A critical overview of 19 case-control studies that have investigated the relationship between prone sleeping position and sudden infant death syndrome (SIDS) is presented. Issues relating to the non-comparability of the studies are described in terms of: (i) case definition; (ii) selection of controls; (iii) quality of the sleeping position data; (iv) recall bias; and (v) adjustment for confounding factors. All studies showed a positive association (2 out of the 19 studies were not significant) between prone sleeping position and SIDS. Meta-analysis techniques applied to six of these studies, based on 'usual' sleeping position in cases and population representative controls, has confirmed an overall higher risk of SIDS in infants who usually sleep prone. The most common odds ratio for an association between prone sleeping position and SIDS was 2.72 (95% confidence interval 2.27-3.26). The extent to which the methodological problems of retrospective case-control studies interfere with our interpretations of this association are discussed.