Perinatal mortality in sibships has been examined using the Medical Birth Registry of Norway. Using linear logistic regression, parity specific perinatal mortality in the period 1967-1981 has been analysed, controlling simultaneously for maternal age, year of birth and survival of earlier births. The risk of a perinatal loss was increased by a factor of up to 6.0 following one earlier perinatal loss, and with three earlier losses the increase was 17.0. The general reduction in mortality between the different five year periods varied according to parity and maternal age. No secular reduction in risk, however, was demonstrated in sibships where the first birth died perinatally; for some subgroups of women the risk actually increased between the first and the second five year period. The effect of maternal age changed with parity, being strong in the first birth order, but of no effect either for the third or the fourth birth orders once the perinatal survival of earlier births was controlled for. This study shows the need to take heterogeneity of risk between women into account in studies of perinatal loss; the overall improvement in perinatal mortality does not apply to every woman. As care improves, perinatal losses become less and less random, demonstrated by increased risks of recurrence.