BACKGROUND: Sexual behaviour surveys are widely used, but under-reporting of particular risk behaviours is common, especially by women. Surveys in whole populations provide an unusual opportunity to understand the extent and nature of such under-reporting. METHODS: All consenting individuals aged between 15 and 59 within a demographic surveillance site in northern Malawi were interviewed about their sexual behaviour. Validity of responses was assessed by analysis of probing questions; by comparison of results with in-depth interviews and with Herpes simplex type-2 (HSV-2) seropositivity; by comparing reports to same sex and opposite sex interviewers; and by quantifying the partnerships within the local community reported by men and by women, adjusted for response rates. RESULTS: 6,796 women and 5,253 men (83% and 72% of those eligible) consented and took part in sexual behaviour interviews. Probing questions and HSV-2 antibody tests in those who denied sexual activity identified under-reporting for both men and women. Reports varied little by sex or age of the interviewer. The number of marital partnerships reported was comparable for men and women, but men reported about 4 times as many non-marital partnerships. The discrepancy in reporting of non-marital partnerships was most marked for married women (men reported about 7 times as many non-marital partnerships with married women as were reported by married women themselves), but was only apparent in younger married women. CONCLUSIONS: We have shown that the under-reporting of non-marital partnerships by women was strongly age-dependent. The extent of under-reporting of sexual activity by young men was surprisingly high. The results emphasise the importance of triangulation, including biomarkers, and the advantages of considering a whole population.