AIM--To assess the validity of negative cervical smear reports in women who subsequently developed cervical cancer; and to determine means of improving the screening process. METHODS--One hundred and forty cervical smears, initially reported as negative from 103 women, and taken up to 12 years before diagnosis of cervical cancer, were reviewed. RESULTS--Ninety two smears contained dyskaryotic cells. Analysis showed that these smears formed several well defined patterns. False negative reports were likely to occur if fragments of neoplastic tissue rather than dissociated dyskaryotic cells were present or if the smear contained few dyskaryotic cells. Screening fatigue appeared to be a factor in others. It was also considered important that smears contained cells from the endocervix. These were deficient in 64% of the 47 smears confirmed as negative on review and in 69% of smears containing only a few dyskaryotic cells. CONCLUSIONS--Current methods of quality assurance will not remedy these defects in the screening process. It is the responsibility of laboratories to identify sources of poor smears and liaise with smear takers to ensure an improvement in quality. Assessment of the quality of smears received by a laboratory should become an important part of audit. Staff training should place more emphasis on the interpretation of "microbiopsies". The adoption of a quick scanning technique before conventional screening would probably also substantially reduce false negative results.