Measurements of human attitudes and perceptions have traditionally used numerical point judgments. In the present study, we compared conventional point estimates of weight with an interval judgment method. Participants were allowed to make step by step judgments, successively converging towards their best estimate. Participants estimated, in grams, the weight of differently sized boxes, estimates thus susceptible to the size-weight illusion. The illusion makes the smaller of two objects of the same weight, differing only in size, to be perceived as heavier. The self-selected interval method entails participants judging a highest and lowest reasonable value for the true weight. This is followed by a splitting procedure, consecutive choices of selecting the upper or lower half of the interval the individual estimates most likely to include the true value. Compared to point estimates, interval midpoints showed less variability and reduced the size-weight illusion, but only to a limited extent. Accuracy improvements from the interval method were limited, but the between participant variation suggests that the method has merit.