Two groups of normal subjects were submitted to repeated automated static threshold perimetry. Perimetric results were strongly affected by the level of experience in some subjects; in the majority, however, the effect of experience was small. Initial field tests often showed high numbers of depressed points. Sensitivity increased with perimetric training, particularly between the first sessions. Those subjects who improved most started low, gradually approaching normal levels with experience. Learning effects were more pronounced peripherally than paracentrally and "untrained" fields characteristically showed concentric contraction with numerous points with low sensitivity peripherally. An important practical conclusion is to allow repeated testing of all inexperienced patients in whom initial fields do not agree with clinical findings. A chart showing a concentrically narrowed field should be viewed with particular suspicion. Furthermore, a single initial field may constitute an inadequate baseline for clinical follow-up.