In 1992–1993, we investigated possible cross-contamination of Mycobacterium tuberculosis cultures as part of a study of tuberculosis in Arkansas by using DNA fingerprint analysis. Of patients whose isolates were matched, those for whom smears were negative and only one culture was positive were identified from laboratory records. Clinical, laboratory, DNA fingerprint, and epidemiological data were reviewed. Of 259 patients, nine (3.5%) were judged to be due to cross-contamination. None of these patients had a clinical course consistent with tuberculosis. All nine specimens were processed with another isolate with a matching DNA fingerprint, and epidemiological connections were not identified among any of the patients. To avoid erroneous diagnoses and unnecessary therapy and public health investigations, specimens from patients with tuberculosis whose smears are negative and only one culture is positive should be investigated for cross-contamination. An inconsistent clinical course and a DNA fingerprint that matches those of other culture-positive specimens processed concurrently, coupled with the lack of an epidemiological connection, suggest cross-contamination.