Background: Scanning laser polarimetry is a tool for measuring the retinal nerve fiber layer: both its cross-sectional surface (polarimetric data analysis) and its thickness (mean sector values). Methods: Two observers examined 56 normal volunteers twice by means of scanning laser polarimetry (Nerve Fiber Analyzer type 1, software version 1.6). Measurements of the retinal nerve fiber layer were carried out in four equal sectors of a circle around the optic nerve head. The measured values of the sectors and the calculated ratios among them were used for statistical analysis. Interobserver and intraobserver reproducibility were analyzed following a balanced random three-way cross classification with interactions. Interobserver reproducibility was defined as the part of variance not influenced by the observers. Intraobserver reproducibility was defined as the part of variance not influenced by the time. Results: For repeated measurements of the retinal nerve fiber layer, better intraobserver than interobserver reproducibility was found (0.57–0.79 vs 0.11–0.44). The interobserver reproducibility was improved (0.24–0.65) in comparison to the intraobserver reproducibility (0.32–0.68) by calculating the ratios of the measured values. Conclusion: The instrument is clinically useful only if used by the same observer. If measurements are performed by different observers the ratios of the measurements must be used. Further development in the apparatus is needed to improve interobserver reproducibility.