Questionnaires and diaries have complementary biases and advantages for obtaining information on sexual behaviour but self-completed sexual diaries have the advantage of reducing retrospective bias. In a validation study of homosexual behaviour, sexual diary counts and subsequent questionnaire estimates (together with ratings of the certainty of the estimates) referring to the same month are compared and the discrepancies analyzed. Main findings include: questionnaire data yield consistently higher average estimates than diary counts, but have the same ordinal profile; individual difference (diary-questionnaire) scores show that 55% of questionnaire estimates of acts are higher than diary counts, 20% are identical and 25% are under-estimates; discrepancies are differentially located in different sexual acts. Masturbation and fellatio are systematically over-estimated in questionnaires and anal intercourse without a condom is the major source of inaccuracies, but in different directions: active partners under-estimate and passive partners over-estimate the amount of highest-risk sex. A strategy of joint use is discussed.