The potential for stressors to interact or elicit different responses depending on ecological setting can confound the use of contaminant guidelines or predictions from laboratory-based ecotoxicological studies. However, predicting and identifying multiple stressor effects is challenging because of the wide range of treatment combinations needed for laboratory experiments and the difficulty of conducting field experiments across a range of conditions that may influence stress responses. In this study we analyse density responses apparent in the natural populations of 46 macrofaunal taxa to stressors representing habitat requirements (sediment particle size), food (sediment organic content) and heavy metal contamination (copper, zinc and lead). Simple responses to either a single stressor or stressors acting additively were less apparent in our regression-based models than multiplicative effects. We also investigated whether species that exhibited a unimodal abundance distribution in response to a habitat factor were more likely to exhibit differential responses to contaminants than species that exhibited monotonic responses to habitat factors. This proved to be the case, although not all unimodal abundance distributions were related to differential responses. The results of this study suggest that regression analysis, together with well-designed, large-scale, field surveys, offer a good method for describing and predicting effects of multiple stressors and helping to both target experiments and scale-up their results.