The validity and performance of the Papanicolaou (PAP) smear, as a diagnostic test, was retrospectively analysed in 806 patients who presented with at least one abnormal smear. Histological tissue examination was available in 620 of these patients. It is shown that the definition of errors, selection criteria of smears, time intervals between examinations, definition of 'disease' and the values separating 'positive' and negative' test results must be clearly defined before comparison with other published error rates is possible. Consequently, a standardized method of reporting is recommended. Exact agreement between the results of cytology and histology was poor. The use of sensitivity, specificity and predictive values is preferred, but it is shown that precise definitions are still necessary. The optimal choice of the value separating positive and negative test results, i.e., the choice between an expectant and an active management of patients, must be tailored to the anticipated prevalence of the disease.