Summaryobjective To identify the determinants of skilled and unskilled birth attendance. method Population-based survey in a rural area in Cambodia, of women aged 15-49 years who had delivered during the previous 3-month period. An analytical framework based on Andersen's behavioural model served to identify determinants according to delivery place (facility vs. non-facility), birth attendant at home births (skilled vs. unskilled), and change of birth attendant during delivery (changed vs. unchanged). We used logistic regression to analyse the data.results Of 980 women included in the analyses, 19.8% had skilled attendants present during delivery. The determinants of facility delivery were different from those for having skilled attendants assisting in home births. In case of facility deliveries, previous contact with a skilled attendant through antenatal care was a significant determinant. In case of home births, the type of birth attendant (i.e. skilled or unskilled) at the preceding delivery was a significant determinant.conclusion Community-based programmes need to reach primiparas, because once a woman has delivered with the aid of an unskilled attendant, she is five to seven times less likely to seek skilled help than a primipara.keywords skilled attendant, birth attendant, health seeking behaviour, maternal health, maternal mortality, antenatal care