We have prospectively examined 46 consecutive unrelated children with JRA, 28 females and 18 males, for HLA‐A, ‐B, ‐Dw1, and ‐Dw3 antigens. The data were analyzed according to mode and age of onset, as well as presence of extraarticular features. Patients with JRA were found to have an elevated frequency of HLA‐B27 (26% vs. 8% of controls, P < 0.01). This elevated frequency of HLA‐B27 was noted in patients with either a pauciarticular or polyarticular onset and consistent with previous observations, was more common in males than females. However, the most striking associations in JRA were apparent when the MLC‐defined locus HLA‐Dw3 was studied. The total patient population had a 46% frequency of HLA‐Dw3 as compared with 24% in controls (P < 0.01). Moreover, when the group of 46 children was subdivided, the HLA‐Dw3 frequency of patients with polyarticular disease was 64% as compared with 21% in pauciarticular disease (P < 0.01). Further, although the frequency of HLA‐B8 was similar in JRA to that in the control panel, a significant degree of linkage disequilibrium was shown to exist between HLA‐B8 and HLA‐Dw3. No such linkage was shown to exist between HLA‐B27 and either HLA‐Dw1 or Dw3. Although we have not multiplied the P value by the number of HLA‐A, ‐B and ‐D antigens, we suggest that the presence of the MLC‐defined locus HLA‐Dw3 is significantly elevated in children with polyarticular mode of onset of JRA and suggest that further definition of lymphocyte‐dependent loci may aid in understanding the mechanism of pathogenesis as well as the classification and prognosis of childhood onset arthritis.