A modified version of the bubbles masking paradigm was used in three experiments to determine the key areas of the body that are used in self-estimates of body size. In this paradigm, parts of the stimuli are revealed by several randomly allocated Gaussian “windows” forcing judgements to be made based on this partial information. Over multiple trials, all potential cues are sampled, and the effectiveness of each window at predicting the judgement is determined. The modified bubbles strategy emphasises the distinction between central versus edge cues and localises the visual features used in judging one’s own body size. In addition, eye-movements were measured in conjunction with the bubbles paradigm and the results mapped onto a common reference space. This shows that although observers fixate centrally on the torso, they are actually directing their visual attention to the edges of the torso to gauge body width as an index of body size. The central fixations are simply the most efficient way of positioning the eye to make this estimation. Inaccurate observers are less precise in their central fixations and do not evenly allocate their attention to both sides of the torso’s edge, illustrating the importance of efficiently sampling the key information.