The global mean defect (GM) is probably the most useful visual field index for glaucoma follow-up. We compared 50 regional subsets of test locations to estimate the GM. Using the data on 424 automated fields of 257 patients with either primary open-angle glaucoma or ocular hypertension, we calculated the partial sample error of the regional mean defect as compared with the GM. In many regions, the measured sample error was greater than expected for a representative sample of points. Some sample errors were up to 5 times larger (regions "upper hemifield" and "lower hemifield"). Only a few subsets proved to be representative of the whole field, namely, regions "ring 3" (10 degrees-15 degrees), "ring 4" (15 degrees-20 degrees) and "ring 5" (20 degrees-25 degrees). The use of such programs for follow-up is very accurate for staging. Moreover, theoretical calculations reveal that trend analysis is even more significant if the reduction in examination time is combined with a proportional increase in examination frequency.