818 F.2d 943 (1987)2
United States Court of Appeals, District of Columbia Circuit.
Argued December 19, 1986.
Decided May 15, 1987.
 William B. Schultz, with whom Alan B. Morrison and Katherine A. Meyer, Washington, D.C., were on the brief for appellants.5
Michael J. Ryan, Asst. U.S. Atty., with whom Joseph E. diGenova, U.S. Atty., Royce C. Lamberth, Craig Lawrence, Asst. U.S. Attys., Washington, D.C., Thomas Scarlett, Chief Counsel, Rockville, Md., and Michael M. Landa, Associate Chief Counsel for Enforcement, Food and Drug Admin., were on the brief for appellee. Patricia J. Kenney, Asst. U.S. Atty., Washington, D.C., Stephen D. Terman, Counsel, Food and Drug Admin., Rockville, Md., Richard K. Willard, Atty. Gen., Leonard Schaitman and John M. Rogers, Attorneys, Dept. of Justice, Washington, D.C., entered appearances for appellee.6
 Philip C. Olsson, Washington, D.C., was on the brief for amicus curiae in support of appellee, State of South Carolina, urging affirmance of the Dist. Court decision.7
Before MIKVA, EDWARDS and STARR, Circuit Judges.8
Opinion PER CURIAM.9
Opinion concurring in part and dissenting in part filed by Circuit Judge STARR.10
This case makes its second appearance before this court. It presents a challenge by a consortium of organizations and private citizens (collectively referred to as CNI) to the Food and Drug Administration's regulation of certain unavoidable contaminants in food, most particularly, aflatoxins in corn. Pursuant to its statutory mandate to limit the amount of "poisonous or deleterious substances" in food, see 21 U.S.C. § 346, FDA establishes "action levels" informing food producers of the allowable levels of unavoidable contaminants such as aflatoxins. Producers who sell products that are contaminated above the action level, which for aflatoxins in corn is currently set at 20 parts per billion, are subject to enforcement proceedings initiated by FDA.12
CNI filed suit in federal district court, launching a three-pronged attack on FDA's action level for aflatoxins in corn: (1) in issuing the action level, FDA failed to comply with the rulemaking requirements of the Food, Drug and Cosmetic Act (FDC Act), see 21 U.S.C. § 346; (2) the action level violated the Administrative Procedure Act because it constitutes a legislative rule issued without the requisite notice-and-comment procedures, see 5 U.S.C. § 553; and (3) FDA's decision to permit adulterated corn to be blended with unadulterated corn to bring the total contamination within the action level violated the FDC Act. The District Court granted summary judgment in favor of FDA on each issue.13
In our initial opinion, we confined ourselves to CNI's first argument. We concluded that the FDC Act, by stating that FDA "shall promulgate regulations," 21 U.S.C. § 346, required that FDA issue formal regulations or "tolerances," rather than informal action levels. Having invalidated the action level on this ground, we concluded that CNI's APA argument was thus rendered moot and that the blending issue stood in need of reevaluation on remand. See 757 F.2d 354 (D.C.Cir.1985).14
The Supreme Court reversed our decision, ___ U.S. ___, 106 S.Ct. 2360, 90 L.Ed.2d 959, holding that the FDC Act was not so clear as to preclude FDA's interpretation of the statute under which the agency could lawfully proceed by way of action levels. 106 S.Ct. 2360 (1986). Since the Court did not reach the APA or blending issues, it remanded the case to this court for "further proceedings consistent with [its] opinion." Id. at 2366. Thus, with the first issue resolved by the High Court, we must now address the still pending APA and blending issues.15
Under the APA, agency rules may be issued only after the familiar notice-and-comment procedures enumerated in the statute are completed. See 5 U.S.C. § 553. It is undisputed that the action level at issue here was promulgated sans those procedures. FDA, however, argues that notice-and-comment requirements do not apply by virtue of subsection (b)(3)(A) of section 553, which carves out an exception for "interpretative rules [and] general statements of policy." According to the  FDA, action levels represent nothing more than nonbinding statements of agency enforcement policy. CNI, on the other hand, argues that the action levels restrict enforcement discretion to such a degree as to constitute legislative rules.17
The distinction between legislative rules and interpretative rules or policy statements has been described at various times as "tenuous," Chisholm v. FCC, 538 F.2d 349, 393 (D.C.Cir.), cert. denied, 429 U.S. 890, 97 S.Ct. 247, 50 L.Ed.2d 173 (1976), "fuzzy," Pacific Gas & Electric Co. v. FPC, 506 F.2d 33, 38 (D.C.Cir.1974), "blurred," Saunders, Interpretative Rules With Legislative Effect: An Analysis and a Proposal For Public Participation, 1986 Duke L.J. 346, 352, and, perhaps most picturesquely, "enshrouded in considerable smog." Noel v. Chapman, 508 F.2d 1023, 1030 (2d Cir.), cert. denied, 423 U.S. 824, 96 S.Ct. 37, 46 L.Ed.2d 40 (1975), quoted in American Bus Association v. United States, 627 F.2d 525, 529 (D.C. Cir.1980). As Professor Davis puts it, "the problem is baffling." 2 K. Davis, Administrative Law Treatise 32 (2d ed. 1979). By virtue of Congress' silence with respect to this matter, it has fallen to the courts to discern the line through the painstaking exercise of, hopefully, sound judgment. Guardian Federal Savings & Loan Ass'n v. FSLIC, 589 F.2d 658, 667 (D.C.Cir.1978).18
Despite the difficulty of the terrain, prior cases do provide some useful guideposts. In this circuit, we are particularly guided by American Bus. There, in speaking for the court, Judge McGowan identified "two criteria" that courts have used in their efforts to fathom the interpretative/legislative distinction:19
First, courts have said that, unless a pronouncement acts prospectively, it is a binding norm. Thus ... a statement of policy may not have a present effect: "a `general statement of policy' is one that does not impose any rights and obligations"....
The second criterion is whether a purported policy statement genuinely leaves the agency and its decisionmakers free to exercise discretion.
627 F.2d at 529 (quoting Texaco v. FPC, 412 F.2d 740, 744 (3d Cir.1969)).21
In conducting our analysis of these two criteria, we consider and give some, albeit "not overwhelming," deference to an agency's characterization of its statement. Brock v. Cathedral Bluffs Shale Oil Co., 796 F.2d 533, 537 (D.C.Cir.1986). As befits a principled exercise in interpretation, courts are to give far greater weight to the language actually used by the agency; we have, for example, found decisive the choice between the words "will" and "may." Compare American Bus, 627 F.2d at 532 (use of "will" indicates statement is in fact a binding norm) with Guardian Federal, 589 F.2d at 666 (use of "may" indicates statement is a "general statement of policy").22
Applying these principles to the case at hand, we are persuaded that the FDA action levels are legislative rules and thus subject to the notice-and-comment requirements of section 553. While FDA  now characterizes the action levels as policy statements, see, e.g., Supplemental Brief for Appellee at 7 ("In sum, action levels do not bind courts, food producers or FDA."), a variety of factors, when considered in light of the criteria set out in American Bus, indicate otherwise.23
First. The language employed by FDA in creating and describing action levels suggests that those levels both have a present effect and are binding. Specifically, the agency's regulations on action levels explain an action level in the following way:24
[A]n action level for an added poisonous or deleterious substance ... may be established to define the level of contamination at which food will be deemed to be adulterated. An action level may prohibit any detectable amount of substance in food.
21 C.F.R. § 109.4 (1986) (emphasis added). This language, speaking as it does of an action level "defin[ing]" the acceptable level and "prohibit[ing]" substances, clearly reflects an interpretation of action levels as presently binding norms. This type of mandatory, definitive language is a powerful, even potentially dispositive, factor suggesting that action levels are substantive rules. Cf. Cathedral Bluffs, 796 F.2d at 538. Moreover, the regulations provide that an action level may appropriately be established whenever there exist the same conditions required to establish a formal tolerance, if, in addition, the "appropriateness" of the tolerance may change in the near future because, for example, of technological changes. 21 C.F.R. § 109.6(c); see also id. § 109.6(b) (requirements for establishing a tolerance).26
Second. This view of action levels — as having a present, binding effect — is confirmed by the fact that FDA considers it necessary for food producers to secure exceptions to the action levels. A specific regulatory provision allows FDA to "exempt from regulatory action and permit the marketing of any food that is unlawfully contaminated with a poisonous or deleterious substance" if certain conditions exist. Id. § 109.8(a). This language implies that in the absence of an exemption, food with aflatoxin contamination over the action level is "unlawful." This putatively unlawful status can derive only from the action level, which, again, indicates that the action level is a presently binding norm. If, as the agency would have it, action levels did indeed "not bind courts, food producers or FDA," Supplemental Brief for Appellee at 7, it would scarcely be necessary to require that "exceptions" be obtained.27
Third. On several occasions, in authorizing blending of adulterated with unadulterated corn, see infra Part II, the FDA has made statements indicating that action levels establish a binding form. For example, in a telegram to the Commissioner of the South Carolina Department of Agriculture,  in which it indicated its approval of a blending plan, the FDA stated that "[a]ny shipments made independent of this plan would, if found to exceed the 20 ppb level, be considered adulterated and subject to condemnation." Joint Appendix (J.A.) at 112 (emphasis added). But we need not resort to informal communications, cf. Ciba-Geigy Corp. v. EPA, 801 F.2d 430 (D.C.Cir.1986), where precision of draftsmanship may understandably be wanting. For, in a formal notice published in the Federal Register of a decision to permit blending and interstate shipment, FDA wrote:28
Any food that contains aflatoxin in excess of 20 ppb ... is considered by FDA to be adulterated under section 402(a)(1) of the Federal Food, Drug, and Cosmetic Act (21 U.S.C. 342(a)(1)), and therefore may not be shipped in interstate commerce.
46 Fed.Reg. 7447 (1981) (emphasis added), reprinted in J.A. at 49. Both statements, one informal and the other elaborately formal, indicate that, when judged under the American Bus criteria, action levels constitute substantive rules. The agency's own words strongly suggest that action levels are not musings about what the FDA might do in the future but rather that they set a precise level of aflatoxin contamination that FDA has presently deemed permissible. Action levels inform food producers what this level is; indeed, that is their very purpose.30
We are not unmindful that in a suit to enjoin shipment of allegedly contaminated corn, it appears that FDA would be obliged to prove that the corn is "adulterated," within the meaning of the FDC Act, rather than merely prove non-compliance with the action level. See United States v. Boston Farm Center, Inc., 590 F.2d 149, 151 (5th Cir.1979). The action level thus does not bind food producers in the sense that producers are automatically subject to enforcement proceedings for violating the action level. This factor, accordingly, points in favor of the agency's characterization. But the fact that action levels do not completely bind food producers as would a more classic legislative rule (where the only issue before the court would be if the agency rule were in fact violated) is not determinative of the issue. For here, we are convinced that FDA has bound itself. As FDA conceded at oral argument, it would be daunting indeed to try to convince a court that the agency could appropriately prosecute a producer for shipping corn with less than 20 ppb aflatoxin. And this type of cabining of an agency's prosecutorial discretion can in fact rise to the level of a substantive, legislative rule. Cf. Nader v. CAB, 657 F.2d 453 (D.C.Cir.1981); Guardian Federal, 589 F.2d at 666-67 ("If it appears that a so-called policy statement is in purpose or likely effect one that narrowly limits administrative discretion, it will be taken for what it is — a binding rule of substantive law.") (emphasis added). That is exactly what has happened here.31
In sum, consideration of a variety of factors leads us to conclude that the FDA's action levels are not within the section 553(b)(3)(A) exception to notice-and-comment requirements. Since all agree that  those procedures were not followed, the action level at issue here cannot stand.32
We pause to observe that nothing in our decision today generally precludes FDA from proceeding by way of informal action levels. The Supreme Court has, of course, decided that such a course is permissible under the FDC Act. Our limited holding is that the current action levels are treated as substantive rules by FDA and, as such, can only be permitted if notice-and-comment procedures are employed. If it so chooses, FDA could proceed by action levels that are pure policy statements. But in order to do so, FDA must avoid giving action levels the kind of substantive significance that it now so plainly attaches to them.33
We add one additional caveat. Our holding today in no way indicates that agencies develop written guidelines to aid their exercise of discretion only at the peril of having a court transmogrify those guidelines into binding norms. We recognize that such guidelines have the not inconsiderable benefits of apprising the regulated community of the agency's intentions as well as informing the exercise of discretion by agents and officers in the field. It is beyond question that many such statements are non-binding in nature and would thus be characterized by a court as interpretative rules or policy statements. We are persuaded that courts will appropriately reach an opposite conclusion only where, as here, the agency itself has given its rules substantive effect.34
In sum, our holding today is narrow. We conclude that in the circumstances of this case, FDA by virtue of its own course of conduct has chosen to limit its discretion and promulgated action levels which it gives a present, binding effect. Having accorded such substantive significance to action levels, FDA is compelled by the APA to utilize notice-and-comment procedures in promulgating them.35
The FDA has, on several occasions, granted exemptions to permit the blending of contaminated corn with uncontaminated corn. See, e.g., 43 Fed.Reg. 14,122 (1978), reprinted in J.A. at 41. This blending is typically permitted to help food producers cope with seasonal weather variations which may result in a particular harvest yielding corn with a higher level of aflatoxin. See Supplementary Brief of the State of South Carolina as Amicus Curiae in Support of Appellee at 1-4. In such circumstances, the FDA may permit blending to bring the total level of aflatoxin contamination in the final, blended corn within acceptable levels.37
CNI contends that this blending practice causes the final product to be "adulterated" within the meaning of 21 U.S.C. § 342(a)(2)(A). This section reads, in relevant part, as follows:38
A food shall be deemed to be adulterated — ... if it bears or contains any added poisonous or added deleterious substance ... which is unsafe within the meaning of section 346 of this title.
Section 346, in turn, states that40
[a]ny poisonous or deleterious substance added to any food, except where such substance ... cannot be avoided ... shall be deemed to be unsafe for purposes of the application of clause (2)(A) of section 342(a).
21 U.S.C. § 346. The intentional blending of contaminated corn obviously cannot in reason be considered unavoidable. Surely there can be little doubt that blended corn therefore stands branded as "adulterated" for purposes of the FDC Act; indeed, in its initial brief FDA conceded as much:42
The Government does not dispute [CNI's] contention that use of an adulterated feed as an ingredient in another feed causes the finished (blended) feed to be adulterated.
Brief for Appellee at 36. But as FDA goes on to point out, a conclusion that a particular food product is "adulterated," in the abstract, means little other than that FDA  could choose to initiate enforcement proceedings. CNI does not, and could not, point to any provision in the FDC Act requiring FDA to initiate enforcement action against every food falling within the Act's definition of "adulterated." Rather, the Act makes clear that these enforcement decisions are vested, as they traditionally are, with the agency. Section 336, for example, expressly recognizes that FDA need not "report for prosecution, or for the institution of libel or injunction proceedings, minor violations of this [Act]." 21 U.S.C. § 336.44
Upon analysis, therefore, the gravamen of CNI's complaint is that FDA failed to initiate enforcement proceedings. But as the Supreme Court held in Heckler v. Chaney, 470 U.S. 821, 105 S.Ct. 1649, 84 L.Ed.2d 714 (1985), FDA enjoys complete discretion not to employ the enforcement provisions of the FDC Act, and those decisions are not subject to judicial review. As this court recently concluded, the "provisions [of the FDC Act] authorize, but do not compel the FDA to undertake enforcement activity; they `commit complete discretion to the [FDA] to decide how and when they should be exercised.'" Schering Corp. v. Heckler, 779 F.2d 683, 686 (D.C.Cir.1985) (quoting Heckler v. Chaney, 105 S.Ct. at 1658). In light of this principle, CNI's arguments on the blending issue must fall.45
In conclusion, we find FDA's action levels to be invalid in that they were issued without the requisite notice-and-comment procedures, but we reject CNI's challenge to FDA's disinclination to initiate enforcement action against certain blended corn. Accordingly, the case is remanded to the District Court for further proceedings not inconsistent with this opinion.47
It is so ordered.48
I fully concur in section II of the panel opinion, holding that FDA's decision to permit blending fell within the agency's enforcement discretion. I am also persuaded that the majority's careful treatment of the substantive rules-policy statements distinction is not unfaithful to the teachings of this circuit's more recent precedents. Nonetheless, I am constrained to conclude that the correct rule was that laid down by our court thirteen years ago in Pacific Gas & Electric Co. v. FPC, where we held:50
The critical distinction between a substantive rule and a general statement of policy is the different practical effect that these two types of pronouncements have in subsequent administrative proceedings.... A properly adopted substantive rule establishes a standard of conduct which has the force of law.
506 F.2d 33, 38 (D.C.Cir.1974). The Pacific Gas-enunciated factor — whether the pronouncement has the force of law in subsequent proceedings — should be deemed determinative of the issue. Because there is no doubt that the agency pronouncements at issue here have no such effect, I respectfully dissent from section I of the panel opinion.52
The abiding characteristic of a legislative rule is that it is law. It defines a standard of conduct that regulated individuals or entities ignore at their peril, in the face of possible enforcement action. Significantly, the only issue in any such proceeding is whether the rule applies to the facts at hand. "The underlying policy embodied in the rule is not generally subject to challenge before the agency." Id., at 38.54
Before such rules are sanctioned one would think that they should be carefully crafted, with the "underlying polic[ies] embodied in the rule" recognized, openly discussed, and deliberately weighed. To return to basic civics for a moment, statutes passed by Congress have been refined in this manner by the very nature of the legislative process. Bills are considered by committees, hearings are held, debate is conducted, compromises are reached, and votes are taken. And all this is carried on in a bicameral legislative body with the final result presented to the President for  his approval or rejection. See INS v. Chadha, 462 U.S. 919, 103 S.Ct. 2764, 77 L.Ed.2d 317 (1983). A statute therefore possesses important attributes justifying placement of the coercive power of the state behind a particular policy framed in accordance with the process ordained at the Founding.55
In the modern administrative state, many "laws" emanate not from Congress but from administrative agencies, inasmuch as Congress has seen fit to vest broad rulemaking power in the executive branch, including independent agencies. See Synar v. United States, 626 F.Supp. 1374, 1383-84 (D.D.C.) (three judge court), aff'd, ___ U.S. ___, 106 S.Ct. 3181, 92 L.Ed.2d 583 (1986); K. Davis Administrative Law Treatise § 3.8 (1982 Supp.); cf. Ticor Title Insurance Co. v. FTC, 814 F.2d 731, 732 (D.C.Cir.1987) (dismissing on procedural grounds a challenge to the delegation of law enforcement powers to an independent agency). This rulemaking power is obviously cabined by whatever requirements may exist in the particular statute delegating rulemaking authority, a subject which we treated in our initial decision in this case. See 757 F.2d 354 (D.C.Cir.1985). But Congress has also provided in the APA for certain procedural protections before that which achieves the lofty status of "law" is promulgated by an agency acting in its Congressionally authorized lawmaking capacity. Chief among these protections are the notice-and-comment requirements laid down in the familiar provision of 5 U.S.C. § 553. In a sense, notice-and-comment procedures serve as a Congressionally mandated proxy for the procedures which Congress itself employs in fashioning its "rules," as it were, thereby insuring that agency "rules" are also carefully crafted (with democratic values served by public participation) and developed only after assessment of relevant considerations. It is thus, in theory, important for APA procedures to be followed before an agency pronouncement is deemed a binding legislative rule not merely because the APA says so, but because in saying so the APA is protecting a free people from the danger of coercive state power undergirding pronouncements that lack the essential attributes of deliberativeness present in statutes. Because of the value inhering in such procedures, it is well-established that "`only reluctantly should courts recognize exceptions therefrom.'" American Bus Association v. United States, 627 F.2d 525, 528 (D.C.Cir.1980) (quoting Humana of South Carolina v. Califano, 590 F.2d 1070, 1082 (D.C.Cir.1978)).56
Nonetheless, in crafting the APA, Congress directed that courts should recognize certain exceptions to the statute's notice-and-comment requirements. Specifically, Congress recognized that not all agency pronouncements, even those of considerable moment, rise to the dignity of law. Thus, the APA excepts, as the panel opinion recounts, interpretative rules and general statements of policy from the general notice-and-comment requirements. While it is no doubt true, and indeed is frequently recognized, that such agency pronouncements may have a direct effect on the regulated community, and may even be judicially reviewable, see, e.g., AFL-CIO v. Donovan, 757 F.2d 330, 341-43 (D.C.Cir.1985); cf. Abbott Laboratories v. Gardner, 387 U.S. 136, 87 S.Ct. 1507, 18 L.Ed.2d 681 (1967), these pronouncements still lack the dignity of "law." Before that status can be achieved, the agency must run its policies through the notice-and-comment gantlet. Perhaps in part because the agency here has avoided testing its pronouncements in this way, it must in any future proceeding defend and justify its chosen standard in the face of a challenge to that standard. Cf. United States v. Boston Farm Center, Inc., 590 F.2d 149, 151 (5th Cir.1979).57
The majority is quite correct when it chronicles the difficulty courts have found in attempting to fathom the distinction between legislative or substantive rules on one hand, and interpretative rules or policy statements on the other. Inasmuch as our  decisional law over the last decade avowedly reflects considerable uncertainty in discerning the line between agency pronouncements that are "law" and those that are "policy," see Majority Opinion at 946, it seems advisable to return to the pristine teaching of Pacific Gas. In that case, this court articulated a rule which is clearly preferable to the present muddy state of the law. The wisdom of the Pacific Gas principle is in no small measure found in the fact that the case reflects the "core principles" which I sought briefly to adumbrate in the preceding section of this opinion. As I read the case, Pacific Gas deems as "critical" the effect of the agency pronouncement in future proceedings. This is as it should be, for as I discussed above, it is this element that is the essence of "law." Not only is the Pacific Gas approach therefore the most principled manner in which to draw the legislative-interpretative line (in view of the fact that the determination is whether a pronouncement is "law" or not), but it has the not insignificant practical benefit in an unclear world of providing great clarity where previously there has been "considerable smog." See American Bus, 627 F.2d at 529.59
We should reembrace our Pacific Gas test as the determinative factor in analyzing whether a particular pronouncement is legislative or interpretative in nature. If the pronouncement has the force of law in future proceedings, it is a legislative rule. Unless that critical feature is present, however, the agency statement should be considered to be a lower form of pronouncement, a "non-law" as it were, or in APA terms an "interpretative rule" or "general statement of policy." The correct measure of a pronouncement's force in subsequent proceedings is a practical one: must the agency merely show that the pronouncement has been violated or must the agency, if its hand is called, show that the pronouncement itself is justified in light of the underlying statute and the facts.60
Application of this test can readily be illustrated by the case at hand. Action levels offer guidance to the regulated community with respect to what products FDA deems adulterated within the meaning of the FDC Act. But in an enforcement proceeding in which FDA seeks either to impose sanctions for shipment of an adulterated product or to enjoin shipment of an adulterated product, the agency must prove the product is "adulterated." That is, FDA cannot merely show that the product at issue fails to comply with the action level. See Boston Farm Center, 590 F.2d at 151. Rather, FDA must offer scientific or other probative evidence to support its contention that the product is adulterated. Thus, the action level does not have the force of law in the subsequent proceeding. Indeed, it has no "force" at all.61
Let me quickly observe that I am not encouraging rebellion and sedition within the circuit. While the law has certainly evolved since Pacific Gas, at no point has this court disavowed that decision. Quite to the contrary, the leading case since Pacific Gas, American Bus, expressly embraced the earlier decision. See 627 F.2d at  529. Moreover, post-American Bus decisions grappling with the legislative rule-interpretative rule distinction have often emphasized the Pacific Gas factor. Just the other day, for example, Judge Bork, in writing for the court in The National Latino Media Coalition v. FCC, 816 F.2d 785, 788 (D.C.Cir.1987), emphasized that "an interpretative rule does not have the force of law and is not binding upon anyone" in holding that the agency pronouncements at issue there were not legislative rules and thus not subject to the APA's notice-and-comment requirements. To my eye, this represents an application, albeit not explicit, of Pacific Gas. Thus, I view Pacific Gas as not only "good" law, in the sense that it has never been overruled, but can discern in our precedents examples, including very fresh ones, of its being heartily applied. I would therefore urge that the "force of law" factor from Pacific Gas be explicitly recognized anew as the polestar guiding the legislative-interpretative determination. Such a modest step would not only be faithful to fundamental principles informing the nature of legislative rules but would have the happy practical consequence of bringing clarity to a troubled area.62
While I thus view this step as warranted, I recognize a potential danger lurking in the embrace of a single-factor, Pacific Gas test. Agencies may yield to temptation and seek to shield their regulations from the scrutiny occasioned by notice-and-comment procedures, choosing instead to cast would-be regulations as interpretative rules. The rule would still, of course, be subject to scrutiny in a subsequent proceeding, but this fact may be of little comfort to prospective commentors, given the deference accorded agency views in any such proceedings. See generally Chevron U.S.A. Inc. v. NRDC, 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). But upon analysis, the danger is more theoretical than real. Indeed, Congress not atypically provides agencies with a direct command to promulgate regulations, cf., e.g., Industrial Union Department, AFL-CIO v. American Petroleum Institute, 448 U.S. 607, 100 S.Ct. 2844, 65 L.Ed.2d 1010 (1980), thereby imposing a duty that would not be satisfied with issuance of an humble interpretative rule. In this case, of course, the Supreme Court has now ruled that FDA is not required by the organic statute to issue formal regulations, but can instead proceed by way of informal action levels. ___ U.S. ___, 106 S.Ct. 2360, 90 L.Ed.2d 959 (1986). But if Congress prefers a traditional model of command to the agencies to issue regulations, it can embrace that model through legislation, informed by the need, reiterated by the Supreme Court in this very case, to avoid dangling participles and other forms of that omnipresent occupational hazard of the draftsman's workshop, ambiguity. But see INS v. Cardoza-Fonseca, ___ U.S. ___, 107 S.Ct. 1207, 94 L.Ed.2d 434 (1987); Board of Governors v. Dimension Financial Corp., 474 U.S. 361, 106 S.Ct. 681, 88 L.Ed.2d 691 (1986). Because Pacific Gas, and the basic principles upon which it stands, points the way to a quite different result than that reached by my colleagues, I am constrained to dissent in part.63
 Aflatoxins are by-products of certain common molds that grow on various crops, including corn.64
 The APA defines "rule" as "the whole or a part of an agency statement of general or particular applicability and future effect designed to implement, interpret, or prescribe law or policy...." 5 U.S.C. § 551(4). FDA does not dispute that its action levels are indeed "rules." The notice-and-comment requirements of section 553 are thus unquestionably triggered and the only question before us is whether one of the statutory exceptions to notice and comment is applicable.65
 See also Brock v. Cathedral Bluffs Shale Oil Co., 796 F.2d 533, 536-37 (D.C.Cir.1986) ("[T]here is no axiom to distinguish between a regulation and general statements of policy.").66
 The first criterion from American Bus could be read as facially inconsistent with the APA. Specifically, criterion one appears to indicate that interpretative rules must be solely prospective: "[C]ourts have said that, unless a pronouncement acts prospectively, it is a binding norm." 627 F.2d at 529. Since under the APA, interpretative rules can be developed and offered as present indications of the agency's enforcement policy, this "present-is-legislative" reading of American Bus would create a conflict with the statutory definition. In our view, however, criterion one from American Bus is addressed to the binding effect of the agency pronouncement; if a statement has a present-day binding effect, it is legislative. Mere pronouncements of what the agency intends, whether for the present or for the future, which do not have a binding effect, are properly classified as interpretative rules. Thus, it is the binding effect, not the timing, which is the essence of criterion one.67
 As now-Justice Scalia put it when writing for this court: "[T]here is deference and there is deference — and the degree accorded to the agency on a point such as this is not overwhelming." Cathedral Bluffs, 796 F.2d at 537.68
 The conditional "may" in this provision refers only to the agency's discretion in deciding whether to establish an action level. It does not suggest that, once established, an action level is non-binding. Rather, with respect to binding effect, the regulation speaks in non-conditional terms.69
 See also Action Levels for Poisonous or Deleterious Substances in Human Food and Animal Feed (FDA Publication 1980) ("Action levels and tolerances represent limits at or above which FDA will take legal action to remove adulterated products from the market.") (emphasis added), reprinted in Joint Appendix (J.A.) at 76, 77.70
 The specific action level for aflatoxin in corn was published in the Federal Register. In Cathedral Bluffs, this court indicated that such publication did not necessarily suggest that "the matter published was meant to be a regulation, since the APA requires general statements of policy to be published as well." 796 F.2d at 539 (emphasis in original). Rather, the Cathedral Bluffs court indicated that the "real dividing point between regulations and general statements of policy is publication in the Code of Federal Regulations." Id. This factor, among others, led the court to conclude that the pronouncement at issue there was a general statement of policy. In our case, however, the line cannot be so simply drawn. While the specific action level itself was not published in the Code of Federal Regulations, FDA did publish in the C.F.R. regulations authorizing establishment of action levels which, as noted above, indicate that action levels are to have a binding effect. Thus, we deal with a mixed Federal Register-C.F.R. publication, and accordingly find ourselves unable to divine persuasive indications, one way or another, from what may be called the "publication factor."71
 Boston Farm Center could be read to indicate that the Fifth Circuit viewed action levels as mere general statements of policy. In reviewing the Government's argument that a court should defer to the action level as positively indicating a violation of the Act, the Boston Farm Center court noted that "[t]he deference principle is less compelling when the agency threshold is a matter of prosecutorial discretion instead of rule-making." 590 F.2d at 149. But see United States v. Ewig Bros., 502 F.2d 715, 725 (7th Cir.1974) (An action level is "meant to operate more like a rule of general applicability than a mere prediction of how the agency would choose to exercise its prosecutorial discretion."), cert. denied, 420 U.S. 945, 95 S.Ct. 1324, 43 L.Ed.2d 423 (1975). We may well prefer the Seventh Circuit's characterization, but neither court was addressing the issue before us: whether action levels are within the interpretative rules/general statements of policy exception to APA notice-and-comment requirements. Thus, these casual characterizations are of limited help in our inquiry.72
 Our conclusion is buttressed by the Supreme Court's opinion in this case, which could be read to reflect an understanding of the action levels as substantive rules. Specifically, the Court described action levels in the following terms: "In setting an action level, the FDA essentially assures food producers that it ordinarily will not enforce the general adulteration provisions of the Act against them." 106 S.Ct. at 2363. This characterization of the action levels as an agency assurance is consistent with our holding that the action levels are substantive rules subject to notice-and-comment procedures.73
 In addition, the difficulty can be seen in attempts by courts to quite understandably make sense of this area by seeking simple, easily-applied rules. In Brock v. Cathedral Bluffs Shale Oil Co., 796 F.2d 533, 539 (D.C.Cir.1986), now-Justice Scalia intimates that "[t]he real dividing point between regulations and general statements of policy is publication in the Code of Federal Regulations." Such an approach would certainly be simple, if arguably inconsistent with the notion that the deference entitled to an agency's own characterizations of its statements is "not overwhelming." Id., at 537. I prefer the Pacific Gas approach, which has a much firmer rooting in the core principles at issue.74
 I recognize that in many ways, nothing can be "readily" demonstrated by the case at hand; action levels, as the FDA has fashioned them, are a rather curious hybrid. Under a functionalist analysis, this form of agency action comes tantalizingly close to a substantive rule. I find particularly compelling the fact that FDA considers it necessary to grant "exceptions" to its action levels. See 21 C.F.R. § 109.8 (1986). Thus, I must confess in candor that I view the majority's approach not only as faithful to our circuit's more recent pronouncements, but as buttressed by common sanse.75
 What seems to have taken place as the law has evolved (and arguably become obfuscated) is that the Pacific Gas factor has been consigned to a supporting role, as one aspect of what seems to be a multi-factor test.
82 F.3d 165 (1996)2
United States Court of Appeals, Seventh Circuit.
Argued February 9, 1996.
Decided April 25, 1996.
  William J. Tabor (argued), Terre Haute, IN, for Patrick D. Hoctor.5
Margaret M. Breinholt, M. Bradley Flynn (argued), Dept. of Agriculture, Office of Gen. Counsel, Washington, DC, for U.S. Dept. of Agriculture.6
Before POSNER, Chief Judge, and DIANE P. WOOD and EVANS, Circuit Judges.7
A rule promulgated by an agency that is subject to the Administrative Procedure Act is invalid unless the agency first issues a public notice of proposed rulemaking, describing the substance of the proposed rule, and gives the public an opportunity to submit written comments; and if after receiving the comments it decides to promulgate the rule it must set forth the basis and purpose of the rule in a public statement. 5 U.S.C. §§ 553(b), (c). These procedural requirements do not apply, however, to "interpretative rules, general statements of policy, or rules of agency organization, procedure, or practice." 5 U.S.C. § 553(b)(A). Distinguishing between a "legislative" rule, to which the notice and comment provisions of the Act apply, and an interpretive rule, to which these provisions do not apply, is often very difficult — and often very important to regulated firms, the public, and the agency. Notice and comment rulemaking is time-consuming, facilitates the marshaling of opposition to a proposed rule, and may result in the creation of a very long record that may in turn provide a basis for a judicial challenge to the rule if the agency decides to promulgate it. There are no formalities attendant upon the promulgation of an interpretive rule, but this is tolerable because such a rule is "only" an interpretation. Every governmental agency that enforces a less than crystalline statute must interpret the statute, and it does the public a favor if it announces the interpretation in advance of enforcement, whether the announcement takes the form of a rule or of a policy statement, which the Administrative Procedure Act assimilates to an interpretive rule. It would be no favor to the public to discourage the announcement of agencies' interpretations by burdening the interpretive process with cumbersome formalities.9
The question presented by this appeal from an order of the Department of Agriculture is whether a rule for the secure containment of animals, a rule promulgated by the Department under the Animal Welfare Act, 7 U.S.C. §§ 2131 et seq., without compliance with the notice and comment requirements of the Administrative Procedure Act, is nevertheless valid because it is merely an interpretive rule. Enacted in 1966, the Animal Welfare Act, as its title implies, is primarily designed to assure the humane treatment of animals. The Act requires the licensing of dealers (with obvious exceptions, for example retail pet stores) and exhibitors, and authorizes the Department to impose sanctions on licensees who violate either the statute itself or the rules promulgated by the Department under the authority of 7 U.S.C. § 2151, which authorizes the Secretary of Agriculture "to promulgate such rules, regulations, and orders as he may deem necessary in order to effectuate the purposes of [the Act]." The Act provides guidance to the exercise of this rulemaking authority by requiring the Department to formulate standards "to govern the humane handling, care, treatment, and transportation of animals by dealers," and these standards must include minimum requirements "for handling, housing, feeding, watering, sanitation," etc. 7 U.S.C. § 2143(a).10
The Department has employed the notice and comment procedure to promulgate a regulation, the validity of which is not questioned, that is entitled "structural strength" and that provides that "the facility [housing the animals] must be constructed of such  material and of such strength as appropriate for the animals involved. The indoor and outdoor housing facilities shall be structurally sound and shall be maintained in good repair to protect the animals from injury and to contain the animals." 9 C.F.R. § 3.125(a).11
Enter the petitioner, Patrick Hoctor, who in 1982 began dealing in exotic animals on his farm outside of Terre Haute. In a 25-acre compound he raised a variety of animals including "Big Cats"—a typical inventory included three lions, two tigers, seven ligers (a liger is a cross between a male lion and a female tiger, and is thus to be distinguished from a tigon), six cougars, and two snow leopards. The animals were in pens ("primary enclosures" in the jargon of the administration of the Animal Welfare Act). The area in which the pens were located was surrounded by a fence ("containment fence"). In addition, Hoctor erected a fence around the entire compound ("perimeter fence"). At the suggestion of a veterinarian employed by the Agriculture Department who was assigned to inspect the facility when Hoctor started his animal dealership in 1982, Hoctor made the perimeter fence six feet high.12
The following year the Department issued an internal memorandum addressed to its force of inspectors in which it said that all "dangerous animals," defined as including, among members of the cat family, lions, tigers, and leopards, must be inside a perimeter fence at least eight feet high. This provision is the so-called interpretive rule, interpreting the housing regulation quoted above. An agency has, of course, the power, indeed the inescapable duty, to interpret its own legislative rules, such as the housing standard, just as it has the power and duty to interpret a statute that it enforces. Stinson v. United States, 508 U.S. 36, 42-46, 113 S.Ct. 1913, 1918-19, 123 L.Ed.2d 598 (1993).13
On several occasions beginning in 1990, Hoctor was cited by a Department of Agriculture inspector for violating 9 C.F.R. § 3.125(a), the housing standard, by failing to have an eight-foot perimeter fence. Eventually the Department sanctioned Hoctor for this and other alleged violations, and he has sought judicial review limited, however, to the perimeter fence. He is a small dealer and it would cost him many thousands of dollars to replace his six-foot-high fence with an eight-foot-high fence. Indeed, we were told at argument that pending the resolution of his dispute over the fence he has discontinued dealing in Big Cats. The parties agree that unless the rule requiring a perimeter fence at least eight feet high is a valid interpretive rule, the sanction for violating it was improper.14
We may assume, though we need not decide, that the Department of Agriculture has the statutory authority to require dealers in dangerous animals to enclose their compounds with eight-foot-high fences. The fence is a backup fail-safe device, since the animals are kept in pens, cages, or other enclosures within the compound, in an area that is itself fenced, rather than being free to roam throughout the compound. Since animals sometimes break out or are carelessly let out of their pens, a fail-safe device seems highly appropriate, to say the least. Two lions once got out of their pen on Hoctor's property, and he had to shoot them. Yet, when he did so, they were still within the containment fence. The Department's regulations do not require a containment fence, and it is unclear to us why, if that fence was adequate—and we are given no reason to suppose it was not — Hoctor should have had to put up an additional fence, let alone one eight-feet high. But we lay any doubts on this score to one side. And we may also assume that the containment of dangerous animals is a proper concern of the Department in the enforcement of the Animal Welfare Act, even though the purpose of the Act is to protect animals from people rather than people from animals. Even Big Cats are not safe outside their compounds, and with a lawyer's ingenuity the Department's able counsel reminded us at argument that if one of those Cats mauled or threatened a human being, the Cat might get into serious trouble and thus it is necessary to protect human beings from Big Cats in order to protect the Cats from human beings, which is the important thing under the Act. In fact Hoctor had shot the two lions because they were dangerously close to one of his employees. Since tort liability for injury caused by a wild  animal is strict, Burns v. Gleason, 819 F.2d 555 (5th Cir.1987); Behrens v. Bertram Mills Circus Ltd.,  2 Q.B. 1; W. Page Keeton et al., Prosser and Keeton on the Law of Torts § 76, p. 542 (5th ed. 1984), the common law, at least, is solicitous for the protection of the citizens of Terre Haute against escapees from Hoctor's menagerie even if the Animal Welfare Act is not. The internal memorandum also justifies the eight-foot requirement as a means of protecting the animals from animal predators, though one might have supposed the Big Cats able to protect themselves against the native Indiana fauna.15
Another issue that we need not resolve besides the issue of the statutory authority for the challenged rule is whether the Department might have cited Hoctor for having a perimeter fence that was in fact, considering the number and type of his animals, the topography of the compound, the design and structure of the protective enclosures and the containment fence, the proximity of highways or inhabited areas, and the design of the perimeter fence itself, too low to be safe, as distinct from merely being lower than eight feet. No regulation is targeted on the problem of containment other than 9 C.F.R. § 3.125, which seems to be concerned with the strength of enclosures rather than their height. But maybe there is some implicit statutory duty of containment that Hoctor might have been thought to have violated even if there were no rule requiring an eight-foot-high perimeter fence.16
We need not decide. The only ground on which the Department defends sanctioning Hoctor for not having a high enough fence is that requiring an eight-foot-high perimeter fence for dangerous animals is an interpretation of the Department's own structural-strength regulation, and "provided an agency's interpretation of its own regulations does not violate the Constitution or a federal statute, it must be given `controlling weight unless it is plainly erroneous or inconsistent with the regulation.'" Stinson v. United States, supra, 508 U.S. at 44-46, 113 S.Ct. at 1919. The "provided" clause does not announce a demanding standard of judicial review, although the absence of any reference in the housing regulation to fences or height must give us pause. The regulation appears only to require that pens and other animal housing be sturdy enough in design and construction, and sufficiently well maintained, to prevent the animals from breaking through the enclosure — not that any enclosure, whether a pen or a perimeter fence, be high enough to prevent the animals from escaping by jumping over the enclosure. The Department's counsel made the wonderful lawyer's argument that the eight-foot rule is consistent with the regulation because a fence lower than eight feet has zero structural strength between its height (here six feet) and the eight-foot required minimum. The two feet by which Hoctor's fence fell short could not have contained a groundhog, let alone a liger, since it was empty space.17
Our doubts about the scope of the regulation that the eight-foot rule is said to be "interpreting" might seem irrelevant, since even if a rule requiring an eight-foot perimeter fence could not be based on the regulation, it could be based on the statute itself, which in requiring the Department to establish minimum standards for the housing of animals presumably authorizes it to promulgate standards for secure containment. But if the eight-foot rule were deemed one of those minimum standards that the Department is required by statute to create, it could not possibly be thought an interpretive rule. For what would it be interpreting? When Congress authorizes an agency to create standards, it is delegating legislative authority, rather than itself setting forth a standard which the agency might then particularize through interpretation. Put differently, when a statute does not impose a duty on the persons subject to it but instead authorizes (or requires — it makes no difference) an agency to impose a duty, the formulation of that duty becomes a legislative task entrusted to the agency. Provided that a rule promulgated pursuant to such a delegation is intended to bind, and not merely to be a tentative statement of the agency's view, which would make it just a policy statement, and not a rule at all, the rule would be the clearest possible example of a legislative rule, as to which the notice and comment procedure not followed here is mandatory, as distinct  from an interpretive rule; for there would be nothing to interpret. American Mining Congress v. Mine Safety & Health Administration, 995 F.2d 1106, 1109 (D.C.Cir.1993); Robert A. Anthony, "`Interpretive' Rules, `Legislative' Rules and `Spurious' Rules: Lifting the Smog," 8 Admin. L.J. of Am. Univ. 1 (1994). That is why the Department must argue that its eight-foot rule is an interpretation of the structural-strength regulation-itself a standard, and therefore interpretable, in order to avoid reversal.18
Even if, despite the doubts that we expressed earlier, the eight-foot rule is consistent with, even in some sense authorized by, the structural-strength regulation, it would not necessarily follow that it is an interpretive rule. It is that only if it can be derived from the regulation by a process reasonably described as interpretation. Metropolitan School District v. Davila, 969 F.2d 485, 490 (7th Cir.1992). Supposing that the regulation imposes a general duty of secure containment, the question is, then, Can a requirement that the duty be implemented by erecting an eight-foot-high perimeter fence be thought an interpretation of that general duty?19
"Interpretation" in the narrow sense is the ascertainment of meaning. It is obvious that eight feet is not part of the meaning of secure containment. But "interpretation" is often used in a much broader sense. A process of "interpretation" has transformed the Constitution into a body of law undreamt of by the framers. To skeptics the Miranda rule is as remote from the text of the Fifth Amendment as the eight-foot rule is from the text of 9 C.F.R. § 3.125(a). But our task in this case is not to plumb the mysteries of legal theory; it is merely to give effect to a distinction that the Administrative Procedure Act makes, and we can do this by referring to the purpose of the distinction. The purpose is to separate the cases in which notice and comment rulemaking is required from the cases in which it is not required. As we noted at the outset, unless a statute or regulation is of crystalline transparency, the agency enforcing it cannot avoid interpreting it, and the agency would be stymied in its enforcement duties if every time it brought a case on a new theory it had to pause for a bout, possibly lasting several years, of notice and comment rulemaking. Besides being unavoidably continuous, statutory interpretation normally proceeds without the aid of elaborate factual inquiries. When it is an executive or administrative agency that is doing the interpreting it brings to the task a greater knowledge of the regulated activity than the judicial or legislative branches have, and this knowledge is to some extent a substitute for formal fact-gathering.20
At the other extreme from what might be called normal or routine interpretation is the making of reasonable but arbitrary (not in the "arbitrary or capricious" sense) rules that are consistent with the statute or regulation under which the rules are promulgated but not derived from it, because they represent an arbitrary choice among methods of implementation. A rule that turns on a number is likely to be arbitrary in this sense. There is no way to reason to an eight-foot perimeter-fence rule as opposed to a seven-and-a-half foot fence or a nine-foot fence or a ten-foot fence. None of these candidates for a rule is uniquely appropriate to, and in that sense derivable from, the duty of secure containment. This point becomes even clearer if we note that the eight-foot rule actually has another component — the fence must be at least three feet from any animal's pen. Why three feet? Why not four? Or two?21
The reason courts refuse to create statutes of limitations is precisely the difficulty of reasoning to a number by the methods of reasoning used by courts. Hemmings v. Barian, 822 F.2d 688, 689 (7th Cir.1987). One cannot extract from the concept of a tort that a tort suit should be barred unless brought within one, or two, or three, or five years. The choice is arbitrary and courts are uncomfortable with making arbitrary choices. They see this as a legislative function. Legislators have the democratic legitimacy to make choices among value judgments, choices based on hunch or guesswork or even the toss of a coin, and other arbitrary choices. When agencies base rules on arbitrary choices they are legislating, and so these rules are legislative or substantive and  require notice and comment rulemaking, a procedure that is analogous to the procedure employed by legislatures in making statutes. The notice of proposed rulemaking corresponds to the bill and the reception of written comments to the hearing on the bill.22
The common sense of requiring notice and comment rulemaking for legislative rules is well illustrated by the facts of this case. There is no process of cloistered, appellate-court type reasoning by which the Department of Agriculture could have excogitated the eight-foot rule from the structural-strength regulation. The rule is arbitrary in the sense that it could well be different without significant impairment of any regulatory purpose. But this does not make the rule a matter of indifference to the people subject to it. There are thousands of animal dealers, and some unknown fraction of these face the prospect of having to tear down their existing fences and build new, higher ones at great cost. The concerns of these dealers are legitimate and since, as we are stressing, the rule could well be otherwise, the agency was obliged to listen to them before settling on a final rule and to provide some justification for that rule, though not so tight or logical a justification as a court would be expected to offer for a new judge-made rule. Notice and comment is the procedure by which the persons affected by legislative rules are enabled to communicate their concerns in a comprehensive and systematic fashion to the legislating agency. The Department's lawyer speculated that if the notice and comment route had been followed in this case the Department would have received thousands of comments. The greater the public interest in a rule, the greater reason to allow the public to participate in its formation.23
We are not saying that an interpretive rule can never have a numerical component. See, e.g., American Mining Congress v. Mine Safety & Health Administration, supra, 995 F.2d at 1108, 1113; St. Mary's Hospital v. Blue Cross & Blue Shield Ass'n., 788 F.2d 888, 889-91 (2d Cir.1986). There is merely an empirical relation between interpretation and generality on the one hand, and legislation and specificity on the other. Especially in scientific and other technical areas, where quantitative criteria are common, a rule that translates a general norm into a number may be justifiable as interpretation. The mine safety agency in the American Mining case could refer to established medical criteria, expressed in terms of numerical evaluations of x-rays, for diagnosing black-lung disease. 995 F.2d at 1112-13. Even in a nontechnical area the use of a number as a rule of thumb to guide the application of a general norm will often be legitimately interpretive. Had the Department of Agriculture said in the internal memorandum that it could not imagine a case in which a perimeter fence for dangerous animals that was lower than eight feet would provide secure containment, and would therefore presume, subject to rebuttal, that a lower fence was insecure, it would have been on stronger ground. For it would have been tying the rule to the animating standard, that of secure containment, rather than making it stand free of the standard, self-contained, unbending, arbitrary. To switch metaphors, the "flatter" a rule is, the harder it is to conceive of it as merely spelling out what is in some sense latent in a statute or regulation, and the eight-foot rule in its present form is as flat as they come. At argument the Department's lawyer tried to loosen up the rule, implying that the Department might have bent it if Hoctor proposed to dig a moat or to electrify his six-foot fence. But an agency's lawyer is not authorized to amend its rules in order to make them more palatable to the reviewing court.24
The Department's position might seem further undermined by the fact that it has used the notice and comment procedure to promulgate rules prescribing perimeter fences for dogs and monkeys. 9 C.F.R. §§ 3.6(c)(2)(ii), 3.77(f). Why it proceeded differently for dangerous animals is unexplained. But we attach no weight to the Department's inconsistency, not only because it would be unwise to penalize the Department for having at least partially complied with the requirements of the Administrative Procedure Act, but also because there is nothing in the Act to forbid an agency to use the notice and comment procedure in cases in  which it is not required to do so. We are mindful that the court in United States v. Picciotto, 875 F.2d 345, 348 (D.C.Cir.1989), thought that the fact that an agency had used notice and comment rulemaking in a setting similar to the case before the court was evidence that the agency "intended" to promulgate a legislative rule in that case, only without bothering with notice and comment. The inference is strained, and in any event we think the agency's "intent," though a frequently cited factor, is rather a make-weight. What the agency intends is to promulgate a rule. It is for the courts to say whether it is the kind of rule that is valid only if promulgated after notice and comment. It is that kind of rule if, as in the present case, it cannot be derived by interpretation. The order under review, based as it was on a rule that is invalid because not promulgated in accordance with the required procedure, is therefore25