Paradox of Self-Amendment by Peter Suber law. All rules deducible from those premises would already be law, although perhaps tacitly, and no rules not so deducible could be made law. New enactments could only make explicit that which was already law implicitly. If a proposed rule were not a strict implication of existing rules, then it could not be inferred from them and therefore could not be enacted. From any initial position, therefore, all possible law is already actual, even if we do not know it yet, just as "Socrates is mortal" is already implied when we have asserted "All humans are mortal" and "Socrates is human". But this absurdity will be avoided by a strict inference model —in the name of a greater absurdity. Law is only changeless if there is law. But a strict inference model implies that there is no law. If legal validity requires that a law be the conclusion of a valid inference in which the premises are already existing valid laws, then we may say more briefly that every valid law is derived from prior valid law. We can take what appears to be present law and locate its authorizing antecedents, but those antecedents are only valid if they have their own valid premises, and so on in an infinite regress. If we allow no exceptions in the inference model, then no rule in the history of authority can be self-validating or validated by an alegal source. The only alternative is an infinite regress of authorities; every valid law requires an infinite genealogy. But no law has an infinite genealogy. Therefore there is no valid law. The exceptionless inference model must say either that there is no law, or if there is, that it is all immutable. It is no accident that this is what systems of formal logic must say about logical theorems, for Ross reduces legal validity to logical validity. If every step of every inference must be concluded from premises in a prior inference, then we cannot begin with axioms because they have no antecedent authority. Hence we fall into an infinite regress. But if we do begin with axioms, then once the rules of inference are specified, all permissible conclusions are instantly determined, even if human beings never finish discovering what they are. We may avoid these absurdities only by admitting exceptions in the inference model as applied to law. These may take the form of self-validation, which Ross directly opposes, or an alegal and alogical source of authority, such as acceptance, for at least some rules, which he indirectly opposes. Note that the alegal source of authority can be many things other than acceptance, such as moral law, military might, or divine command. But it cannot be logic. Ross and the legal formalists represent a new form of natural law theory: one that rests the validity of civic or positive law on logical law rather than moral law. But if formal logic is the alegal source of authority for law, then we encounter the dilemma that laws are only as valid as their premises, and these either recede infinitely or toward some exceptional premise that is self-validating or validated outside the logical and legal systems. This dilemma does not confront the moral version of natural law theory, or an acceptance theory, because these do not make legal validity depend on the logical validity of an inference; hence the pedigree of premises may stop at any point. Natural law theorists and others may nevertheless fall into the trap by searching for the authority of fundamental laws as if for premises of conclusions; but they need not do this. They may say that legal validity comes from moral law, period, or from acceptance, period. If they defend this claim with premises, they will confront the age-old epistemological dilemma of circularity versus infinite regress, but this affects only their theory, like any theory, not the phenomenon of legal validity itself. The acceptance model does not allow logical rules to interfere with processes determined by legal rules and supervised by mortal judges. Whatever legislators and the people make in the way of new law is new law, if it satisfies the mortal judges and the people themselves; whatever is not regarded by anyone as valid law is not valid law even if it follows infallibly from rules that are accepted as valid law.[Note 9] New law must conform to the rules of making new law, but this conformity is not determined in ideality by logicians reducing the rules to some core of univocal purity. It is determined by judges using a complex web of non-formal methods. While judges try to follow both prior law and their own notions of eternal logic, their errors are as authoritative as their other decisions, unless some other judges overturn them or acceptance wanes through neglect or disobedience. The inference model blurs the distinction between logic and law, while the acceptance model blurs the distinction between law and history. The inference model subtly identifies logical and with legal validity, or contradiction and illegality, while the acceptance model subtly identifies de jure and de facto validity, or legality and acceptance. B. Ross's solution: the invisible, immutable amendment clause Ross concludes that the contradiction of self-amendment means that valid amendments to ACs cannot derive their authority from the ACs themselves. Either they are merely accepted, which to Ross is to be legally unauthorized and invalid, or they derive their authority from a tacit rule transcending the constitution posited to validate precisely such changes. Logically Ross has no grounds to choose between these options, but he prefers the second, as if acknowledging that amendments to ACs do occur (which he never explicitly admits) and must be explained. Hence, attempts to amend an AC are not invalid exercises in futility, but neither when they succeed are they cases of self-amendment. Amendments to ACs derive their authority from a tacit rule, transcending the legal system, which states that the AC may be used to create its own successor, but that the successor is made valid only by the tacit, transcendent rule. This is a rudimentary theory of types. The highest positively enacted rule of change in the system is evidently mutable, but logically it cannot change itself or be changed by any other positive rule of change. Hence it cannot really be supreme and must have a tacit superior. The tacit rule of change might be called a tertiary and the constitutional AC a secondary rule, but of course Ross does not use these terms. The positing of higher levels could go on forever in principle except that Ross sees no need to think of the transcendent rule at the first tacit level (the tertiary rule) as mutable. If it were mutable we would have to go to the second tacit level (to a quaternary rule) to find the authority for changing the tertiary. But if the first tacit rule is immutable, then the tacit series stops with only one transcendent rule, necessarily immutable, and conveniently "discovered" transcending every mutable constitutional AC. 31