The human visual system is capable of making spatial discriminations with extraordinary accuracy. In normal foveal vision, relative position, width or size can be judged with an accuracy much finer than the size or spacing of even the smallest foveal cones. This remarkable accuracy of spatial vision has been termed 'hyperacuity'. Almost a century ago Ewald Hering proposed that the accuracy of Vernier acuity could be accounted for by averaging of discrete samples along the length of the lines comprising the targets; however, the discovery that Vernier acuity of a few arc seconds could be achieved with dots has rendered the nature and role of sampling in spatial discrimination unclear. We have been investigating the sampling of spatial information in central and peripheral vision (the perifovea) of normal human observers and in observers with strabismic amblyopia. Our results, presented here, show that peripheral vision and central vision of strabismic amblyopes differ qualitatively in their sampling characteristics from those of the normal fovea. Both the periphery and the central visual field of strabismic amblyopes demonstrate marked positional uncertainty which can be reduced by averaging of spatial information from discrete samples.