1. Control of zoonotic disease is difficult to achieve when populations of multiple hosts, particularly wildlife, become persistently infected. Bovine tuberculosis (TB) is one such disease: its causative agent, Mycobacterium bovis, infects cattle, humans and multiple wildlife species including European badgers Meles meles. 2. In Britain, from 1974 to 1998 various strategies for the control of cattle TB involved culling badgers in the immediate vicinity of TB-affected herds. However, patterns of association between cattle and badgers had not been investigated at a local scale. 3. Using data from the Randomized Badger Culling Trial, an ongoing large-scale study of TB dynamics and control, we investigated local geographical associations between M. bovis infection in badgers and cattle. 4. Mycobacterium bovis infections were locally clustered within both badger and cattle populations. 5. We show, for the first time, that M. bovis infections in badgers and cattle are spatially associated at a scale of 1-2 km. Badgers and cattle infected with the same strain type of M. bovis are particularly closely correlated. These observational data support the hypothesis that transmission occurs between the two host species; however, they cannot be used to evaluate the relative importance of badger-to-cattle and cattle-to-badger transmission. 6. Synthesis and applications. The close associations between M. bovis infections in cattle and badgers suggest that localized badger culling could reasonably be expected to reduce the risks of cattle TB infection; however, experimental culls have found no such beneficial effects over the time-scale on which they were tested. This demonstrates the difficulty of predicting the outcome of management interventions, and reinforces the need for well-designed empirical assessments of future control strategies.