Abstract
The perinatal mortality rate (PNMR) was calculated for each rural parish in Gloucestershire for the years 1968-79. The characteristics of high- and low-rate parishes (each set comprising about 10% of all births) were contrasted according to routinely recorded fetal, maternal and environmental variables. The overall PNMR in the high-rate set was six times higher than in the low-rate set; congenital malformations and low birth weight distribution could explain about one-third of this difference but the residue was not attributable to fetal gestation, maternal age or parity or social class as may have been expected. However, there was a strong association with population and housing density. It is suggested that comparisons within districts may provide more sensitive associations with PNMR than comparisons between districts, regions and countries.