Abstract

Human observers can bisect a space with an accuracy which far exceeds the retinal grain. How spatial information is sampled to provide such hyperacuity thresholds has been of interest since Hering’s (1899) proposal that local signs are averaged along the length of the lines. We investigated this question in central and peripheral vision of human observers by measuring bisection thresholds for stimuli comprised of discrete samples, each ~1′ long. The number of samples and the intersample spacing were systematically varied. The results show that (1) the critical variable is not the total length of the pattern (i.e., length of the samples plus interspaces) but rather the number of samples; (2) central and peripheral vision differ qualitatively in their sampling characteristics. In central vision, only a small number of samples (5) are needed to provide optimum performance; increasing the number of samples from 1 to 5 results in only a slight improvement in bisection thresholds. The periphery, on the other hand, requires more samples for optimal performance, and thresholds improve in proportion to the square root of the number of samples. The data of strabismic amblyopes show the same $\sqrt{n}$ behavior as the normal periphery. The results suggest that in strabismic amblyopia, and in the normal periphery, spatial information is undersampled.

© 1985 Optical Society of America