When single-molecule fluorescence localization techniques are pushed to their lower limits in attempts to measure ever-shorter distances, measurement errors become important to understand. Here we describe the non-Gaussian distribution of measured distances that is the key to proper interpretation of distance measurements. We test it on single-molecule high-resolution colocalization data for a known distance, 10 nm, and find that it gives the correct result, whereas interpretation of the same data with a Gaussian distribution gives a result that is systematically too large.