In this paper a method is described in which retinal contour is derived from measurements of peripheral refraction, keratometry and A-scan ultrasonography. Theoretical aspects are discussed first, followed by a closer examination of the computing scheme and its evaluation. Evaluation involved investigation of inherent errors arising from the computing scheme's methods of compensating for unmeasured human ocular complexities such as crystalline lens asphericity and gradient index optical structure, finite aperture pupils, ocular surface misalignment and toricity. It is found that computed retinal contours are little affected by these inherent errors.