Infants that died suddenly and unexpectedly were studied as part of the European Concerted Action on sudden infant death syndrome (SIDS). Three paediatric pathologists, first independently of each other and later in a consensus meeting, classified 63 cases into 3 groups: SIDS (19 cases), borderline SIDS (30 cases) and non-SIDS (14 cases). The interobserver agreement among the pathologists before the consensus meeting was moderate (Kappa = 0.41) and jointly it was higher (Kappa = 0.83). The distribution of epidemiologically determined risk factors was studied over these three groups. Maternal smoking after birth, low socioeconomic status and thumb sucking were found more often in SIDS than in the other cases. Inexperienced prone sleeping was a determinant for SIDS, but not for non-SIDS. Previous hospital admission, low birthweight and/or short gestation were associated with borderline SIDS. Non-SIDS cases received more breastfeeding, the parents hardly smoked during pregnancy and after birth, a firm mattress had been used, and more often signs of illness had been reported by the parents, compared with the SIDS and borderline SIDS cases. Bedding factors and both primary and secondary prone sleeping were equally distributed over the three groups which supports the hypothesis that, in SIDS and borderline SIDS, as well as in non-SIDS cases, some similar external and preventable factors might influence the events leading to death. Research should therefore focus on all sudden unexpected deaths, after which subgroups such as SIDS cases can be separately analysed. The postmortem is an essential part of the whole work-up of each case and the results should be interpreted with all other available data to arrive at a sound evaluation of cases and thus form the basis for the prevention of all sudden unexpected infant death.