Recent years have witnessed important advances in the analysis of spatially referenced health data. This paper uses GIS and point pattern modelling to address previously unanswered questions regarding the spatial epidemiology of Legionnaires' disease. We demonstrate a distance effect for proximity of residence to cooling towers; mixed support for a directional effect; and some evidence relating to multiple sources. In uncovering complex conceptual and technical problems in the spatial modelling of infection risk we also extended the limits of existing point pattern techniques. We advocate further multidisciplinary research to advance methodological developments for understanding spatial environment-health relationships.