Section 14
Amendment by Sunset Clause
Peter Suber, Paradox of Self-Amendment Table of Contents

A. Sunset clauses in American constitutions

A constitutional provision could say of itself that it automatically expires at a certain date. Because repeal is a form of amendment, such "sunset clauses" are effective rules of amendment. If an AC is used to adopt a new rule with a sunset clause, then the subsequent expiration (repeal) of the rule may be said to have been authorized by the AC. But I prefer to say that it was directly authorized by the sunset clause, a valid constitutional rule in its own right, and indirectly authorized by the AC. This usage is consistent with present practice, for we say that an existing constitutional amendment, for example, is authorized by the AC, not (except indirectly) by the authority that authorized the AC.

The AC of the federal constitution, Article V, came into being with three limitations on its power, two of which were subject to sunset clauses. Those two limitations provided that no amendments to Article I, Section 9, clauses 1 and 4, could be made until 1808. Substantively these limitations protected the importation of slaves and prohibited some capitation taxes for about 20 years. When the sunset clauses protecting these clauses expired in 1808, a form of self-amendment occurred. Amendment by sunset clause had affected the AC itself, removing two limitations on its powers.

The expiration of the two limitations on Article V was authorized directly by the language of Article V, and indirectly by the power that made the federal constitution. The latter point is not trivial. The sunset clauses were written into the original constitution, not added by amendment. Hence only through direct authority in the AC do the expirations represent self-amendment; through their indirect or ultimate authority, they represent irreflexive amendment of the AC.

Because the sunset clauses were written into the original constitution, they share the status of the AC itself as original, official rules of change applicable to constitutional rules. They have only one subject apiece, unlike the AC. Unlike most sunset clauses, which merely declare the expiration date of some rule of law, the two in the federal AC served a second, equally important function during the period before their expiration: they entrenched constitutional rules against amendment. Also unlike most sunset clauses, they are not part of the rules they govern, and therefore may operate as intended without paradoxical self-repeal. Because the federal sunset clauses are rules of change independent of the AC, the appearance of self-amendment in their application to the AC may be reinterpreted as one step of an irreflexive see-saw method.

The federal sunset clauses have another important difference from most: instead of dictating the repeal of a rule at the expiration date, they dictate the change of status of rules from entrenched to disentrenched. Once disentrenched, the rules are still valid; they are merely rendered liable to amendment. This distinction may be blurred if change of status is considered a repeal of the rule defining the difference of status. But the distinction, even if provisional, highlights the flexibility of sunset clauses. They may doom experimental regulations, ensure that a privilege is temporary, protect a rule for a time without repealing it at the end of the protected period, or conversely leave a rule subject to amendment for a time after which it automatically becomes entrenched, perhaps for another finite time. Sunset clauses may therefore contain "sunrise" clauses, and require change of a status more than once, or the change of a status and back again, before finishing their jobs. They could require an oscillation of status without end. But most interestingly, they may dictate changes, including the repeal, of the very provisions in which they lie, thereby repealing or suspending, reviving, or amending themselves.

A clause that required the complete repeal of the statute of which it was a part, and then its complete revival one year later, presents a paradox of continuing validity (the further paradox of) self-repeal. If some kind of continuing validity is possible between periods of self-governed repeal, then an indefinitely long and complex series of changes could be "programmed" into a statute or constitution, some requiring the temporary, but absolute, suspension of the most fundamental premises of law, or of law itself. One thinks of the annual Roman festival on which all law was, by popular understanding, suspended for three days. If a rule of law defined this aspect of the festival, then either it was suspended during the festival like all other law, and there was no law to summon law back again, or it was the exceptional law that was not suspended. In the former case, the resumption of law would be revolutionary; in the latter case the festival would merely be a regulated party, a much less interesting affair. The equivalent on a computer would be a program that alternately turned the entire machine on and off and on again, without, like remote control televisions, leaving the machine partially on at every moment. Truly, law can transcend "mechanical jurisprudence"!

Nothing like this has yet been programmed into an actual constitution. But the Alabama constitution of 1901 was interpreted by the Alabama Supreme Court to have invested its AC with the power to "suspend" parts of the constitution and "to take [a] feature out of the constitution during [a] period and to put it back at the end of the period". Downs v. City of Birmingham, 240 Ala. 177, 198 So. 231 (1940).

The most comprehensive sort of reflexive sunset clause might be a clause of the New Jersey constitution of 1776 (discussed in Section 9 above) which conditionally nullified the entire constitution of which it was a part, if and when the colonies reconciled with England. Can we imagine an even more comprehensive clause —one that conditionally nullified all law in a given legal system if a certain condition came to pass?

The constitutionality of the sunset clauses in the federal AC was never questioned, although the proponents of slavery concocted many other ingenious theories to forestall abolition. Nevertheless, the Supreme Court of Missouri has found a constitutional defect in the device of amendment by sunset clause. In 1910 a group of Missouri citizens submitted a proposed constitutional amendment to the Secretary of State for his ultimate submission to the voters at the general election of 1910. They were using a method of amendment by popular initiative (which had earlier been added by self-amendment). Their proposed amendment was a new map of state senatorial districts. It contained a sunset clause directing that the map expire when replaced by a revised map using the 1920 census. The Secretary of State refused to submit the proposal to the voters. In a mandamus action brought by the citizens to force the Secretary of State to put the proposal on the ballot, the Supreme Court held for the Secretary, declaring the proposal to be "legislative" rather than "constitutional" on account of its "very temporary" provisions. State ex rel. Halliburton v. Roach, 230 Mo. 408, 130 S.W. 689 (1910).[Note 1]

The Halliburton court did not apparently disapprove sunset clauses in constitutions per se, but only those providing too short a life for a constitutional rule. Indeed the proposal in this case would have amended a provision that was itself subject to a sunset clause, and that had expired in 1881. The court voiced no regret that the framers had used the device. If the court had declared sunset clauses nullities per se, then the 19th century rule to be amended would have been resurrected, rendering it liable to the 1910 amendment proposal, removing one defect from that proposal. But as soon as the 1910 proposal was rehabilitated from this direction it would have to be quashed from another on account of its sunset clause.

The narrow ruling in Halliburton seems to be that amendment proposals whose content is very particular and by nature subject to periodic change, smelling legislative from two causes, cannot also be made subject to sunset clauses of comparatively short duration, for to add this third cause would definitely push the proposal over the line separating constitutional rules from legislation.[Note 2]

B. Self-repealing sunset clauses

If a sunset clause is included in the rule to expire, call it "reflexive". Call a sunset clause outside the rule it governs "irreflexive". If the two sunset clauses in the federal AC are interpreted to govern the importation of slaves and capitation taxes, then they are irreflexive; but if they are interpreted to govern the scope of the amending power, then they are reflexive, for they are located in Article V. Most statutory sunset clauses are reflexive, for there is no single amendment rule for statutes in which they might be put. Moreover, because statutes are drafted and adopted one after another, if one is to be subject to a sunset clause it must usually contain the clause itself.

Note a peculiar feature of any reflexive sunset clause, for example, clause C in statute S, saying "statute S will expire at time T". If one were to ask, some time after T, whether S were valid, the answer would clearly be no. But how do we know that? We know by reference to the wording of C. These words bind us even after they expire. Those who act as if S were still valid may be opposed in court on the ground that present law recognizes that S has expired. The sunset clause itself may be cited as authoritative on the past effectiveness of the repeal of S. But to do so means that we believe the sunset clause did not completely swallow itself. Statute S, including sunset clause C, did expire, but somehow a meta-statement about this expiration did not expire. But there was no such meta-statement in S. If the meta-statement is law, then it appears that it expired; and if did not expire it appears that it is not law. But if we insist after time T that valid legal authority can be cited for the invalidity of S, then we are appealing to such a metaphysical mystery. The cat is gone but the grin lingers on.

This raises the most interesting question whether a properly reworded reflexive sunset clause could utterly repeal itself, cat, grin, and all, leaving no legal authority behind on which to assert the invalidity (expiration) of the laws it took with it. Suppose the following sentence appeared after the substantive portion of a statute or constitutional rule, but in the same section:

This section shall be utterly null and void, for all intents and purposes whatsoever, at 12:01 A.M., January 1, 2000 A.D., and forever thereafter.

After the expiration date, how will we know that the section has expired? Has the expiration-directive expired too? If it has, then the section is paradoxically valid and invalid at once: valid because the expiration-directive is ineffective (having expired), invalid because the whole section is valid, including the expiration-directive (because the expiration-directive is self-nugatory). In fact, the validity of the section would imply its invalidity, and its invalidity (at least its invalidity solely on account of the expiration-directive) would imply its validity. After the expiration date the section becomes a genuine paradox of the Liar-type as defined in Section 1 —true if false and false if true.

In the hypothetical case of the self-repeal of an entire legal system, we would worry less, for we are less likely to want to cite a currently valid legal authority for the occurrence of legicide after it had supposedly occurred. But a court would be enwrapped in paradox if a citizen challenged the validity of the total self-destruction of law. The court could decide that it had jurisdiction only if it had already decided the question at issue against the plaintiff. The court could not hold for the validity of the legicide without subverting itself, nor against its validity without at least a suspicion of illegitimate bootstrapping, nor could it even authoritatively dismiss the case without presupposing a position on the merits.[Note 3]

If utter self-repeal leads to the paradox of the liar, what about partial self-repeal? If the substantive portion of a statute expires but the expiration-directive survives, then this clearly gives effect to the intent of the framers, but at the expense of the "plain meaning" of the language and the appearance of a logically arbitrary preference for the expiration-directive over the substance. In law, the evident intent of the framers might well supersede the logic of the situation, especially when the latter leads to paradox. The expiration-directive could be read as self-excepting. The appeal to the framer's intent lessens the arbitrariness of the interpretation; but even if arbitrary, avoidance of paradox may comprise a legally sufficient reason to read the statute and sunset clause in this way. Of course, under the acceptance model, partial self-repeal and arbitrariness may be accepted, and avoidance of paradox is not urgent. Even total, paradoxical self-repeal may be accepted, as we will see.

The appeal to the intent of the framers loses its force as the intent shifts (as in our hypothetical) from those common to actual sunset clauses to the mischievous purpose of complete self-repeal. For the latter purpose the language of the clause should be tightened:

This section, including this very sentence, shall be utterly null and void...

Similar issues arise in philosophy where the magical elements of law cannot support an answer. The ancient Greek skeptics faced the dilemma of speaking or not speaking. If they remained silent, they would persuade no one; if they spoke, others might construe their statements as assertions of knowledge and accuse them of self-refutation. The Greek skeptics chose to speak and, among the things they chose to say, to argue that skepticism and speech are compatible. If they said that they did not know something, and an opponent retorted, "Do you know that you do not know?" then the skeptics replied that their expressions of ignorance and uncertainty were "self-canceling".[Note 4] The expressions announced the speaker's uncertainty on the irreflexive topic, and then reflexively announced their own uncertainty, and so "canceled themselves". Sextus Empiricus compared such expressions to "aperient" drugs; Montaigne compared them to rhubarb:[Note 5] they washed out the system and then washed themselves out. A fair summary of the philosophical learning on the rhubarb theory of complete self-repeal is that non-skeptics ignore it, skeptics don't know, and logic and grammar are still far from the sophistication needed to rule on the question.

But law is not limited to logic. A Rossian inference model that hews closely to the forms and procedures of formal logic would declare the rhubarb theory as paradoxical as self-amendment. Indeed, the rhubarb method of speaking and all other forms of self-repeal are but species of self-amendment.

However a theory grounding legal authority in acceptance and usage can tolerate the phenomenon of a clause repealing the section of which it is a part, and thereby repealing itself. The section may be known to be repealed by virtue of an inconsistent appeal to the repealed expiration-directive, as if it were a valid rule after it became invalid. Or the section may be known to be repealed without any doctrinal justification in the face of paradox. These are the boldest ways in which the acceptance theory could domesticate the paradox.

More sophisticated justifications could be contrived if needed. The inference model requires that a rule's superior, authorizing rule be in effect as long as the inferior rule is to be authorized. When the authorizing rule is repealed, either the authorized rule is repealed or it receives "transtemporal validation" (see Section 10). Ross denies that a repealed rule can validate anything, but we might find that a rule survives repeal in a limited sense, the sense needed to tell us that it is gone, for example, if it departed through self-repeal. This is not completely arbitrary, for our actual practice suggests that we appeal to the "repealed" expiration-directive as authority or evidence for the validity of self-repeal. That represents a form of transtemporal validation. On the other hand, for this sort of self-repeal or transtemporal validation we might even be comfortable with a Rossian tacit, transcendent rule that could tell us that the repeal had been effective. Alternately, the procedural model could justify the outcome (complete self-repeal) if the terms of the procedure (sunset clause) were valid and obeyed. Any contradiction between outcome and process is ignored by the procedural model. This avenue could justify self-repeal but would leave no sign for us to cite as authority for the effectiveness of the repeal.

Or, most simply, the repeal could be authorized directly by acceptance, not the sunset clause, eliminating the contradiction and the strictly reflexive self-repeal as well. The best explanation of what actually happens, however, may be that genuine self-repeal occurs, and subsequently the repealed expiration-directive is cited as authority for the fact of repeal, with no thought given to the presupposed present validity of that language and its effect on self-resurrecting the repealed clause. Difficulties, especially if they are genuinely paradoxical, are simply ignored.[Note 6]

The acceptance theory, in short, could tolerate partial self-repeal, and live with the idea of a grin outliving its cat even if the idea is inconsistent, metaphysically incoherent, or exegetically arbitrary. The acceptance model can also permit complete self-repeal, and live with the notion of an expiration-directive valid enough to cite after it invalidates itself, or perennially effective despite its expiration or invalidity, or totally invalid but not subjected to doctrinal scrutiny. If this is messy, self-contradictory, paradoxical, or even magical, nevertheless it may be law.

C. Effective-date clauses

Before leaving this section we should note the temporally symmetrical problem. If sunset clauses seem to require a validity after invalidation, then effective date clauses seem to require a validity before validation. A clause that says,

This section shall become effective to all intents and purposes at 12:01 A.M., January 1, 2000 A.D.

is known to be yet invalid in 1990, and that knowledge derives from the clause as if it were effective before it were effective. It must take effect —in some sense— sooner than it says if it is to tell us, authoritatively, that it is not yet effective. It is impermissible under the inference model either way, however, for it is either effective before it is effective, which is a contradiction, or it is self-applicable, which violates the theory of types.

The effective date puzzle is a time-based or temporal paradox, but it has a spatial variant. If a clause of a contract said, "this contract shall be construed under Illinois law", then in one sense it applies before it applies, that is, before we know the content of Illinois law and independently of what that content turns out to be. If the contract was made in Arizona, and Illinois law said to use the law of the state where the contract was made, then the contract's clause would have led us on a circuitous route to Arizona law. But it authoritatively led us to Illinois law, and to Illinois' "pointer" to Arizona law, even before we knew whether Illinois law would in turn send us somewhere else.[Note 7]

One may object that there is a slight but significant difference between the effective date and the sunset clauses. Effective date clauses belong to statutes already passed by legislatures (or other types of law already adopted in the proper way) and may be considered valid from the moment of approval and as governing only the rest of the statute. This is a makeshift, however. It is available to judges, certainly, and attractive because it actually does provide logical coherence, but it is a judicial response to paradox, not a solution latent in the language. Its inadequacy lies in the fact that, like totally self-repealing sunset clauses, effective date clauses frequently occur in the very sections they control. And as with sunset clauses, we can tighten this language in order to examine the logic of the problem:

This section, including this very sentence, shall become effective to all intents and purposes only at...

But even if the solution can be made applicable to statutes, it cannot apply to effective date clauses, or ratification clauses, in constitutions.[Note 8] Article VII of the federal constitution states that the document of which it is a part shall become effective when ratified in a certain way. In the two years between its drafting in 1787 and its ratification in 1789, Article VII authoritatively told courts and all interested parties that the new constitution was not yet effective. One cannot argue that Article VII had already been approved in the proper way and governed only the rest of the constitution.

An even stronger example may be found in Article 13, Section 1, of the Missouri constitution of 1865:

The preceding parts of this instrument shall not take effect unless this constitution be adopted by the people, at the election to be held as hereinafter directed...

The Article then proceeds to direct an election as if already binding law.[Note 9] The Article clearly excepts itself as if it could be valid before the rest of the constitution, but of course that is the sheerest fiction. If the fiction is taken at face value, then we have a very clear case of constitutional self-justification (see Section 7.B). The federal ratification clause, Article VII, left the ratification elections in each state to the initiative of proponents and did not presume to order them as if it were already binding law. One may ask whether a Missouri court in 1864 would have heard a writ of mandamus ordering the election required by the ratification clause of the unratified constitution.

Approved statutes and unratified constitutions awaiting effectiveness are clearly different. The statutes may be assigned some provisional validity, or their effective date clauses assigned some exceptional validity, while this cannot be done with constitutions. The reason is that the statutes have already been approved in the manner appropriate for statutes, and need only the condition stipulated in the effective date clause in order to become effective. Falling in between statutes and constitutions in this regard are treaties that have been signed by representatives but not yet ratified by the signatory states. Article III of the Kellogg-Briand Pact of 1928 required its signatories to ratify the pact in accordance with their various domestic procedures for ratification. This "requirement" suggests that the Pact was already binding law; but if it was, then ratification would be unnecessary. If a signatory state failed to ratify, would it violate the treaty, or would it for that very reason exempt itself from the treaty's provisions?

Finally we may observe that effective date clauses may do some of the work of sunset clauses and cause self-repeal. The Seventeenth Amendment, which provided for the direct election of senators, contains the following clause:

This amendment shall not be construed as to affect the election or term or any Senator before it becomes valid as part of the Constitution.

When the amendment was ratified in 1913, this clause became generally inapplicable. Only senators who had a grievance arising from the pre-ratification period, or bodies intending to interfere with senatorial elections or terms in the pre-ratification period, would ever have to appeal to the clause prior to ratification. If they did appeal to the clause prior to ratification, however, they would find an unratified proposal, not law. The unratified proposal would still have some use as a reliable guide to how courts would act after ratification. But if a senator did have a grievance under the clause, she would have to wait for the amendment to be ratified for this clause to be authoritative. And of course some remedy would be available even if the amendment were never ratified, since even without the clause it remains the case that unadopted proposals are not law and should not affect the elections or terms of senators. So by becoming explicit in an amendment proposal, the clause is generally useful only upon ratification (so that it is lawful) and is generally useless upon ratification (because the period of its intended application is over). By the time of its effective date, it has canceled itself.


1. The amendment proposal had two other, more substantial defects that led to the Court's judgment: (1) it violated the procedures of amendment by initiative by failing to give the full text of the proposed amendment, and (2) it purported to amend Article 4, Section 11, which was itself subject to a sunset clause and had expired in 1881, while the existing procedures on redistricting were contained in Article 4, Section 7. People v. Potter, 47 N.Y. 375 (1872) asserted a commonly held view of temporary provisions of constitutions: they are not favored, they cannot be interpreted to clash with the "general design" of the constitutions, and in conflicts with other provisions they must yield. [Resume]

2. On the issue of particularity the court said at id. 694 that a proper constitutional rule on redistricting

simply points out a plan or method for such redistricting, and prescribes permanent rules and principles for carrying out such method or plan. On the other hand, the matter of actually redistricting is a matter of very temporary character.

On the need for changing the provision the court said at id. 694 that senatorial districts depend on population, which changes, and that

it was never contemplated under our present constitutional scheme to incorporate as a part of the permanent and fundamental law of the state a provision which must of necessity demand frequent alterations and changes.

The holding of Halliburton was criticized by Henry Rottschaefer, Handbook of American Constitutional Law, West Pub. Co., 1939, at 398. For some surprising data on very frequent constitutional amendment, see note 6 to Section 15, below. [Resume]

3. These paradoxes are similar to the question, which has actually arisen, whether the constitution that establishes a certain court may be declared invalid by that court. Some courts have held it logically impossible to declare such invalidity. Luther v. Borden, 48 U.S. 1, 39-40 (1849), Coleman v. Miller, 307 U.S. 433, 455, 457 (1939), Carpenter v. Cornish, 83 N.J.Law. 254, 83 A. 31 (1912), State v. Starling, 15 Rich.Law. (South Carolina, 1867). Other courts have held that a judgment of invalidity could be competent. Loring v. Young, 239 Mass. 349, 132 N.E. 65 (1921). [Resume]

4. Sextus Empiricus, Works, vol. 1, Outlines of Pyrrhonism, trans. R.G. Bury, Loeb Classical Library, Harvard University Press, 1933, at i.14-15, i.206-07, ii.187-88. [Resume]

5. Michel Eyqem de Montaigne, The Complete Works of Montaigne [the Essais], trans. Donald Frame, Stanford University Press, 1948, "The Apology for Raymond Sebond," II:12.392-93. [Resume]

6. Self-repeal could be avoided if, seeing the expiration date imminent, an irreflexive repeal (or extension) were passed by the appropriate body. If the sunset clause in the federal AC protecting the importation of slaves, which expired in 1808, had been preempted in this way, say, by adopting the Thirteenth Amendment in 1807 instead of 1865, then a new problem would arise: the irreflexive repeal of the provision protecting slave trade would directly violate the entrenchment function of the sunset clause. The slavery sunset-entrenchment clause was not self-entrenched, except perhaps impliedly, and therefore probably could have been repealed (see Section 8). But to violate the clause without repealing it is much less acceptable; and this seems to be the effect of supposing that the Thirteenth Amendment had been adopted in 1807. If that had happened, we would exchange the paradox of self-repeal of a sunset clause for the paradox of the self-disentrenchment of the AC. [Resume]

7. More on the reflexivity of renvoi (the circuitous quest for applicable law) and a larger family of "know before we know" paradoxes in law may be found in Section 20.H. [Resume]

8. Constitutions that are already ratified but not to become effective for a certain time fall into the same category as the statutes mentioned in the original objection. Such constitutions, incidentally, were common in states anticipating admission to the Union. [Resume]

9. Compare McCullen v. Williamson, 221 Ga. 358, 144 S.E. 2d 911 (1965), discussed above in Section 8.B, in which an amendment said of itself that its effective date was to be determined by the school boards that it would consolidate. In all the controversy over the permissibility of deferring a constitutional question to local government, and permitting contingent delay of effectiveness after ratification, no one seems to have noticed that the problematic effective date clause was being considered effective before it was effective. Even its opponents found it defeasible only by general constitutional principles, not by paradox. [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.