Three successive fields of 136 eyes (86 patients) were extracted from our data base containing over 3000 visual fields performed on the Humphrey visual field analyser with program 30-2. Series of fields in which the second field was depressed relative to the first were selected for analysis to determine how much change between the first two fields was required to predict a downward trend as confirmed by the third field in the series. The data were stratified with respect to initial field damage. Seven regions of the visual field were analysed including the upper and lower temporal and nasal quadrants, the superior and inferior Bjerrum regions, and the whole field. Minimally damaged regions required between 4.7 and 5.6 dB change in mean sensitivity, whereas more damaged regions required between 5.5 and 7.2 dB change in mean sensitivity to have 95% confidence that the negative trend would be confirmed by the third field. The superior Bjerrum region was the most sensitive for the detection of change, and the lower temporal region was the least sensitive. We conclude from this series of data that large changes between two successive fields are required to be sure that the changes are due to disease rather than chance fluctuation. Where any doubt exists, the field should be repeated to confirm the reality of change.