Background: There are basically two possibilities to measure cylindrical refractive errors by eccentric photorefraction. The first is to determine the size and the tilt of the light crescent in the subject's pupil. Sphere, cylinder, and axis can be obtained from two pictures with the knife edge at two different orientations by using equations derived by Wesemann et al. In natural eyes, the procedure has limitations because undetermined factors (not considered in the theory) affect size, shape, and intensity of the light crescent. A second possibility is to perform eccentric photorefraction separately in at least three different meridians.
Methods: We have tested the power of the second possibility. The three critical parameters (sphere, cylinder, and axis) were calculated from Euler's law, which describes curvatures (or refractions) at any given angle. The procedure relied only on empirical calibration and not on a theoretical treatment of the optics. Therefore, it was not necessary to identify all factors that determine the path of light.
Results: The procedure compared favorably with subjective refractive (first population: students, age 26-30 years, N = 7 (14 eyes); correlations: sphere, r = 0.983; cylinder, r = 0.867; axis, r = 0.935) and with a Canon R-1 Autorefractor (second population: children, age 4-14 years, N = 48 (96 eyes); correlations: sphere, r = 0.955; cylinder, r = 0.600; axis, r = 0.846).
Conclusions: Because it is fast, the technique may be suitable for screening in children. The refractions in the different meridians are performed in real time (25 to 30 Hz) and a single reading (the average from 4-6 refractions in each of the 6 meridians) is obtained in 1-2 s. It constitutes a major improvement to commercially available videorefractors which use measurements only in two meridians in conjunction with the formula by Wesemann et al., although it is still not precise enough to permit spectacle prescription.