SummaryWe studied 7300 singleton births in the highlands and 4881 in coastal Papua New Guinea in order to examine the separate contribution of anaemia or malaria to low birthweight. The highland sample was selected from a non-malarious area (Goroka) and the coastal sample from an area with perennial malaria transmission (Madang). There was an approximately three-fold increased risk of low birthweight (<2500 g) in live-births in Madang compared to Goroka. The prevalence of anaemia in the two areas was strikingly different, with 29·2% of Goroka and 89·0% of Madang women anaemic. There was a trend towards increased low birthweight with decreasing haemoglobin levels in both areas, but this was significant only for Madang. It was assumed that for a given haemoglobin level the increased low birthweight percentage in Madang compared to Goroka was due to malaria exposure, and on this basis relative risk values were estimated for the effect of malaria exposure on low birthweight. Using this approach separate estimates for anaemia and malaria population-attributable risk for low birthweight in Madang were calculated. These indicated that up to 40% of low birthweight babies born in malarious areas may be attributable to malaria and less than 10% attributable to severe anaemia (Hb < 7·0 g dl−1). The magnitude of the malaria effect estimated in this analysis places a high priority on malaria control in pregnancy as a strategy for improving birthweight and child survival.