Paradox of Self-Amendment by Peter Suber Hart gets into this circle by his theory that official practice discloses (otherwise tacit) rules that state that such practice is lawful in that system. I will call this view the normative practice doctrine. If officials appeal to acceptance, then by the normative practice doctrine that appeal to acceptance hides a rule authorizing the appeal to acceptance. If officials appeal to acceptance, and not merely to rules authorized by acceptance, then Hart is bound to this circularity. Acceptance, which was supposed to ground the legal system alegally, and sit inertly outside the entanglements of rules, is caught up by the normative practice doctrine in a rule-bound web of self-justification. What was supposed to be social and historical, not rule-like, becomes rule-like under the normative practice doctrine. As we will see, there is a sense in which officials for Hart do appeal to acceptance, not merely to rules authorized by acceptance; there is also a sense in which Hartian acceptance is circularly justified even if officials do not appeal to acceptance. Is there a rule holding that what is accepted and used as law thereby is law? (1) If there is such a rule, then its authority may be questioned, and Hart may want to find its authority in acceptance. In that case acceptance would justify the rule that justified acceptance, which might be fatally circular. If there is such a rule but its authority does not lie in acceptance, then the acceptance model does not explain that rule and therefore does not explain the ultimate source of legal authority. (2) But if there were no such rule, then either acceptance is not justified in authorizing law, or else it justifies itself without a mediating rule. Moreover, it would be untrue, Hart notwithstanding, that a genuine rule of recognition could be inferred from the practice of officials in deciding what is law. If there is a rule to the effect that what is accepted as law is law,[Note 3] then in Hart's terms such a rule would be a secondary rule of recognition —the type of ultimate rule that is only validated or authorized by acceptance. Hart says we can tell what the rule of recognition is in a society by the empirical means of looking and seeing how officials identify valid rules (98, 106, 107).[Note 4] If acceptance in the appropriately complex sense is the criterion of validity in a society, at least for ultimate rules, which Hart believes is the case, then empirical investigation should reveal the fact, and by Hart's terms we must call it a rule of recognition for recognizing the other rules of recognition. But abandoning the normative practice doctrine would not allow Hart to preserve his empirical approach to law, that wants to identify tacit rules by observing practice. Without it, tacit rules might only be discerned as Ross came upon his tacit, transcendent, immutable, and universal rule: by a priori speculation on what is needed to make sense of what happens, without regard for what people and courts accept as making sense. Another way to avoid the circularity would be to distinguish two species of acceptance. The first is the complex sort actually articulated by Hart which is used to authorize the rule of recognition. The second, simpler type could authorize the former type through passive toleration and acquiescence. This remedy looks suspiciously like meaningless hairsplitting, inventing a distinction without a difference just to avoid a circularity. It might commit us, as well, to an infinite regress of species of acceptance in which each new one is invoked to rescue its predecessor from circular justification. Or the circularity might be acknowledged and defended. It is not vicious. Hart's theory is not circular; it is an irreflexive theory of a circular or reflexive phenomenon. As such any charge of fallacy on account of that circularity will beg the question of the nature of the explanandum: for acceptance might truly be authoritative only because it is accepted as authoritative. The statement that acceptance is the ultimate criterion of legality because it is accepted as such, or because it authorizes a rule that authorizes it, is in content simply an expansion of Hart's statement that the ultimate rule of recognition is the ultimate criterion of validity about which it is irrelevant to speak of validity and invalidity (105). If we accept this formulation, then the circularity is as harmless as that in the statement (also arguably to be found in Hart at 45) that customary law is law because it has customarily been law. One might object that acceptance is not law, and therefore need not be authorized, at least not by a rule for recognizing law such as a rule of recognition. This objection may be sound jurisprudence, but it ignores the fact that acceptance is the rule of recognition for the other rules of recognition, according to Hart's normative practice doctrine. We can avoid the self-authorization of acceptance without abandoning the normative practice doctrine by two minor and plausible moves that Hart could well endorse. First, we must explicitly state that acceptance is not law. Second, we must modify the normative practice doctrine to hold that rules of recognition may be inferred from the various practices of officials in deciding what is law except for their practice in deciding what the rule of recognition is (that is, in deciding how to decide what is law). The second move is the important one. Even if plausible, it might well weaken the theory, not strengthen it, for it would leave acceptance unjustified, on a par with a thousand and one social phenomena that neither authorize law nor receive authority from law. Those who are happy that Hart made the ultimate authority of law a matter of observable fact ignore the consequence that Hart has made it impossible to ask whether the admittedly actual, authorized law of a given system is acceptable. If it is accepted, then his inquiry stops. If acceptance is authoritative only because it is accepted, then its authority is self-justifying. Hart did not explicitly argue to this effect, but one may point out that a self-justifying ultimate authority is preferable to the alternatives: an infinite regress of authorities and an unjustified authority. Moreover, the self-justification of acceptance is preferable, on moral and political grounds, to the self-justification of oligarchy, monarchy, or the ipse dixit of the first aspirant to seize the reins of this logic. In the inference model rules are justified or validated by other rules. If every rule is to be justified, there must be an infinite regress, a self-justified rule, or at some point higher rules must be validated by lower rules. To avoid an infinite regress and a counter-hierarchical "stooping" for justification (which is another form of circular justification), the ultimate rule must be self- justifying or unjustified.[Note 5] The inference model cannot abide self-justification and in strict forms cannot admit exceptions; hence it must rest 37