Dr Herbert makes a number of points about our paper that we shall address in turn. Her basic thesis is that when coverage is high, comparison of cervical smear histories in women with and without cancer does not provide useful estimates of the effectiveness (or even the relative effectiveness) of screening at different ages. We strongly disagree. Although observational studies are subject to bias, we do not believe that this can explain the substantial differences in effectiveness that we observed across different age groups. The percentage of women with a screening smear is almost identical in cases and controls aged 20 -29 years (74.3 vs 74.8%), underlying the lack of effectiveness of screening in this group compared to women aged 45 -64 years, in whom coverage is much greater among controls (81.5 vs 59.0%).We agree that some of the screening in our study dates back to the late 1980s and that a variety of quality assurance measures have been introduced since. However, in general, one would expect any advantage of 3-yearly screening over 5-yearly screening to diminish as the quality of screening improves. Table 1b provides an analysis restricted to cases (and their controls) diagnosed since January 1995. The results are quantitatively similar (and qualitatively unchanged) to those presented in our paper (Table 1a).Although it is true that screening coverage improved considerably between 1987 and 1993, the claim that, in 1994 -1995, over 90% of women had been screened is erroneous. This figure is based on the proportion who had been (or were due to be) sent an invitation, not the proportion actually screened. No data exist for the proportion of women aged 25 -64 years in 1994 who had ever been screened. However, it is unlikely to be that much greater than the proportion screened in the previous 5 years (85.7%), since coverage was low prior to 1990 and women screened then were likely to return for screening when invited. Additionally, the relevance of this to comparing 3-vs 5-yearly screening is dubious.We do not believe that cases with unknown stage were more likely to be fully invasive. Most Health Authorities provided information on stage from all or none of the cases for a given year. Restricting analysis to sources that provided complete staging on all cases submitted made little difference to the results (Table 1c).We agree that due to the poor sensitivity of cervical cytology, the number of previous recent smears is also important. This was illustrated in Tables 3 and 4, in which the added affect of having more than one previous smear was estimated. The results further support our main finding that screening is less effective in younger women. The additional relative benefit of a second negative smear in women aged 20 -39 years was only 1.03 compared to 1.32 and 1.35 for women aged 40 -54 and 55 -69 years, respectively.Dr Herbert points out that in 1998 over 4000 cases of CIN3 were detected in UK women under 25 years. What she does not say is that this compares with fewer than 50 cases of invasive cancer u...