Population-based cancer registries are widely used to provide key information about cancer incidence, survival, determinants of progression and clues about pathogenesis. To substantially expand the limited data on diagnostic accuracy and completeness for lymphoproliferative (LP) tumors in such databases, we conducted a retrospective investigation of close to 1,000 cases diagnosed during 4 decades in Sweden. We identified and reviewed medical records for 494 LP tumor patients reported to the population-based Swedish Cancer registry and 503 LP tumor patients identified from hospitals in Sweden among patients with LP tumors diagnosed during 1964–2003. The stratified samples were randomly selected from patients according to LP subtype, over 4 equal calendar periods, and among 6 selected hospitals of diverse size and from different geographic regions. We found 97.9% of reported LP tumor cases to fulfill current diagnostic criteria for having an LP tumor and observed 89.7% to have accurate LP tumor subtype. The overall completeness of non-Hodgkin lymphoma, Hodgkin lymphoma and multiple myeloma cases in the Cancer registry was 95–99% but was lower for the more indolent tumors, chronic lymphocytic leukemia (87.9%) and Waldenström's macroglobulinemia (68.1%). We observed increased overall under-ascertainment for patients diagnosed above age 80 (27%) and among individuals diagnosed before 1973 (12%). In conclusion the diagnostic accuracy and completeness for classically defined LP tumor entities in the Swedish Cancer registry is high. However, we found under-ascertainment of patients with indolent LP tumors, particularly among patients diagnosed at older ages, with early-stage disease and diagnosed in earlier years. © 2007 Wiley-Liss, Inc.