The study was performed to evaluate whether the severity of diabetic retinopathy as assessed by three alternative methods was concordant with the severity of retinopathy as determined from 30 degrees stereoscopic photographs. The three methods were direct ophthalmoscopy through an undilated pupil, nonstereoscopic 45 degrees retinal photography through a pharmacologically undilated pupil and nonstereoscopic 45 degrees photography through a dilated pupil. A single 45 degrees photograph centered between the disc and fovea was taken and direct ophthalmoscopy was performed on 99 persons prior to pharmacological dilation of the pupil. After dilation, another 45 degrees photograph was taken of the same field, as well as 30 degrees stereoscopic color photographs of DRS fields 1, 2 and 4 (modified). Corresponding photographic fields were graded by masked, trained graders for the severity of retinopathy and for the presence of specified diabetic lesions using the Modified Airlie House Classification scheme. For three levels of severity of retinopathy (none, nonproliferative or proliferative) exact agreement between direct ophthalmoscopy and grading of retinopathy from stereoscopic photographs taken with the standard 30 degrees camera was 54.3% (n = 94). For four levels of severity of retinopathy (none, microaneurysms only, all other nonproliferative retinopathy and proliferative retinopathy), exact agreement between gradings of retinopathy of the 45 degrees photographs taken through undilated pupils and 30 degrees photographs taken through dilated pupils was 82.5% (n = 63); and for 45 degrees photographs and 30 degrees photographs taken through dilated pupils it was 86.5% (n = 74). These data suggest that 45 degrees nonstereoscopic fundus photographs, when graded according to a standard classification scheme, provide reasonably reliable photographic representation of the severity of retinopathy when broad overall categories are used.