Over-estimation of body size is a cardinal feature of anorexia nervosa (AN), usually revealed by comparing individuals who have AN with non-AN individuals, the inference being that over-estimation is pathological. We show that the same result can be reproduced by sampling selectively from a single distribution of performance in body size judgement by comparing low BMI individuals with normal BMI individuals. Over-estimation of body size in AN is not necessarily pathological and can be predicted by normal psychophysical biases in magnitude estimation. We confirm this prediction in a dataset from a morphing study in which 30 women with AN and 137 control women altered a photograph of themselves to estimate their actual body size. We further investigated the relative contributions of sensory and attitudinal factors to body-size overestimation in a sample of 166 women. Our results suggest that both factors play a role, but their relative importance is task dependent.