The aim of this study was to clarify the reproducibility of the Japanese Ministry of Health, Labor and Welfare (JMHLW) type classification for osteonecrosis of the femoral head. We performed inter-observer and intra-observer trials using 40 sets of magnetic resonance imagings, 20 of which were produced by a 0.5 Tesla (T) superconductive unit and the other 20 produced by a 1.5 T unit, in patients with non-collapsed and asymptomatic osteonecrosis of the femoral head (JMHLW stage 1 or 2). The JMHLW type classification (A, B, C1, or C2) was determined from T1-weighted coronal images at the center of the femoral head. Six orthopedic surgeons independently assessed all 40 images twice, with an interval of 4-5 weeks between sessions. Regarding inter-observer reliability, the percent agreement was 85% and weighted kappa was 0.709 for 0.5 T, versus a percent agreement of 82% and weighted kappa of 0.724 for 1.5 T. Regarding intra-observer reliability, the percent agreement was 82% and weighted kappa was 0.780 for 0.5 T versus a percent agreement of 80% and weighted kappa of 0.800 for 1.5 T. Inter-observer and intra-observer reliabilities did not differ significantly between the 0.5 and 1.5 T units. The JMHLW type classification provided high inter-observer and intra-observer reliabilities.