Objectives: The 1990-1 British national probability sample survey of sexual attitudes and lifestyles (Natsal 1990) was repeated in 1999-2001 (Natsal 2000) to update population estimates of risk behaviours, and assess change over time. We examine whether changes in prevalence estimates may partly result from changes in measurement accuracy. Methods: Taking Natsal 2000 (11 161 respondents) and Natsal 1990 (13 765 respondents aged 16-44) we compared the response rate, sample representativeness, reporting of abortion last year (relative to official statistics), and selected attitudes. Among the common birth cohort eligible for both surveys (aged 16-34 Natsal 1990, 26-44 Natsal 2000), we compared reporting of experiences before 1990. Results: The response rate (66.8% Natsal 1990, 65.4% Natsal 2000) and completeness of reporting abortion were unchanged (84% Natsal 1990, 86% Natsal 2000). Attitudes were significantly changed in Natsal 2000 relative to Natsal 1990 - for example, increased tolerance of male homosexual sex, OR (95% Cl) 2.10 (1.93-2.29) men and 2.95 (2.74 to 3.18) women. In the common birth cohort reporting of heterosexual intercourse before 16 (OR 1.15 (1.02 to 1.29) men, 1.49 (1.31 to 1.69) women), and homosexual experience (OR 1.80 (1.46 to 2.21) men, 2.00 (1.61 to 2.48) women) were significantly increased. Conclusions: The results are consistent with improved reporting accuracy for some sensitive behaviours in Natsal 2000, in line with greater social tolerance and improved survey methodology. However, the evidence is not conclusive, and may not be generalisable to all such behaviours. The increase found in the reported prevalence of STI risk behaviours between Natsal 1990 and Natsal 2000 is likely to be somewhat overstated.