DNA fingerprinting of Mycobacterium tuberculosis is used to study the epidemiology of tuberculosis, but the specificity of the widely used IS6110 technique has not been validated. Isolates from Denver, Colorado from December 1988 through June 1994 were fingerprinted with the IS6110 technique. Available records were reviewed for patients whose isolates were within IS6110-defined clusters, and these isolates were fingerprinted with an independent technique (pTBN12). Of 189 isolates, 86 (46%) were in IS6110-defined clusters. Clustering was inversely related to the number of copies of IS6110, ranging from 12 of 12 (100%) to 37 of 48 (77%) and 37 of 129 (29%) for isolates having one, two to five, and more than five copies (p < 0.001). Of the 86 isolates clustered with the IS6110 technique, 35 (41%) had unique pTBN12 fingerprints. Discordant results with the two fingerprinting techniques were more common among isolates having five or fewer copies of IS6110. Epidemiologic links were identified among four of 35 (11%) patients whose isolates had discordant fingerprinting results, as compared with 40 of 51 (78%) of those whose isolates matched by both IS6110 and pTBN12. DNA fingerprinting with the IS6110 technique was not a specific marker of DNA clonality, particularly among isolates having fewer than five copies of IS6110. The use of a supplemental DNA fingerprinting technique decreased clustering and improved the correlation between the transmission links predicted by molecular techniques and epidemiologic investigation.