A geometric classification of fault breccia borrowed from the cave-collapse literature has been suggested as an alternative to available genetic classifications. Here, image analysis is used to explore geometric discrimination between the visually assigned classes of crackle breccia, mosaic breccia and chaotic breccia, using samples from the well-understood Dent Fault, northwest England. Clast sphericity and surface roughness show some correlation with the breccia classes, but particle size distributions and their fractal dimension show none. A more useful parameter is the percentage of sample area occupied by clasts. Crackle breccia has >75% clasts, mosaic breccia 60–75% clasts, and chaotic breccia has <60% clasts. The eye is also good at judging the tessellation (goodness-of-fit) of clasts, and a semi-quantitative approach to assessing this parameter is explored. The average degree of rotational misfit of the clasts is strongly related to breccia class: crackle breccia involves less than 10° average rotation, mosaic breccia 10–20° and chaotic breccia more than 20° rotation. Comparison charts are provided for semi-quantitative classification of fault breccias.