The study used DNA fingerprint typing (spoligotyping and Heminested-lnverse-PCR) of Mycobacterium tuberculosis from all culture-confirmed inner London patients over a 12-month period to describe transmission. The methodology was evaluated by comparison with standard IS6110 typing and by examining its ability to identify known household clusters of cases. Isolates sharing indistinguishable typing patterns using both techniques were defined as clustered. Clusters were investigated to identify epidemiological links. The methodology showed good discriminatory power and identified known household clusters of cases. Of 694 culture-confirmed cases, 563 (81%) were typed. Eleven (2%) were due to laboratory cross-contamination and were excluded. Of the remaining 552 isolates 148 (27%) were clustered. Multivariate analysis indicated that clustering was more common in those with pulmonary smear positive disease (P < 0.02); those born in the United Kingdom (P < 0.0003) and in patients living in south London (P = 0.02). There was also a trend towards clustering being more common in those not known to have HIV infection (P = 0.051). The results suggest that in inner London, recent local transmission makes an important contribution to notification rates.