Summary Two cohorts of women born in 1914-18 and 1929-33 who participated in a cervical screening programme have been followed for over 40 years. Age-specific incidence rates of squamous carcinoma of the cervix by rank of smear and length of interval between smears are reported. The younger cohort, who had undergone more frequent screening, had lower rates of invasive disease. From these incidence rates, estimates of false-negative rates and regression rates for carcinoma in situ have been made. The false-negative rate was estimated to be about 15%. Regression seemed more frequent in younger than in older women. For the younger cohort it was estimated to be 72% and in the older 47%. Before the updating of the cohort data was attempted it was realised that, for many of the women, no post-1969 records would exist, since, before or soon after 1970, some of the women would have died from causes other than cervical cancer, some would have had a hysterectomy and some would have left the province. During the 1970s some culling of the files in the cytology laboratory took place. The files of women who had had nothing but negative smears, no histology and had not had a smear taken in the previous 7 years, or were known to be dead, were deleted. Interpolating from the data on annual terminations of surveillance, it would appear that the files of approximately 10% of the women were removed.Before proceeding with the update an attempt was made to estimate from the smear and pathology records of the original cohorts that were on magnetic tape the number of women who had a very low probability of being linked. They fell into two groups. The first group were women without histology records and only a single smear which was taken before 1967. The authors considered that women below age 55 with a gap without a smear in the years 1967-70 (when many families returned to the eastern provinces) and with only a single smear preceding, was a strong indication of a transient resident. The second group were those who, before 1970, had had a hysterectomy for reasons other than cervical cancer or its precursive abnormalities, or who had died of cervical cancer. In cohort I, 12 923 women fell into these two groups, and in cohort II 14 537. These women, however, were included in all matching attempts since it was easier to leave their records in the files and because there was a slight probability that a linkage might be made.Two methods of matching were used. One was to search the files at the cytology laboratory for each unique smear number included in the original cohort study. A successful match on these numbers enabled us to identify the woman in the cytology files and, from this, update our records. The Correspondence: BJ Morrison