Interest in the effects of neighbourhood or local area social characteristics on health has increased in recent years, but to date the existing evidence has not been systematically reviewed. Multilevel or contextual analyses of social factors and health represent a possible reconciliation between two divergent epidemiological paradigms-individual risk factor epidemiology and an ecological approach. Keyword searching of Index Medicus (Medline) and additional references from retrieved articles. All original studies of the effect of local area social characteristics on individual health outcomes, adjusted for individual socioeconomic status, published in English before 1 June 1998 and focused on populations in developed countries. The methodological challenges posed by the design and interpretation of multilevel studies of local area effects are discussed and results summarised with reference to type of health outcome. All but two of the 25 reviewed studies reported a statistically significant association between at least one measure of social environment and a health outcome (contextual effect), after adjusting for individual level socioeconomic status (compositional effect). Contextual effects were generally modest and much smaller than compositional effects. The evidence for modest neighbourhood effects on health is fairly consistent despite heterogeneity of study designs, substitution of local area measures for neighbourhood measures and probable measurement error. By drawing public health attention to the health risks associated with the social structure and ecology of neighbourhoods, innovative approaches to community level interventions may ensue.