A disturbance in the perception of personal body size and shape is a key feature of both anorexia and bulimia nervosa, but it has proved difficult to quantify. Previous attempts have used methods like the distorting video technique (DVT), which alters an image by stretching the figure in either the X- or Y-axis. This is a poor representation of the way fat is added to or lost from the body, and the pattern of distortion provides a host of cues to the degree to which the image has been altered. To overcome these problems we have used a specially designed software system that uses biometric data based on real body shapes, instead of simply stretching or compressing images of bodies. This technique also allows individual body parts to be altered separately, so we can determine whether a specific body part is overestimated relative to others. We can also calculate the apparent body mass index (BMI) of our modified pictures, using the perimeter-area ratio (PAR). This allows us to compare an observer's actual BMI with that calculated for their estimated and ideal bodies. We tested 30 anorexic, 30 bulimic and 137 control observers. Our results show that all three observer groups tend to overestimate their body size, but not significantly so. Both the control and bulimic observers prefer an ideal body with a BMI of 20, which is at the lower end of the 'normal' BMI range. However, the anorexies ideal BMI is 15, which is on the border between the emaciated and underweight BMI categories.