Section 11
Attempts to Dissolve the Paradox:
Self-Embracing Omnipotence and Specific Authorization
Peter Suber, Paradox of Self-Amendment Table of Contents

A. The authorization fallacy

In one of the first modern accounts of the paradox of omnipotence, J.L. Mackie proposed the following solution as applicable equally to deities and sovereigns.[Note 1] We must distinguish laws governing behavior from laws governing other laws. Mackie calls these first and second order laws; Hart calls them primary and secondary rules. Mackie continues:

Correspondingly, we should distinguish two orders of sovereignty, first order sovereignty (sovereignty (1)) which is unlimited in its authority to make first order laws, and second order sovereignty (sovereignty (2)) which is unlimited in its authority to make second order laws. If we say that parliament is sovereign we might mean that any parliament at any time has sovereignty (1), or we might mean that parliament has both sovereignty (1) and (2) at present, but we cannot without contradiction mean both that the present parliament has sovereignty (2) and that every parliament at every time has sovereignty (1), for if the present parliament has sovereignty (2) it may use it to take away the sovereignty (1) of later parliaments. What the paradox shows is that we cannot ascribe to any continuing institution legal sovereignty in any inclusive sense....

[O]mnipotence must in any case be restricted in one way or another.

Mackie's distinction between first and second order sovereignty is very similar to Hart's distinction between continuing and self-embracing omnipotence (or sovereignty). In fact, in this passage Mackie has shown a connection that Hart himself failed to draw between Hart's distinction between continuing and self-embracing omnipotence and Hart's distinction between primary and secondary rules. Continuing omnipotence is the unlimited power to make primary rules. It is continuing because primary rules themselves cannot limit the power that makes them; that power continues, therefore, because it cannot diminish itself. Self-embracing omnipotence is the unlimited power to make secondary rules. It is self-embracing because one group of secondary rules (the rules of change) governs the making of rules, and an unlimited power to make rules of change could diminish the power to make them. Mackie is also more explicit than Hart in inferring that the unlimited power to make secondary rules must therefore be self-applicable.

Time must be brought into the picture, because an initially self-embracing omnipotent parliament or deity could limit itself immutably in its first act. Continuing omnipotence by its nature preserves that nature over time. Self-embracing omnipotence may by its nature lose its omnipotence and even its self-embracing character.

There is nevertheless a difference between Hart's and Mackie's distinctions. While Mackie's first order sovereignty can make no second order rules at all, Hart's continuing omnipotence can make any secondary rules consistent with its continuing character. Similarly Mackie's second order sovereignty can make no first order rules at all, while Hart's self-embracing omnipotence can make any primary rules whatsoever until it limits its power to do so. Another difference between Hart's and Mackie's distinctions is that no being can simultaneously possess both continuing and self-embracing omnipotence, without contradiction, but a being could simultaneously possess first and second order sovereignty. Self-embracing and continuing omnipotence are mutually exclusive because some self-limitations within the power of the former would violate the latter. But first and second order sovereignty are not mutually exclusive. A being could have, at a given moment but not continuously, unlimited power to make both primary and secondary rules. In the next moment the unlimited power to make secondaries could be used to limit itself or the power to make primaries. This is why Mackie believes there are only two consistent distributions of these powers: (1) that the being now has and always will have only first order omnipotence, and (2) that the being now has both first and second order omnipotence, but need not keep either power forever. A third theory is contradictory: (3) that the being presently has second order omnipotence and continuously has first order omnipotence.

Mackie and Hart use their distinctions as if they could solve or dissolve the paradox of omnipotence.[Note 2] If "omnipotence" (they would say) is equivocally thought to embrace both species (either both first and second order omnipotence, or both continuing and self-embracing omnipotence), then and only then could we speak of an omnipotent entity limiting itself as a contradiction or paradox. If we keep the two species separate, then we must recognize that an omnipotent being can either limit itself freely and without contradiction, because its power it second order or self-embracing, or that it cannot limit itself at all, because its power is first order or continuing.

Is this solution adequate? It has strong intuitive appeal because both Hart's and Mackie's distinctions are distinctions with a difference that help us classify slippery phenomena; both expose an equivocation on "omnipotence" common to many descriptions of sovereigns and deities, and even to many statements of the paradox of omnipotence; and both show the sense in which self-limitation is and is not self-contradictory, thereby dispelling a cloud of confusion that all too frequently hovers over the paradox. But I believe the solution is inadequate nevertheless. Not only does it fail to dissolve the paradox, but the very idea of self-embracing omnipotence is inadmissible for both the inference and acceptance models of legal change.

First we must bear in mind that the acceptance and procedural models of legal change eliminate, or more precisely, ignore and excuse the paradox of self-amendment. It is only under the inference model or formal logic generally that attempted dissolutions of the paradox are sought, for example by separating the new and old AC's temporally, or by distinguishing the continuing from the self-embracing senses in which they might be omnipotent. Hence, the proposed Hart-Mackie solutions will be tested under the inference model, which is how a logician would test them, free from any magical propensities of law and theology to absorb and forgive contradiction. Therefore, the question is whether continuing and self-embracing omnipotence can each, when considered separately and without equivocation, give rise to the paradox of self-amendment.

(1) If the AC is thought to have continuing omnipotence, then irrevocable self-limitation and self-repeal (but not revocable self-limitation or non-limiting self-amendment) are precluded as ultra vires. However, self-amendment that does not limit either the omnipotence or the continuing character of the

omnipotence of the AC would still be possible. We must ask whether such self-amendment can give us a new, amended AC that is inconsistent with the present AC; if so, then under the inference model the paradox will still arise. The answer to the question is yes, if by "inconsistent" we mean that the new AC permits what was impermissible under the old AC, or prohibits what it permitted (see Section 12.C). For example, self-amendment by the addition of a new, complete method of amendment (say, by popular referendum) would not limit the continuing omnipotence of the old AC, yet is "inconsistent" with the old AC for permitting what was once impermissible. In such cases the paradox of self-amendment would still arise under the inference model. In the inference that models that act of amendment, the conclusion (new AC) would still contradict a premise (the old AC). Therefore, if the old AC is univocally considered to have continuing omnipotence, then the paradox is not routed or dissolved.

(2) If the old AC is considered to have self-embracing omnipotence, then the paradox is not routed either, although this is more difficult to show. Indeed, self-embracing omnipotence seems by definition able to perform any self-amendment whatsoever without self-contradiction. But we should beware of solving this problem by appeal to a kind of power that is simply defined as able to solve the problem. It turns out that, while self-embracing omnipotence authorizes all self-amendments, it cannot prevent contradiction. That is, we should not hastily infer that what is authorized is thereby freed from contradiction, for we can always posit an authority for bizarre acts or rules. One of the acts authorized by self-embracing omnipotence, apparently, is irrevocable self-limitation. We may be more specific: one act authorized by it is (to use Ross's language) the derivation of norms inconsistent with the source. If this is really a self-contradiction (Sections 5, 10), then self-embracing omnipotence does not remove the contradiction but only "authorizes" it or makes it somehow legitimate, or even acceptable. But if, on the other hand, the authority provided by self-embracing omnipotence is enough to solve the paradox or to remove the inconsistency of what is authorized, then it turns out that self-embracing omnipotence is really the power to override the principle of non-contradiction. Even if we say that it is, we don't understand it any better for defining it that way; instead we impute it, in all its mystery, to selected sovereigns and deities to absolve their contradictions. It is a way of using words to avert our eyes.

One of the defining characteristics of self-embracing omnipotence is the power to limit itself irrevocably. By contrast, continuing omnipotence can only limit itself revocably. The power of self-embracing omnipotence, so long as it is self-embracing omnipotence, is unlimited but limitable. Any self-amendment that immutably limits self-embracing omnipotence is inconsistent with its source, in the sense that it forbids what was once permissible. The inference model cannot tolerate such self-amendment, nor can formal logic generally. So long as our test of inconsistency is whether the new AC forbids what the old AC permitted, or permits what the old AC forbade, then self-embracing omnipotence will be unable to escape self-contradiction in self-amendments that limit it immutably.

One might object that no amendment is inconsistent with the AC that authorized it if it is really authorized; immutable self-limitation is concededly authorized by self-embracing omnipotence, and therefore is consistent with it. This objection confuses the vice of inconsistency with the vice of proceeding ultra vires (beyond authority). If a rule specifically authorized the derivation of norms inconsistent with itself, regardless how we define "inconsistent", then the new norms could actually be inconsistent with the source despite the authorization. Authorization does not remove inconsistency even if it legitimates it. Even for Alf Ross what purports to authorize inconsistency is ipso facto not a valid legal authority. For those who do not hold the inference model, the principle is even clearer: one of the differences between law and logic is that laws may violate logical rules and still be valid laws, even if they are for that reason "bad" laws that should be amended or repealed. Self-embracing omnipotence is not even a law, but merely an hypothetical concept of omnipotence that is experimentally ascribed to an AC. But even if an AC clearly possessed self-embracing omnipotence, it would merely authorize amendments inconsistent with itself, not eliminate the inconsistency from all authorized amendments.

Because "inconsistency" may be defined in any way whatsoever and still not be overcome by specific authorization,[Note 3] this conclusion does not depend upon the particular test of inconsistency that I have used (to forbid what was once permitted or to permit what was once forbidden). Other criteria of inconsistency for normative propositions will be examined in Section 12.C. But whichever we pick, a rule of law, valid by all legal tests, could permit the derivation from itself of norms "inconsistent" with itself by the chosen criterion. This is a flat denial of Ross's most fundamendal premise, that forbids all rules of all kinds from permitting the valid derivation from themselves of norms inconsistent with themselves. While Ross's principle is obviously imported from logic, not native to law, and imposed on law as if with a priori title, my counter-proposition is derived from the empirical experience of law of the sort highlighted by the procedural model of legal change, in which procedures often permit outcomes inconsistent with their antecedent authorities.

In Hart's version of the Hart-Mackie thesis, self-embracing omnipotence could be valid law even if it were concededly self-contradictory. In that sense Hart has actually dissolved the paradox with his distinction between continuing and self-embracing omnipotence. But his distinction does not eliminate the inconsistency of self-amendments that would be legally permissible for both species of amendments, and in that sense has failed to dissolve the paradox for the inference model or formal logic generally.

Regardless whether omnipotence is continuing or self-embracing, therefore, legally permissible self-amendment of some type will be paradoxical or self-contradictory under the inference model. The distinction between continuing and self-embracing omnipotence does not dissolve the paradox of self-amendment, but merely directs our attention two distinct species of omnipotence in each of which the paradox is fully replicated.

B. Against self-embracing omnipotence

In short, the Hart-Mackie thesis fails because self-embracing omnipotence, whatever its merits in theology, cannot be made plausible under any coherent concept of law. This applies especially to the half of the Hart-Macke thesis that would allow any self-amendment under self-embracing omnipotence to be consistent with its source. Under the inference model or formal logic it is self-contradictory. It is admittedly and essentially a source of norms inconsistent with itself. Under the acceptance model, the contradiction of deriving immutable self-limitations from an unlimited power could easily be tolerated. But as I argued in Sections 8 and 9, any limitations that are thought by one generation to be immutable could justifiably be thought by the next generation to be mutable. Shifts in acceptance can render mutable even the best entrenched clauses. Therefore, an AC thought to have self-embracing omnipotence may still not enact immutable self-limitations except in the estimation of those believing in its self-embracing omnipotence. It is possible that acceptance will never shift, and that what was once imagined immutable will in fact never change, and will be perceived, accepted, and used as if it were immutable. But whether acceptance shifts is contingent. Because there is no preventing such shifts by law, the contingent possibility of them suffices to make all limitations on the amending power revocable in principle, except the limitation against irrevocable limitations. In short, a supreme AC always has continuing omnipotence and cannot deprive itself of the power to validate and invalidate law. Under the inference model as well, no AC can have self-embracing omnipotence because it is self-contradictory, or formal logic cannot quiet its disposition to forbid contradiction wherever it occurs.

Self-embracing omnipotence, then, and any allegedly immutable self-limitations depend for their defining characteristics on contingent future events, under the acceptance model. The concept of a power able to limit itself immutably becomes incoherent when the effectiveness of its self-limitation, the essence of its unique power, is made contingent upon a future and external force. This is not true of deities or other entities to which omnipotence might be ascribed, for the acceptance model is not as applicable to them. But in law the idea that a supreme amending power may have self-embracing omnipotence, when acceptance is the ultimate source of legal authority, is incoherent, not merely a violation of democratic values. The acceptance model does not permit any legal power to make legally immutable rules, even if it permits "contingent immutability" ratified by each successive generation for its own reasons (see Section 21.C). Self-embracing ominpotence is hardly omnipotent and hardly self-embracing if it cannot, by its own exertions, limit itself immutably.

In short, then, self-embracing omnipotence is self-contradictory for the inference model, and incoherent or unable to live up to its definition under the acceptance model.

The distinction between continuing and self-embracing omnipotence is still meaningful, for self-embracing omnipotence may be a meaningful and coherent concept under a different model of law. Indeed, it appears to be coherent under the procedural model. If a power were postulated with self-embracing omnipotence, under the procedural model, then it could limit itself immutably merely by applying the procedures of self-amendment to a proposal to limit the power and to self-entrench the limitation. The contradiction with its antecedents, fatal under the inference model, is immaterial here; the threat of contingent future events repealing or transmuting the immutable limitation, fatal under the acceptance model, is impossible here. If procedures must be followed, then acceptance cannot override or supersede procedures that make no provision for amending a certain rule or that positively forbid it.

But while self-embracing omnipotence is coherent for the procedural model, the distinction between continuing and self-embracing omnipotence does not resolve the paradox for this model either. If the paradox is that self-amendment (when the new and old AC's are inconsistent with each other) is a contradiction, then the procedural model, like the acceptance model, can permit self-amendment without removing its paradoxical nature or making it free of contradiction. The only rules inconsistent with self-embracing omnipotence are immutable limitations on its power. At least these are inconsistent by the test of forbidding what was once permitted. Immutable self-limitations are permitted under the procedural model, unlike the acceptance model, but the contradiction is not removed, just ignored.

Other models of legal authority and change are undoubtedly possible, but they are much more likely to admit and excuse the paradox than to dissolve it. Dissolution requires the elimination of the contradiction in the self-amendment of an AC to a form inconsistent with its original form. The hope of self-embracing omnipotence was that by authorizing such self-amendments, their inconsistency would disappear. This turned out to be false. In fact, I believe that the attempt to erase inconsistency by appeal to specific authorization is to shift subtly to the procedural model from the inference model, to make authority primary and consistency negligible. In order to make self-amendment legally permissible or legally harmless, it gives up the goal of dissolving the paradox for the inference model or for formal logic generally.


1. J.L. Mackie, "Evil and Omnipotence," Mind, 64 (1955) 200-21, at pp. 211-12. [Resume]

2. Mackie is explicit throughout his essay that his distinction is a solution. Hart is explicit only in his later essay, "Self-Referring Laws," Festskrift Tillägnad Karl Olivecrona, Stockholm: Kungl. Boktryckeriet. P.A. Norstedt & Söner, 1964, pp. 307-16, at p. 315, where he even cites Mackie. [Resume]

3. There is an exception, of course, for the definition of "inconsistency" as something eliminable by authorization. But this is to resort to fiction, much as a legislature will declare time stopped in order to finish its business by midnight on a certain day. An inconsistency that can be overcome merely by the say-so of one of the propositions involved is not the kind of inconsistency that underlies serious problems. [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.