How do most predicativists justify their acceptance of an impredicative definition of the set of natural numbers? Is there a philosophical (not practical) argument that justifies this exception? GL