Section 10
Attempts to Dissolve the Paradox: Time
Peter Suber, Paradox of Self-Amendment Table of Contents

A. Ross's answer to the time-based objection

Most people's first response to the paradox of self-amendment is that, even under the inference model, it is spurious and need not arise. One such avenue of attack is an objection based on time. If the validity and supremacy of the old AC exist under a time limit, or if the new and old ACs are never in effect at the same time, then perhaps self-amendment will be permissible under the inference model. Perhaps the old and new ACs are not inconsistent if they never overlap in time, especially if each says that they do not overlap. If the validity of the old AC exists under a time limit, then the limit may be pre-determined, as by a "sunset clause" which brings about the expiration of the whole section at a given time (see Section 14), or it may float and become fixed whenever the AC is amended. Most attempts to solve or dissolve the paradox depend on the fact (if it is a fact) that the old and new ACs were never simultaneously valid. No such objection, however, suffices to validate self-amendment under the inference model, as we will see.

First, it is clear that there is nothing problematic in supposing that an AC's period of validity does not overlap that of its successor. On the contrary, the alternative is difficult to conceive. If an AC and an amended form of itself meant to be its successor did overlap in time, then for the length of the overlap we could not say that the original AC had been amended. The new AC could stipulate that it took effect upon its adoption and that its predecessor would be repealed five minutes later. Even if there is no intolerable contradiction in this idea, it would be legally pointless; during those five minutes two contradictory AC's would enjoy equal validity. Could we use either one we liked, or both, or neither?[Note 1]

Similarly, it is difficult to conceive a gap between the expiration of the old AC and the effective date of the new AC. If there were such a gap, then the usual reasons for the expiration of the old AC would be lacking. Again, the new AC could stipulate that it would take effect upon its adoption and five minutes after the expiration of its predecessor. But it seems that it could only declare the expiration of the prior AC if it were itself already valid; its predecessor would have to be valid at least up to the moment that its successor became valid.

So let us agree that there is nothing problematic about the old AC becoming invalid and the new AC becoming valid at the same dimensionless point of time, with neither gap nor overlap. The core of the time-based objection to Ross's paradox of self-amendment is that, even if there is some kind of contradiction between the new and old AC's, the temporal separation of their periods of validity eliminates it, or tames it, just as two countries on a map can be drawn in the same color if they meet at only one point.

Ross meets this objection directly by distinguishing legal and logical contradiction. Legal contradiction exists between inconsistent laws valid at the same time. If one is repealed or invalid, then the legal contradiction disappears. Logical contradiction is an inconsistency between propositions taken more abstractly. If they contradict one another by their form or content, regardless of the accidents of human history that give some propositions legal effect, then they are logically inconsistent. If two propositions of law contain a logical contradiction, and if they are valid at the same time, then repealing one of them will eliminate the legal contradiction but not the logical contradiction between them.

Ross's first answer to the objection is that temporal separation can only cure legal contradiction, not logical contradiction. It can prevent inconsistent rules (the new and old AC's) from being valid at the same time; but it cannot remove their inconsistency in the logical sense. Indeed, the desire to remove the legal contradiction between them presupposes an acknowledgment of their logical contradiction. Yet it is the logical contradiction that invalidates the inference that models the act of amendment. And if the deduction corresponding to the amendment is logically invalid, then the amendment is legally invalid. (For the possibility that a time-based response could satisfy a formal logic even if it did not satisfy Ross's inference model, see 10.B below.)

The attempt to dissolve the paradox by temporal separation of the old and new ACs misconceives the essence of the inference model. The validity of deductive inference is the criterion of the legal validity of amendments, and the validity of deductive inferences is not a temporal phenomenon. The premises and conclusions are examined for consistency without regard to time, because any inconsistency between them will eternally lie between them. If the premises pass out of effect as the conclusion is established, that legal adjustment does not bear on their logical relations.

This is not only the approach of Ross and the inference model, but also the approach of logicians who must try to address the issue with logical tools only, without the benefit of a legal or divine capacity to tolerate contradiction.

Ross puts it this way: the time-based objection[Note 2]

confounds legal with logical contradiction. There is no contradiction in law, because [the new AC] supersedes [the old AC]. But why does [the new AC] supersede [the old AC]? Precisely because [the new AC] logically, that is corresponding to its content, contradicts [the old AC]. This follows from the well-known lex posterior principle according to which in case of conflict between two equivalent norms (i.e., norms at the same level in the hierarchy) the later precedes the earlier one.

If we grant Ross the distinction between legal and logical contradiction, then he is certainly right to say that temporal separation can avoid the former but not the latter. (One sense in which the logical contradiction can be removed is discussed in 10.B below.)

But for those who reject the distinction, Ross has another reply to the time-based objection. If a rule of law is authorized by another rule of law, then usually the authorizing rule must be legally valid in order to lend its authority to the authorized rule. For example, if a court is defined by statute and the statute is repealed, the court will usually be dissolved at the same time. If a rule of law remains authorized even after its authority is repealed, then it possesses what J.M. Finnis called "transtemporal validation". Ross rejects the possibility of transtemporal validation for reasons central to the inference model: it does not work in valid deductive inference, so it cannot work in valid legal change. When in a given inference, a conclusion relies on a certain premise, then if the premise is denied, the conclusion is no longer proved. The conclusion may be inferred anew from other premises, of course, but in any given inference the premises that authorize the conclusion must be affirmed, not denied. So in law, if a rule is repealed, everything it authorized (that is not authorized by something else) must cease to be authorized.

So if an act of amendment requires transtemporal validation, then it violates the inference model. But temporal separation in the case of strict self-amendment requires transtemporal validation. If the new AC is authorized by the old AC so that strict self-amendment occurs, and if the old AC is invalidated just as the new AC is validated, then the new AC is either unauthorized or transtemporally authorized.

The only alternative under the inference model is for the new AC to be validated by something like Ross's tacit, transcendent rule, which remains in effect as long as we like. But this is to deny that strict self-amendment has occurred (which is part of Ross's strategy). In strict self-amendment the authority for the new AC must be the old AC. So if the new AC is not authorized "from the grave" of the repealed old AC, transtemporally, then it is not authorized at all —for the inference model.

In short, temporal separation will not cure strict self-amendment for the inference model. If we accept the distinction between legal and logical contradiction, then temporal separation only prevents a legal contradiction. The logical contradiction survives between the conclusion (new AC) and the premise (old AC) from which it derives its validity. This suffices to invalidate the inference that models self-amendment (when the premises are consistent with one another). On the other hand, if we reject the distinction between legal and logical contradiction, then temporal separation forces us to appeal to transtemporal validation for strict self-amendment. This violates the inference model just as much as contradiction.

B. "Valid until amended", temporal indexing, universal self-entrenchment

What if the old AC says that it is valid only until amended or repealed? Wouldn't that prevent both legal and logical contradiction between the new and old AC's? The old AC need not assert its exclusive and permanent validity; if it does not, then it is compatible with its replacement in the new AC.

This proposal is more complicated than it appears. There are at least four points to make about it. First, it will probably work as advertised, if done with sufficient care. But the inference model requires more than simple care in draftsmanship. It requires a model of deductive inference than can take the proposed nuance into account. So far, temporally indexed deontic logics have only been sketched;[Note 3] they must be fully fleshed out and applied to self-amendment. However, the success of other temporal logics in coping with time-dependent truth values[Note 4] gives me confidence that this project can be fulfilled.

Second, the natural way to fulfill it would be to make the old and new AC's logically consistent with each other (by temporal qualification of their claims to validity), while enforcing their temporal separation. Even if this method succeeds in avoiding both legal and logical contradiction, it will require transtemporal validation. Hence, it will satisfy our carefully constructed formal logic without satisfying the inference model.

Third, there are new paradoxes attaching to "sunset clauses" that declare time-limits on their own validity. In particular, after they expire, if we want to claim that they are really invalid, the only obvious legal authority to which we might appeal for their invalidity is their own statement of a time-limit on their validity. But they are presumably invalid. If they remain perpetually valid in some dormant, low-level way, sufficient to tell us authoritatively that they are invalid, then at the very least they are self-excepting ontological monsters; at worst, they contradict the time limitations they assert either by becoming wholly invalid or by remaining partially valid. (These paradoxes will be explored further in Section 14.)

Fourth, despite first impressions, the logical contradiction between the old and new AC's is not removed simply by rewording the old AC so that it specifically authorizes its replacement or its self-amendment. (More on this in Section 11.)

Ross says that the old AC is invalidated or repealed by an inconsistent new AC, adopted under the tacit, transcendent rule. The old AC is repealed by virtue of the legal rule that favors the most recent rule in conflicts between two rules of the same hierarchical level (see Section 16). This rule is often called the lex posterior principle. Ross anticipated the objection that if the old AC incorporated the lex posterior principle, in effect authorizing its own repeal as its successor is established, then the old and new ACs would no longer be inconsistent. This he denies. An AC that incorporated the principle might say,[Note 5]

The rules of the constitution are amendable by process P and only by this process, until by this process it is decided otherwise.

If such an AC amended itself by changing P for Q, when P and Q were inconsistent in content, then the inference by which the new AC would have come into being would still contradict one premise, the old AC, which would still create an insurmountable invalidity in the inference.

Would the incorporation of the lex posterior principle still fail to dissolve the paradox if the old AC said it was to expire at time T (a dimensionless point in time), and if the new AC said it was to take effect at T? The question suggests that some sophisticated logical move is to be made, requiring a different answer from the original situation. But that is illusory. To "incorporate the lex posterior principle" is simply to make it explicit within the new and old ACs that their periods of effectiveness do not overlap, or that the new one repeals the old one to the extent of their irreconcilable inconsistency. Clearly, making this rule explicit in the ACs themselves does not change the fact that they are inconsistent with one another. Moreover, as before, it even presupposes their inconsistency. If one's criterion of inconsistency is the prohibition of what was once permitted, or the permission of what was once prohibited (see Section 12.C), then clearly the new and old ACs could declare different periods of effectiveness without thereby overcoming their inconsistency. If they are inconsistent, then the inference that models the adoption of the new from the old will be invalid, and the formalist model cannot permit it. The old AC is still being asked to authorize a norm inconsistent with itself.

What Ross calls the incorporation of the lex posterior principle may also be called the implied limitation on the AC that it is valid only until amended. As noted in Section 8.C, such a "limitation" does not diminish the amending power in any way, at least not while the AC is still valid. What it does do is insure that the old and new ACs never overlap in time, which might be desirable if one could not rely on the courts to apply the lex posterior principle exogenously. A strong argument could be made that any AC is impliedly valid only until amended, notwithstanding the Kelsenian super-entrenchment of ACs noted in Section 9.D. This is especially so if "valid only until amended" means that an AC in its current form binds only in that form until it is changed to another form. This is nearly a tautology, but not quite: if self-amendment is permitted, it is even closer to tautology, and only becomes a tautology if irrevocable self-amendment is permitted. If an AC has self-embracing omnipotence, then it may limit or repeal itself irrevocably, and therefore invalidate itself. If it has continuing omnipotence, then all self-limitations are revocable; but while they last they genuinely change the AC and bind in another form, even if they cannot totally change the AC. Moreover, the very immutability of the continuing character of continuing omnipotence shows that, if it could be changed irrevocably, then its original form would be invalidated.

Because the "valid until amended" limitation is simply the incorporation of the lex posterior principle, it therefore cannot be read into an entrenchment or self-entrenchment clause, insofar as the latter are temporally symmetrical correlates of implied repeal and the lex posterior principle (Section 9.D). If "valid until amended" means that newer rules impliedly repeal all inconsistent older rules on their own hierarchical level, pro tanto or to the extent of their irreconcilable conflict, and if entrenchment is the reversal of this rule of priority, allowing the existing rule to repeal the newcomer pro tanto, then entrenchment clauses, taken at face value, reverse and do not incorporate the lex posterior principle. If they are mutable and repealable, then it is by special effort, not by the sort of change so routine that its permissibility may automatically or presumptively be read into them.

In one of the few general studies of the logic of rules, Joan Safran Ganz concluded that one of the recurrent features of rules is that they have been adopted and "remain in force until unadopted".[Note 6]

As noted in Section 9.D, the Rhode Island constitution is the only one in the nation that says of itself that it is valid until amended: a clause entrenching the right to alter or abolish government adds that "the constitution which at any time exists, till changed by an explicit and authentic act of the whole people, is sacredly obligatory upon all" (emphases added). Most states are content to presume that their constitutions are valid only "till amended". But Rhode Island is the only state to endure the upheaval of a revolution in which insurgents with strong popular support set up a constitution and government in opposition to the incumbent. After its bitter experience in which the Supreme Court refused to decide which was the legitimate government of the state, and in which President Tyler supported the incumbent with national troops, Rhode Island was understandably careful in its subsequent constitutional convention to assert both the lex posterior principle (to satisfy the insurgents) and a rough test of revolutionary legality (to satisfy the incumbents).[Note 7]

With the memories of the new nation's revolution still bright in his mind, Alexander Hamilton wrote in similar vein in Federalist #78. The most basic laws are mutable, but only by the forms and procedures the people have imposed upon themselves in their constitution.

Until the people have, by some solemn and authoritative act, annulled or changed the established form, it is binding upon themselves collectively, as well as individually; and no presumption, or even knowledge of their sentiments, can warrant their representatives in a departure from it prior to such an act.

This is a strong statement of the "self-entrenched" character of all law, which requires that its own forms be followed. But for the same reason it is a strong statement of the principle that law binds until changed —in some way the opposite principle. Hamilton is asserting the balance between stable old law that demands respect, and the rights of new law to take over when solemn formalities have indicated consent. It is strange and wonderful that this play of opposites should sound nearly tautologous to our ears. It is no accident that this balanced view is made explicit in the turmoil and long memories following revolution, and set up as a warning. It is important at those moments to reassert the unexceptionable rule of law, but equally important to reassert the mutability of law.

C. The procedural model

Under the procedural and acceptance models the paradox need never arise, either because the new AC is validated directly by acceptance, not by the old AC, or because the contradiction between the new and old ACs is tolerable under the procedural and acceptance models. No effort need be made, theoretically or practically, to insure that the new and old ACs are never in effect at the same time, for any overlap and contradiction may be excused or ignored.

We can distinguish several theories of the source of the validity of a newly changed AC. With Ross, the source of legal validity may be the logical validity of a certain inference from a prior, higher rule of change. With Hart, it may be the acceptance of the people and the usage of officials, or conformity to a rule of recognition that traces its own authority to acceptance. These are the inference and acceptance models of authority and validity.

A third model lies between the inference and acceptance models in many ways. It may be called the procedural model. It derives the validity of a newly changed rule from compliance with the demands of a valid procedure. Procedures are legal entities, not logical ones, and may perform whatever valid law permits them to perform, not only those things that can be validated by sound deductive inferences from prior rules.

If the procedure for amendment requires x, y, and z for a valid, final, and complete act of amendment, and if x, y, and z have certifiably been performed, then amendment has occurred and is valid under the procedural model. The importance of the procedural model, and what distinguishes it from the inference model, is that the amendment will be valid even if there is reason to find the outcome inconsistent with x, y, or z, with a rule of change, or with itself. The procedural model envisions outcomes abstractly, as separate and separable from their mode of becoming, and without any "backward-looking" consideration for the procedure itself or its premises in the particular case. Under the procedural model, the outcome of a judicial procedure may deny the jurisdictional basis of the proceeding without undermining the outcome itself.[Note 8] The outcome of a legislative investigation may reveal the total absence of the required basis for a legislative investigation without thereby invalidating the outcome itself.[Note 9] The outcome of a jury deliberation may logically imply the opposite outcome without subverting itself or legally requiring that the opposite outcome be given effect.[Note 10]

Logically the procedural model is anomalous, if not scandalous. But that is just to say that it differs from the inference model. If procedures are pictured as inferences, then the procedural model permits invalid inferences to make valid law. For the procedural model, the validity of the outcome does not depend on the logical validity of any inference, and therefore does not depend on consistency with any antecedents; it depends solely on the legal validity of the procedure itself and the satisfaction of its requirements. In short, under the procedural model to picture procedures as inferences is inaccurate, for it suggests that some permissible procedures are impermissible.

Hart's exposition of the acceptance model does not go into sufficient detail to allow us to determine its exact relations with the procedural model. Both models will allow the outcomes of procedures to contradict indispensable premises, although for different reasons. For the acceptance model, procedures may be defectively applied, or bypassed for invention; for the procedural model, acceptance is irrelevant once the procedure is known to be lawful.

D. Other views

The inadequacy of the time-based objection is the main reason why I prefer Hart's acceptance model to Hart's own explicit answer to Ross. In his later essay, "Self-Referring Laws,"[Note 11] Hart uses the time-based objection and did not notice that a solution wrought from his own acceptance theory applies elegantly to the problem and far surpasses the adequacy of temporal separation. Speaking of the American constitution and its AC, Hart said,[Note 12]

It may be true, at least in some cases, that a norm cannot be derived from 'a source of derivation' with which it conflicts. But there is no conflict if Article V is amended in accordance with its own provisions. For the original Article V and the amended Article relate to different periods of time: the original procedure is to be used until it is replaced by the new, and the new procedure is to be used thereafter.

This argument is unavailing against the inference model, and unnecessary for other models. Hart seems to sense this, and after quoting Ross on the inferential structure of legal change, replies,[Note 13]

To this it might be objected that the exercise of legislative power to introduce new norms is not a deductive inference and it is not clear how this logical principle applies to a legislative act.

Hart has suggested an adequate solution, I contend, simply by rejecting the inference model. But this is far different from dissolving the paradox for the inference model or for formal logic generally. Hart continues, saying that an act of self-amendment that led to an inconsistent new AC is in any event perfectly possible and coherent if the old AC had self-embracing omnipotence. This proposal will be considered in Section 11.

Ross's statement of the paradox of self-amendment evoked many responses. The Scandinavian comments all relied on the time-based objection.[Note 14] In English I have found only six responses in addition to Hart's: by Geoffrey Marshall, Joseph Raz, Norbert Hoerster, J.M. Finnis, J.C. Hicks, and William F. Harris II.[Note 15] Of these Marshall, Raz, and Hoerster all urge the time-based objection (with other objections), and Finnis shows its insufficiency; Hicks generally supports Ross. Harris does not decide the adequacy of Ross's case.

Marshall argues that a "phased" rule of amendment that explicitly said that its power will be limited by its successor will avoid the self-referential aspect of the paradox.[Note 16] It will not avoid the logical contradiction between premise and conclusion for the inference model.

Raz believes that at a certain moment, T1, the old AC may be valid and a new AC may be adopted in accordance with it, while at T2 the old AC is invalid and the new AC is valid.[Note 17] This is certainly true, or could be true, but Raz is wrong to say that in that form, "the conclusion does not contradict the assumptions and the argument is both valid and accurately describes the situation."[Note 18] Raz deals only with legal, not logical contradiction, and ignores the problem of transtemporal validation for the inference model. The argument or inference will be invalid so long as one premise asserts the validity of the old AC and the conclusion asserts the validity of the inconsistent new one.[Note 19]

Hoerster argues that the new AC does not contradict any valid law, but only contradicts the old AC and never when both are valid.[Note 20] Here Hoerster also confuses legal and logical contradiction. The contradiction vanishes legally when the old AC passes out of effect, but the logical contradiction endures.

Finnis is the only commentator to appreciate that the time-based objection ignores Ross's insistence that no valid rule can authorize any rule inconsistent with itself.[Note 21] This principle applies (for formalists) regardless of the events that temporally intervene between the rule of change and one of its products, such as the repeal of the rule of change in self-amendment. Ross was worried that the believers in self-amendment not only believed in the valid derivation of norms inconsistent with the source, but also in the transtemporal validation of rules. Finnis coined the term "transtemporal validation" to designate the sense in which the old AC continues in force to validate its successor.[Note 22] Those who believe in it appeal to the validity-bestowing power of a repealed and invalid rule, as if the old AC arose from the grave in rarefied form to support the validity of its successor.

Because Ross believed that no conclusion remains established unless its premises remain true, he ascribed the doctrine of transtemporal validation to all those who believed in self-amendment, overlooking that they might justify self-amendment by a source of authority external to the constitution, such as acceptance, consent, or military power, or a source heedless of the death of the authority, as under the procedural model. Ross denied that transtemporal validation could occur, and accused proponents of self-amendment of presupposing it. Finnis apparently agrees with Ross on both counts, and therefore favors a Hartian (acceptance) solution precisely because "in Hart's analysis...the transtemporal validation of basic rules, relied on by Kelsen and recognized as troublesome by Ross, is simply and silently omitted."[Note 23] Hart avoids the necessity of transtemporal validation by denying that all rules are validated by antecedent rules. Some rules, on the contrary, are validated directly by present acceptance.[Note 24] Hart also rejected transtemporal validation explicitly in his later essay:[Note 25] Ross's argument, he wrote, is "questionable" to the extent that it relies on the statement that a derivation of a new norm presupposes not only the validity of the superior norm but its continued existence after the creation of the new norm.

The solution provided by the acceptance theory does not require the denial of transtemporal validation, although every rule validated by acceptance makes that rule's transtemporal validation unnecessary. I will argue in Section 14 that some rules that authorize their own repeal may be considered to persist afterwards in order to confirm the fact of repeal, and that this sort of transtemporal validation may be accepted. Indeed, in Section 21 I will argue that any absurdity that can be accepted in the rigorous sense can be law. Finnis, in short, did not have to deny transtemporal validation in order to deny Ross, but neither on the other hand did he have to concede that all proponents of permissible self-amendment, except possibly Hart, require transtemporal validation.

Ross avoids transtemporal validation and self-amendment by deriving the authority of the amended AC from a postulated tacit, immutable rule superior to the constitution,[Note 26] not from the old AC. I have argued (in Section 6.B) that this is ad hoc and fictitious. Indeed, Ross only gets into a position from which ad hoc and fictitious remedies will suffice because he has insisted on judging law by the principles of formal logic, a course not justified by any legal principle but only by the pretension of formal logic itself to rule the world.

Raz answers Ross, not through the time-based objection, but by arguing that self-reference is unnecessary in self-amendment, and in any case is not meaningless or problematic. Moreover, if Ross would have us avoid self-reference, then we will face many more difficulties interpreting law than we now face. For the avoidance of self-reference leads to crude paraphrases in which at least two rules are needed to transcribe one, and even then do not capture the original sense perfectly.[Note 27]

Marshall objects, quite rightly, that if some new AC cannot become valid because it cannot be derived from its predecessor, then it is not at all clear how the present AC could have become valid.[Note 28] Ross would answer that the tacit transcendent rule validated the current AC. But in Ross's formulation of the terms of that tacit rule, it applies only to AC's that already exist and point to their successors; it does not apply not to unstructured amending power in constitutional conventions or in the people creating the first AC. We can trace the authority of the current AC back a few constitutions, but then Ross would leave us without an answer. But this is damaging to Ross's position, for if the endpoint AC had no authority, then its offspring have no authority. This is a variation on the argument that Ross commits himself to an infinite genealogy for every valid legal system.

Hoerster has several arguments against Ross other than the time-based argument. First, he argues that there is no contradiction in the Rossian inference because the old and new ACs equivocate on the word "constitution". The old AC authorizes amendment of the constitution that contains it, while the new AC authorizes amendment of the different constitution that omits the old AC and contains the new.[Note 29] This argument is specious and self-defeating. If the equivocation is admitted, then the inference is still fallacious, though now for equivocating. On Ross's inference model, the conclusion is still far from established. But the equivocation may be avoided by implied adjustments to meaning or explicit rewording, even if the appropriate rewording occurs in no actual AC. If the rules in the constitution other than the AC are called set S, then the AC may be read to say, "this AC authorizes amendment of set S and this AC." The new AC could be read to authorize the amendment of set S and the new AC, eliminating all equivocation.

Hoerster also objects to Ross's tacit, transcendent rule because it always authorizes amendment of the AC, when in fact the people may wish to amend it to give it a continuing omnipotence that would resist some amendments.[Note 30] Or, the immutability of the tacit, transcendent rule prevents the immutable self-limitation and self-repeal of the AC, and therefore does not capture the legal possibilities inherent in self-amendment. He also objects that if a nation amended its AC every generation for 100 generations, without reference to such a tacit, transcendent rule, then that nation cannot be said to defer to such a rule, or such a rule cannot be said to govern the amendment of the AC. He rightly notes that Ross would not allow 100 generations of acceptance and usage to cure an original defect. Ross and other formalists must worry that an unknown defect buried in the past makes the present AC and entire legal order illegal.

Hoerster adopts an acceptance theory as his final answer to Ross. There may have been a defect in the adoption of the current AC or the ancient ancestor of the current AC, under the terms of the tacit, transcendent rule:[Note 31]

But the important thing is, we don't ever care to know, if we want to find out how to settle basic constitutional questions: To this task political history is simply irrelevant. All that is relevant is whether [the current AC] in our community is accepted as basic.

E. One more try at satisfying the inference model

This section was mislabeled "D" in the print edition.

If Ross takes pains to deny transtemporal validation, then it seems he believes that the old AC can authorize nothing after it loses validity, not even the norms consistent with it. But suppose one argued that the old AC authorizes its successors while it is still valid, and that it passes out of validity just as its successor becomes effective. It is true, as Hoerster argued, that the new AC never contradicts valid law, but might it also be the case that it never contradicts the old AC for the same reason that the old AC is unable to authorize any new norms? Are the old and new ACs really inconsistent, logically, as opposed to legally, if they are never thought to bear the status of valid law at the same time? The procedural model denies transtemporal validation, and permits new rules to become and remain valid even though their authorizing rules have been repealed. If the vanishing of the old AC is not likened to the denial of a premise in an inference, as Ross has it, but is likened to the discontinued assertion of a premise, then might the inference work? If the old AC is not positively denied, but simply dropped from the set of asserted premises, before or as the conclusion comes to be asserted, then does the conclusion contradict any premise?

This is like asking whether p and ~p are inconsistent if only one of them is asserted to be true. The answer may be yes or no depending on one's metaphysic of asserted and unasserted meanings. A surer route to an answer is to note that the inference model requires integrity of inference in general, and that consistency is just one element of this. The inference is also unsuccessful in establishing its conclusion if we cease asserting an indispensable premise just before, or just as, we assert the conclusion.

Logic has no name for this defect because logicians evaluate inferences under the assumption that all the premises are true. That all are asserted to be true is thereby also assumed, but tacitly. The consequences for the proof of denying a premise are well known: the conclusion is no longer proved. The consequences of ceasing to assert a premise midway in the inference are novel. Logicians also consider inferences to occur instantaneously. A set of premises implies or determines a set of conclusions without any lapse of time; only the human logician is under the sway of time in discovering (or proving or writing) those conclusions. A set of premises asserted to be true determines a set of conclusions at once; there is literally no time to change the status of a premises after an inference has begun and before it has ended, unless we consider inferences to be psychological processes. But if do that, then the concept of validity used by Ross and other logicians also vanishes.

Logically, the jurist who defends the inference model must choose between (1) asserting all and only the conclusions determined by the joint assertion of all her premises, and forgoing the derivation of all conclusions inconsistent with any premise, and (2) asserting all and only the conclusions determined by the joint assertion of those premises she is willing to leave asserted throughout the inference, and forgoing all the premises inconsistent with the conclusions she desires to derive. (The third option of using inconsistent premises deliberately is omitted because it clearly would not solve Ross's problem.)

In short, the paradox of self-amendment will still arise, even under the inference model, when we assume that the old and new ACs were never in effect at the same time. However, under the procedural model we may accomplish something very similar to the derivation of a conclusion after ceasing to assert an indispensable premise. For procedures do take time, and do not fail merely because the technical rules of inference are not satisfied. The old AC specifies the procedures of change, and ex hypothesi of its own change. Under the procedural model we may rely on the validity of the procedure, and just as we assert the validity of the new AC deny or suspend our assertion of the validity of the procedure. (While this is so, it is also the case that under the procedural model such maneuvers to avoid contradiction are unnecessary.) Because procedures are not instantaneous, there is time between between the steps to change the status of a "premise". Moreover, the logical difficulty or impossibility of asserting and ceasing to assert the same premise in the same inference becomes irrelevant. If it seems preposterous and illogical, it may be because one has not fully appreciated the freedom from logic that the procedural model finds in law.

The fact that the old and new ACs are never in effect (valid and supreme) at the same time, does not dissolve the paradox for the inference model. If one thinks it does, one may unconsciously have exchanged the procedural for the inference model, and forgotten that the inference model requires logically valid deductive inferences, not merely well-run, valid procedures. One may also have forgotten that the inference model requires the continuing validity (or "truth") of the premises in order to provide continuing validity (or "truth") to the conclusion.

In sum, under the inference model self-amendment of an AC to a form inconsistent with its original form requires either (1) the inconsistency between a premise and the conclusion, when the premises are consistent with one another, which makes the inference logically invalid and the conclusion legally invalid, or (2) the abstract separation of steps in the inference, as if in time, and the arrest of our assertion of an indispensable premise just as we conclude the conclusion, which is either impossible or a dressed up description of contradictory attitudes toward a premise. The attempt to avoid the former fallacy by keeping the old and new ACs separate in time, while still deriving the new from the old, leads to the latter fallacy.

In law we may freely and fictitiously suppose that the premises survive self-repeal only in the minimal sense needed to continue authorizing their successors without transtemporal validation. Or we may fictitiously suppose that the continuing validity of the premises is unnecessary to the continuing validity of the conclusion. Or we may fictitiously separate the steps of an inference, and start with the assertion of all premises and end with the assertion of only those premises consistent with the conclusion. But the insistence that legal logic shall be identical with formal logic precludes the use of these saving devices.

Hence, whatever the attraction of these or other stratagems, they cannot be used by Ross's inference model, and therefore cannot dissolve the paradox of self-amendment for that model of legal change.

Ross's analysis of self-amendment is the correct logical analysis (but for, perhaps, the fulfillment of the promise of temporally indexed deontic logic). The proposed solutions, even those that dissolve the paradox, invariably introduce a legal, extra-logical element. If Ross is correct on the logic of the problem, and absurd on the law, then that is an argument to look for another, non-formalist, model of legal change.


1. See the entry for Georgia in Appendix 2 for a possible example of overlapping periods of validity. Several amendments to the Georgia constitution of 1945 were ratified at the same election that established the entirely new constitution of 1968. The amendments were incorporated into the new constitution. This meant that the 1945 AC applied to the 1968 constitution, when the 1968 constitution contained a new AC that was presumably valid from the date of the election and that (otherwise) displaced the 1945 AC. See the entries for North Carolina and New Mexico for similar cases. [Resume]

2. Alf Ross, "On Self-Reference and a Puzzle in Constitutional Law," Mind, 78 (1969) 1-24, at 20. [Resume]

3. See e.g. L. Thorne McCarty, "Permissions and Obligations: An Informal Introduction," forthcoming, North-Holland. [Resume]

4. See e.g. Arthur Prior, Past, Present, and Future, Oxford University Press, 1967, 1978; Nicholas Rescher and Alasdair Urquhart, Temporal Logic, Springer-Verlag, 1971. [Resume]

5. Ross, op. cit. at 20-21. [Resume]

6. Joan Safran Ganz, Rules: A Systematic Study, The Hague: Mouton, 1971, at p. 5. [Resume]

7. See entry under Rhode Island in Appendix 2, and Luther v. Borden, 48 U.S. 1 (1849). [Resume]

8. For example, if a court's jurisdiction in a case is based only on the supposed domicile of the defendant, and if the court rules that the defendant is not in fact domicilied in the jurisidiction, then the ruling subverts its own jurisdictional basis and validity, but still stands. To one who believes the holding is self-subverting, J.D.I. Hughes replies that no one can believe in the self-subverting effect without presupposing that the defendant is not domiciled in the jurisdiction, which is to presuppose or acknowledge the validity of the holding. J.D.I. Hughes, "Judicial Method and the Problem in Ogden v. Ogden," Law Quarterly Review, 64 (1948) 217-26, at 226.

Another example occurs in ex parte annullments. Because annullments adjudicate the rights of both husband and wife, both should be in court. But a fiction has grown up that marriage is a thing (res) that travels with each spouse. By virtue of this fiction a court may have in rem jurisdiction over an ex parte annullment (one at which only one spouse is present). But because annullments declare that the marriage was void ab initio (from the beginning), the decree removes its own jurisdictional basis, but nevertheless remains valid. See Robert H. Gerdes, "Conflict of Laws: Jurisdiction to Annul a Marriage," California Law Review, 16 (1927) 38-44, at 40; and Herbert F. Goodrich, "Jurisdiction to Annul a Marriage," Harvard Law Review, 32 (1918) 806-24, at 810-11, 814-15. [Resume]

9. Congressional committees may only investigate areas on which they may legislate. Congress may not legislate on wholly intra-state commerce. But it may investigate a business to determine whether it is wholly intra-state. U.S. v. DiCarlo, 102 F.Supp. 597 (1951). Such an investigation could well discover that the business is wholly intra-state, which would normally mean that Congress could not have investigated it, even to discover this fact. Either the procedure and its outcome are kept apart, unable to form an explosive mixture, as the procedural model demands, or Congress has a second, independent legal basis for investigations. There is no evidence for the latter view. [Resume]

10. The neatest example is the ancient Greek case of Protagoras v. Euathlus, in which Euathlus asked Protagoras for lessons in argumentation —the Greek education of a lawyer— provided that he could postpone payment until he had won his first legal case. Progatoras agreed. After the lessons were complete and before Euathlus took a case, Protagoras sued for payment. Protagoras argued that he should be paid whether he wins or loses, for if he wins, Euathlus must pay by the judgment of the court, and if he loses, Euathlus must pay under the contract. Euathlus learned his lessons well and replied that he need not pay whether he wins or loses, for if he wins, he need not pay by the judgment of the court, and if he loses, he need not pay under the contract. The court of Areopagus in Athens was said to have been so puzzled that it adjourned for 100 years. The only American case that has cited Protagoras v. Euathlus, according to Lexis (a computer search service) is State v. Jones, 80 Ohio App. 269 (1946). Jones is a fascinating example of a jury verdict that implies its own negation. For the story of Jones, and more on Protagoras v. Euathlus, see Section 20. [Resume]

11. H.L.A. Hart, "Self-Referring Laws," Festskrift Tillägnad Karl Olivecrona, Stockholm: Kungl. Boktryckeriet, P.A. Norstedt & Söner, 1964, pp. 307-16, passim, but esp. 312-15. [Resume]

12. Hart, ibid. at 314 (emphasis in original). [Resume]

13. Hart, ibid. at 315. [Resume]

14. Ross, op. cit. at 7.n.1, and 20. [Resume]

15. Geoffrey Marshall, Constitutional Theory, Oxford University Press, 1971, pp. 46-48; Joseph Raz, "Professor A. Ross and Some Legal Puzzles," Mind, 81 (1972) 415-21; Norbert Hoerster, "On Alf Ross's Alleged Puzzle in Constitutional Law," Mind, 81 (1972) 422-26; J.M. Finnis, "Revolutions and Continuity of Law," in A.W.B. Simpson (ed.), Oxford Essays in Jurisprudence, Second Series, Oxford Univesity Press, 1973, pp. 44-76, at 52ff; J.C. Hicks, "The Liar Paradox in Legal Reasoning," Cambridge Law Journal, 29 (1971) 275-91; and William F. Harris II, The Interpretable Constitution, Chapter Four, "Revising the Constitutional Polity: The Limits on Textual Amendability," esp. pp. 238-39, Ph.D. Dissertation, Department of Politics, Princeton University, 1985;. I thank Professor John Vile for bringing Harris's dissertation to my attention. [Resume]

16. Marshall, op. cit. at p. 48. [Resume]

17. Raz, op. cit. at 420. [Resume]

18. Raz, ibid. [Resume]

19. How the consistency of such inferences is affected by the fact that the old AC in the first premise authorizes amendments in general, and specifically authorizes self-amendment, will be examined in Section 12.C. [Resume]

20. Hoerster, op. cit. at 423. [Resume]

21. Ross, op. cit. at 21: "Any attempt at solution must stand by the principle that from the validity of a norm it is impossible to derive the validity of any norm in conflict with [it]." [Resume]

22. Finnis, op. cit. at 55. [Resume]

23. Finnis, ibid. [Resume]

24. Finnis, ibid. cites Hart, The Concept of Law, Oxford University Press, 1961, at 103-04. [Resume]

25. Hart, "Self-Referring Laws," op. cit. at 314. [Resume]

26. Ross, op. cit. at 21-22, 23-24. [Resume]

27. Raz, op. cit., passim. [Resume]

28. Marshall, op. cit. at p. 47. [Resume]

29. Hoerster, op cit. at 422-23. [Resume]

30. Hoerster, ibid. at 424. [Resume]

31. Hoerster, ibid. at 426. [Resume]

This file is one section of the book, The Paradox of Self-Amendment. Return to the Table of Contents.

Ribbon] Peter Suber, Department of Philosophy, Earlham College, Richmond, Indiana, 47374, U.S.A. Copyright © 1990, Peter Suber.