The orthogonality of subduction: An empirical rule?