The visual fields of 30 patients (subjects) with glaucoma were sent to six experienced clinicians (observers). Each subject had at least four visual field examinations on the OCTOPUS 201 automated perimeter spanning at least one year. Each observer was asked to review the visual field data of each subject and determine whether the visual fields were stable, improved, or worse over time. The visual field data were then analyzed using six different statistical models. In only 15 of the 30 subjects did at least five of the six human observers agree on the behavior of the visual field. Agreement among the statistical models was better, with at least five of the six models agreeing on 22 of the 30 subjects. It was concluded that there is, at present, no validated technique for detecting progressive visual field loss in glaucoma using automated perimetry when relatively few visual fields are available for analysis.