Environmental pollution is traditionally classified as either localized or diffuse. Local pollution comes from a point source that emits a well-defined cocktail of chemicals, distributed in the environment in the form of a gradient around the source. Diffuse pollution comes from many sources, small and large, that cause an erratic distribution of chemicals, interacting with those from other sources into a complex mixture of low to moderate concentrations over a large area. There is no good method for ecological risk assessment of such types of pollution. We argue that effects of diffuse contamination in the field must be analysed in the wider framework of stress ecology. A multivariate approach can be applied to filter effects of contaminants from the many interacting factors at the ecosystem level. Four case studies are discussed (1) functional and structural properties of terrestrial model ecosystems, (2) physiological profiles of microbial communities, (3) detritivores in reedfield litter, and (4) benthic invertebrates in canal sediment. In each of these cases the data were analysed by multivariate statistics and associations between ecological variables and the levels of contamination were established. We argue that the stress ecology framework is an appropriate assessment instrument for discriminating effects of pollution from other anthropogenic disturbances and naturally varying factors. © 2008 Elsevier B.V. All rights reserved.