatozoa compared with spermatids. The fertilization rate was not different between round and elongated spermatids, although the fertilization rates for round and elongated spermatids in the ejaculate were 33 and 18% respectively, compared with 22 and 38% respectively when testicular spermatids were utilized. In three patients sibling oocytes were used to compare round and elongated spermatids found in the ejaculate with spermatozoa extracted from seminiferous tubules. The fertilization rate was 24% for spermatids and 79% for testicular spermatozoa. This result suggests that, should only spermatids be available in the ejaculate, a testicular biopsy in the hope of obtaining testicular spermatozoa would be worth while.