Paradox of Self-Amendment by Peter Suber requirement of a referendum could only be repealed by a referendum. Hence, the existence of the Legislative Council was incompletely self- entrenched. The N.S.W. parliament, however, wanted to abolish the Legislative Council without resort to a referendum, and believed that its sovereignty and omnipotence allowed it to do so through an ordinary amendment. The highest court in N.S.W. struck down the attempt, and thereby upheld the validity of "manner and form" limitations on the amending power, or incomplete self-entrenchment, even against a parliament supreme and omnipotent in the English tradition. Attorney General for New South Wales v. Trethowan, (1931) 44 C.L.R. 394; [1932] A.C. 526.[Note 17] Another case, just as famous and more dramatic, occurred in South Africa in 1952. When the English parliament wrote the South African constitution in 1909 it included one Section (number 35) that prohibited racial discrimination, and one (number 137) that assured the equal status of the English and Dutch languages. A third section (number 152) incompletely entrenched both of these provisions and itself, requiring for their amendment a two-thirds vote of each house of the South African parliament sitting in joint session. In order to institute apartheid, the South African parliament tried to repeal all three of these sections by a statute, using only simple majorities. It knew that the 1909 clauses stood in the way of this plan, but argued that its intervening independence from Britain made it sovereign and capable of amending its constitution without British interference. The South African Supreme Court rejected the attempt, partly on the ground that South Africa was not completely sovereign so long as Britain could repeal the Statute of Westminster that granted former colonies their independence (see Section 20.I). Harris v. Dönges, (1952) 1 T.L.R. 1245. The South African government eventually established apartheid "legally" by packing the parliament until the two-thirds votes could be obtained.[Note 18] The rule in the United Kingdom and its former colonies and dominions is that incomplete self-entrenchment clauses are binding and can only be amended or repealed by the procedures they lay down for themselves. In other words, a parliament cannot bind its successors irrevocably, but it can make things difficult. Continuing omnipotence, however, is not threatened by mere difficulty or by revocable limitations on power. Unfortunately no case has apparently arisen in which the "manner and form" of permissible amendment set out in the incomplete self- entrenchment clause is so difficult as to amount to virtually complete self-entrenchment. Such a case might reveal that complete and incomplete self-entrenchment differ only in degree, not in kind, and that an extremely difficult amendment procedure will be treated like a prohibition on amendment, not like a manner and form limitation.[Note 19] If "revocability" is a matter of degree, matching the degrees of difficulty of amendment, then a parliament could approach the irrevocable binding of its successors by increasing the difficulty of the manner and form limitation. If courts strike down would-be irrevocable limitations on future power, should they strike down extremely difficult manner and form limitations? What about manner and form limitations that are not so difficult for some entrenched rules but that would be virtually impossible to overcome for certain popular rules by virtue of their popularity? For example, if an article permitting the institution of slavery and an article asserting the freedom of religion were each to be incompletely self-entrenched so that only 80% supermajorities of the national population in a referendum could repeal them, then the first would still be repealed while the second would become virtually immutable. Should courts take this into account when deciding what constitutes an irrevocable limitation on future generations? When an incomplete self-entrenchment clause is to be violated by a proposed amendment, as attempted in Australia and South Africa, then the violation may possibly be cured by acceptance. In such cases the perceived fairness and difficulty of the procedures specified in the entrenchment clause would have some bearing on the acceptability of circumventing them. Circumvention would almost always become more acceptable as the limitations on the amending power were perceived as impositions of a foreign power or former imperial sovereign. When the specified procedure is not intolerably difficult, however, or perceived as a remnant of colonial subservience, then the acceptance model will usually agree with the inference model in finding manner and form limitations valid against the amending power. The acceptance model provides, however, that when acceptance in the appropriately complex sense is present, then it suffices to override even the most "reasonable" entrenchment clauses to cure violation or circumvention. Under the inference model complete self-entrenchment clauses are immutable, providing that we tolerate the self-reference involved. A clause that said, "this section may not be amended," when it is part of the entrenched section, plainly forbids its own amendment. If it is an original limitation, then it supersedes the general language of the AC as specific language customarily supersedes general language (the lex specialis principle). If it is a self-imposed limitation, then it supersedes both as the more specific rule and as the more recent voice of the sovereign (the lex posterior principle) which impliedly repeals all existing portions of the text not reconcilable with it (see Section 16). However, note that self-imposed complete self-entrenchment will almost always impliedly amend the AC, again both as the more specific and as the more recent language of the constitution. As an act of the AC that impliedly amended (limited) the power of the AC, it would constitute self- amendment, which the inference model could not tolerate. The inference model might solve the problem by dropping the lex posterior principle that gives recent law priority over older law in cases of irreconcilable conflict, since it is by virtue of this principle that the AC "transmutes" the "immutable" and amends the complete self-entrenchment clause. But this solution requires a change of law —for the lex posterior principle is actual law already— that the inference model cannot accomplish from its merely logical standpoint. Nor can it say that because the lex posterior principle has logically regrettable consequences some of the time that it is already repealed or limited in its applicability. But the mere idea of this solution points up an important ambivalence of the inference model about the status of temporal priority of law in its relation to logical priority. On the one hand, the inference model requires that new law be authorized by temporally and logically prior law. That seems to imply deference to the lex posterior principle. So does the method of avoiding contradiction in law by invalidating (impliedly amending or 45