Epidemiologic relatedness of Mycobacterium tuberculosis isolates from Arkansas residents diagnosed with tuberculosis in 1992-1993 was assessed using IS6110- and pTBN12-based restriction fragment length polymorphism (RFLP) and epidemiologic investigation. Patients with isolates having similar IS6110 patterns had medical records reviewed and were interviewed to identify epidemiologic links. Complete RFLP analyses were obtained for isolates of 235 patients; 78 (33%) matched the pattern of > or = 1 other isolate, forming 24 clusters. Epidemiologic connections were found for 33 (42%) of 78 patients in 11 clusters. Transmission of M. tuberculosis likely occurred many years in the past for 5 patients in 2 clusters. Of clusters based only on IS6110 analyses, those with > or = 6 IS6110 copies had both a significantly greater proportion of isolates that matched by pTBN12 analysis and patients with epidemiologic connections, indicating IS6110 patterns with few bands lack strain specificity. Secondary RFLP analysis increased specificity, but most clustered patients still did not appear to be epidemiologically related. RFLP clustering in rural areas may not represent recent transmission.