The convoluted pattern of normal cerebral cortex resists description by Euclidean geometry. However, subtle abnormalities often unrecognised in vivo may underlie a wide range of neurological disorders. Abnormalities of cortical morphology frequently characterise these lesions, and we applied fractal geometry in an attempt to quantify these abnormalities. We have demonstrated typical fractal scaling properties for the cortical-white matter interface on axial and coronal MR images, with a fractal dimension (Df) of 1.45 +0.06 (mean +1 standard deviation) in normal subjects. We have also applied this to 16 patients with frontal lobe epilepsy, a condition associated with subtle disruptions of the cortical ribbon, all of whom had no obvious abnormality on visual inspection of the images. Nine of these 16 had a Df < 1.27 (mean -3 standard deviations). A control group of 10 patients with temporal lobe epilepsy with a focal foreign tissue lesion (vascular or tumoural) all had Df in the normal range. Analysis of shape using this method identifies subtle abnormalities of the cortical ribbon, and has potential application to images of the human brain in a wide range of clinical and pathophysiological settings.