[Margin of error in subjective assessment of visual acuity]

The frequencies for correct answers to individual Landolt-Cs were measured as a function of their size in 183 eyes ("psychometric function", 48 tests per acuity step, computer screen). An integrated Gauss function modified for assessing bias was fit (least squares) to the values of each subject. The mean value of the fitting function is a measure of visual acuity; its standard deviation is a measure of the reliability of the acuity measurements. The standard deviations are subject to a logarithmic normal distribution centered at a width of 1.5 steps. The distribution does not depend on age, sex or diagnosis (except for visual field defects). The width of the psychometric function is a universal number that is independent of the visual acuity itself (greater than 0.1). The distribution of acuity results measured according to DIN 58220 can be computed using the values of the psychometric function. For criteria 3/4, 3/5, 5/8 and 6/10, specified in the recently revised version of DIN 58220, one obtains standard deviations of the acuity measurements of 0.89, 0.84, 0.71 and 0.67 acuity steps. Compared to 3/4 the new criteria on average yield acuity values that are 0.55, 0.36, and 0.44 steps higher than previously. Given an actual acuity measurement value, confidence intervals are computed using Baye's formula. The results are presented as tables.