Paradox of Self-Amendment by Peter Suber repealing) the older rule in cases of conflict. But on the other hand, the lex posterior principle is denied by Ross's preferred method of avoiding contradiction in law, namely, to invalidate every attempt to derive a conflicting norm from an existing one, never allowing a conflicting rule to come into a position to impliedly amend or repeal an older rule. Moreover, if a pre-existing entrenchment clause forbids certain amendments, then the inference model would probably read any attempt to ratify an amendment of the forbidden type as a violation, not a repeal, of the language of the constitution. Most models of legal change would read such an attempt this way, which suggests that the old AC could neither transmute nor amend a complete self-entrenchment clause. It would be prevented at least by contradiction, if not also by incompetency. The acceptance model is free to read it otherwise, however, if the attempt to ratify a forbidden amendment is allowed to run its course. For the dynamics of acceptance by the people and officials of the system, as opposed to whatever might be happening in ideality according to logicians, may validate the amendment despite its inconsistency with antecedent authority. This would make the amendment what it purports to be, an amendment, not a violation, of the language of the constitution. It is the lex posterior principle in action. The inference model, then, probably forbids the enactment of complete self-entrenchment clauses on the ground that through them the AC limits its own power and performs self-amendment. But if such a clause is an original limitation on the AC, not self-imposed, then the limitation on the AC is not an amendment and hence not a case of self-amendment. Therefore it may limit the AC immutably for the inference model. For the acceptance model, by contrast, an amendment could repeal, rather than violate, the language of the constitution, since its authority need not rest exclusively on existing provisions of the constitution, but might arise instead from present or future acceptance. The acceptance of the people is a source of authority that could justify the repeal of a complete self-entrenchment clause by an AC that was general in its language and silent on this special application. Under the acceptance model the violation or contradiction involved in allowing the AC to amend a provision that it is forbidden to amend can be cured if it contingently meets with the approval of the people and officials. The act still violates written law, but under the acceptance model that is not decisive; acceptance overrules written law, both to validate and to invalidate changes of law, even if it does so only rarely. If the people want to repeal a complete self-entrenchment clause, in short, logic will not stop them, and law will not stop them in the name of logic. Law may stop them on other grounds, of course, and as noted, the very provisions of the constitution that the amendment violates will have their effect in determining what is accepted. Under the inference model an original complete self-entrenchment clause must be immutable, and a self-imposed self-entrenchment clause cannot exist. If, however, a self-imposed self-entrenchment clause could be enacted under the inference model, one might argue that it would be mutable, even if the self-entrenchment were complete. To make the argument work one must appeal to a principle of customary or "natural" law, often found in dicta, that a legal power can unmake anything that it can make. Such a rule is not at all unlikely, for it is a variant of the rule that one generation cannot bind its successors except revocably or in manner and form. If a proponent of the inference model finds this rule available to her, then she might allow some self-imposed complete self-entrenchment clauses to be mutable, if only they could exist to begin with. Note, however, that this principle that legal powers can unmake anything they can make may give the people a legal right to abolish the constitution (see Section 18). This could result in the repeal of the AC, although not by the procedure of the AC. Proponents of the inference model, if they pick up the principle at all, are likely to qualify it by holding that a legal power can unmake anything it has made, but only by following the procedures then in effect, including any procedures made by the power to be unmade. If this qualification is not to reinstate the immutability of complete self-entrenchment clauses it must give an AC continuing omnipotence, or at least sufficient power to transmute or amend completely self-entrenched rules, even if no other existing procedure would allow it. The qualification could not give the AC self-embracing omnipotence, for then it could make something that it could not unmake, namely, a genuinely immutable limitation on the AC. Enacting an immutable limitation on the AC would be impossible under the principle that legal powers can unmake what they make. It would also be impossible under the inference model stripped of that principle, since it would amount to self-amendment. Note that the qualified principle directly gives the AC the power that was sought in intermediate rules; it makes complete self-entrenchment clauses mutable by stipulating that the AC can amend them, not by appeal to a rule independent of the AC that could justify the result. If the AC was never limited by an immutable rule, then its repeal of a complete self-entrenchment clause should not count as an augmentation of its power, even if it is self-amendment. However, if even the mutable complete self-entrenchment clause is considered a limitation on the AC, albeit revocable by the AC, then repealing it would take us from a temporary period of limited power to a succeeding period of unlimited power, which is certainly a kind of self-augmentation of the power of the AC. It is like a person unlocking a room in which she had locked herself, or resigning from a game she had agreed to play. That would be a kind of self-amendment, revoking revocable self-limitations, and so would disqualify it under the inference model. The acceptance model may authorize the repeal of a complete self-entrenchment clause by virtue of the capacity of the people and officials to overlook or tolerate contradiction. But there is a neater way. Suppose the AC amended itself[Note 20] by addition, appending to itself the following clause: Through the procedures described in this Article the people may amend or repeal any provision now or ever to be part of this Constitution, without exception, notwithstanding any protective language, any supposed inherent or implied limitations on the amending power, any prohibition against self-application, any conjectured or actual reservations intended by any maker of this Constitution, or any future amendment to this Constitution not made in convention. 46