The width of retinal vessel images on fundus photographs was determined by projection micrometry and microdensitometry. These methods were evaluated on 12 retinal vessels using the negatives of 570-nm monochromatic fundus photographs. For projection micrometry, the intraobserver reproducibility of vessel width measurements was 1.6%-2.9%, depending upon the experience of the observer. Significant interobserver differences in the measured widths were demonstrated. For microdensitometry, three distinct measurement criteria were used. Significant differences in width as measured by the three criteria were found, but there were no interoperator differences for each criterion. The intraoperator reproducibility of vessel width measurements by microdensitometry was 2.1%-2.5%. Significant differences were found in the vessel widths determined by the micrometry and densitometry methods, and results obtained by micrometry are discussed in terms of edge-detection phenomena.