Repeatability indices for the Farnsworth D-15 test.

The repeatability of the D-15 color-vision test is considered to be excellent. However, this conclusion is based on a subject pool which contained a large percentage of color-normals. This type of sampling could bias the repeatability results because color-normals rarely fail the test. Furthermore, color-normals usually do not perform the D-15 in the clinical setting. To establish the repeatability of the D-15 for a relevant clinical population, we examined the D-15 results from two different sessions for 116 subjects who had a congenital red-green color-vision defect. The kappa coefficient for intersession agreement indicated that approximately 84% of the subjects obtained the same pass/fail results at both sessions. The type of defect was repeatable on approximately 80% of the subjects. Although the repeatability of the D-15 for color-defective subjects was good, it was lower than the near-perfect agreement reported previously. The coefficients of repeatability for the crossings show that if a person makes less than five crossings then the test should be administered again in order to ensure that the test result is repeatable.