The role of similarity and bias in letter acuity measurements: a noisy template model
Hatem Barhoom, Gunnar Schmidtmann, Mahesh R. Joshi, Paul H. Artes, Mark A. Georgeson
The poster (#59) will be presented on Thursday, 1st September, in the Peripheral Vision session at Refter in the Erasmus building.
Higher resolution version of the poster can be downloaded here.
Previous studies have demonstrated that similarity and bias are major causes of errors in the identification of Sloan letters in visual acuity testing. However, these two factors and their relative contribution have not been investigated extensively at central and paracentral visual field locations. Using the method of constant stimuli, visual acuity was measured in 10 observers at central and paracentral (±3˚ vertical meridian) visual field locations. A “noisy template” model was adapted to distinguish biases and similarities from random errors in letter identification. At all three test locations, we found that the best model was one that combined the effects of both bias and similarity. The relative contribution of bias was higher than similarity, at all three test locations. Additionally, biases may exclusively explain variations in letter identification performance for all letter sizes at the three test locations. Although there was no significant effect of bias or similarity on the estimated letter-size acuity, we observed a substantial increase in the spread of the psychometric function (mainly at the periphery and at the upper part of the psychometric function) in the models that included only similarity but not bias. . In clinical vision tests, most letter stimuli are supra-threshold, so it is plausible to attribute differences in identification performance, especially in peripheral vision, to similarity alone. However, it will be important to investigate this assumption and to examine whether bias and similarity are likely to have a clinically significant effect on measures of visual performance and letter acuity in the periphery.