The aim of this study was to measure the reliability, in test score units, of several clinical tests which use high- and low-contrast letters, and to provide an estimate of what constitutes a significant change in performance over time. Patients with normal vision and with early or subtle eye disease were recruited so that the results would be representative of the population likely to present for primary vision screening. Patients were tested on the Bailey-Lovie logMAR chart, the Regan low-contrast letter charts and the Pelli-Robson low-contrast letter chart on two occasions; the two test sessions were separated by at least four weeks to give an estimate of reliability appropriate for the conditions under which the tests are likely to be used. A 'significant change', i.e. one which would be observed in only about 5% of patients with stable visual performance, was about +/- 2 'steps' of the measurement scale, i.e. +/- 2 lines for the Bailey-Lovie and Regan charts and +/- 2 letter groups for the Pelli-Robson chart.