A small group of glaucoma patients with scotomata were compared with normal coevals. Plots were obtained of the error of each measurement as a function of retinal illumination. A statistical analysis shows the limitations of perimetry as a monitor of the progress of glaucoma. The significance of the data is discussed.