We sought to determine whether nonrandom association of IS6110 with Mycobacterium tuberculosis could result in false-positive clustering in unselected collections of isolates. We typed 196 strains of M. tuberculosis from an unselected community-based study in northern Tanzania by IS6110 and polymorphic GC-rich repetitive-sequence (PGRS) methodologies. The strains were analyzed by Gelcompar computer software. Analysis of 13 out of 25 groups showed that isolates with identical IS6110 and PGRS patterns were likely to be the same strain. Some IS6110 groups containing strains with identical PGRS patterns had similar IS6110 patterns that differed only by movement of the element. Isolates assigned to a single group (i.e., group 11) on the basis of sharing an identical IS6110 fingerprint pattern did not share identical PGRS fingerprint patterns. Six out of the nine bands in these isolates were in hot-spot locations, as previously defined. This indicates that nonrandom association may result in false-positive clustering in unselected community-based studies. Only strains with identical PGRS and IS6110 patterns are likely to be recently transmitted.