Evaluation of numerical displays of visual field data by human observers--the definition of scotomas and clusters by human interpreters.

Numerical printouts of 23 visual field examinations performed with glaucoma program G1 were analyzed by 3 perimetrists. The task consisted of defining the existence of clusters or scotomas. Considerable discrepancy with regard to the existence or the size of clusters and/or scotomas was found among the three perimetrists. The median difference in number of clusters as compared for each pair of interpreters was found to differ from zero at a significance level of 5% (Wilcoxon's signed-rank nonparametric test). These findings suggest that the interpretation of a visual field status by human observers will always remain subjective and to a large extent contradictory. This situation calls for the implementation of interpretative computer software to supplement the currently available algorithms.