The reliability of self-reported answers about sexual behaviour is assessed by means of a survey of couples in rural Senegal, Africa, carried out in 1992. A total of 51 married couples were followed weekly over a five-week period. Reports for a recall period of seven days are reliable: dates of sexual acts differed by no more than a day in 73 per cent of cases. Reports over longer periods are less reliable. Retrospective recall of coital frequency for a four-week period yields much higher estimates than those obtained with a seven-day recall, particularly for men.