The association of tuberculosis (TB) with poverty has long been recognized, yet it may reflect not only characteristics of poor individuals, but also housing and neighborhood features which promote airborne spread. We sought to determine whether dwelling and building features, residential density and crowding are independently associated with TB occurrence in a low-incidence setting. We used residential addresses to geocode active TB cases reported in Montreal in 1996-2000. These "case dwellings" were linked to the municipal dwelling geodatabase from 2000, and to Canadian census data from 1996. We compared them with randomly selected Montreal dwellings ("controls," in a 1:10 ratio), using the same data sources. From multivariate logistic regression, the 595 case dwellings were more likely than the 5950 control dwellings to be in buildings >5 stories tall (adjusted odds ratios [OR] 1.6; 95% CI: 1.0-2.5), constructed since 1970 (adjusted OR 2.5; 1.8-3.6), in the lowest quartile for resale valuation (adjusted OR 1.3; 1.0-1.6), and on blocks where lot coverage exceeded the median value (adjusted OR 1.3; 1.0-1.6). Case dwellings were also more often found in census tracts with more persons per room, and a higher proportion of inhabitants who had arrived in Canada within the last 5 years. We conclude that dwelling and building features-notably dwellings in taller and new buildings, with lower resale value, and dwellings on blocks with high residential density-as well as crowding, were associated with TB occurrence, after adjustment for sociodemographic factors.