Full opinion text
Opinion for the Court filed by Circuit Judge HARRY T. EDWARDS. HARRY T. EDWARDS, Circuit Judge: The appellants, a group of electrical and mechanical engineers employed by an electric utility company, seek attorneys’ fees from the National Labor Relations Board (“NLRB” or “Board”) for legal expenses incurred in the course of their ultimately successful quest for a separate representation election. The appellants rely on section 204(a) of the Equal Access to Justice Act (“EAJA” or “the Act”), 28 U.S.C. § 2412 (Supp. V 1981), which provides, inter alia, that, in a suit brought by or against the United States, the court shall award attorneys’ fees to a private prevailing party who satisfies certain financial eligibility requirements, unless “the position of the United States was substantially justified or ... special circumstances make an award unjust.” The District Court, 548 F.Supp. 256, found that the appellants met the eligibility requirements and had “prevailed” within the meaning of the Act. The court denied their application for attorneys’ fees, however, on the ground that the position taken by the Board was substantially justified. The appellants contest the latter ruling. Resolution of this appeal necessitates an inquiry into the theory and practice of the EAJA. We begin by construing two crucial, ambiguous phrases used in the Act— “the position of the United States” and “substantially justified.” Next, we consider the standard of review by which a court of appeals should scrutinize a trial judge’s determination of a party’s entitlement to fees. We then bring our analysis to bear on the District Court’s judgment in the case before us. We conclude that the court’s findings and rulings were proper and accordingly affirm. I. Background In 1938, the Utah Power and Light Company (“the Company”) voluntarily recognized Local 57 of the International Brotherhood of Electrical Workers (“the Union”) as the exclusive bargaining representative of its employees. The Company and Union agreed that the bargaining unit would consist of all employees except management personnel and supervisory officials with the authority to hire and fire. Included in the unit, consequently, were non-supervisory engineers. The NLRB assented to the parties’ “stipulation” and certified the Union. By 1978, there were 2,600 employees in the bargaining unit, approximately 100 of whom were highly educated electrical or mechanical engineers. During that year, 72 of the engineers filed a petition with the Board, seeking decertification of the Union as their bargaining representative. In support of their request, they argued that, as professional employees, section 9(b) of the National Labor Relations Act (as amended) guaranteed them a right to vote as a separate group regarding union representation. Regional Director Francis Spearandeo dismissed the engineers’ petition, on the ground that “[t]he Board’s established policy and general rule is that the unit appropriate in a decertification election must be coextensive with the unit previously certified or the unit recognized.” The Board denied the engineers’ petition for review of the dismissal. In 1979, the engineers amended their petition to request clarification of the scope of the bargaining unit. The Regional Director dismissed this petition as well, on the ground that the engineers constituted neither a labor organization nor an employer and therefore lacked standing under the Board’s regulations to petition for unit clarification. Having failed to secure relief from the NLRB, the engineers sought the aid of the judiciary. In September 1979, they brought suit in the District Court for the District of 'Columbia against both Spearandeo and the Board. They requested a declaration that they were indeed “professionals” within the meaning of section 2(12) of the National Labor Relations Act and an injunction compelling the Board to provide them a separate representation election. The Board admitted most of the factual allegations in the complaint, but responded that the engineers nevertheless had failed to state a claim on which relief could be granted and that the District Court lacked subject matter jurisdiction over the controversy. The parties then filed cross-motions for summary judgment. In 1980, while the case was pending in the District Court, two of the engineers filed a second decertification petition with the Board. This time the Board acceded to the petitioners’ request. Utah Power & Light Co., 258 N.L.R.B. 1059 (1981). The Board reasoned that the situation was “unique” insofar as “the professional employees seeking decertification have never had an opportunity to vote in a self-determination election”; under these special circumstances, it decided, “the policies inherent in Section 9(b)(1)” warranted making an “exception” to the general rule (which it reaffirmed) that a decertification election will not “normally” be ordered “in a unit not coextensive with the existing unit.” Id. at 1061 (footnote omitted). The long-sought-after election was held on October 29, 1981, and the engineers voted overwhelmingly against continued representation by the Union. Eight days later, the Board moved to dismiss as moot the engineers’ complaint. The engineers did not contest the motion, responding instead with an application for costs and attorneys’ fees under the EAJA. The Board filed a memorandum opposing the application. On May 28, 1982, the District Court granted the Board’s motion to dismiss. Finding that the engineers qualified as “prevailing parties,” the court granted their motion for costs. The court ruled, however, that the engineers were not entitled to attorneys’ fees under either of the two arguably relevant provisions of the EAJA: first, since the engineers had not shown that the Board had acted in “bad faith,” they were not entitled to an award of fees under section 2412(b); second, since the Board had sustained its burden of showing that its position was “substantially justified,” appellants were precluded from receiving an award of fees under section 2412(d)(1)(A). On appeal, no party contests the dismissal of the suit itself, the finding that the engineers constituted “prevailing parties,” or the allocation of costs. The engineers challenge the denial of attorneys’ fees. II. The Equal Access to Justice Act A. Introduction The background against which the Equal Access to Justice Act must be viewed is the so-called “American Rule” pertaining to the allocation of costs of counsel. Under that longstanding doctrine, each litigant ordinarily must pay his own lawyer. Alyeska Pipeline Service Co. v. Wilderness Society, 421 U.S. 240, 247, 95 S.Ct. 1612, 1616, 44 L.Ed.2d 141 (1975). The rule has only two narrow exceptions: when a loser “has acted in bad faith, vexatiously, wantonly, or for oppressive reasons,” he may be obliged to reimburse the winner for his attorneys’ fees, F.D. Rich Co. v. United States, 417 U.S. 116,129,94 S.Ct. 2157,2165,40 L.Ed.2d 703 (1974) (dicta); accord Lipsig v. National Student Marketing Corp., 663 F.2d 178, 180 (D.C.Cir.1980) (per curiam); and, when an individual litigant, by successfully maintaining a suit, has conferred a benefit on a group of persons, the court may allow him to recover his attorneys’ fees from the beneficiaries, Mills v. Electric Auto-Lite Co., 396 U.S. 375, 392-97, 90 S.Ct. 616, 625-28, 24 L.Ed.2d 593 (1970). Traditionally, the United States was even less vulnerable to an award of attorneys’ fees than a private litigant; in the absence of an express statutory provision, even the aforementioned “bad faith” and “common benefit” rules could not be invoked against the federal government. Pealo v. Farmers Home Administration of the United States Department of Agriculture, 562 F.2d 744,748 (D.C. Cir.1977) (alternative holding); Rhode Island Committee on Energy v. General Services Administration, 561 F.2d 397, 405 (1st Cir.1977). The protection from liability for attorneys’ fees enjoyed by the United States under these circumstances derived from two sources: the general doctrine of sovereign immunity; and 28 U.S.C. § 2412, which, prior to its amendment by the EAJA, was “consistently construed as immunizing the United States against attorney’s fees awards absent clear or express statutory authority to the contrary.” NAACP v. Civiletti, 609 F.2d 514, 516 (D.C. Cir.1979), cert. denied, 447 U.S. 922, 100 S.Ct. 3012, 65 L.Ed.2d 1114 (1980). Since at least the 1920s, the restrictive American doctrine pertaining to the allocation of attorneys’ fees has been subjected to persistent attack. In the 1960s and early 1970s, federal courts sensitive to these criticisms began to experiment with various alternative, more liberal, fee-shifting rules. The most important of these developments was the so-called “private attorney general” theory, under which prevailing parties were allowed to recover attorneys’ fees when their suits resulted in enforcement of “important societal rights.” In 1975, in the Alyeska case, the Supreme Court called a halt to this judicially managed doctrinal innovation. The Court concluded that, in view of the importance and complexity of the field and of the history of legislative involvement, “it is apparent that the circumstances under which attorneys’ fees are to be awarded and the range of discretion of the courts in making those awards are matters for Congress to determine.” 421 U.S. at 262, 95 S.Ct. at 1624 (footnote omitted). Consequently, it held that, in the absence of specific statutory authorization, federal courts sitting in non-diversity cases could award attorneys’ fees to prevailing parties only under the established common law exceptions to the American Rule. Id. at 247, 257-60, 271, 95 S.Ct. at 1616, 1621-23, 1628. The Supreme Court’s message was not lost on the legislature. A year after the Alyeska decision, Congress passed the Civil Rights Attorney’s Fees Awards Act, which allows prevailing parties in suits brought under specified civil rights statutes to recover “a reasonable attorney’s fee as part of the costs.” The Equal Access to Justice Act, passed four years later, constitutes an even more comprehensive response to the Supreme Court’s invitation and self-abnegation. The Act applies to all civil actions brought by or against the United States and allows private prevailing parties to recover attorneys’ fees from the government in a wide variety of circumstances. Two provisions of the statute are crucial in the present case. Section 2412(b) authorizes courts to award private prevailing parties “reasonable fees and expenses of attorneys,” except when such awards are “expressly prohibited by statute,” and declares that “[t]he United States shall be liable for such fees and expenses to the same extent that any other party would be liable under the common law or under the terms of any statute which specifically provides for such an award.” The most important effect of this provision is to codify and make applicable to the government the “bad faith” and “common benefit” exceptions to the American Rule. Section 2412(d)(1)(A) is more sweeping. It provides that a court “shall” award an eligible private prevailing party attorneys’ fees and other litigation expenses unless some other statute specifically provides otherwise or “the court finds that the position of the United States was substantially justified or that special circumstances make an award unjust.” Ordinary tort suits are excluded from the coverage of the provision, but the legislative history makes clear that constitutional tort actions come within its purview. Section 2412(d) is avowedly experimental. Unlike section 2412(b), it contains a “sunset” provision; unless reenacted before October 1, 1984, it will be automatically repealed. Because of the large bodies of case law defining the contours of the two traditional exceptions to the American Rule, interpretation and application of section 2412(b) usually is relatively straightforward. Section 2412(d)(1)(A), unfortunately, is much more difficult to apply. It is not grounded in any body of common law doctrine. Because of its recent vintage, it has not yet acquired a discernible gloss of judicial interpretation. And, most importantly, several of its crucial terms are distressingly ambiguous. For reasons that will become apparent, two of those terms — “the position of the United States ” and “substantially justified” — loom large in the case before us. Before proceeding, therefore, we must bring those phrases into sharper focus. And, of course, we must strive to lend them shapes that will make them meaningful vehicles, not only for disposing of the case at hand, but for resolving other controversies that come within the broad coverage of the provision. B. “The Position of the United States” The statute permits the government, when it loses a case, to avoid liability for attorneys’ fees if it can show that its “position” was substantially justified. The government’s “position,” for these purposes, might mean one of two things. First, the term could refer to the governmental action that precipitated the lawsuit. Second, it could refer to the posture assumed by the government in litigation. These two interpretations have come to be known, respectively, as the “underlying action ” and the “litigation position ” theories. The judiciary has yet to settle on one or another construction. As of the end of March of 1983, the issue had been addressed (explicitly or implicitly) in twenty-two decisions. In twelve of those cases, the court opted for the “underlying action” theory. In the other ten, the court opted for the “litigation position” theory. This circuit has not yet been confronted with the issue. The case law, in sum, affords us little aid in attempting to construe “the position of the United States.” We are obliged, therefore, to conduct our own inquiry. 1. Legislative Intent The statute itself contains no clear indication of which of the readings is correct. Section 2412(d)(1)(A) does not explicate the term. And section 2412(d)(2), the definitional provision, makes no reference to “the position of the United States.” Only two provisions of the Act offer any guidance, and they point in inconsistent directions. First, section 2412(d)(2)(C) indicates that the term “‘United States’ includes any agency and any official of the United States acting in his or her official capacity.” As the Third Circuit observed in a recent case, this encompassing definition provides some support for the “underlying action” theory, both because of its use of the word “acting” and because of the fact that most of the “officials” referred to by the provision would be incapable of taking a position in litigation. Natural Resources Defense Council v. U.S. Environmental Protection Agency, 703 F.2d 700, 707 (3d Cir.1983). However, the support that this definitional provision lends to the “underlying action” theory is at best modest. As Judge Hunter pointed out in his dissent in NRDC, the definitional section might plausibly be read as nothing more than an acknowledgment of the fact “that the government as an entity can only take a ‘position,’ whatever the meaning of that term, through its agencies and the individuals who administer its agencies.” Id. at 718 n. 2. Second, section 2412(d)(3) directs a court, when reviewing adversary adjudications before administrative agencies, to award to a private party who prevails on appeal the attorneys’ fees he incurred in the course of his litigation before the agency, unless the court finds “that during such adversary adjudication the position of the United States was substantially justified, or that special circumstances make an award unjust.” As the highlighted language indicates, the provision presumes that the government’s “position,” for the purpose of assessing its liability for attorneys’ fees resulting from administrative adjudication, is the stance it adopted in litigation before the agency. It remains possible, of course, that the phrase “position of the United States” was intended to have a different meaning when used in the context of litigation before the courts, but if Congress contemplated such a difference, it failed to indicate as much in the statute. In short, the unequivocal adoption of the “litigation position” theory in section 2412(d)(3) suggests that the identical phrase, when used without qualification in section 2412(d)(1)(A), should be interpreted as the posture adopted by the government before the court. On balance, we think that the support lent the “litigation position” theory by section 2412(d)(3) is somewhat stronger than that lent the “underlying action” theory by the definitional provision in section 2412(d)(2)(C). We concede, however, that the scales do not tip decisively one way or the other. One would hope that the legislative history would make possible a clean resolution of this issue. Unfortunately, study of relevant statements made by proponents and opponents of the bill proves even more inconclusive. Neither the committee reports nor the debates contain an explicit statement purporting to construe “the position of the United States.” The legislative history is, however, rife with arguments that take for granted one or the other option. Unfortunately, the number that seem founded on the “underlying action” theory is roughly the same as the number that seem founded on the “litigation position” theory. Those various references have been canvassed elsewhere; for present purposes, it will suffice to note the most important. Strong support for the “litigation position” theory is provided by the language used in both committee reports to describe the standard of liability under section 2412(d)(1)(A): “Where the Government can show that its case had a reasonable basis both in law and fact, no award [of attorneys’ fees] will be made.” H.R.Rep. No. 1418, supra note 20, at 10; S.Rep. No. 253, supra note 20, at 6 (emphasis added) U.S. Code Cong. & Admin.News 1980, at 4989. Other statements, also appearing in both reports, suggest (at least when read in isolation) that Congress was exclusively concerned with irresponsible governmental decisions to: initiate or continue litigation: A court should look closely at cases ... where there has been a judgment on the pleadings or where there is a directed verdict or where a prior suit on the same claim had been dismissed. Such cases clearly raise the possibility that the Government was unreasonable in pursuing the litigation. The standard, however, should not be read to raise a presumption that the Government position was not substantially justified, simply because it lost the case. Nor, in fact, does the standard require the Government to establish that its decision to litigate was based on a substantial probability of prevailing. H.R.Rep. No. 1418, supra note 20, at 11; S.Rep. No. 253, supra note 20, at 6-7 (emphasis added) U.S.Code Cong. & Admin. News 1980, at 4989. One can find equally evocative statements, however, supportive of the “underlying action” theory. For example, in defending the decision to place on the government the burden of demonstrating that its position was substantially justified, both committee reports argue that “it is far easier for the Government, which has control of the evidence, to prove the reasonableness of its action than it is for a private party to marshal the facts to prove that the Government was unreasonable.” H.R.Rep. No. 1418, supra note 20, at 11; S.Rep. No. 253, supra note 20, at 6 (emphasis added), U.S. Code Cong. & Admin.News 1980, at 4989. The conclusions to both reports declare: “Thus, by allowing an award of reasonable fees and expenses against the Government when its action is not substantially justified, S. 265 provides individuals an effective legal or administrative remedy where none now exists.” H.R.Rep. No. 1418, supra note 20, at 12; S.Rep. No. 253, supra note 20, at 7 (emphasis added), U.S.Code Cong. & Admin. News 1980, at 4991. And, in the floor debates, both supporters and critics of the bill frequently manifested their assumption that liability for fees would turn upon the strength of the government’s justification for its regulatory behavior, rather than the colorability of the legal arguments subsequently advanced in support of that conduct. We have thus arrived at an impasse; we must acknowledge that Congress failed clearly to resolve the question before us. To answer it, consequently, we must enlarge our field of vision; we must try to discern the underlying purposes of the Act as a whole, and then determine which of the two possible definitions of “the position of the United States” would best serve those ends. 2. A Practicable Definition The central objective of the EAJA, and of section 2412(d)(1)(A) in particular, was to encourage relatively impecunious private parties to challenge unreasonable or oppressive governmental behavior by relieving such parties of the fear of incurring large litigation expenses. Achievement of that end, it was believed, would promote three more general goals. First, Congress hoped to provide relief to the victims of abusive governmental conduct, to enable them to vindicate their rights without assuming enormous financial burdens. Second, it sought to reduce the incidence of such abuse; it anticipated that the prospect of paying sizeable awards of attorneys’ fees when they overstepped their authority and were challenged in court would induce administrators to behave more responsibly in the future. Third, by exposing, a greater number of governmental actions to adversarial testing, Congress hoped to refine the administration of federal law — to foster greater precision, efficiency and fairness in the interpretation of statutes and in the formulation and enforcement of governmental regulations. Single-minded pursuit of the foregoing goals would have induced Congress to enact a law providing for the automatic award of attorneys’ fees to private parties who prevailed in suits against the government. Congress did not go that far, however, because of its sensitivity to two other considerations. First, it did not wish to inhibit legitimate efforts by the executive to enforce the law. Second, it feared the potentially huge cost to the government of an automatic fee-shifting provision. Accordingly, it selected an intermediate rule. In the words of the committee reports: Under S. 265, fees will be awarded unless the Government can show that its action was substantially justified or that special circumstances make an award unjust. This standard balances the constitutional obligation of the executive branch to see that the laws are faithfully executed against the public interest in encouraging parties to vindicate their rights. H.R.Rep. No. 1418, supra note 20, at 10; S.Rep. No. 253, supra note 20, at 6, U.S. Code Cong. & Admin.News 1980, at 4989. Which of the two possible interpretations of “the position of the United States” would comport better with the foregoing complex of concerns? To answer that question, we first need to determine when and how it will matter which theory we adopt. Examination of the variety of kinds of controversies covered by the Act reveals that, in the large majority of contexts, it makes no functional difference how one conceives of the government’s “position.” In actions brought by the United States, the governmental action that precipitates the controversy almost invariably is its litigation position. Most suits brought against the United States entail a similar correspondence. In the usual case, the government acts in a particular fashion and then defends its conduct before an administrative agency and/or a court. Under such circumstances, the litigation position of the United States will almost always be that its underlying action was legally justifiable. Only in a minority of cases does it matter whether a court, reviewing a petition for fees brought under the EAJA, looks to the government’s original behavior or to its subsequent legal arguments. Assessment of the competing theories, therefore, must be made with reference to those unusual contexts. Relying on our experience, we believe it not unreasonable to assume that most such cases will take one of five forms: Example One: Appeals from Agency Actions Under Deferential Standards of Review The standards by which courts review most kinds of administrative action embody some principle of deference. An agency rulemaking decision, for instance, must be upheld if it contravenes no statutory or constitutional provision, was reached in compliance with pertinent procedural requirements, is supported by substantial evidence, and is not arbitrary, capricious, or an abuse of discretion. 5 U.S.C. § 706(2) (1976). A decision by an agency not to institute a rulemaking proceeding is accorded even more deference: if the agency adequately explains the facts and policy concerns it relied upon, and if those facts and concerns have some basis in the record, a court will not overturn its judgment. WWHT, Inc. v. FCC, 656 F.2d 807, 817 (D.C.Cir.1981). To take a more specific example especially germane to the instant case, a decision by the NLRB regarding the appropriate scope of a bargaining unit will not be reversed by a court unless it was “made in excess of [the Board’s] delegated powers and contrary to a specific prohibition in the [National Labor Relations] Act.” See Leedom v. Kyne, 358 U.S. 184, 188, 79 S.Ct. 180, 183, 3 L.Ed.2d 210 (1958); Boire v. Greyhound Corp., 376 U.S. 473, 480-82, 84 S.Ct. 894, 898-99, 11 L.Ed.2d 849 (1964). When an appellate court, applying one of these deferential standards of review, overturns an agency’s decision, how should the agency’s liability for attorneys’ fees under the EAJA be assessed? Assume, for example, that the promulgation of a novel regulation is held to have been “arbitrary and capricious.” If the ensuing EAJA petition submitted by the prevailing private party were evaluated under the “underlying action” theory, the agency’s liability for fees would turn upon whether its adoption of the rule was “substantially justified.” It is very unlikely that the agency would prevail under such a test; a court that has just concluded that the agency’s action was “arbitrary and capricious” would be hard pressed to rule that its action was nevertheless “substantially justified.” The net effect would be that the EAJA would become, for all practical purposes, an automatic fee-shifting provision in these circumstances. That result seems plainly inconsistent with Congress’ general objective in enacting section 2412(d)(1)(A): to establish an “intermediate” standard for the allocation of attorneys’ fees, one falling somewhere between the English Rule and the current American Rule. More consistent with the purposes of the Act would be a rule whereby a court, when reviewing an EAJA petition arising out of a case of the sort described above, would ask: Was the agency’s litigation position (i.e., its argument on appeal that its promulgation of the regulation was not arbitrary and capricious) substantially justified? Example Two: Standards of Liability Linked to the Justification for Conduct The government’s liability, in several kinds of suits, turns upon its ability to offer a reasonable justification for its own or its officials’ behavior. The category that springs most readily to mind are suits brought against the United States under the Eighth Amendment alleging that conditions in federal prisons constitute “cruel and unusual punishment.” Though the standard of liability governing such actions remains somewhat in flux, a crucial element in almost all modern formulations of the test is the absence of a defensible rationale for the conditions at issue. Thus, the Supreme Court recently summarized the governing law as follows: Today the Eighth Amendment prohibits punishments which, although not physically barbarous, “involve the unnecessary and wanton infliction of pain,” or are grossly disproportionate to the severity of the crime. Among “unnecessary and wanton” inflictions of pain are those that are “totally without penological justification.” Rhodes v. Chapman, 452 U.S. 337, 346, 101 S.Ct. 2392, 2398, 69 L.Ed.2d 59 (1981) (citations and footnote omitted). Clearly, a prerequisite to a finding of a constitutional violation, under the foregoing standard, is a determination that the prison conditions in question are not “substantially justified.” The close parallel between the standard of liability, in suits of this sort, and the criterion for the award of attorneys’ fees under the EAJA gives rise to a set of problems similar to those discussed in the preceding example. If, when evaluating an EAJA petition brought by a plaintiff (or class of plaintiffs) who prevailed in a case of the kind just described, the judge focused on the justification for the government’s “underlying action,” he or she would find it virtually impossible to deny the plaintiff(s) attorneys’ fees. The net result would be that the EAJA would again become something approaching an automatic fee-shifting provision for a significant category of cases. As indicated above, this seems not to have been contemplated by the congressmen who enacted the statute. Thus, in this context as well, it seems preferable for the court to attend to the strength of the arguments advanced by the government in support of its denial of liability, rather than to the justification for the behavior itself. Example Three: Defenses Unrelated to the Merits of the Government’s Behavior When civil actions are brought against the United States, it is sometimes able to resist on the basis of defenses unrelated to the merits of the underlying actions of its officials. For example, if, as a result of a change in the government’s own position or of unrelated events, the potential for impairment of a plaintiff’s interests has been reduced, the government may be able to argue that the case is moot. Or, if the plaintiff delayed bringing suit, the government may be able to rely on a statute of limitations or the doctrine of laches. Each of these defenses is founded on important policy considerations. For example, the mootness doctrine reduces the likelihood that courts will decide cases unaided by the kind of thorough adversarial scrutiny of the relevant issues that only a genuine controversy between the^ parties can create. Statutes of limitation and the laches doctrine promote repose and encourage aggrieved persons to bring suits while evidence is still fresh and reliable. The choice between the “underlying action” and “litigation position” theories will significantly affect the government’s willingness to assert viable defenses of this sort. An example might be the situation of a government attorney who must decide whether to contest a moderately stale suit against the United States. Because of the passage of time, he may be able to collect only scant evidence to dispute the plaintiff’s allegations, but he may nevertheless have a colorable (but not “sure-fire”) laches defense. If the governing interpretation of “the position of the United States,” within the meaning of the EAJA, were the government’s “underlying action,” the attorney might well decide not to contest the suit (i.e., not to assert the laches argument). He would realize that, if he did not prevail, the government would automatically be liable, not only for the value of the claim, but also for the plaintiff’s attorneys’ fees (because government counsel would be unable, at this late date, to show that the original behavior by the alleged perpetrators of the plaintiff’s harm was substantially justified). In sum, tying EAJA awards to the strength of the justification for the government’s original action would discourage counsel for the government from asserting defenses unrelated to its officials’ conduct that have a significant chance of prevailing. Such an outcome seems both unfortunate, from a policy standpoint, and inconsistent with Congress’ desire not to chill legitimate efforts by the executive to enforce the law. Thus, here again, it seems more sensible for courts, when reviewing EAJA petitions, to attend to the strength of the legal arguments advanced on behalf of the United States. Example Four: Supervening Change in the Law Frequently, during the time in which a suit brought against (or by) the United States is pending, the pertinent legal rules are altered in a manner adverse to the government. The effect of such a change often is that, while the government’s action was “substantially justified” when originally undertaken, persistence in defending that action in court is unjustifiable. In such a situation, if counsel for the United States stubbornly continues litigating the case, compelling the private party to incur attorneys’ fees before eventually prevailing, it seems clear that the objectives underlying the EAJA would be served by shifting those costs to the government. Application of the “underlying action” theory would not have that effect; application of the “litigation position” theory would. Example Five: Surrender by the Government Assume that a government contractor believes that the United States has failed to abide by the terms of an agreement (e.g., has misapplied a provision designed to adjust contract prices to accord with inflation). The contractor hires a lawyer who (without consulting the government contracting officer) conducts an investigation and then drafts and files a civil complaint. The government’s attorney, upon receiving the complaint, conducts his own inquiry and concludes that the claim is meritorious. (E.g., he finds that, though the government’s error was inadvertent, it nevertheless constituted a clear violation of the terms of the contract.) The government attorney informs the responsible official of his findings, who pays the plaintiff the money to which he is legally entitled. The plaintiff then brings a second action, under the EAJA, to recover his attorneys’ fees. If the petition is evaluated under the “underlying action” theory, the contractor will prevail. (Even though the government never took a litigation position hostile to the plaintiff, the plaintiff will qualify as a “prevailing party,” and the conduct that gave rise to the suit will fail the “substantial justification” test.) If it is assessed on the basis of the “litigation position” theory, the contractor will lose. It is not immediately obvious which of the foregoing results is preferable. In favor of the first option, it could be argued that the prospect of being compelled to pay attorneys’ fees in such situations would make government administrators more careful and (perhaps more importantly) would reduce the incidence of less innocent “mistakes.” In favor of the second option, it could be argued that routine payment of attorneys’ fees under these circumstances would only foster collusion between lawyers and putative victims of governmental errors and would result in unnecessary expenditures of public funds. Most such disputes can be resolved quickly and easily through informal discussion between the aggrieved party and the responsible government official. Awarding attorneys’ fees to private parties who, instead of availing themselves of such avenues of relief, file civil complaints would simply result in unnecessary — and unnecessarily costly — formalization of the processes whereby the competing interests of state and citizen are clarified and accommodated. Though the issue is not clear-cut, we think that, on balance, the latter considerations are the more compelling, in view of the purposes of the Act as a whole. Accordingly, the “litigation position” theory seems preferable in this context as well. To summarize, in each of the five contexts we can imagine in which the definition of “the position of the United States” would make a difference, it seems more sensible and consistent with the purposes of the EAJA to interpret the phrase as the stance taken by the United States in litigation than to interpret it as the governmental behavior that precipitated the suit. One potential objection remains to be considered: it might be argued that our analysis has been flawed by our reliance on a false dichotomy. Why not award fees if either the government’s litigation position or its underlying action was not substantially justified? Review of the scenarios described above, however, reveals the emptiness of this argument. Of the five situations we examined, only one — the supervening change in the law — could be handled sensibly through application of the suggested alternative definition of “position of the United States.” Finally, we note that exclusive use of the “litigation position” theory would have one important incidental advantage: it would permit and encourage courts to award to prevailing private parties the fees they incurred in combatting unreasonable arguments advanced by the government, while denying fees incurred in defeating substantial arguments. Such a practice, if it became widespread, would induce government counsel to evaluate carefully each of the various claims they might make in a particular controversy, and to assert only those that are substantially justified. The net result would be more sensitive and effectual promotion of the objectives of the EAJA. For the foregoing reasons, we hold that “the position of the United States,” for the purposes of the Act, means the arguments relied upon by the government in litigation. C. “Substantially Justified ” The second of the two phrases whose meaning is crucial to the case at bar is “substantially justified.” Only by showing that “the position of the United States” (as we have now defined it) satisfied this standard can the government avoid liability for attorneys’ fees. Fortunately, the legislative history of the Act casts considerable light on the meaning of this phrase. For example, the committee reports definitively resolve the potentially troublesome question of how burdens of proof are to be allocated: once the private party has shown that he “prevailed” in the litigation, the government bears the burden of demonstrating that its position was “substantially justified.” This principle thus far has been consistently recognized by the courts. It is somewhat more difficult to determine what exactly Congress intended the government should be required to prove, but much insight can be gleaned from the following passages of legislative history: The test of whether or not a Government action is substantially justified is essentially one of reasonableness. Where the Government can show that its case had a reasonable basis both in law and fact, no award will be made.... Certain types of case dispositions may indicate that the Government action was not substantially justified. A court should look closely at cases, for example, where there has been a judgment on the pleadings or where there is a directed verdict or where a prior suit on the same claim had been dismissed. Such cases clearly raise the possibility that the Government was unreasonable in pursuing the litigation. The standard, however, should not be read to raise a presumption that the Government position was not substantially justified, simply because it lost the case. Nor, in fact, does the standard require the Government to establish that its decision to litigate was based on a substantial probability of prevailing. H.R.Rep. No. 1418, supra note 20, at 10-11; S.Rep. No. 253, supra note 20, at 6-7, U.S. Code Cong. & Admin.News 1980, at 4989. Analysis of one other aspect of the legislative history enables us to refine the standards suggested above. The Senate Judiciary Committee considered and rejected an amendment to the bill that would have changed the pertinent language from “substantially justified” to “reasonably justified.” S.Rep. No. 253, supra note 20, at 8. That refusal suggests that the test should, in fact, be slightly more stringent than “one of reasonableness.” The foregoing generalizations regarding the requisite strength of the government’s case should enable courts to dispose of most EAJA petitions. Borderline cases, however, can be fairly resolved only through the application of more particularized criteria. For guidance in developing such guidelines, we must refer once again to the objectives underlying the Act as a whole. Such an inquiry makes possible the identification and elaboration of several useful principles. To begin with, it seems clear that the purposes of the Act would not be promoted by treating the question whether the position taken by the United States in a particular case was “substantially justified” as equivalent to the question whether it was “reasonable” for government counsel to pursue the litigation. To be more specific, it would be improper for a court, when evaluating an EAJA petition, to take into account either of two circumstances that might make the decision of the government’s attorney to pursue a very weak case appear “justifiable.” First, the government’s lawyer might have been bound by bureaucratic constraints. For example, his office might have exclusive authority to represent the United States in a particular kind of suit, and departmental policy might require defense, on appeal, of all favorable decisions rendered in adversary adjudication. Such constraints, though they may excuse the conduct of individual attorneys, are precisely what the EAJA was designed to change. Awards of attorneys’ fees in such circumstances would have a salutary effect: agencies that at present adhere to such policies likely would alter their rules in ways designed to promote more careful evaluation of the merits of individual cases. Second, the importance of a legal issue may justify a decision by government counsel to “take a long shot” — for example, to argue on appeal for the overruling of a controlling precedent unfavorable to the United States, even though the likelihood of obtaining such a judgment is slight. Several considerations suggest that, when the government loses such a case, it should be obliged to reimburse the private party for his attorneys’ fees. To begin with, in controversies of this sort, it is especially important that the private litigant not be deterred, by the prospect of high litigation expenses, from defending his interests; only if they are subjected to vigorous adversarial testing will such important issues receive the attention they deserve. That concern is reinforced by consideration of the causes and implications of the rarity with which such cases arise. Suits of this kind are often “test cases”; the government has decided that, though hitherto it routinely abided by the dictates of the controlling precedent, the time has come to venture a challenge to the rule. The result is that the private party, through no fault of his own, is forced to expend an unusually large amount of time and effort defending his claims. The value of periodic reexamination of legal norms may make us willing to tolerate the resultant inequity, but it is difficult to justify forcing the private party, who has been randomly selected by the government to be the target of its test case, to pay the large cost of litigating the issue if he ultimately prevails. In short, this is precisely the sort of situation in which the public at large, not the individual litigant, should pay the cost of “refining the administration of the law.” Finally, there is no good reason to suppose that fee-shifting under these circumstances will deter the government from bringing such test cases. If the issue is important enough, government officials, who of course are not personally liable for the payment of fees, shojild not be dissuaded by the prospect of an award of fees to a private party’s counsel. And in those rare instances when government officials are dissuaded from challenging established law, we cannot assume that the result would be inconsistent with Congress’ purposes in enacting the EAJA. What factors, then, are appropriately considered by a court deciding a hard case? Reference to the complex of statutory objectives described above suggests the following three criteria. This list is not intended to be exhaustive, but attention to these matters seems likely to advance Congress’ stated ends. (1) Clarity of the Governing Law The reasons for the relevance of this first factor are implicit in the foregoing discussion. For the purposes of the EAJA, the more clearly established are the governing norms, and the more clearly they dictate a result in favor of the private litigant, the less “justified” it is for the government to pursue or persist in litigation. The government is always free, of course, to seek modification or repudiation of established doctrine, but individual private litigants should not be compelled to subsidize such reevaluations of controlling doctrine. In sum, a court assessing an EAJA petition should attend to the strength of the government’s arguments) that extant law permits a result in its favor, not to the strength (or probability of success) of its arguments) that the law should be changed. (2) Foreseeable Length and Complexity of the Litigation The longer and more complex the course of litigation necessary to vindicate his position, the more hesitant a private party will be to defend his interests, even if an eventual finding that the government’s position was not substantially justified would result in reimbursement of his costs of counsel. The rational litigant will realize that, the more protracted the proceeding, the more he stands to lose if either he does not prevail on the merits or his petition for fees is denied, while the money he stands to gain (or the point of principle he hopes to make) remains constant. Sensitivity to the central objective of the Act — reduction of the deterrents to challenges of unreasonable government conduct — thus suggests that, in categories of cases in which substantial investments of effort and money commonly are required to prosecute suits to their ultimate conclusions, the government should be obliged to make an especially strong showing that its persistence in litigation was justified. (3) Consistency of the Government’s Position A problem frequently cited by congressional proponents of the EAJA was the capacity of the government with impunity to “single out” particular private parties. Too often, they believed, agencies and officials adopt a general policy in dealing with cases of a given variety but take a different position (inexplicably or maliciously) in one or a few cases. One of the systemic sources of inequity advocates of the Act sought to eliminate was the ability of the government to use its superior resources to beat into submission the hapless victims of such deviations from customary practices. The relevance of inconsistency of this kind to the EAJA calculus becomes even more apparent when one recalls the two “public policy” objectives underlying the Act; awarding attorneys’ fees to private parties who successfully challenge governmental departures from established policy would both foster the “refinement” of the administration of the law and provide a valuable deterrent to officials contemplating making gratuitous exceptions to general rules. In sum, a court reviewing an EAJA petition should demand that the government make a particularly strong showing of justification for its position when there is evidence that the petitioner was subjected to atypically harsh treatment. We acknowledge that our holding that the court should take into account whether the government, in its treatment of the plaintiff, departed from established policy stands in some tension with our holding that “the position of the United States” means the government’s litigation position, not its underlying action. Our willingness to tolerate that tension derives from our belief that both holdings are required in order most effectually to promote the underlying objectives of the Act. To clarify our position: when the government acts inconsistently, and subsequently loses a civil suit challenging its behavior, it should be obliged to make an especially strong showing that its legal arguments were substantially justified in order to avoid liability for fees under the EAJA. D. The Standard of Appellate Review Our task in this case is not to determine de novo the appellants’ entitlement to attorneys’ fees; it is to review the District Court’s dqnial of their petition. Disposition of the appeal thus requires that we define the standard by which trial judges’ rulings on EAJA applications should be evaluated by courts of appeals. The case law affords us little aid in deciding this issue; to date, no circuit has seriously addressed the question of the proper standard of review. And, unfortunately, the likely repositories of evidence of legislative intent prove empty. The statute itself contains no hint as to the proper standard, and the legislative history is barren of reflection on the issue. Some insight might be gained from a study of the standards appellate courts have developed for reviewing trial judges’ rulings on other issues that frequently arise in attorneys’ fees cases. Unfortunately for our purposes, those standards vary considerably. Examples include: the deferential “abuse of discretion” standard by which district court determinations regarding the appropriate amount of attorneys’ fees are examined; the much closer scrutiny generally accorded lower court findings concerning whether litigants qualify as “prevailing parties”; and the distinctive criteria by which trial judges’ decisions that “special circumstances” warrant denying otherwise appropriate fee awards are evaluated by courts of appeals. In sum, the diversity of the arguably relevant analogous standards of review, combined with the unusual character of the issue before us, disables us from gaining significant guidance from the case law. To determine the appropriate standard of review, consequently, we must advert to a combination of the general principles governing appellate review of district court judgments in civil nonjury cases and the objectives underlying the EAJA. A few aspects of the issue may be settled quickly. First, any findings of fact made by a trial judge in the course of his ruling on a petition for attorneys’ fees may be overturned only if “clearly erroneous.” Fed.R.Civ.P. 52(a). The trial judge’s conclusions on questions of law, by contrast, are not protected by Rule 52(a) and are subject to de novo review. See Pullman-Standard v. Swint, 456 U.S. 273, 287, 102 S.Ct. 1781, 1789, 72 L.Ed.2d 66 (1982). Moreover, any determinations made by the trial judge that derive from application of legal standards to the facts of the case should be subjected to similar scrutiny; if the appellate court concludes that the judge’s view of the law is incorrect, it may overturn the conclusions founded thereon. United States v. Singer Manufacturing Co., 374 U.S. 174,194 n. 9, 83 S.Ct. 1773,1784 n. 9, 10 L.Ed.2d 823 (1963). Unfortunately, not all judgments made by district courts in the course of decisions regarding EAJA petitions fall into the foregoing familiar categories. Frequently, such judgments assume one of two unusual forms. First, they often consist of determinations, not whether a particular interpretation of the law is correct, but whether it is plausible or colorable. Judgments of this sort are somewhat anomalous in the appellate scheme. We have little difficulty, however, in concluding that they should be regarded as findings of law and reviewed as such. It would appear that, if anything, the special expertise and experience of appellate courts in assessing the relative force of competing interpretations and applications of legal norms makes the case for de novo review of judgments of this order even stronger than the case for such review of paradigmatic conclusions of law. Any uncertainty we might have regarding the appropriate classification of rulings of this kind is alleviated by consideration of the central purpose of the EAJA. The disincentive to challenge unreasonable governmental action caused by fear of incurring large litigation expenses will be significantly reduced only if potential litigants are able to predict reasonably accurately whether they ultimately will be able to recover their attorneys’ fees. Their ability to make such predictions, in turn, is dependent on the consistency with which the EAJA is interpreted — i.e., with which the justifications for (losing) legal arguments are assessed. Nondeferential appellate review of such assessments will best achieve a goal of consistency and thus enable the Act more effectually to serve its intended social functions. The second atypical type of finding that commonly appears in EAJA decisions involves evaluations of the probative force of evidence submitted by the government. Frequently, the governing legal principles in a civil suit brought by or against the United States are undisputed, and the controversy revolves around competing characterizations of the underlying facts. When the government loses such a case and the prevailing party requests attorneys’ fees, the trial judge must assess the plausibility of the government’s original depiction of the situation that gave rise to the suit. Determinations of this order are difficult to classify. Insofar as they entail judgments concerning whether the government’s submissions would have been sufficient to satisfy the pertinent legal standard had they not been overborne by the evidence presented by the private litigant, they bear some resemblance to rulings on so-called “mixed questions of law and fact,” whose vulnerability to review by appellate courts is extraordinarily confused. On balance, however, we think such judgments are more properly classified as factual findings. They involve the same kind of evaluation of the credibility of witnesses and weighing of evidence entailed by ordinary findings of fact — and at which trial judges are especially experienced and skilled. Moreover, considerations of judicial economy counsel strongly against subjecting such judgments to close scrutiny. Trial judges’ explanations in EAJA cases of their assessments of the probative value of the government’s submissions frequently are sketchy. If they knew that their determinations were to be examined de novo, judges would no longer confine themselves to such vignettes; to assist the reviewers (or to protect their judgments against reversal), they would feel obliged to describe and analyze the government’s evidence in much more detail. The net result would be substantial waste of time and judicial resources. In sum, we conclude that, to the extent that trial judges’ rulings regarding the strength of positions taken in litigation by the United States are based upon assessments of the probative value of the evidence offered by the government, they should be reversed only if “clearly erroneous.” An analogy may be helpful to clarify our holding regarding the appropriate standard of review. The problem in judicial administration that most closely resembles the one before us is appellate review of grants or denials of preliminary injunctive relief. Because such determinations most often are made prior to full-scale assessment of the merits of the case, the factual findings are frequently sketchy. See United States Steel Corp. v. Fraternal Association of Steelhaulers, 431 F.2d 1046, 1048 (3d Cir. 1970). And the rulings on questions of law incorporate a speculative element analogous to that embodied in EAJA determinations; the trial judge’s findings are confined to assessment of “the validity, or at least the probable validity, of the legal premise underlying the claim of right in jeopardy of impairment.” Delaware & Hudson Railway Co. v. United Transportation Union, 450 F.2d 603, 620 (D.C.Cir.) (emphasis added), cert. denied, 403 U.S. 911, 91 S.Ct. 2209, 29 L.Ed.2d 689 (1971). Courts of appeals, sensitive to the unusual aspects of judgments regarding preliminary relief, have developed a standard for reviewing them that would work equally well in the present context. That standard is frequently described as one permitting reversal only for an “abuse of discretion.” See, e.g., Prendergast v. New York Telephone Co., 262 U.S. 43, 50-51, 43 S.Ct. 466, 469, 67 L.Ed. 853 (1923); U.S. Steel v. Steelhaulers, 431 F.2d at 1048; A Quaker Action Group v. Hickel, 421 F.2d 1111,1115 (D.C.Cir.1969). But that ubiquitous phrase has a special meaning in this area. On one hand, it has been interpreted to require highly deferential review of district courts’ tentative findings of fact; such determinations are not overturned unless “clearly erroneous” and, in practice, are rarely subjected to critical examination. See Engine Specialties, Inc. v. Bombardier Limited, 454 F.2d 527, 530 (1st Cir.1972); U.S. Steel v. Steelhaulers, 431 F.2d at 1048; United States v. Ingersoll-Rand Co., 320 F.2d 509, 523 (3d Cir.1963). On the other hand, it allows for close scrutiny of district courts’ rulings on questions of law. See California ex rel. Younger v. Tahoe Regional Planning Agency, 516 F.2d 215, 217 (9th Cir.), cert. denied, 423 U.S. 868, 96 S.Ct. 131, 46 L.Ed.2d 97 (1975); Northeast Construction Co. v. Bird Associates, 485 F.2d 752, 756 (D.C.Cir.1973); Delaware & Hudson Railway Co. v. United Transportation Union, 450 F.2d at 620-21. The same composite standard, we hold, should govern appellate review of district court dispositions of EAJA petitions. III. The Merits of the Appellants’ Petition Now at last we can bring our analysis to bear on the case at bar. The appellants’ principal contention on appeal is that the District Court erred in denying them attorneys’ fees under section 2412(d)(1)(A) on the ground that the position taken by the NLRB was “substantially justified.” In addition, they challenge the District Court’s judgment that the Board did not act in “bad faith” in pursuing the litigation, and therefore was not liable for fees under section 2412(b). We consider those contentions in order below. A. Section 2412(d)(1)(A) For the reasons indicated in Part II.B. supra, “the position of the United States,” whose justification must be assessed, is the position taken by the Board in litigation before the District Court. That position was stated succinctly in the penultimate paragraph of the memorandum submitted by the Board in support of its motion for summary judgment: In summary, since the Board failed to contravene any express statutory command in denying plaintiffs’ decertification petition or unit clarification petition, this Court should decline to exert jurisdiction to review the Board’s actions under Leedom v. Kyne. To assess the strength of this argument, we must retrace the analytical path the Board itself traveled in its memorandum. As the Board recognized, the touchstone of this controversy is the doctrine enunciated by the Supreme Court in Leedom v. Kyne, 358 U.S. 184, 79 S.Ct. 180, 3 L.Ed. 210 (1958). Ordinarily, determinations by the NLRB in representation proceedings are not subject to direct judicial review; only if such determinations form the basis of final orders issued in subsequent unfair labor practice proceedings can they be challenged in the courts. Boire v. Greyhound Corp., 376 U.S. 473, 476-77, 84 S.Ct. 894, 896, 11 L.Ed.2d 849 (1964); May Department Stores v. NLRB, 326 U.S. 376, 379, 66 S.Ct. 203, 206, 90 L.Ed. 145 (1945); AFL v. NLRB, 308 U.S. 401, 409, 60 S.Ct. 300, 304, 84 L.Ed. 347 (1940). In Leedom v. Kyne, the Court carved out a very narrow exception to this general principle: when the Board acts in a manner that contravenes an express statutory prohibition and the aggrieved party has no alternative way of securing relief, a district court has the authorit