Objectives: To determine whether the three mailings routinely used by researchers for epidemiological surveys are useful and appropriate, by comparing social characteristics and selected disease histories to find out if there are differences between individuals responding at different times. Methods: Social characteristics and selected disease histories of 11,797 British women still under general practitioner observation in the Royal College of General Practitioners' Oral Contraception Study were compared. The women aged 40-78 at December 1992 were sent health survey questionnaires via their general practitioners between November 1994 and July 1995. Results: Significant differences in the characteristics of first, second and third mailing respondents and non-respondents were found for smoking habits, social class, parity, area of residence, further contact, bronchitis, hysterectomy and mental illness. First mailing respondents were more likely to be from a non-manual social class, have a parity of less than three, live in England, and be happy to be contacted again, but were less likely to have had bronchitis or mental illness and were less likely to have been smokers at the time of recruitment than second mailing respondents. A comparison of second and third mailing respondents on the above factors showed no significant differences between the two groups. Conclusions: The inclusion of third mailing respondents did not significantly change the social or health characteristic profile of the cohort and suggests that the effort and resources expended in carrying out a third mailing may not be justified.