Independent review of slides of 668 cases of non-Hodgkin's lymphoma (NHL) by a panel of 4 experienced pathologists using the Working Formulation (WF) allowed determination of the agreement between reported diagnoses and panel review of slides. The panel agreed with the reported diagnosis of NHL in 93% of cases, but with the NHL subtype in only 55% of cases overall. The ability of the panelists to agree among themselves, however, was only slightly better than the panelists' agreement with the reported diagnosis (60% vs. 54%, respectively). Agreement of the panel with the reported subtype diagnosis varied from 14% to over 90%. The best agreement was with small lymphocytic lymphoma and follicular subtypes. Conclusions from this study are: 1) The WF functions well as common language for translation and comparison of diagnoses of subtypes of NHL. 2) Relative to time and cost involved, panel review using only light microscopy may not be useful in epidemiologic studies of NHL. 3) Small lymphocytic and follicular subtypes of NHL can be used more confidently in epidemiologic studies than can other subtypes whether the subtyping is done from abstracted reports or by panel review.