We determined the cross-sectional natural history of retinopathy by prospective study of 461 insulin-dependent juvenile-onset diabetics. In so doing, we compared the sensitivity of ophthalmoscopy, photography, and fluorescein angiography in detecting retinopathy. Photography was far more reliable than ophthalmoscopy in detecting early retinopathy and equivalent to angiography. Retinopathy was not present at diagnosis of diabetes. After a lag period, the prevalence of retinopathy rose in sigmoidal fashion, reaching 50% at just over seven years duration, and asymptotically approaching 90% at 17--50 years. Proliferative retinopathy was first seen at 13 years duration, and its prevalence rose to 26% at 26--50 years. From the natural history we computed the dimensions of a proposed clinical trial to test the effect of tight metabolic control in prevention of retinopathy.