To examine and quantify any change in the dark adaptation (DA) function of normal subjects due to learning effect on test-retest. Sixteen normal subjects (12 women, four men) whose ages ranged between 24 and 52years (mean 34.6 ± 6.7years) were studied. The interval period between test and retest ranged between 0.92 and 2.37months (mean 1.38 ± 0.40months). DA was measured with a Goldmann-Weekers (GW) dark adaptometer, and subjects were pre-adapted using a light intensity of 2,700cd/m(2) Ganzfeld background for 5 minutes. Exponential non-linear regression analysis was used to determine seven parameters of DA function. These were time of cone-rod break, cone and rod final thresholds, and magnitude of change and time constant of the cone and rod limbs. The mean cone-rod break time with 95% confidence intervals (CI) was 0.098 (CI: -0.527, 0.330) minutes faster on retest (p = 0.630)). Fourteen of the 16 subjects demonstrated an increase or 'worsening' of their final cone and rod thresholds on the second visit. The mean final threshold differences on retest for the cone limb was 0.105 (CI: 0.032, 0.179) log cd/m(2) (p = 0.008) and 0.093 (CI: -0.039, 0.225,) log cd/m(2) (p = 0.155) for the rod limb. The magnitude of change for the cone limb was 0.016 (CI: -0.122, 0.155) log cd/m(2) (p = 0.805) and -0.196 (CI: -0.435, 0.827) log cd/m2 (p = 0.518) for the rod limb, while the time constant on retest for the cone limb was -0.021 (CI: -0.128, 0.169) minutes, (p = 0.770) and 0.276 (CI: -0.424, 0.976) minutes (p = 0.410) for the rod limb. None of the DA parameters that were examined demonstrated a learning effect of clinical significance between test and retest. None of the changes in mean from test to retest for the seven parameters were found to be statistically significant, and the changes were clinically negligible. Therefore, any change among patients that may occur in dark adaptation between a visit interval may be considered real, and not due to the effect of learning.