We see the forests before the trees but how do they combine? Shape information combines linearly across multiple scales