Information from the Jamaican Perinatal Mortality Survey was used to identify features of mothers and their pregnancies that were independently associated with perinatal death. Social, biological, environmental, life style and medical aspects of mothers and their pregnancies were collected on two inter-locking subsamples: (1) all births on the island of Jamaica in the 2 months of September and October 1986, the 'cohort months', and (2) all fetal deaths of weight 500 g or more, together with all neonatal deaths, in the 12-month period from 1 September 1986 to 31 August 1987. Singleton survivors from the cohort months were compared with all perinatal deaths in the 12-month period using logistic regression. The first model omitted items concerning past obstetric history, but these were included in the second model. In total, 21 variables entered the first model and 24 the second. The only item that became non-significant when past obstetric history was included was maternal age. The final model compared 1017 perinatal deaths with 7672 survivors. It consisted of the following: union (marital) status (married being at lower risk, P < 0.01), maternal employment status (housewives at lowest risk, P < 0.001), number of adults in household (the more the higher the risk, P < 0.05), the number of children aged < 11 (the more the lower the risk, P < 0.0001), use of toilet facilities (shared with other households increased risk, P < 0.001), maternal height (tall women at reduced risk, P < 0.001), mother's report that she was trying to get pregnant (P < 0.001), maternal alcohol consumption (drinkers had lower risk, P < 0.05), maternal syphilis (higher risk, P < 0.0001), bleeding before 28 weeks (higher risk, P < 0.0001), bleeding at 28 weeks or more (higher risk, P < 0.0001), first diastolic blood pressure (80 mm + at higher risk, P < 0.0001), highest diastolic blood pressure (100 mm + at increased risk, P < 0.0001), highest proteinuria (++ or more at increased risk, P < 0.0001), vaginal discharge/infection (untreated at increased risk, P < 0.001), pre-eclampsia diagnosed in antenatal period (increased risk, P < 0.01), maternal diabetes (increased risk, P < 0.05), start of antenatal care (first trimester at reduced risk, P < 0.01), iron taken (reduced risk, P < 0.0001), type of perinatal care available in parish of residence (reduced risk if consultant obstetricians and paediatricians available at all times, P < 0.0001), number of miscarriages and terminations (the more the higher the risk, P < 0.0001), previous stillbirth (higher risk, P < 0.0001), previous early neonatal death (higher risk, P < 0.001), previous Caesarean section (higher risk, P < 0.01). The implications for reduction in perinatal mortality rates are discussed.