BACKGROUND: This paper provides a systematic, quantitative review of the epidemiological evidence relating parental smoking and sudden infant death. METHODS: Thirty two relevant publications were identified after consideration of 692 articles selected by electronic search of the Embase and Medline databases using keywords and Mesh headings relevant to passive smoking in children. Eleven further articles were identified from reviews and by talking to authors. The search was completed in April 1997 and identified 39 studies. RESULTS: The unadjusted pooled odds ratio for prenatal maternal smoking was 2.77 (95% CI 2.45 to 3.13). After adjustment for a variety of confounders the pooled odds ratio was reduced to 2.08 (95% CI 1.83 to 2.38) and was similar in cohort and case-control studies. Four studies reported on maternal postnatal smoking after controlling for prenatal maternal smoking (pooled odds ratio 1.94 (95% CI 1.55 to 2.43)). Of three studies reporting on the risk of paternal smoking where the mother was a non-smoker, two found significant effects while one found no effect. Dose-response relationships with both prenatal and postnatal maternal smoking were present in most studies which provided data. CONCLUSIONS: Maternal smoking doubles the risk of sudden infant death syndrome. The relationship is almost certainly causal. There is good evidence that postnatal exposure to environmental tobacco smoke from both mother and father are important. Because prenatal smoking is almost invariably associated with postnatal smoking, the role of prenatal smoking per se will be difficult to resolve using epidemiological studies.