On the Inverse Separation Problem