OBJECTIVE This study examined quantitatively the commonly held assumption that the reflectance of the retinal nerve fiber layer is proportional to its thickness. METHODS The reflectance of the retinal nerve fiber layer of an in vitro vertebrate cyecup preparation was measured with an imaging microreflectometer, an instrument that provided quantitative images of a small retinal area and allowed precise control of the directions of illumination and observation. After an experiment, the eyecup was fixed and embedded for histology. A direct correspondence between the measured images and the histologic sections was established by serial reconstruction of the retinal blood vessel pattern. The thickness of the previously imaged nerve fiber bundles was measured from photomicrographs of the histologic sections. RESULTS The reflectance of the retinal nerve fiber layer was proportional to its thickness in areas where nerve fiber bundles were parallel. When bundles were not parallel, the directional reflectance of the retinal nerve fiber layer caused a failure of proportionality for a single illumination angle. When multiple angles of illumination were used to find the peak reflectance, this peak reflectance was again proportional to the thickness of the retinal nerve fiber layer. CONCLUSIONS These results indicate that the reflectance of the retinal nerve fiber layer arises from throughout its thickness, and have implications for assessment methods designed to identify the surface of the retinal nerve fiber layer. In addition, they indicate that control of observation geometry may be an important means to reduce variability in clinical assessment systems.