Between August 1987 and July 1991, 22 patients with acute nonlymphocytic leukemia have received allogeneic bone marrow transplantation (BMT) with non-T-lymphocyte-depleted marrow obtained from matched sibling donors. Of these patients, 12 patients were in first complete remission (CR) and 10 patients in second CR or greater or in relapse. All patients were treated with a preparative regimen consisting of cyclophosphamide (CTX, 60 mg/kg) or combined drugs, and 850 cGy single-dose or $150~200$ cGy fractionated total body irradiation (TBI) administered twice daily for a total dose of $1200~1320$ cGy. Survivors have been followed from 8 to 64.5 months (median, 24 months). The overall 2 year survival rate, relapse rate and incidence of radiation pneumonitis and graft versus host disease (GVHD) have been evaluated by age, phase of disease, initial WBC count, modality of TBI or conditioning chemotherapy. Overall 2 year survival was $58{\%}$. The median survival was 31 months and mean survival was 23.2 months. Overall survival have significant impact in patients of age >19 years old (p=0.008), patients in first CR (p=0.09). Two year survival rate is significantly correlated with age ( >19 vs $leqq$19, $79.4\%$ vs $14.3\%$, p=0.0008), regimen of chemotherapy (CTX vs combined drug, $76.9\%;vs;33.3\%$, p=0.04), phase of disease (1st CR vs geqq2nd$ CR or relapse, $83.3\%;vs;30\%$, p=0.01) and method of TBI (fractionated vs single dose, $70.7\%;vs;37.5\%$, p=0.05). The influence of French-American-British (FAB) subtypes on relapse rate is not significant, but initial WBC count > 20000/$mm^3$ is associated with increased relapse rate. There is difference in the rate of radiation pneumonitis ($14.3\%;vs;25\%$), GVHD ($14.3\%;vs;50\%$) and relapse ($21.4\%;vs;50\%$) according to fractionated versus single-dose TBI. As mentioned above, fractionated TBI is compatible for the preparative regimen combined with chemotherapy En allogeneic BMT of first CR patients under 41 years of age with suitable donor. Those results from a retrospective, non-randomized study clearly need additional clinical data, ideally from a randomized study.