Bisection thresholds are approximately proportional to the separation/eccentricity of the targets. This "Weber's law" for position has been invoked over the past century. It is the separation of the reference targets, or their eccentricity which determines the threshold? In previous studies separation and eccentricity are confounded. In the present report we have pitted separation against eccentricity. Bisection thresholds were measured for stimuli presented on an isoeccentric arc, so that separation could be varied while holding the eccentricity of the test lines constant. We used a 5-fold range of separations from 2-10 deg. In this regime, the present results provide strong evidence against Weber's law. When separation is varied but eccentricity held constant, there is no Weber's law. Rather the thresholds are approximately constant. Our results suggest that the judgement of the separation of widely separated objects is similar to a distance measurement using a ruler on the cortex, in that the error of measurement is independent of the separation between objects. The results imply that when we attempt to gauge the distance between widely separated objects it is unlikely that we do so on the basis of the outputs of large spatial filters; rather it appears that we make such judgements by estimating the cortical distance which separates the targets of interest.