This suit under the Employee Retirement Income Security Act of 1974,. 29 U.S.C.[*1049] §§ 1001 et seq., challenges the denial by a company’s pension trust of a claim for a pension. The district court granted summary judgment for the trust but denied its request for an award of attorney’s fees. The parties have cross-appealed.

Albert Van Boxel went to work for the Journal Company (the publisher of the Milwaukee Journal) as a printer in 1951. In 1959 the company granted him a leave of absence, without pay, to enable him to take a full-time position as Secretary-Treasurer of the local of the International Typographical Union that represents the company’s printers. (He still holds the position, although he no longer works full time.) A series of six-month extensions of his leave of absence from the Journal Company came to an end in 1963. Although he did not return to work for the company upon the expiration of his last leave of absence, in 1975 the company and the union signed a collective bargaining agreement that required the company to prepare a list of employees who were guaranteed their jobs until they reached the age of 65 — and Van Boxel’s name appeared on the list and on successive lists prepared in subsequent years. In 1984, exercising his rights under the collective bargaining agreement, Van Boxel returned to work for the Journal Company for the first time since he had left in 1959 (25 years earlier) — though only for one day; the next day he returned to his union job. A year later Van Boxel, who by this time was 53 years old, applied to the Journal Company’s pension trust fund for a pension based on his having put in 20 years of service with the company. To get up to 20 years he had to count time during which he was working full time for the union. The trustees refused to let him do this and consequently rejected his claim for a pension. Pursuant to the collective bargaining agreement in force at this time, all the trustees had been appointed by the Journal Company.

The black-letter rule is that a decision by a pension trust to deny benefits to an individual claimant can be set aside in a suit under ERISA only if the decision is “arbitrary and capricious.” See, e.g., Pok-ratz v. Jones Dairy Farm, 771 F.2d 206, 208-09 (7th Cir.1985); Adcock v. Firestone Tire & Rubber Co., 822 F.2d 623, 626 (6th Cir.1987); Varhola v. Doe, 820 F.2d 809, 812-13 (6th Cir.1987); Holland v. Burlington Industries, Inc., 772 F.2d 1140, 1148-49 (4th Cir.1985), aff’d without opinion, — U.S. -, 106 S.Ct. 3267, 91 L.Ed.2d 559 (1986); Miles v. New York State Teamsters Conference Pension, Etc., Plan, 698 F.2d 593, 599 (2d Cir.1983); Dennard v. Richards Group, Inc., 681 F.2d 306, 313-14 (5th Cir.1982). In Allen v. United Mine Workers of America 1979 Benefit Plan & Trust, 726 F.2d 352, 354 (7th Cir.1984), we equated this term to “totally unreasonable,” and in Teskey v. M.P. Metal Products, Inc., 795 F.2d 30, 32 (7th Cir.1986), to “whimsical, random, or unreasoned.” For scholarly discussion and critique see Fischel & Langbein, ERISA’s Fundamental Contradiction: The Exclusive Benefit Rule, pt. V (Wkg. Paper 29, Program in Law and Economics, Univ.Chi.Law School, Oct. 1987); Note, Judicial Review of Fiduciary Claim Denials Under ERISA: An Alternative to the Arbitrary and Capricious Test, 71 Cornell L.Rev. 986 (1986); Comment, The Arbitrary and Capricious Standard Under ERISA: Its Origins and Application, 23 Duquesne L.Rev. 1033 (1985).

Van Boxel challenges the “arbitrary and capricious” standard. We are not entirely unsympathetic to the challenge, and notice that although the weight of authority is against him there is growing skepticism about the orthodox approach. The skeptical tendency, illustrated by dicta in Varho-la v. Doe, supra, 820 F.2d at 813, has culminated in the Third Circuit’s recent decision in Bruch v. Firestone Tire & Rubber Co., 828 F.2d 134, 137-45 (3d Cir.1987), which holds that, whenever someone who is not a plan beneficiary is in a position to benefit from the rejection of a claim — the company, for example — no deference should be given the trustees’ decision; the case should be treated as an ordinary contract dispute between the claimant and the trust. See id. at 145. Further muddying the waters, many cases that still adhere to the “arbitrary and capricious” standard[*1050] add “bad faith” to the grounds for upending the trustees’ denial of benefits. See, e.g., Allen v. United Mine Workers of America 1979 Benefit & Trust Plan, supra, 726 F.2d at 354; Sly v. P.R. Mallory & Co., 712 F.2d 1209, 1211 (7th Cir.1983). And some eases, such as Dockray v. Phelps Dodge Corp., 801 F.2d 1149, 1152 (9th Cir.1986) (quoting earlier cases), expand the normal standard even further, to “arbitrary, capricious or made in bad faith, not supported by substantial evidence, or erroneous on a question of law.” See also Brown v. Retirement Committee of Briggs & Stratton Retirement Plan, 797 F.2d 521, 525 (7th Cir.1986). And sometimes “arbitrary, capricious or made in bad faith” is defined as lacking substantial evidence or resting on an error of law, see, e.g., Wardle v. Central States, Southeast & Southwest Areas Pension Fund, 627 F.2d 820, 824 (7th Cir.1980); Dennard v. Richards Group, Inc., supra, 681 F.2d at 314, thus seeming to eliminate (but no doubt unintentionally) all judicial deference to the exercise of discretion by the trustees.

The “arbitrary and capricious” standard is familiar from administrative law, where it is used to guide judicial review of discretionary decisions by administrative agencies. See 5 U.S.C. § 706(2)(A) (Administrative Procedure Act). Pension fund trusts are not administrative agencies and most of the decisions they make are not discretionary in the sense, familiar from administrative law, of decisions that make policy under a broad grant of delegated powers. Certainly in a case such as the present one, pension fund trustees are not policy-makers; they are interpreters of contractual entitlements. The closest analogy might seem to be to arbitration. Like arbitrators, pension fund trustees asked to approve an application for a pension are private persons contractually authorized to decide the merits of a claim under a contract. If pension fund trustees can be equated to arbitrators, the arbitrary and capricious standard might be thought to confine their discretion too closely. Decisions by arbitrators can be set aside only for fraud or conflict of interest, or because the arbitrators failed to interpret the agreement they were appointed to interpret and instead embarked on a frolic of their own. See, e.g., Brotherhood of Locomotive Engineers v. Atchison, Topeka & Santa Fe Ry., 768 F.2d 914, 921 (7th Cir.1985). The excruciatingly narrow judicial review of arbitrators’ contractual interpretations lends an element of paradox to the Third Circuit’s statement in Bruch that “the validity of the claim is likely to turn on a question of law or of contract interpretation. Courts have no reason to defer to private parties to obtain answers to these kinds of questions.” 828 F.2d at 144 (emphasis added; footnote deleted).

But how close really is the analogy between arbitrators and pension fund trustees? The fact that often (and here) the trustees are appointed by the company (which is perfectly proper under ERISA, see Brown v. Retirement Committee of Briggs & Stratton Retirement Plan, supra, 797 F.2d at 535) rather than by both the company and the claimant, or by a neutral third party such as the American Arbitration Association, is only superficially inconsistent with the analogy to arbitration. This is not only because standards of neutrality are more relaxed in arbitration than in adjudication, see, e.g., Merit Ins. Co. v. Leatherby Ins. Co., 714 F.2d 673 (7th Cir.1983), and because, as also emphasized in Merit, the underlying relationship is contractual and hence consensual (dramatically so in this case, where the complainant is an official of the union that negotiated the collective bargaining agreement which authorized the company to appoint all the pension fund trustees); it is also because the company really may be neutral, despite appearances. Often, if the applicant’s claim for a pension is denied, it is not the company but the pension trust fund that is the richer for the denial; and the company has no access to that fund. However, in the case of defined-benefit pension plans (the vast majority of ERISA plans and the type involved in this case), the company has contractual obligations that it must honor whether or not the pension trust is adequately funded; hence, the wealthier the[*1051] fund is, the less likely is the company to be called on to ante up pension money out of its own pocket. Moreover, some employee-benefit plans subject to ERISA are unfunded, including one of those in Bruch.

Yet even in such a case, the impact on the company’s welfare of granting or denying an individual application for a pension will usually be too slight to compromise the impartiality of the trustees, even if all are appointed by the company. Nor is it in a company’s long-run best interest to alienate employees by dealing unfairly with pension claims; for the less an employee’s pension rights are worth, the higher are the wages that he will demand (assuming he is rational and well informed) so that he can buy supplementary retirement protection. But if the denial of an application is on grounds disqualifying a vast number of applicants, or if the company is not interested in the long run because it is about to go out of business, any presumption of neutrality may fail. Some courts do not apply the arbitrary and capricious standard at all in such cases, while others apply it in words only. See, e.g., In re White Farm Equipment Co., 788 F.2d 1186, 1191 (6th Cir.1986); Jung v. FMC Corp., 755 F.2d 708, 711-12 (9th Cir.1985); Struble v. New Jersey Brewery Employees’ Welfare Trust Fund, 732 F.2d 325, 333-35 (3d Cir.1984); Harm v. Bay Area Pipe Trades Pension Plan Trust Fund, 701 F.2d 1301, 1305 (9th Cir.1983) (citing cases).

From the foregoing analysis one could argue (we do not say how persuasively) for a dual approach to the review of decisions by pension fund trustees. (1) The very low standard applicable to arbitrators, rather than the “arbitrary and capricious” standard applicable to agency decisions under the Administrative Procedure Act, would be used when the circumstances were such as to make the company a neutral party in the decision by the trustees, or the trustees were neutral third parties rather than employees of the company, or half the trustees had been appointed by the union rather than the company. (2) Where none of these conditions obtained, and especially where the trustees had an actual conflict of interest, the “arbitrary and capricious” standard would be jettisoned in favor of de novo judicial review — the approach in Bruch. (These are also cases where the normal standard for review of arbitrators’ decisions would not be followed even in an arbitration case, though as we said the criteria for conflict of interest are looser in the setting of arbitration than in that of conventional adjudication, for reasons explained in Merit.)

All this could presumably be left to contract; and it is left to contract, in a sense: parties to a pension plan can always insert an arbitration clause, and maybe when they do so they automatically curtail the scope of judicial review, though this question was left open in Plumbers’ Pension Fund, Local 130 v. Domas Mechanical Contractors, Inc., 778 F.2d 1266, 1268 (7th Cir.1985), and we leave it open today. Given the uncertainties in the administration of the “arbitrary and capricious” standard in the pension setting, however, it is unclear what the parties’ expectations concerning judicial review are when they decide not to insert an arbitration clause or decide to allow the company to appoint all the trustees.

If we are right that the “arbitrary and capricious” standard, at least if thought to lay down a unitary standard of “total unreasonableness” (see Allen, supra), is too lax in some pension cases and too stringent in others, the question arises why it is the dominant standard in ERISA cases. History points us to the answer. Common law trustees (a term used loosely, since trusts are “equitable” rather than “legal” devices) are forbidden to engage in self-dealing; and with their impartiality thus assured it makes sense for a court to defer to their discretionary judgments. The deference is not so complete as in the case of decisions by arbitrators but that is mainly because courts of equity traditionally exercise broad supervision over trustees, see, e.g., Bogert & Bogert, Handbook of the Law of Trusts, ch. 18 (5th ed. 1973), a tradition having no counterpart in judicial review of arbitration — an alternative to adjudication against which both common law courts and equity courts long fought, see Kulukundis Shipping Co. v. Amtorg [*1052] Trading Corp., 126 F.2d 978, 982-85 (2d Cir.1942); Hayes v. Allstate Ins. Co., 722 F.2d 1332, 1339 (7th Cir.1983) (dissenting opinion).

The Taft-Hartley Act authorized employers to set up pension plans jointly with unions, see 29 U.S.C. § 186(c)(5), but made no provision for judicial review of the decisions of the plan administrators. The tradition of equitable supervision of trustees made this omission seem anomalous; so when a plan provision as interpreted had the effect of denying an application for benefits unreasonably, or, as it came to be said, arbitrarily and capriciously, courts would hold that the plan as “structured” was not for the sole and exclusive benefit of the employees, so that the denial of benefits violated the statute. See Comment, supra, 23 Duquesne L.Rev. at 1035-41. This standard was taken over for use in reviewing benefit denials under ERISA (which does not define the standard of judicial review of trustees’ decisions on benefit claims), apparently without the courts’ noticing that employers often held the whip hand in ERISA trusts as they did not with the joint employer-union trust funds authorized by Taft-Hartley. Or perhaps the courts believed (without saying) that the parties to ERISA plans had contracted in the expectation that trustees’ decisions would be reviewed under the arbitrary-and-capricious standard, no matter what. But, despite what we said earlier, this may be implausible; pension rights are too important these days for most employees to want to place them at the mercy of a biased tribunal subject only to a narrow form of “arbitrary and capricious” review, relying on the company’s interest in its reputation to prevent it from acting on its bias. Nor is it clear that the contractual perspective is the correct one in which to view claims under ERISA. A Congress committed to the principles of freedom of contract would not have enacted a statute that interferes with pension arrangements voluntarily agreed on by employers and employees. ERISA is paternalistic; and it seems incongruous therefore to deny disappointed pension claimants a meaningful degree of judicial review on the theory that they might be said to have implicitly waived it. Cf. Lee v. Dayton Power & Light Co., 604 F.Supp. 987, 1001 n. 11 (S.D.Ohio 1985).

Transposed to the ERISA setting, the arbitrary and capricious standard may be inapt, a historical mistake, or a mechanical extrapolation from different settings, at once too lax and too stringent, but even if it is any or all of these things it is saved from doing serious harm by its vagueness and elasticity. There are more verbal distinctions among the standards of judicial review than there are real differences. It is easier to multiply standards than actually to differentiate among them — to keep them from overlapping — in the setting of a particular case. Associated Industries of New York State, Inc. v. United States Dept. of Labor, 487 F.2d 342, 349-50 (2d Cir.1973) (Friendly, J.); 5 Davis, Administrative Law Treatise §§ 29:6-7 (2d ed. 1984). The fundamental difference in the depth or penetration or exactingness of judicial review is between deferential and nondeferential review, that is, between reversing a tribunal’s decision because it is unreasonable and reversing it merely because it is wrong. Sometimes even this difference blurs. When the members of the tribunal — for example, the trustees of a pension plan — have a serious conflict of interest, the proper deference to give their decisions may be slight, even zero; the decision if wrong may be unreasonable. The less likely it is that the trustees’ judgment was impaired by their having a stake, however indirect, in the outcome, the less inclined a reviewing court will be to override their judgment unless strongly convinced that they erred.

As this example shows, flexibility in the scope of judicial review need not require a proliferation of different standards of review; the arbitrary and capricious standard may be a range, not a point. There may be in effect a sliding scale of judicial review of trustees’ decisions (cf. Roland Machinery Co. v. Dresser Industries, Inc., 749 F.2d 380, 387-88 (7th Cir.1984) (standard for granting a preliminary injunction)) — more penetrating the greater is the suspicion of partiality, less penetrating the smaller that[*1053] suspicion is. Compare our discussion in Roland of the “abuse of discretion” standard of judicial review: “the words describe a range ... in which the amount of deference given the trial judge’s rulings varies with the nature of the ruling — a range that overlaps clear error, and even error, period.” Id. at 390 (citation omitted). See also Friendly, Indiscretion About Discretion, 31 Emory L.J. 747, 764, 784 (1982). The existence of a sliding scale in judicial review of ERISA trustees’ decisions is suggested by the cases that, while purporting to apply a uniform “arbitrary and capricious” standard, in fact give less deference to a decision the more the trustees’ impartiality can fairly be questioned. See, e.g., Holland v. Burlington Industries, Inc., supra, 772 F.2d at 1148-49; Dockray v. Phelps Dodge Corp., supra, 801 F.2d at 1152-53.

Flexibly interpreted, the arbitrary and capricious standard, though infelicitously— perhaps even misleadingly — worded, allows the reviewing court to make the necessary adjustments for possible bias in the trustees’ decision. So there is no urgent need to throw it overboard and cast about for an alternative verbalization. Where, as in Allen and Pokratz, the claimant does not argue or is unable to show that the trustees had a significant conflict of interest, we reverse the denial of benefits only if the denial is completely unreasonable. The greater the conflict of interest of a majority of the trustees, the less we defer to a denial of benefits that appears to be wrong.

The present case may involve a conflict of interest, though how serious a one we shall not have to determine. Van Boxel bases his claim for a pension primarily on the years in which he was working full time for the union that represents the Journal Company’s printers. Labor relations in this period were tense and adversarial, and Van Boxel describes himself with some plausibility as a thorn in the company’s side. Cf. Dockray v. Phelps Dodge Corp., supra, 801 F.2d at 1152-53 (claim by striker advanced in midst of bitter strike). And Van Boxel’s claim should not be viewed in isolation. If the company’s pension fund is obligated to award pensions to full-time union employees, this will make it easier for the union to recruit high-quality workers, whom the company would prefer to see working for it rather than for an adversary. More important, it may send a “signal” of company approval of the union that the company may not want to send. For these reasons the trustees of the pension fund, in deciding whether to grant Van Boxel’s claim for a pension, had a conflict of interest. This argues, if not for their disqualification (a course not suggested by Van Boxel), still for careful judicial scrutiny to make sure their action was reasonable.

Clearly it was. Even when we interpret the pension documents ourselves rather than review the trustees’ interpretation to decide whether it was reasonable (the latter being the usual approach in such cases, see Quinn v. Burlington Northern Inc. Pension Plan, 664 F.2d 675, 678 (8th Cir.1981), but not the proper one where the trustees’ bona fides have reasonably been drawn into question), we find that Van Boxel is not entitled to a pension. The collective bargaining agreement in force in 1963 when the last of Van Boxel’s leaves of absence expired provided that “every employee who is granted a leave of absence for a specified period shall be deemed to have ceased to be an employee at the expiration of such period if he does not then return to service.” Nothing could be clearer. Although from 1975 on Van Boxel was listed by the company as one of its employees — and moreover one with a guaranteed status, albeit he received no wage from the company — the purpose of the listing is apparent and has nothing to do with pensions. It is to protect employees of the union by guaranteeing them an opportunity to return to the company if their employment with the union terminates, as well it might for any one of a number of reasons, including the union’s ceasing to be collective bargaining representative or its officers’ being voted out of office by the membership.

From the way Van Boxel’s name climbed up the list in subsequent years until it reached the second position it appears that[*1054] the company was letting him earn seniority on the company’s employee rolls while . working full time for the union. But this is still a far cry from earning pension benefits. The company was not paying him a wage, and from the company’s standpoint a fringe benefit such as pension rights is a form of wage, that is, a labor expense. A company naturally is loath to pay a worker either directly or indirectly when he is working full time for another employer. It is true that the pension plan did not require employer contributions as such—it was, as we said, a defined-benefits plan—but with every passing year the employee earned pension credits that the company would have to make good on eventually; it was a real, if contingent, cost.

The clincher is the method by which the amount of the pension is determined. As is usually the case it depends on the wage that the pensioner earned in his last years of working for the company. Computation of the pension is straightforward for any employee actually working for the company, but Van Boxel was not. How then was the amount of his pension to be determined? By the wage he received back in the 1950s before he left to work for the union and ceased to receive any wages from the company? By the wage he received for his one day’s work for the company in 1984? By the wage he received from the union—a wage over which the company had no control? No doubt some equitable formula could be devised for determining an appropriate level of company pension for one in Van Boxel’s position, but the fact that the pension documents contain no such formula is telling evidence that the parties never contemplated a pension for persons in that position.

The pension trust argues that Van Boxel’s claim is so weak that he should be made to pay its attorney’s fees. In Bittner v. Sadoff & Rudoy Industries, 728 F.2d 820, 830 (7th Cir.1984), we interpreted the attorney’s fee provision of ERISA as authorizing an award of attorney’s fees to the defendant where the plaintiff’s claim not only fails but lacks “substantial justification,” even though it is not frivolous in the sense of utterly groundless. See also United Independent Flight Officers, Inc. v. United Airlines, Inc., 585 F.Supp. 1206 (N.D.Ill.1984); Waukesha Engine Division v. Department of Industry, Labor & Human Relations, 619 F.Supp. 1310, 1318 (W.D.Wis. 1985). Our analysis was rejected by the First Circuit in Gray v. New England Tel. & Tel. Co., 792 F.2d 251, 257-60 (1st Cir.1986), which decided to adhere to the conventional five-factor test, even though that test is oriented toward the case where a prevailing plaintiff rather than a prevailing defendant is asking for fees. See also Firestone Tire & Rubber Co. v. Neusser, 810 F.2d 550, 557-58 (6th Cir.1987). Speaking with great respect, we find nothing in the First Circuit’s opinion that persuades us to withdraw from Bittner, and we shall therefore adhere to it. See Chicago Painters & Decorators Pension, Etc., Funds v. Karr Bros., Inc., 755 F.2d 1285, 1292 (7th Cir.1985). We are not as enamored of mul-ti-factored balancing tests as some of our colleagues in other circuits.

The district court in this case concluded that the consistent listing of Van Boxel’s name on the company’s employee roll from 1975 on raised sufficient question concerning the correctness of the company’s claim that he had ceased to be an employee in 1963 to give Van Boxel’s claim “a sufficiently solid, though not correct, basis” to ward off an award of attorney’s fees under the Bittner standard. We are not disposed to disturb the district judge’s conclusion, especially given the uncertainty about the proper standard for judicial review of decisions by pension plan trustees under ERISA.

AFFIRMED.