Subjects misjudge distances considerably when forced to rely on extra-retinal information. Nevertheless, they can reproducibly set a target to the same distance as a reference, or to double or half that distance, even when they have to look back and forth between them because they are prevented from seeing one when looking at the other. Our explanation for this apparent discrepancy is that people have access to reasonably accurate extra-retinal information on changes in ocular convergence, but can only use this information to judge distances if they had reliable information about the orientation of the eyes before the convergence changed.