Conflicting views have been expressed in the literature regarding the sign and magnitude of the difference between the values of ocular correction deduced by retinoscopy and by subjective refraction. Possible causes for such a discrepancy are briefly reviewed. Calculations based upon experimental data show that neither the reddish colour of the retinoscopic reflex nor the position of the layer in the retina at which the reflex originates are likely to introduce any important discrepancy. The probable significance of other factors is assessed.