Paradox of Self-Amendment by Peter Suber Adding a method with self-embracing omnipotence would solve all problems if it were a coherent idea. However I stand by my arguments in Section 11.B that it is not, at least for the inference and acceptance models of legal change and validity. In any case the possibility may be eliminated for sport, since it would undoubtedly allow any AC to reach any content. If one has taken the short-cut of adding a method of continuing omnipotence, B, and then used it to put the other methods into desired shape, then one is stuck if one then desires to leave the entire AC finite in power. The method of adding an omnipotent, specially armed assassin will never accomplish its goal without an infinite regress of assassins of assassins. Still, there are a few paths out. If one is willing to await and risk future contingencies, then B might repeal itself, Alf Ross and formal logic notwithstanding. Or it might institute a rule of desuetude and later fall to it. But no strategy that requires future contingencies can guarantee the desired result. One possibility is that B could be repealed by C when C is not itself omnipotent. If A adds C carefully, and C has the ordinary competence to repeal clauses of the constitution, then C could repeal B without itself being omnipotent; even under the inference model, there is no requirement that the rule of change be "more powerful" than its prey, insofar as power is distinct from hierarchical priority. Under the acceptance theory, there is no requirement that a rule of change be prior to, or higher than, its prey. If finite A can add omnipotent B, the apotheosis of the AC, then finite C could repeal B, the Götterdämmerung. Method B has continuing omnipotence; therefore its continuing character purports to be immutable. Under the acceptance model, if any rule of change can disentrench and amend an immutable rule, or transmute an immutable rule into a mutable one, then a finite rule of change can do so. Omnipotence need not be arrayed against immutability, for the outcome does not require the infinite power of formal validity, only the actual acceptance of the people and officials. To put it another way, the omnipotence needed to undo immutability need not lie in the legal power as formally described in legal terms if it lies in the acceptance of the people and officials who intend to use that power to undo immutability. Another reason is that some immutable rules are made immutable by entrenchment, not self-entrenchment; and even if an entrenchment clause is complete, it is itself mutable and repealable so long as it is not itself entrenched. There is nothing inherent in the idea of continuing omnipotence that requires us to think that its continuing character is self-entrenched rather than irreflexively entrenched. The continuing character must be entrenched against the continuing power, but that can be accomplished by irreflexive entrenchment. If continuing omnipotence is made continuing by irreflexive entrenchment, then it cannot directly limit itself, but it can repeal the (perhaps tacit) entrenchment of its continuing character and then limit itself. However, as noted in Section 11.B, even without an immutable continuing character, no AC can be thought to limit itself irrevocably. But if B, a method of continuing omnipotence, is made continuing by irreflexive entrenchment, then A, B, C, or any other method can repeal B's entrenchment clause and then repeal B. Any strategy that requires transmutation or the repeal of a limitation that purports to be immutable is to that extent impossible under the inference model. If such a strategy depends on the acceptance model, then it relies on future contingencies. If it relies on the procedural model, then it need not rely on future contingencies, shifts of acceptance, or invalid inferences. Therefore, if one is uncomfortable pinning the success of the see-saw method on contingencies, then the procedural model may save the technique. There are other ways to save the technique without resorting to the procedural model or relying on contingencies. Note, however, that one of the purposes of developing the see-saw method was to enable a people to change its AC without self-amendment, and therefore consistently with the inference model. The inference model, it turns out, can be satisfied only if the short-cut through continuing omnipotence is not taken, for the inference model does not tolerate the repeal of immutable rules such as the immutable character of continuing omnipotence. But the acceptance model can tolerate this, and can do so without resort to future contingencies. (However, if we will appeal to the acceptance, we need not resort to the clumsy see-saw method, but may turn straightaway to self-amendment.) For example, method A may add B with a sunset clause that asserts that B will expire after a certain time. Then B could not be said to possess "continuing omnipotence", strictly speaking, for ab initio its character is mutable and temporary. But there is nothing to prevent B from having omnipotence for a finite time. If it is limited during that time only in its power to repeal or limit itself before its expiration date, then we could even say that B had "continuing omnipotence" for a finite time. Its continuing character would be defined after, and relative to, its expiration date, so that even for the inference model there is no "creation of Frankenstein" with a power to overwhelm its creator. Use of B under these circumstances would resemble algebraic manipulation with the square root of negative one, which is an impossible quantity that drops out before the final product is determined, facilitating a process that might otherwise be impossible. Similarly, the sunset clause governing B need not repeal B but merely change it into a finite, mutable amending power at the expiration date. Then the sunset clause would resemble an incomplete more than a complete entrenchment clause. If A is not capable of apotheosis, or of adding an omnipotent B, or if the short-cut through omnipotence is simply avoided, for sport or logic, then the possibility that any AC could reach any content is probably indemonstrable and irrefutable. If the limitation on A is "hereditary" and applies to every method added by A, then some limitations might never be overcome by the see-saw method except through future contingent shifts of acceptance. One important hereditary characteristic for the inference model is consistency with the premises or axioms (provided they are consistent with one another). This "genotype" would prevent any amendment or repeal of any immutable rule, or possibly (as noted in Section 12.C ) any substantial change at all. The acceptance and procedural models allow valid "mutation" at every generation, in principle, affecting every limitation on the rule of change and thereby every limitation on every rule of law. 98