2. Logic

Continuing sovereignty has often been considered the most natural understanding of sovereignty. The fact that it has for so long remained the dominant interpretation of the powers of the United Kingdom Parliament probably has something to do with an assumption that other interpretations might be logically unsound.^{Footnote 7} Dicey, for one, seemed to think that continuing sovereignty was somehow dictated by logic:

In other words, the incoherence of the popular concept of “omnipotence”, the fact that one cannot have both continuing and self-embracing powers, was believed to involve that self-embracing sovereignty was logically impossible.

Ross succeeded in raising doubts about the logical status of self-embracing powers when he wrote that a typical pattern of constitutional amendment may involve a logical contradiction.^{Footnote 9} He was referring to constitutional amending provisions that are commonly interpreted as providing for their own replacement.^{Footnote 10} He used the example of the amending provision of the Danish constitution to introduce his puzzle, which he explains as follows:

Ross believed the reasoning to be irremediably flawed by the apparent contradiction between conclusion and first premise as to how the constitution can be amended. “It is not possible,” he writes, “that anything can appear in the conclusion of a valid deductive inference which is in conflict with the premises.”^{Footnote 12}

The strategy most often invoked to dissolve the apparent paradox consists in showing that art. 88 and art. 88' do not apply over the same time period, and so do not come into conflict.^{Footnote 13} This strategy embodies the common-sense reaction and, certainly, the reaction most lawyers would have. But Ross was not satisfied with that. He believed that the argument “confounds legal with logical contradiction.”^{Footnote 14} The inference could not be made logical by stipulation; the contradiction remained even though the law might have tools to deal with it, such as the lex posterior principle.

Another strategy has been to point out that the first premise does not refer to the same constitution as the one mentioned in the conclusion. This is the explanation offered by Goldstein, who refers explicitly to Ross’s formulation of the inference:

The contradiction apparently vanishes because the word “constitution” has a different referent in premise and conclusion. One can readily imagine how Ross would have reacted to this. Applying his system of logic, Ross would have been entitled to say that Goldstein’s strategy merely shows how the inference pattern of self-amendment can be interpreted as involving the modification of the first premise in “mid-inference”. In Goldstein’s version of the example, the referent of “constitution” is modified just as what is taken to be the conclusion is being asserted, thus irremediably vitiating the inference.^{Footnote 16} On this plane, Goldstein’s claim that no contradiction is involved because “conclusion and first premise predicate contrary things about different objects” would seemingly amount to a claim that the conclusion can simply not be said to follow from premises.^{Footnote 17}

Hart doubted, perhaps more helpfully, that the exercise of legislative power to create new norms could be modelled as a deductive inference of that type. “[I]t is not clear,” he writes, “how this logical principle [that ‘nothing can appear in the conclusion of a valid deductive inference which is in conflict with the premises’] applies to a legislative act.”^{Footnote 18} Although the point was not dwelt upon by Hart, Ross makes the following remark about his suggestion:

The problem, I would suggest, is not whether constitutional self-amendment is, or is believed to be, based on an inference. In one sense, self-amendment certainly is based on some form of inferential pattern. The problem seemingly lies, rather, in the sophistication of the logic used to model the real-world phenomenon of constitutional self-amendment. Ross gives a clear indication of what he takes logic to be when he writes that a “logical inference … knows of no sequence in time.”^{Footnote 20} This suggests that his whole argument is based on the most basic first-order logic where the only connectives (or operators) allowed are those of propositional logic (not, and, or, implies, if and only if, and exclusive or).^{Footnote 21} Hart’s reaction to Ross’s puzzle—pointing out that logic perhaps does not apply to legal reasoning in that way, or that self-amendment does not really result in a contradiction—may be a good reflection of the fact that a formal representation of the type of legal reasoning here envisaged requires a more complex system of logical representation.

The basic system used by Ross is known to show very serious shortcomings when it comes to modelling instances of real-life reasoning. The most important limit of that system in the representation of law is probably the fact that it provides no tools to handle statements about sentences.^{Footnote 22} If one takes a close look at Ross’s example, one will notice that the conclusion of his inference is not the new amending provision. The conclusion is a statement about the sentence taken to be that new amending provision. Consider Ross’s formulation of the conclusion:

And a more accurate version:

The general modelling of legal change offered by Ross shows this clearly:

The conclusion here is simply “The norm N is valid” and not “The norm N is valid, that is, N.” The conclusion is that N is valid whatever its content and not the supplementary predication of that content. Basic first-order logic does not provide the relation between the “statement about the sentence” and the “sentence” itself.^{Footnote 23} Any attempt to express “statements about sentences” directly in that logical system is very likely to lead to trouble such as apparent contradiction.^{Footnote 24} Ross’s modelling of self-amendment was such an attempt.^{Footnote 25}

Nothing in the foregoing means, of course, that self-amendment in law is illogical or absurd. What it means is that more powerful tools are certainly needed if a logical representation of the phenomenon is to be provided. Artificial intelligence research has had to deal with the numerous problems involved in the logical representation of real-life reasoning. It has explored various ways in which basic first-order logic can be augmented or extended to that purpose. One of the more obvious strategies has been to provide additional logical connectives or operators for sentences, so that statements about sentences could adequately be dealt with.^{Footnote 26} The most important of those connectives in the representation of real-life, or common-sense, reasoning, are probably the time connectives.^{Footnote 27} This is so because “[v]ery few commonsense problems can be formulated in purely static terms.”^{Footnote 28} And if that is the case for common-sense problems in general, it is probably even more so with respect to legal problems. In Ross’s example of self-amendment, for instance, the present tense used in “the constitution may be amended”, in both conclusion and first premise, is not the timeless present it appears to be and that basic first-order logic would require it to be.^{Footnote 29}

Let me now go back to where our enquiry into logic started in order to point out that the orthodox British understanding of sovereignty cannot be said to be dictated by logic. Consider a British version of Ross’s puzzle. For that purpose, suppose a formal expression of parliamentary sovereignty were found by some lucky analyst in a secret document used by judges only. Article 88 of that document, along the lines of Ross’s puzzle, reads thus:

Suppose Parliament, in some unprecedented effort to modernize British political life, enacted a statute stating that the law and constitution of the United Kingdom can be modified by Acts of, say, the new “House of Citizens”, and by no other process. Before even starting to imagine inference patterns that could yield a conclusion, conscientious judges might want to have a second look at art. 88 and wonder whether the sovereignty is continuing or self-embracing. Like Danish judges, they would find that the answer is not really there: art. 88 can be interpreted either way in both cases.^{Footnote 30} Ross assumed for his popular puzzle that art. 88 allowed self-amendment; put in British terms, for our puzzle, the assumption is that the body designated or formed by the amendment procedures enjoys self-embracing powers. The attempt to formulate self-embracing sovereignty in British terms gives a choice between two tactics:

1. (1) Parliament can make any law including a law changing this rule

2. (2) Parliament can make any law until it provides otherwise

Since, as already noted, basic first-order logic does not handle time-frames and has no representation for “this” used self-referentially, Ross concluded that a contradiction appears whenever one starts from such premises. So my imaginary British court might want to say that in (basic first-order) logic, “Parliament” cannot be a time-varying term in any valid inference, that sovereignty must therefore be continuing, and conclude that, notwithstanding its purported transfer of power to the House of Citizens, the same (old) Parliament can still make law exclusively. But to be consistent, the court would have to assume that continuing sovereignty can validly be expressed in a first-order logical construction. And it cannot. Consider the two corresponding formulations:

1. (1) Parliament can make any law excepting a law changing this rule

2. (2) Parliament can make any law at all times

Here again, one gets the same choice between a self-referential “this” and the time operator. And both involve a second-order predicate.^{Footnote 31}

Basic first-order logic is fundamentally a static system, and reality is not static. A common problem in the treatment of “sovereignty” in terms of logic, it seems, has been that of confusing the necessarily static directions of basic first-order logic with the contextual assumption that may have for a long time favoured, in British theory, static answers to real-life legal questions.^{Footnote 32} Very few legal questions can be solved without the help of time-frames, which basic first-order logic ignores.^{Footnote 33} Pointing to the “staticity” of that logic to support the assumption that Parliamentary sovereignty cannot be self-embracing, therefore, would be very much like using a four-equations pocket calculator to disprove Gödel’s theorem.

I have assumed so far that reasoning from norms can perhaps be translated into an extended system of logic commensurable with the principles underlying first-order logic.^{Footnote 34} Under that assumption, I showed that there is no reason to believe that continuing sovereignty is more (or less) “logical” than self-embracing sovereignty.^{Footnote 35} In both cases, a satisfactory representation would clearly require extensional operators on sentences, which basic first-order logic does not provide. Leaving that aside, what I wish to stress now is the fact that constitutional change is already intelligible in the form law presents it. The reason lies in context, which clearly informs the way lawyers envisage the common occurrences of constitutional change, and which logicians must look to in their attempt to offer accurate representations of this real-life phenomenon. This is what is next considered.

3. “Pragmatics”

If lawyers often know that an older amending provision does not conflict, in any relevant sense, with a newer amending provision derived from it, it is because the context in which they find constitutional self-amendment tells them so. It is because in some sense “in” that context there are principles about how words are used around constitutional amendment and, more generally, around legal change.^{Footnote 36} Apart from syntax and semantics, this is the third source of information that logicians have to look to in attempting to provide adequate representations of language-based, real-world reasoning. This is where Ross found a basis for choosing his interpretation of art. 88, which was one of at least two possibilities. For as I suggested, on the basis of syntax and semantics only, he could have taken art. 88 to be immutable, just like Parliamentary sovereignty can be, and has been, viewed as being forever “continuing”. But the context directed him to choose the self-embracing interpretation even though he a priori thought that it could lead to logical contradiction. The information provided by context is difficult to handle because it is but rarely expressed in the form of words.

In order to solve his puzzle, Ross suggested a principle he meant to serve as the immutable “basic norm” of the legal system, from which the amendment provisions could derive their validity:

This, he thought, would solve his self-reference problem because any amending provision in the future could be said to rest on the general principle rather than on the amending provision(s) preceding it.^{Footnote 38} His principle, however, could certainly not solve the problem of logical representation as he conceived it since the basic logic he was using knows of no time connectives on sentences such as “until”. Ross’s “basic norm” is nevertheless valid as an attempt to express the unwritten rules that are followed (often unreflectively) when self-amendment of the type implied in art. 88 is in fact understood and acted upon in legal practice. But as such it has serious limitations.

One of the problems with Ross’s “basic norm” is that it is assumed to be immune from legal change, which is to say that any change to it must happen outside the law. Its modification must be analyzed as some sort of revolution, perhaps peaceful, entailing a break in the continuing identity of the system. From a descriptive standpoint, this is not entirely satisfactory because it fails to explain how a change in a basic norm can be effected, on a principled basis, using normative materials that are part of the legal system and that survive the change.^{Footnote 39} Moreover, observation tells us that, from the point of view of a judge, the acceptance of a new ultimate rule of competence never entails a total rejection of the “former” legal system.^{Footnote 40} Cases of successful revolutions or coups usually show that the only norm necessarily changed in the “new system” is the ultimate rule of competence; all other norms are generally understood to continue in force, subject to repeal under the new rule of competence.^{Footnote 41} Ross’s “immutable basic norm” does not adequately explain the continuing validity of most norms in cases where a break in legal continuity is believed to have occurred. The surviving norms usually include the ultimate rules of adjudication, and rules of identification envisaged separately from the rule of competence.^{Footnote 42} When courts are called upon to settle constitutional matters in cases of revolutions or coups, as they were, for instance, in Pakistan,^{Footnote 43} Uganda,^{Footnote 44} and Grenada,^{Footnote 45} it is generally assumed by all that former rules of adjudication are to continue in force and that a rule of identification points to enactments of the former regime (under the former rule of competence) as valid law for the adjudication of ordinary disputes.^{Footnote 46} Some of those surviving rules may perhaps be changed under the new rule of competence, but it is clear that a new rule of competence does not otherwise affect them. The case of legal devolution shows the same phenomenon where a British-style independence statute is understood to bring about a “new” (independent) legal system. Ceylon provides a good example. The Privy Council found in 1964 that even though the Independence Act, 1947 was “irrevocable”^{Footnote 47} and made the legal system completely independent, it did not “alter the existing corpus of law in Ceylon,”^{Footnote 48} so that appeals to the Privy Council remained possible under the law of Ceylon. Ultimate rules of adjudication and identification had survived the acknowledged change in the ultimate rule of competence.

Finnis helped clarify the unwritten rules surrounding legal change in general, at least in the Anglo-American literature, when he proposed a more convincing formulation of what lawyers actually apply, more or less consciously but quite consistently, when they have to handle the “life” of formal legal norms through time.^{Footnote 49} He managed to shape his conclusion into one rule-like principle (which can be said to include Ross’s basic norm):

Here as with Ross’s principle, it is easy to see how an amendment provision can effectively provide for its own replacement if that principle is taken to be part of the normative context in which self-amendment takes place. The present validity of the new amending provision, which is in one sense “derived” from a provision no longer in force, can be said to rest on the general principle, which has remained untouched throughout the change. Like Ross’s basic norm, the principle solves any problem of self-reference there may be by providing an external ground for the expiration, “according to its own terms”, of the old amending provision. But Finnis’s principle is broader and can be said to apply to any situation of formal legal change, including constitutional self-amendment; it explains the continuing validity of all formal norms adopted under a rule of competence no longer in force, whether or not it is believed that a “new system” was introduced upon a change of the ultimate rule of competence. Any attempt at formalizing the actual reasoning that surrounds legal change would have to take account of this principle, which unlike Ross’s basic norm, is not tied to any particular system.

The degree to which such contextual considerations “surrounding” legal change can usefully be given formal legal expression, however, is not clear. This type of information has an ambiguous status. It can usually be formalized to a certain extent, but because of the dynamic character of context, it can never be reduced effectively to its formal expression.^{Footnote 51} An example of a paradigmatic case of this is afforded by certain rules of statutory construction. Those “rules” are about the ways in which language is used in and around statutes. Case law has worked them out to an extent where it seemed at one point possible to put some of them down in a statute, and this is what many common law jurisdictions have done. Have the Interpretation Acts stopped the possible evolution of the rules of interpretation they express? Surely not. The context has not disappeared, and the Interpretation Acts may themselves need a contextual framework for their interpretation. The evolution of the practice will inevitably affect, among other things, the use of language in legislation and, in turn, the rules of interpretation that will actually be in use.^{Footnote 52} Finnis’s principle itself is also a case in point. At first sight it may appear as a timeless principle quite independent from context. But as Finnis himself points out, a common law rule provided otherwise, in a certain respect, for a very long time, and that which perhaps seems obvious today may have looked odd in a different context.^{Footnote 53}

It would appear, then, that there is a type of normative material surrounding the hierarchical structure through which legislative norms are formally validated, material that resists attempted reduction to formal expression. The formal structure of legal validation, often seen as a pyramid, is embedded in a context that is essentially dynamic and cannot be reduced to a set of rigid rules. The formal structure can attempt to draw up the entire context in which it is embedded only at the price of quickly losing its grip on reality, that is, the social reality for which and in which it exists. It is fair to say that the formal structure can replicate the surrounding or underpinning informal materials but can never appropriate them. It is perhaps at the constitutional level that this is most apparent. Consider the absurdity of a constitution attempting to exclude all unwritten rules or principles that judges may want to use in constitutional adjudication. The attempt would simply fail. Even Kelsen thought that a formal constitutional enactment would be powerless to exclude the paramount application of customary law at the constitutional level.^{Footnote 54} The very notion of the supremacy of written constitutional law over “unexpressed concepts” could only come from the pervasive, unwritten normative surroundings of the formal constitution. The rules of a formal constitution can in no way be “superior” to the reasons why that constitution was adopted and the principles under which it is understood and upheld.

4. Validity, Acceptance, and Reasons

The importance of “context”, broadly understood, is one of the reasons why the “acceptance model” of the foundation of a legal system offered by Hart was so popular. The model relieves ultimate norms from their Kelsenian “hypothesized”, “presupposed”, or eventually, “fictional” character.^{Footnote 55} The model explains that the basic rules of a system ultimately rest on the fact of their present recognition by officials, most importantly judicial organs, given a background of general obedience to the rules of that system. They rest, in other words, on something found outside the formal structure. Hart elaborates this in connection with the emergence and development of independent legal systems in the colonial context:

The ability better to account for the emancipation of British colonies was one of the strengths of Hart’s rule of recognition when it was formulated. The existence of independent legal systems throughout the Commonwealth had become a basic datum to be accounted for in legal theory when Hart published The Concept of Law. One can fairly say, therefore, that Hart’s rule of recognition was a move forward in that it drew attention to the importance of acceptance, that is, the importance of the context in which formal law is embedded. With respect to the hierarchical structure of formal validation, which Hart insisted had to be accounted for, the acceptance model highlighted the fact that formal validity has ultimately to rest on something other than formal.^{Footnote 57}

But Hart’s account also shows well-known weaknesses. Because the model recognizes the possibility of direct acceptance as sufficient for some norms, one is led to ask if it is not the case that all norms are law by virtue of their being directly accepted, whether or not they meet the criterion of validity posed by the rule of recognition.^{Footnote 58} The reliance on acceptance as mere “fact” thus seriously threatens Hart’s overall position against legal realism: if the rule of recognition can be accounted for as mere fact, why should we not treat other rules of the system in the same way?

A related problem is that of the continuing unity and identity of the legal system. At first sight, and envisaged statically, Hart’s rule of recognition may seem to provide a good basis for saying that this or that is an independent legal system. One is tempted to say that a legal system is the set of rules unified by a distinct rule of recognition. But on analysis, the “rule of recognition” inevitably appears, rather, as a set, not fully defined, of various rules held together by nothing in particular.^{Footnote 59} This is already noticeable in The Concept of Law where Hart has to discuss the “rules of recognition” of a system, despite his usual reference to “the” rule of recognition.^{Footnote 60} And there seems to be nothing bringing these rules together but for the fact that they are part of the same system, which begs the question. The explanation is even more problematic with respect to the further problem of unity and identity regarded through time, that is, the unity and identity of dynamic systems. Nothing formal, first of all, brings together a rule of identification of rules and a rule of competence distinct from it.^{Footnote 61} Further, the rule of recognition being on the whole a customary law, it is subject to changes that are, Hart tells us, to be authoritatively determined by courts.^{Footnote 62} Some elements of the rule, therefore, will appear and others will vanish over time, thus generating a series of sets of ultimate rules deprived of a unifying element.

As I said earlier, Hart’s model makes the highest rule of the formal structure rest on acceptance and makes other formal rules derive their validity from that highest rule. Finnis goes further and gives more importance to that which happens outside the formal structure in making the continued validity of all formal rules rest individually on a principle “of the practical and theoretical understanding of law,” which stands outside the formal structure of validation and refers to it. This principle provides an illuminating perspective on the important but clearly limited place actually taken by the hierarchical structure of formal validation in a broader understanding of law. That perspective reverses a familiar picture and treats context as a phenomenon prior to, and as necessarily embedding, formally validated law.^{Footnote 63}

At one end, there can be no doubt, as even positivist analysts agree, that the unity and identity through time of legal systems lie, at least in part, somehow “outside” the structure of formal validation. They ultimately rest on the unity and identity of the society whose legal system it is taken to be.^{Footnote 64} They depend on the maintenance of the ensemble of social relations;^{Footnote 65} they lie in the political system and its continuity^{Footnote 66} and hinge on a smooth evolution of its “dominant ideology”.^{Footnote 67} These inform the context in which the structure of formal validation is embedded.

At the other end, a parallel conclusion should follow with respect to the very notion of legal validity applied to norms. All of the analysts quoted above have, more or less openly, taken the positivist notion of validity as their starting point in showing how the unity and identity through time of legal systems cannot rest exclusively on formal considerations.^{Footnote 68} Finnis, to take the most meticulous, begins with this assertion: “No lawyer will deny that every rule in a legal system is equally valid.”^{Footnote 69} From there, he works his way up to the “radical” conclusion that “the legal system, considered simply as a set of ‘valid rules’, does not exist” because there is nothing (formal) that provides (and secures) its unity and identity through time.^{Footnote 70} He then points out that the unity and identity of a legal system are, like the all-or-nothing notion of legal validity, “a basic phenomenon of legal experience,” and goes on to find that “there seems only one conclusion: the continuity and identity of a legal system [are] a function of the continuity and identity of the society in whose ordered existence in time the legal system participates.”^{Footnote 71} All of this would seem to imply more than our analysts mentioned. For there is apparently no reason that could stop more sceptical analysts who would take the notion of validity as no more than an assumption for discussion, from working their way down from the absence of a unifying formal master rule. Starting from the formal disunity of the system, the absence of a formal test of membership that could give the system an identity through time, one could go all the way down to show, “radically”, that “validity”, considered simply as the characteristic of all the rules making up a “unified and identifiable” legal system, cannot exist. Since all-or-nothing validity may nevertheless be considered “a basic phenomenon of legal experience” and cannot be grounded on a formal master rule, “there seems only one conclusion,” one might say: the legal validity of a rule is a function of the validity of that rule considered in the broader context of the society that it is meant to regulate.

As with the case of the legal system’s identity, however, stepping out of the formal structure means that black-and-white certainty is no longer afforded: all-or-nothing validity turns into relative worth or value; it becomes a matter of justification.^{Footnote 72} Some may perhaps say that, empirically speaking, all-or-nothing legal validity is not, after all, such an important phenomenon of legal experience. They may say that the notion of justification may provide a better account of the actual use of norms in adjudication than does the notion of all-or-nothing validity. But few would deny that there are weighty normative reasons in favour of adopting and maintaining an all-or-nothing notion of legal validity. There are weighty reasons, that is, in favour of adopting and maintaining a framework of formal validation. Showing the limits of that framework, as I tried to do here, is simply pointing to the underlying reasons that sustain it. Those reasons are part of the context in which the whole structure of formal validation moves; they make up the normative aspect of that context. Those reasons often take the form of principles that play a justificatory role in adjudication. The often undeniable “political” status of principles does not disqualify them from also having a status as legal principles.

5. Permeability and Circularity

The need for security and predictability in social relations is an obvious enough reason for maintaining legal forms. But it is perhaps fairness as impartiality that provides the most basic drive toward the adoption of legal forms:

This drive is what brought about such models as Kelsen’s, under which judges are said to derive the validity of standards in a formal way down from the historically first constitution beyond which no court can go. Hart attempted to trade this past-tense cascade of validation for a notion of the acceptance of the present rules of recognition that would not otherwise disturb the “source” thesis at the lower levels. I noted that “informal” considerations cannot be eliminated completely even at the lower levels of the structure of formal validation. As Finnis remarks, from a position that can only be characterized as one of outstanding awareness of the importance of legal forms, the “drive to insulate legal from moral reasoning can never … be complete.”^{Footnote 74}

Hart was concerned with preserving the moral neutrality of the ultimate rules and attempted to treat their existence as a factual matter. Regarding the emancipation of the former colonies, for instance, Hart could thus write, descriptively, that at the end of the process “[t]he new rule [of recognition] rests simply on the fact that it is accepted and used as such a rule in the judicial and other official operations of a local system.”^{Footnote 75} But the external perspective there adopted by Hart makes things look tidier than they are. For the explanation says nothing of the question, a great deal trickier, of what the matter might look like from what Hart himself named the “internal point of view”. Hart simply eludes this problem when he states, in a way reminiscent of Kelsen, that upon reaching “the rule that what the Queen in Parliament enacts is law, we are brought to a stop in inquiries concerning validity,” and that this rule is special because “there is no rule providing criteria for the assessment of its own legal validity.”^{Footnote 76} As was seen earlier, Hart did improve legal theory’s grasp of the phenomenon of legal emancipation. But he did so from outside the legal systems concerned, treating “context” as a purely factual matter, a position that directly threatened his stance against legal realism. If rules cannot be satisfactorily accounted for without due regard being had for their “internal aspect”, as he otherwise so elegantly showed, why should we treat the rule of recognition differently? Surely the rule of recognition also has an intelligible purpose from the internal point of view. Why is Hart’s otherwise fruitful use of the internal viewpoint put to the side here?^{Footnote 77} For our purposes, the typical insider is the judge in a newly independent country who has to explain the switch from one ultimate rule of recognition or competence to another. More is expected of a judge than a “factual” statement of the type “it is so because it is so.”^{Footnote 78} Judges have to justify their decisions; they have to provide reasons that are intelligible intra-systemically.

To the extent that the reasons offered by judges must relate to the system as it exists, the explanation of fundamental change from their perspective will inevitably involve a measure of circularity. In The Concept of Law, Hart acknowledges that the explanation of a judicial decision on the ultimate rules of the system could appear somewhat problematical:

In a sense, this is the inevitable result of having put aside the requirement of derivation down from the historically first constitution. One could say that the unappealing quest for the historically first constitution was basically traded for a form of circularity in the explanation of law.^{Footnote 80} Hart was conscious of this and was of the opinion that the picture thus offered was an improvement. Answering his own question, laid down above, he writes:

The picture here sketched is one that integrates circularity as a useful explanatory tool for law.^{Footnote 82} What is found, I would suggest, at the ultimate level—or in the ensemble of constitutional standards—is a number of interdependent standards that support and shape one another in a complex network of justification that is partly circular. Any one of those standards can be challenged in some of its elements on the basis of other accepted standards. This process being effected mostly through judicial decisions made on a case-by-case basis, each affecting a usually small element of the network, the circular movements of justification are but very rarely exposed.

This is not to say, however, that the system is closed. If it were, there could never be a change in the ensemble of constitutional standards; there could not be a change, for instance, in the ultimate rule of formal validation. And such changes do occur. At all levels, external elements are constantly being brought into the system.^{Footnote 83} Those elements may be provided by unrestricted moral reasoning in the community. The sense in which the legal system’s claim to “closedness” must be understood is that any external element needs intra-systemic grounds to be received into the system. One may refer here to MacCormick’s “toe-hold” criterion: a judicial decision is a legal decision insofar as grounds are found in the existing legal materials to justify it. It is the role which external elements are allowed to play in a legal system’s evolution that keeps the acknowledged circular justificatory processes of that system from being “vicious”.