Paradox of Self-Amendment by Peter Suber approach, and decides that the "purpose" of parliament is to convert the "inchoate will of the nation into a set of concrete rules," and therefore that it could not commit suicide,[Note 38] legally....For the effect would be to deprive the nation, which can only speak through Parliament, of all legal means of expressing itself, and that would be clearly contrary to the purpose for which it exists. Middleton suggests that insofar as power is restrained by the purposes of its existence, parliament cannot commit suicide; but he leaves the door wide open for the interpretation that parliamentary suicide would be improper, not illegal, a betrayal of trust more than a violation of law. Using similar reasoning, William Marbury has argued that the "power to 'amend' the Constitution was not intended to include the power to destroy it."[Note 39] In the theological version of the paradox of omnipotence the parallel argument is that the deity has self-embracing omnipotence, and logically could limit or destroy itself, but will not because some auxiliary blessing, such as infinite goodness, argues in favor of the greater utility of remaining unlimited in power.[Note 40] Like the moral injunction instead of the legal one in the old Virginia statute, this argument is something of a cheat. The deity's goodness is usually described in terms that make its verdicts necessary, not contingent. Hence, the deity whose goodness prevents self-limitation should actually be said to possess continuing, not self-embracing, omnipotence, unless the omnipotence can repeal the goodness: but presumably the goodness itself will prevent that. One may wonder why the self-destruction of parliament or Jehovah, or the self-repeal (without replacement) of a constitution, might not be even more permissible than the repeal or destruction of lesser entities. Wittgenstein noted, in a different context, that one who doubts that Napoleon existed is not of sound mind; but one who doubts that the world exists is on firmer ground.[Note 41] If one accepts the world but doubts Napoleon, then one is accepting a system of evidence that strongly supports the existence of Napoleon. But if one rejects the whole system of evidence, nothing remains to confute one's doubt. Similarly, if an AC repeals itself, rules are left behind that might suffice to declare the act invalid, such as (for Alf Ross) legal rules incorporating the principle of non-contradiction. But if the constitution is repealed, or better, if the British parliament commits suicide, then at the very least the validity of the act will be injusticiable, if not actually beyond the strictures of the rules and criteria it took with it.[Note 42] Moral and logical rules will remain after every such act, except possibly after nuclear annihilation, and while both may find self- destruction objectionable, they cannot find it unlawful. E. Self-repeal Suppose someone were to object that a particular self-entrenched AC could not disentrench itself because it had continuing, not self-embracing, omnipotence. This would be a plausible objection for any AC, for self-disentrenchment presupposes some self-embracing power, although not necessarily self-embracing omnipotence. However, the objection is ill-founded. If the AC is truly characterized by continuing omnipotence, then it must have the capacity to authorize all amendments that are not inconsistent with its continuity or omnipotence. That means it must have the capacity to undo by amendment all limitations on its continuity or on its remaining omnipotence, and that implies the power of self- disentrenchment. All limitations on its power (except the limitation that makes it continuing) are revocable by it. An AC of continuing omnipotence could not entrench itself in the first place, except revocably. If the self-entrenchment clause were original, not self-imposed, then the AC could still undo it through a self-application that merely revoked the revocable limitation. If the original limitation is irrevocable, then the AC does not have continuing omnipotence. Therefore, the inability of continuing omnipotence to amend itself in significant ways is not a barrier to self-disentrenchment. If the limitations to be disentrenched were self-imposed, then they are necessarily revocable, because continuing omnipotence cannot limit itself irrevocably. If they are original, then either they are revocable, or the AC does not actually have continuing omnipotence. A subtler problem lurks here. How do we know whether an AC has self-embracing or continuing omnipotence? Hart says the question is empirical. He does not say that the decisive empirical question is whether self-limiting self-amendment has occurred. He says it is "an empirical question concerning the form of rule which is accepted as the ultimate criterion in identifying the law."[Note 43] This suggests that the empirical inquiry seeks to ascertain whether the rule of recognition sees the AC as self-embracing or continuing. This just postpones the question of what evidence is to count either way. One empirical test that might be the language of the AC. However, no AC that I know of explicitly says whether it is self-applicable. The closest that I have seen is New Mexico's, which implies self-applicability by denying that one section of the AC can be amended except in convention. Many ACs say they apply to "this constitution", which implies self-applicability, but not enough to satisfy Alf Ross, for example. But self-applicability is not really decisive on this question, even if it is suggestive. Self-embracing omnipotence requires self-applicability. But so does continuing omnipotence, in order to repeal the revocable limitations on its continuing power or to add others. Note that if an AC were reworded so that it explicitly permitted all kinds of self-amendment, including self-repeal, self-entrenchment, and self- disentrenchment, then the paradox of self-amendment would not be any closer to solution or dissolution. The reason is that a self-amendment under the old AC inconsistent with its original form would still make an invalid inference. And of course an amended AC could still become inconsistent with its original form by, say, prohibiting what was once permitted or permitting what was once prohibited (see Section 12.C). 69