Six experiments investigated the limiting conditions on and the causes of asymmetries in estimates of euclidean distance. Participants estimated distances between locations on recently learned maps or between buildings on their college campus. Estimates between landmarks and neighboring nonlandmarks were often asymmetric, but estimates between other pairs of locations were typically symmetric. These and other results were inconsistent with the predictions of models that attribute asymmetries to stimulus or to retrieval bias. A contextual scaling model of asymmetry is proposed. According to this model, asymmetries in proximity judgments are caused by general principles of human memory and judgment: (a) Stimuli differ in the contexts they establish in working memory and (b) magnitude estimates are scaled by the context in which they are made.