This appeal concerns the standard courts should use when reviewing a denial of a request for benefits under an ERISA plan by an insurance company which, pursuant to a contract with an employing company, both determines eligibility for benefits, and pays those benefits out of its own funds. This question, and variations thereof, have bedeviled the federal courts since considered dicta in Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101 (1989), gave opaque direction about how courts should review discretionary benefits denials by potentially conflicted ERISA fiduciaries. In Firestone, the Court instructed that the "arbitrary and capricious" standard was appropriate but that a conflict of interest should be considered as a "factor" in applying this standard.

Courts of appeals have taken different approaches to integrating these seemingly incongruous directions when reviewing decisions of insurance companies that fund a plan and are also ERISA plan administrators. Following the lead of five other such courts, we hold that, when an insurance company both funds and administers benefits, it is generally acting under a conflict that warrants a heightened form of the arbitrary and capricious standard of review. In reaching this conclusion, we are cognizant of the previous cases in which we have been highly deferential to decisions of an employer who funds and administers a benefit plan, a practice grounded in the belief that the structural incentives to deny meritorious claims are generally outweighed by the opposing incentives to grant them--such as the "incentives to avoid the loss of morale and higher wage demands that could result from denials of benefits." Nazay v. Miller, 949 F.2d 1323, 1335 (3d Cir. 1991). However, we conclude that these incentives (assuming their existence) do not apply with the same force to an insurance company that pays benefits out of its own coffers. The relationship with the welfare of the beneficiaries is more attenuated, and there are problems of imperfect information. In the insurance company-as-funder-and-administrator context, the fund from which monies are paid is the same fund from which the insurance company reaps its profits. This is in contrast to the actuarially determined benefit funds typically maintained by employers (especially in the pension area) that usually cannot be recouped by the employer or directly redound to its benefit. Our rule is also informed by the understanding that "smoking gun" direct evidence of purposeful bias is rare in these cases so that, without more searching review, benefits decisions will be virtually immunized.

The courts of appeals that have forged the trail in this area have presented different formulations of the heightened standard. Some courts, led by the Eleventh Circuit, have established a standard approaching de novo review, shifting the burden to the defendant company to explain its decisions. However, we side with the majority of courts of appeals, which apply a sliding scale method, intensifying the degree of scrutiny to match the degree of the conflict.

In this case, applying a heightened degree of scrutiny because of the financial conflict, we conclude that there is a genuine issue of material fact as to whether the defendant, Reliance Standard Life Insurance Company, acted arbitrarily and capriciously when it concluded that the plaintiff, Maria Pinto, an employee of Reliance Standard's client Rhone-Poulenc Corporation, was not totally disabled by her cardiac condition and therefore did not deserve long-term disability benefits. Our heightened review allows us to take notice of discrete factors suggesting that a conflict may have influenced the administrator's decision. First, Reliance Standard's reversal of its initial decision to grant benefits was itself questionable. Second, its final report credited the evidence favorable to denial while inadequately explaining why it rejected the contrary evidence--the same evidence on the basis of which it had initially determined to award benefits. Third, while Reliance Standard relies on the fact that two physicians found Pinto not to be totally disabled while two others disagreed, one of the doctors on whom Reliance Standard relied was not a cardiologist but a pulmonologist, and he found Pinto's condition satisfactory only from his (pulmonary) vantage point, whereas the disability dispute is over a condition that is cardiological in nature.

In light of the evidence in the record, we conclude that a factfinder could find that Reliance Standard's actions were arbitrary and capricious. Therefore, we will reverse the grant of summary judgment and remand to the District Court for further proceedings consistent with this opinion.

Pinto was an accounting clerk for Rhone-Poulenc from 1986 to 1991. In July 1991, she stopped working because of a heart condition, which was diagnosed as mitral stenosis and cardiac asthma. After receiving short-term benefits from Rhone-Poulenc, she applied, in June 1992, for long-term disability (LTD) benefits from Reliance Standard, which had contracted to administer and pay LTD benefits under Rhone-Poulenc's ERISA plan. The policy provides benefits for individuals who submit "satisfactory proof " of "Total Disability" to Reliance Standard. In pertinent part, an employee is "Totally Disabled" when, "after a Monthly Benefit has been paid for 24 months, an Insured cannot perform the material duties of any occupation." It is undisputed that Reliance Standard had discretion to interpret the plan.

When Pinto applied for LTD benefits, Dr. Alan Bahler, her treating physician since 1977, sent Reliance Standard a diagnosis of her condition, which was confirmed by a cardiac catheterization. He reported that she had mitral stenosis secondary to rheumatic heart disease, which brings on shortness of breath, and orthopnea (the inability to breathe well without sitting erect) with borderline congestive heart failure. Bahler further attested that Pinto had developed symptoms of mitral valvular dysfunction, including symptoms of cardiac asthma and early congestive heart failure, worsening exercise tolerance, and palpitations. He concluded that "[h]er present condition precludes her from actively working even at a clerical level . . . [h]er only viable option at the present time is continued medical therapy, sedentary life style, and avoidance of high stress situations that could precipitate her cardiac asthma." In December 1992 and April and August 1993, Bahler recertified Pinto's total disability. In the 1993 certification, Bahler indicated that Pinto could not stand for long, could not lift ten pound objects, could not be exposed to stress, and should remain sedentary.

In October 1992, Reliance Standard sent Pinto a letter granting her application for long term benefits. It advised her that periodic medical certification would be required, and requested that she promptly apply for social security disability benefits. In December 1992, Pinto certified, in connection with a disability review by Reliance Standard, that she had not worked in any capacity, and that she remained under treatment. She noted that she had been hospitalized for two days in November of that year, when she had been treated for bronchial asthma and acute bronchitis. In April 1993, Pinto recertified that she was disabled and represented that she had recently been treated by two physicians.

Pinto also applied for Social Security Disability benefits. In May 1993, the Social Security Administration (SSA) denied Pinto's application, finding her not disabled. She forwarded a copy of the determination letter to Reliance Standard. Reliance Standard strongly encouraged Pinto to appeal the adverse decision, which she did. In September 1993, SSA denied Pinto's appeal, concluding that her asthma attacks could be controlled by medication, that her rheumatic heart disease was stable, and that her shortness of breath did not preclude work. One month later, Reliance Standard requested that Dr. Bahler relay to it the specific limitations that prevented Pinto from being an accounting clerk. Dr. Bahler responded by referring to his previous reports.

In November 1993, Reliance Standard terminated Pinto's benefits. Its denial letter cited the SSA denial, and the language tracked that of the denial.1 It also asserted: "Your physician has stated that you can perform the duties of a sedentary occupation within your present physical limitations and restrictions." Reliance Standard appeared to read Dr. Bahler's assertion that Pinto needed to maintain a sedentary lifestyle as a statement that she could perform sedentary work, and the described limitations to define absolutely the limits of her potential (i.e., it apparently read his statement that she could not lift ten pound items to imply that she could regularly lift less weighty items).

Pinto requested a review of this decision. Dr. Bahler wrote Reliance Standard in January 1994, explaining Pinto's medical history and affirming his determination that Pinto's "only viable option at the present time is continued medical therapy, sedentary life style, and avoidance of high stress situations that could precipitate her cardiac asthma. . . . Pinto is totally and permanently disabled at this time and therefore is unfit to perform any task or job in the labor market." Then, in February 1994, the SSA reversed its earlier denial and awarded her benefits. It determined that she had a severe cardiac condition and that she was too disabled to perform any job for which she had the requisite skills.

In the early summer of 1994, Reliance Standard retained Dr. Martin I. Rosenthal, an internist, who examined Pinto and reviewed Dr. Bahler's echocardiogram and cardiac catheterization studies. He recommended pulmonary testing and, after some initial ambivalence, concluded that Pinto was not totally disabled. In November (following Rosenthal's suggestion), Dr. Robert A. Capone, a pulmonologist, examined Pinto, and initially declined to decide whether she had a disabling reactive airways disease because he thought therapeutic intervention might make a difference. However, when pressed by Reliance Standard in December (no therapeutic interaction had occurred between November and December) he indicated that he did not think that any respiratory condition prohibited her from working, and, while noting the inadequacy of the available data, concluded that he did "not believe that there is a strong likelihood of reactive airways disease."

After Dr. Rosenthal's examination but before Dr. Capone's, a Reliance Standard staff worker, in an internal document, recommended reestablishing Pinto's benefits pending the pulmonary testing. However, Reliance Standard decided to do the opposite, holding the resumption of benefits until the pulmonary testing. It is noteworthy that the same staff worker had similarly recommended a resumption of benefits in April because she thought Reliance Standard had misunderstood Dr. Bahler's assertion that Pinto must be sedentary to mean sedentary work instead of sedentary lifestyle. In February 1995, Reliance Standard rejected Pinto's appeal of its earlier benefits reversal. It wrote her that "Dr. Bahler, Dr. Rosenthal, and Dr. Capone have all indicated you retain the physical functional capacity to engage in sedentary work. The subsequent correlation of this activity level with the material duties of your occupation substantiated that you are capable of performing the material duties of your regular occupation."

In January 1996, Pinto was examined by Dr. Rowland D. Goodman, II, a heart and chest specialist who shares offices with Dr. Bahler. Dr. Goodman reviewed her medical records, examined her, and concluded that she suffered from rheumatic heart disease with mitral stenosis and that she was totally disabled. Goodman's report was unavailable to Reliance Standard when making the initial decision, but was used when making the decision in question here (after the remand discussed infra).

In July 1996, Pinto filed the present ERISA suit in the District Court under 29 U.S.C. S 1132(a)(1)(b), which allows for a beneficiary to sue for "benefits due to him under the terms of the plan." In January 1997, Reliance Standard moved for summary judgment on the grounds that the decision was discretionary, and not arbitrary and capricious. The District Court agreed and granted the motion. Pinto appealed. In an unpublished opinion (hence non-precedential under our Internal Operating Procedures S 5.3), we vacated the judgment and remanded, see Pinto v. Reliance Std. Life Ins. Co., 156 F.3d 1225 (Table) (3d Cir. May 28, 1998) (No. 97-5297), concluding that Reliance Standard had apparently misinterpreted Dr. Bahler's diagnosis when it stated that "all" of the physicians who had examined her determined that Pinto had the "functional capacity to engage in sedentary work." Given the disconnect between this interpretation of Dr. Bahler's diagnosis and his own repeated conclusion that Pinto was unfit for any work, we stated that

we cannot confidently rule that Reliance's decision was not arbitrary and capricious. We are unsure whether Reliance properly reviewed Dr. Bahler's reports or whether it misinterpreted his conclusions. Moreover, we do not know whether it would have made the same decision based solely on Dr. Rosenthal and Dr. Capone's evaluations. Therefore, these are matters that will require reconsideration by Reliance.

We also briefly discussed the problem of the standard of review for situations where an insurer administers benefits out of its own funds:

We are not convinced that such a dual role presents the type of conflict of interest that would warrant discarding the arbitrary and capricious standard, but in any event under Firestone such a conflict would merely be a factor in the court's determination whether there has been an abuse of discretion. . . . We . . .review Reliance's determination under an arbitrary and capricious standard, taking into account the circumstances.

Reliance Standard dutifully reconsidered, and affirmed its earlier denial. In August, an Assistant Manager of Quality Review, Richard D. Walsh, issued a letter explaining the rejection. The letter opined that, although Dr. Bahler had stated that Pinto should not work, the limitations that he put on her activity would not preclude her from working. Walsh cited the United States Department of Labor's The Revised Handbook for Analyzing Jobs as evidence that the job of accounting clerk is "sedentary." He also cited the conclusions of Drs. Rosenthal and Capone. As regards Dr. Goodman's examination, Walsh stated that he had "provided no new findings, restrictions, or limitations to substantiate his conclusion," and commented on the fact that he shares a mailing address with Dr. Bahler, implicitly suggesting that Goodman's conclusions might be biased by his association with Dr. Bahler. Walsh did not mention Pinto's successful appeal of the Social Security denial. Although there is no record evidence that Reliance Standard knew of Social Security's reversal, it must have known of it at least after the case was remanded as it is mentioned in the previous panel's opinion.

On remand, the District Court again granted summary judgment for Reliance Standard. Although it purported to apply the arbitrary and capricious standard "shaped by the circumstances of the inherent conflict of interest," it proceeded to explain that "an administrator's decision will only be overturned if it is without reason, unsupported by substantial evidence or erroneous as a matter of law." Pinto v. Reliance Std. Life Ins. Co., No 96-3508 (D.N.J. Dec. 12, 1998). The court concluded that there was not an issue of material fact as to whether Reliance Standard had acted arbitrarily and capriciously. Of the rejection of Dr. Bahler's conclusions in favor of those of its own doctors, the court stated that "[s]uch a determination based on independent medical evaluations is not arbitrary and capricious, even when Reliance Standard's dual role as both insurer and decisionmaker is taken into account." Id. This appeal followed. We have jurisdiction pursuant to 28 U.S.C. S 1291, and our standard of review is plenary.

Our analysis of the issue in this case must begin with Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101 (1989). Prior to Firestone, courts had adopted different approaches to the conflict of interest problem under ERISA, many choosing to vary the degree of deference they gave ERISA benefits administrators operating under a conflict of interest. See Brown v. Blue Cross & Blue Shield of Ala., 898 F.2d 1556, 1560 (11th Cir. 1990) (collecting cases). However, as one court of appeals has stated, "the Supreme Court [in Firestone] . . . swept the standard of review board clear." De Nobel v. Vitro Corp., 885 F.2d 1180, 1185 (4th Cir. 1989).

Firestone began when a group of plaintiffs sued their employer, who was also the ERISA plan administrator, for wrongfully terminating welfare and pension benefits. A panel of this court considered the relevant principles of trust law, with special attention to the rationales for the general deference given to impartial trustees, concluding that those reasons carry little or no force when trustees are in a position to profit from denying trust benefits. See Bruch v. Firestone Tire & Rubber Co., 828 F.2d 134, 145 (3d Cir. 1987). We also considered the incentives and actual relationship of the parties, and the fact that the benefit plan was contracted for and its terms subject to negotiation. Id. We concluded that trust and contract principles both dictated that our review of the conflicted benefits denial should be de novo, giving no deference to either the administrator's or participants' interpretations. We essentially applied "the principles governing construction of contracts between parties bargaining at arms length." Id.

The Supreme Court affirmed the specific holding in that case--that the administrator's decision should be reviewed de novo, giving no deference to either party--but used a significantly different rationale. The Court began by stating that interpretation of ERISA should be governed by the common law of trusts, and then grounded the de novo review on the fact that the plan gave the administrator no discretion to interpret the plan. See Firestone , 489 U.S. at 111. The Court observed that trust principles dictated that fiduciaries should be given no deference when making nondiscretionary decisions, see 489 U.S. at 111; see also supra note 2. It then turned to a brief discussion pertinent to this case, noting that "a deferential standard of review [is] appropriate when a trustee exercises discretionary powers,"2 but that "if a benefit plan gives discretion to an administrator or fiduciary who is operating under a conflict of interest, that conflict must be weighed as a "factor in determining whether there is an abuse of discretion." Id. at 115 (quoting Restatement (Second) of TrustsS 187, cmt. d (1959)) (emphasis added). Since Firestone, courts have struggled to give effect to this delphic statement, and to determine both what constitutes a conflict of interest and how a conflict should affect the scrutiny of an administrator's decision to deny benefits. The next two Sections discuss these problems as applied to an independent insurer who is empowered with discretion to determine who deserves benefits under a plan which it funds.

B. What Constitutes A Conflict?

Employers typically structure the relationship of ERISA plan administration, interpretation, and funding in one of three ways. First, the employer may fund a plan and pay an independent third party to interpret the plan and make plan benefits determinations. Second, the employer may establish a plan, ensure its liquidity, and create an internal benefits committee vested with the discretion to interpret the plan's terms and administer benefits. Third, the employer may pay an independent insurance company to fund, interpret, and administer a plan. While we have previously held that the first two arrangements do not, in themselves, typically constitute the kind of conflict of interest mentioned in Firestone, see infra Section E, today we address the third arrangement for the first time, concluding that it generally presents a conflict and thus invites a heightened standard of review.3 Our sister circuits that have examined this issue have fallen into two basic camps. Most hold that the nature of the relationship between the funds, the decision, and the beneficiary invites self-dealing and therefore requires closer scrutiny, but others allow heightened review only if there is independent evidence that the conflict infected a particular benefits denial.

C. Courts of Appeals Holding that the Independent Insurance Company Administrator is Operating under an Inherent Conflict

The Eleventh Circuit was the first to conclude that an insurance company acts under a "strong conflict of interest" when both administering and paying out benefits under an ERISA plan. Brown v. Blue Cross & Blue Shield of Ala., 898 F.2d 1556, 1561 (11th Cir. 1990). It held that there is

an inherent conflict between the roles assumed by an insurance company that administers claims under a policy it issued. . . . Because an insurance company pays out to beneficiaries from its own assets rather than the assets of a trust, its fiduciary role lies in perpetual conflict with its profit-making role as a business. Id. at 1561 (internal quotations omitted). The Brown court noted that a structural conflict of interest may unconsciously encourage even a principled fiduciary to make decisions that are not solely in the interest of the beneficiary. See id. at 1565. Under this view, although the arrangement is not illegal or inappropriate under ERISA, it warrants heightened scrutiny. "[J]udicial hesitation to inquire into the fiduciary's motives will leave the beneficiaries unprotected unless the existence of a substantial conflicting interest shifts the burden to the fiduciary to demonstrate that its decision is not infected with self interest." Id.

In Doe v. Group Hospitalization & Med. Servs. , 3 F.3d 80 (4th Cir. 1993), the Fourth Circuit, like the Eleventh, concluded that a "conflict flows inherently from the nature of the relationship" when an employer contracts with an insurance company to provide and determine ERISA benefits. Id. at 86.

Undoubtedly, [Blue Cross's] profit from the insurance contract depends on whether the claims allowed exceed the assumed risks. To the extent that Blue Cross has discretion to avoid paying claims, it thereby promotes the potential for its own profit. . . . Even the most careful and sensitive fiduciary in those circumstances may unconsciously favor its profit interest over the interests of the plan, leaving beneficiaries less protected than when the trustee acts without self interest and solely for the benefit of the plan.

Id. at 86-87. See also Bedrick v. Travelers Ins. Co., 93 F.3d 149, 154 (4th Cir. 1996) (citing Doe). The Fifth Circuit, in a recent en banc discussion, also affirmed a commitment to heightened scrutiny of decisions by an insurer who administers benefits from its own funds. In Vega v. Nat'l Life Ins. Serv., Inc., 188 F.3d 287 (5th Cir. 1999), the plan administrator insurance company was a subsidiary of the plan insurer (the court treated the interests as aligned), and while the court recognized that if the company denied meritorious claims, its "reputation may suffer as a result and others may be less willing to enter into contracts where the company has discretion to decide claims," id. at 295 n.8, it concluded that these incentives did not outweigh the strong incentive to self-deal, see id. at 295-98. The Tenth Circuit and Eighth Circuits have also concluded that when an insurance company acts as the administrator for benefits coming from its own funds, the conflict warrants a more searching review. See Armstrong v. Aetna Life Ins. Co., 128 F.3d 1263 (8th Cir. 1997); Pitman v. Blue Cross & Blue Shield of Okla., 24 F.3d 118, 120-22 (10th Cir. 1994) (citing Doe).

D. Courts of Appeals Holding that the Independent Insurance Company Structural Relationship Does not Give Rise to a Conflict That Should Affect Standard of Review

The Seventh Circuit requires a specific demonstration that bias affected a decision before modifying the arbitrary and capricious standard when reviewing the decisions of an insurance company in this posture. See Mers v. Marriott Internat'l Group Accidental Death and Dismemberment Plan, 144 F.3d 1014 (7th Cir. 1998). In Mers, the benefit plan was insured by the American International Group (AIG), an independent insurer, that also was charged with interpreting the plan. The Mers court considered, and rejected, Mers's argument that less deference should be given to AIG's decision because it was operating under a conflict of interest. "We presume," it held, "that a fiduciary is acting neutrally unless a claimant shows by providing specific evidence of actual bias that there is a significant conflict." Id. at 1020. Relying on what it styled as law and economics principles, the court concluded that the requested payout in that case was slight compared to the company's bottom line, and that it is in a company's best long-term interest to award meritorious claims so that employees and employers will think and speak well of it, and seek business with it. Id. at 1021. Neutrality, opined the panel, begets business success, while self-dealing hurts it. Id. Therefore, a claimant bears the burden of providing specific evidence of a "significant conflict" (without suggesting what that would entail), or specific evidence of bias. Id. at 1020.

The Second Circuit, like the Seventh, requires evidence that a conflict actually infected the decision before it uses anything but the most deferential review of a fiduciary's determination. See Whitney v. Empire Blue Cross & Blue Shield, 106 F.3d 475 (2d Cir. 1997); Sullivan v. LTV Aerospace & Defense Co., 82 F.3d 1251, 1255-56 (2d Cir. 1996); Pagan v. NYNEX Pension Plan, 52 F.3d 438, 440-44 (2d Cir. 1995). It reasons not from effect but language, concluding that Firestone simply does not require anything but arbitrary and capricious review unless the plaintiff demonstrates how a conflict biased a fiduciary's decision. See Pagan, 52 F.3d at 440-44. However, it is noteworthy that a recent panel of the Second Circuit, in an opinion by Judge Oakes joined by Judges Newman and Winter, has expressed dissatisfaction with Pagan and Whitney. While recognizing that it was bound by precedent, it stated that "[w]e have numerous concerns regarding Pagan, which we believe reduces Firestone's ruling as to the impact of a conflict of interest." DeFelice v. American Int'l Life Assur. Co. of New York, 112 F.3d 61, 66 n. 3 (2d Cir. 1997).

E. Courts of Appeals in Which the Law is Unclear

The Ninth and Sixth Circuits appear to be unsettled on this issue. The Ninth Circuit sometimes requires something more than the fact that an insurance company administers benefits out of its own funds to trigger heightened review. In Atwood v. Newmont Gold, 45 F.3d 1317, 1322-23 (9th Cir. 1995), the court explained that the traditional abuse of discretion standard applies even in conflicted situations unless there is specific evidence that the conflict infected the process. In Snow v. Standard Ins. Co., 87 F.3d 327, 331 (9th Cir. 1996), the court followed Atwood when an insurance company both funded and administered an ERISA plan, declining to apply heightened review because there was no evidence that the "formal conflict led to a true conflict." 87 F.3d at 331. See also Lang v. Long-term Disability Plan of Sponsor Applied Remote Tech., 125 F.3d 794 (9th Cir. 1997) (only applying heightened review because there were independent indications that the conflict biased the decisionmaking). On the other hand, in Tremain v. Bell Indus., Inc., 196 F.3d 970, 976 (9th Cir. 1999), the court stated that "less deferential" arbitrary and capricious review was in order when the plan administrator was also the insurer. Cf. Kearney v. Standard Ins. Co., 175 F.3d 1084, 1090 n.2 (9th Cir. 1999) (Because we conclude that Kearney is entitled to de novo review, which gives no deference at all to Standard's decision, we do not reach the question whether he would be entitled to less deferential review were he entitled only to review for abuse of discretion.").

The Sixth Circuit's precedent is also unclear. In Miller v. Metropolitan Life Ins., 925 F.2d 979, 984-85 (6th Cir. 1991), the court took the insurance company's conflict of interest into account in the court's review of the insurance company's decision as an administrator, therefore applying a heightened arbitrary and capricious standard. On the other hand, in Yeager v. Reliance Standard, 88 F.3d 376, 381-82 (6th Cir. 1996), the court did not consider the conflicted role of the insurance company when applying the arbitrary and capricious standard.

We have not previously addressed the precise issue involved in this case. There is, however, some cognate discussion of the standard of review in cases where an employer both funded and administered a plan. In the first such case, Nazay v. Miller, 949 F.2d 1323 (3d Cir. 1991), we applied the unmodified arbitrary and capricious standard in reviewing a denial of benefits. While implicitly recognizing that there might be a risk of opportunism, we concluded that this alone did not constitute evidence of a conflict of interest, in part because the employer "had incentives to avoid the loss of morale and higher wage demands that could result from denials of benefits." Id. at 1335. We also commented on the fact that the denial was individual, instead of class-based, implying that when more money was at stake--i.e., when a large class of beneficiaries requested and was denied benefits--the potential conflict might invite closer scrutiny. See id.

In the same year, we decided Kotrosits v. GATX Corp. Non-contributory Pension Plan for Salaried Employees , 970 F.2d 1165, 1173 (3d Cir. 1992), in which a benefits committee within a company administered the company's funded plan. We concluded that the unmodified arbitrary and capricious review was appropriate in the absence of specific, tangible evidence that the structural relationship had tainted the review process. Though we held that a plaintiff urging that we disregard the grant of discretion in a plan "has the burden of showing some reason to believe the exercise of discretion has been tainted," we stated that this burden could be met "where such a party shows the kind of conflict of interest that could realistically be expected to bias the decision makers." Id. at 1173. If such a conflict were present, we suggested, "[Firestone] counsels in favor of withholding deference." Id.4 By way of explaining why we presumed that the fiduciary was not influenced by self-interest in Kotrosits, we compared the assets of the plan (which were substantial) to the potential costs of paying out to the beneficiaries (which were also substantial, but less so), and concluded that it was unlikely that the company would have to replenish the plan. See id. "[T]he record shows no direct impact on the Plan sponsor and only a possibility of future indirect consequences to it." Id.

In Abnathya v. Hoffman-LaRoche, Inc., 2 F.3d 40 (3d Cir. 1993), we recognized that "some degree of conflict inevitably exists where an employer acts as the administrator of its own employee benefits plan," but held that the conflict in that case was insufficiently compelling to "require special attention or a more stringent standard of review under [Firestone]." Id. at 45 n.5. We noted that the company's contributions to the fund were fixed such that it"incurs no direct expense as a result of the allowance of benefits, nor does it benefit directly from the denial or discontinuation of benefits." Id. See also Mitchell v. Eastman Kodak, 113 F.3d 433, 437 n.4 (3d Cir. 1997) (following the reasoning and language of Abnathya).

While Heasley v. Belden & Blake Corp., 2 F.3d 1249 (3d Cir. 1993) is not directly on point, it demonstrates an openness to using heightened scrutiny when a party administers benefits out of its own funds. In Heasley, we reviewed de novo whether an ERISA plan term "experimental procedure" applied to a liver transplant. We recognized that the apparent ambiguity of this term might be cured if the plan provided for a party to interpret it, but noted that under such a scheme, the interpretation should be allocated to an independent party, given the threat of self-dealing. See id. at 1260-61 n.12. This, we said, followed the "general and sensible rule that courts scrutinize more closely decisions by plan administrators acting under a conflict of interest." Id.

The final opinion that bears mention is our earlier unpublished (and therefore non-precedential) opinion in this very case, Pinto v. Reliance Std. Life Ins. Co., 156 F.3d 1225 (Table) (3d Cir. May 28, 1998) (No. 97-5297). In that opinion we stated as follows:

We are not convinced that such a dual role presents the type of conflict of interest that would warrant discarding the arbitrary and capricious standard, but in any event under Firestone such a conflict would merely be a factor in the court's determination whether there has been an abuse of discretion.

An issue similar to that before us here was considered by our sister circuits in Brown v. Blue Cross and Blue Shield of Alabama, Inc., 898 F.2d 1556 (11th Cir. 1990), and Miller v. Metropolitan Life Insurance Corp., 925 F.2d 979 (6th Cir. 1991). The Miller court, following Brown, held that in such a circumstance although the insurance company's "fiduciary role lies in perpetual conflict with its profit making role as a business, and the conflict of interest is substantial . . . the abuse of discretion or arbitrary and capricious standard still applies, but application of the standard should be shaped by the circumstances of the inherent conflict of interest." Miller, 925 F.2d at 984. We too will apply this standard and review Reliance's determination under an arbitrary and capricious standard, taking into account the circumstances.

Under the arbitrary and capricious standard, an administrator's decision will only be overturned if it is "without reason, unsupported by substantial evidence or erroneous as a matter of law. . . . the court is not free to substitute its own judgment for that of the defendants in determining eligibility for plan benefits." It is important to recognize that ERISA does not make the judges the decisionmakers. It merely assures that the appropriate procedure is followed.

The opinion is somewhat delphic. But the citations to Miller, see supra Section II.E, infra Section IV, (and also to Brown, see supra Section II.C, infra Section IV), both of which consider an insurer making decisions out of its own funds to be operating under an inherent conflict (in contrast to citations of cases of a contrary stripe), suggest that a heightened degree of scrutiny is required in this situation, an approach essentially the same as that we adopt in this opinion, but which we refine and clarify. See Part IV, adopting the "sliding scale" approach endorsed by a majority of our sister circuits.5

III. Is Heightened Review Required When an Insurance Company Both Funds and Administers Benefits?

Informed by our canvass of the jurisprudence, we are persuaded that heightened scrutiny is required when an insurance company is both plan administrator and funder. We find especially persuasive the analysis of the Fourth, Fifth, Eighth, Tenth and Eleventh Circuits, and their conclusion that potential self-dealing warrants that fiduciary insurer's decisions be closely inspected. We do not denigrate the Seventh Circuit's suggestion that if a carrier denied clearly meritorious claims on a regular basis and became notoriously unfriendly to claimants, unions and employees might protest and demand that their employer switch to a different insurance plan. Nor do we think that most insurance companies are unmoved by the importance of building a strong reputation and competing successfully for the business of administering plans. An insurance company "can hardly sell policies if it is too severe in administering them." Doe v. Travelers Ins. Co., 167 F.3d 53, 57 (1st Cir. 1999). However, ERISA litigation generally arises only in close cases, and there would seem to be insufficient incentive for the carrier to treat borderline cases (unlikely to become causes celebres) with the level of attentiveness and solicitude that Congress imagined when it created ERISA "fiduciaries." Rather, insurance carriers have an active incentive to deny close claims in order to keep costs down and keep themselves competitive so that companies will choose to use them as their insurers, an economic consideration overlooked by the Seventh Circuit.

To amplify, while in a perfect world, employees might pressure their companies to switch from self-dealing insurers, there are likely to be problems of imperfect information and information flow. Employees typically do not have access to information about claim-denying by insurance companies, and the relationship between employees and insurance companies is quite attenuated; so long as obviously meritorious claims are well-handled, it is unlikely that an insurance company's business will suffer because of its client's employees' dissatisfaction. Additionally, many claims for benefits are made after individuals have left active employment and are seeking pension or disability benefits. Details about the handling of those claims, whether responsible or irresponsible, are unlikely to seep into the collective knowledge of the stillactive employees. If Pinto's claim is denied, few at RhonePoulenc will learn of it, and Reliance Standard will have little motive to heed the economic advice of the Seventh Circuit that "it is a poor business decision to resist paying meritorious claims for benefits." Mers, 144 F.3d at 1020.

We also observe that the typical employer-funded pension plan is set up to be actuarially grounded, with the company making fixed contributions to the pension fund, and a provision requiring that the money paid into the fund may be used only for maintaining the fund and paying out pensions. As we explained in Abnathya and Mitchell, the employer in such a circumstance "incurs no direct expense as a result of the allowance of benefits, nor does it benefit directly from the denial or discontinuation of benefits." Abnathya, 2 F.3d at 45 n.5; Mitchell, 113. F.3d at 437 n.4. In contrast, although there is nothing in the record indicating the precise nature of Reliance Standard's internal structure, the typical insurance company is structured such that its profits are directly affected by the claims it pays out and those it denies.6

We recognize that the preceding section involves implicit assumptions about economic behavior, but such assumptions have become necessary in the post-Firestone era as we, and other courts, must somehow determine when a conflict warrants close scrutiny. Inasmuch as we are making such assumptions, however, they seem less exceptional than those of the Seventh Circuit, which, we believe, has an overly optimistic view of the flow of information and the sophistication of employees. Furthermore, while all circuits that have considered these questions appear to agree that some level of conflict may be unavoidable and not every conflict will heighten the level of scrutiny, the Seventh and Second Circuits alone require evidence of actual self dealing, and hold that the nature of the relationship itself can never, or almost never, affect the standard of review. Needless to say, Firestone contains no such requirement, and its use of the word "conflict" instead of "direct evidence of bias" counsels against the most stern reading. As we opined in Kotrosits, the Firestone court appears, by recognizing the import of a conflict, to have "implicitly adopted the position . . . that, where the claimant demonstrates that it would be inequitable to defer to the plan administrator, stricter scrutiny of his decision is in order." 970 F.2d at 1172. Finally, this is not a scenario where a "smoking gun" is likely to surface, and direct evidence of a conflict is rarely likely to appear in any plan administrator's decision. Our reading, we believe, infuses at least some meaning into the Firestone regime.

Finally, the unique role of insurance companies within ERISA supports our position. ERISA generally requires that assets of a benefits plan be "held in trust by one or more trustees." 29 U.S.C. S 1103(a). However, this requirement is excepted for insurance companies; the requirements that the assets be held in a trust does not apply "to assets of a plan which consist of insurance contracts or policies issued by an insurance company qualified to do business in a State." 29 U.S.C. S 1103(b)(1). Therefore,"[i]nasmuch as `the basis for the deferential standard for review in the first place was the trust nature of most ERISA plans,' the most important reason for deferential review is lacking." Brown v. Blue Cross and Blue Shield of Ala., 898 F.2d 1556, 1561 (11th Cir. 1990) (quoting Moon v. American Home Assurance Co., 888 F.2d 86, 89 (11th Cir. 1989)).

Our own case law in the general area, set forth in Section II.E, supports our conclusion. These opinions are selfconsciously laden with negative pregnants, suggesting that structural bias could heighten the review. For example, we noted that the defendants in those cases did not "incur" a "direct expense as a result of the allowance of benefits," or "benefit directly from the denial or discontinuation of benefits," Abnathya, 2 F.3d at 45 n.5; Mitchell, 113 F.3d at 437 n.4, implying that a company that did profit directly would be subject to a more stringent standard. Likewise, the most deferential review was appropriate when the employer had "incentives to avoid the loss of morale and higher wage demands that could result from denials of benefits," Nazay, 949 F.2d at 1335, and there was only a "possibility of future indirect consequences to it," Kotrostis, 970 F.2d at 1173. By negative implication, a heightened standard of review would appear to be appropriate when a plan funder like an insurance company "incurs a direct expense," the consequences to it are direct and contemporary, and, while it has incentives to maintain good business relationships, it lacks the incentive to "avoid the loss of morale and higher wage demands that result from a denial of benefits." We are also supported by the fact that the great bulk of district courts in this Circuit have interpreted this precedent as mandating heightened scrutiny when the insurance company is the insurer and makes determinations.7

For all the foregoing reasons, we believe that a higher standard of review is required when reviewing benefits denials of insurance companies paying ERISA benefits out of their own funds.

IV. What Standard of Review?

The question remains, then, what should be the higher standard of review? This secondary question is distinct from the first: even those courts that find that there is no conflict in the insurance company context have struggled with how to incorporate a conflict--when theyfind one into the framework of arbitrary and capricious review mandated by Firestone. In Kotrosits v. GATX Corp. Noncontributory Pension Plan for Salaried Employees, 970 F.2d 1165, 1173 (3d Cir. 1992) we stated that had a conflict existed, Firestone "counsels in favor of withholding deference." This suggests de novo review. On the other hand, in Abnathya v. Hoffman-LaRoche, Inc., 2 F.3d 40, 45 n.5 (3d Cir. 1993) we suggested that the circumstances might require "special attention or a more stringent standard of review under Firestone." Again, we turn to the other circuits, where we find three methods of dealing with a conflict: burden shifting, de novo review, and the sliding scale.

We begin with Brown v. Blue Cross & Blue Shield of Ala., 898 F.2d 1556 (11th Cir. 1990), in which the Eleventh Circuit turned to the common law of trusts to determine the appropriate method for reviewing the conflicted discretionary decisions of an insurance company. See 898 F.2d at 1564. It concluded that while an uninterested fiduciary should receive a great deal of deference, common law trust cases dictated that the highly conflicted one should not; even potentially conflicted decisions were closely scrutinized, in part to protect the particular beneficiaries in a given case, and in part "to discourage arrangements where a conflict arises." Id. at 1565. The court determined that a beneficiary need only show a substantial structural conflict of interest in order to shift the burden to the fiduciary to demonstrate that the conflict did not infect a benefits denial. See id. at 1566. It announced the following rule:

[W]hen a plan beneficiary demonstrates a substantial conflict of interest on the part of the fiduciary responsible for benefits determinations, the burden shifts to the fiduciary to prove that its interpretation of plan provision committed to its discretion was not tainted by self-interest. That is, a wrong but apparently reasonable interpretation is arbitrary and capricious if it advances the conflicting interest of the fiduciary at the expense of the affected beneficiary or beneficiaries unless the fiduciary justifies the interpretation on the ground of its benefit to the class of all participants and beneficiaries.

Id. at 1566-67.

To be sure, as a preliminary matter, the court must first determine that the fiduciary's decision was " `wrong' from the perspective of de novo review." Id. at 1567 n.12. But once shifted, the task of justifying the interpretation is by no means insurmountable. If the fiduciary can demonstrate a routine practice or give other plausible justifications-such as the interests of other beneficiaries--deference may be granted. "Even a conflicted fiduciary should receive deference when it demonstrates that it is exercising discretion among choices which reasonably may be considered to be in the interests of the participants and beneficiaries." Brown, 898 F.2d at 1568. The kind of justification that is given as an example is an assertion, supported by evidence, that an insurance company's "interpretation of its policy is calculated to maximize the benefits available to plan participants and beneficiaries at a cost that the plan sponsor can afford (or will pay)." Id. The legitimacy of such an assertion should be ascertained by looking to, among other things, the consistency of the practice, the reasonableness of the "reading" (in that case, interpreting a term), and the internal consistency of the plan with the proferred reading. See id.

The essence of the Eleventh Circuit's approach is that the fiduciary should be accorded deference, but only when deciding between options which are all in the best interest of the beneficiary or beneficiaries. Insurance companies, unlike the typical trustees, may be viewed with some skepticism because of the primacy of their profit-making function. Therefore, given the structural conflict, the administrator of an insurance company funding an ERISA plan has the burden of proving that beneficiary interests motivated a decision which would be "wrong" under de novo review. See id. A version of the Brown approach has been followed by several panels of the Ninth Circuit. See, e.g., Atwood v. Newmont Gold Co., Inc., 45 F.3d 1317, 1322 (9th Cir. 1995) (given material probative evidence of a conflict, the burden is shifted to the denying company to give a legitimate justification for a denial).

The Second Circuit, while stringent in requiring particular evidence that a conflict infected the decisionmaking process, uses de novo review once it credits such evidence. See Sullivan v. LTV Aerospace & Defense Co., 82 F.3d 1251, 1255-56 (2d Cir. 1996). Unlike the Eleventh Circuit, the test outlined in Sullivan does not include burden shifting. "If the court finds that the administrator was in fact influenced by the conflict of interest, the deference otherwise accorded the administrator's decision drops away and the court interprets the plan de novo." Id.

Other courts have rejected the shifting burden and either/or models, and instead use a sliding scale approach, according different degrees of deference depending on the apparent seriousness of the conflict. According to the Fourth Circuit, "the fiduciary decision will be entitled to some deference, but this deference will be lessened to the degree necessary to neutralize any untoward influence resulting from the conflict." Doe v. Group Hospitalization & Medical Services, 3 F.3d 80, 87 (4th Cir. 1993). Despite this divergence from the Eleventh Circuit's burden shifting, we read the Doe court as engaging in a highly demanding exercise when it applies this sliding scale, "review[ing] the merits of the interpretation to determine whether it is consistent with an exercise of discretion by a fiduciary acting free of the interests that conflict with those of the beneficiaries." Id.

The Fourth Circuit's sliding scale approach has been adopted by several other courts. See Vega v. National Life Ins. Service, Inc., 188 F.3d 287, 296 (5th Cir. 1999) (en banc) (using the sliding scale approach); Chambers v. Family Health Plan Corp., 100 F.3d 818 (10th Cir. 1996) ("[T]he arbitrary and capricious standard is sufficiently flexible to allow a reviewing court to adjust for the circumstances alleged, such as trustee bias in favor of a third-party or self-dealing by the trustee."); Miller v. Metropolitan Life Ins. Co., 925 F.2d 979, 984 (6th Cir. 1991) (the arbitrary and capricious standard is "shaped" by the circumstances when there is a conflict of interest). Despite a feint in the direction of adopting the Brown approach, see Armstrong v. Aetna Life Ins. Co., 128 F.3d 1263, 1265 (8th Cir. 1997) (holding that the "perpetual conflict" which exists when an insurer administers benefits from its own plan warrants a de novo standard of review), the Eighth Circuit has settled on the sliding scale. See Woo v. Deluxe Corp., 144 F.3d 1157, 1165 (8th Cir. 1998) (explicitly adopting sliding scale). The First Circuit uses something like the sliding scale, testing a decision by measuring its "reasonableness" in the context it was made, which necessarily includes an awareness of the effects of the decision on the parties. Doe v. Travelers Ins. Co., 167 F.3d 53, 57 (1st Cir. 1999). Reasonableness, notes that court, "has substantial bite itself." Id.

We adopt the approach of the sliding scale cases. That approach allows each case to be examined on its facts. The court may take into account the sophistication of the parties, the information accessible to the parties, and the exact financial arrangement between the insurer and the company. For example, a court can consider whether the insurance contract is fixed for a term of years or changes annually, and whether the fee paid by the company is modified if there are especially large outlays of capital by the insurer.

Another factor to be considered is the current status of the fiduciary. Our previous cases, discussed supra Section II.F, which hold that an employer fiduciary is not conflicted generally assume that the company is stable and will act as a repeat player: The presumed desire to maintain employee satisfaction is based on this premise. When companies are breaking up, or laying off a significant percentage of their employees, or moving all their operations, these incentives diminish significantly. See Langbein, supra note 2, at 216 ("The employer's reputational interest is not likely to be effective when the long term relationship between the firm and the workers is dissolving, as in a plant closing or in a corporate restructuring.").

Furthermore, the sliding scale approach better adheres to Firestone's dictate that a conflict should be considered as a "factor" in applying the arbitrary and capricious standard. 489 U.S. at 115. Following Firestone to the Restatement of Trusts would counsel that a conflict of interest requires tighter review, but not necessarily a shifted burden, when the fiduciary is conflicted. "In the determination of the question whether the trustee in the exercise of a power is acting from an improper motive the fact that the trustee has an interest conflicting with that of the beneficiary is to be considered." Restatement (Second) of Trusts S 187, cmt. g (emphasis added). Comment (d) lists several factors, including conflict of financial interest, to examine whether there is an abuse of discretion, stating that the factors "may be relevant" and including "the existence or nonexistence of an interest in the trustee conflicting with that of the beneficiaries." We think the best way to "consider" these potentially relevant factors (in this case, the structural conflict of interest) is to use them to heighten our degree of scrutiny, without actually shifting the burden away from the plaintiff.

We acknowledge that there is something intellectually unsatisfying, or at least discomfiting, in describing our review as a "heightened arbitrary and capricious" standard. The locution is somewhat awkward. The routine legal meaning of an "arbitrary and capricious" decision is that used, quite understandably, by the district court: a decision "without reason, unsupported by substantial evidence or erroneous as a matter of law." Once the conflict becomes a "factor" however, it is not clear how the process required by the typical arbitrary and capricious review changes. Does there simply need to be more evidence supporting a decision, regardless of whether that evidence was relied upon?

This is unsatisfying. Rather, once "factors" are introduced, arbitrary and capricious stops sounding like arbitrary and capricious and more like some form of intermediate scrutiny, which has no analogue in thisfield. As we have seen, other courts have reconciled the sliding scale and the "arbitrary and capricious" language from Firestone by essentially reformulating the arbitrary and capricious standard for ERISA law, concluding that"the arbitrary and capricious standard may be a range, not a point. . . [it is] more penetrating the greater is the suspicion of partiality, less penetrating the smaller that suspicion is." Wildbur v. ARCO Chem. Co., 974 F.2d 631, 638 (5th Cir. 1992) (quoting Van Boxel v. Journal Co. Employees' Pension Trust, 836 F.2d 1048, 1052-53 (7th Cir. 1987)); Lowry v. Bankers Life & Casualty Retirement Plan, 871 F.2d 522, 525 n. 6 (5th Cir. 1989) (same).

While we also find this explanation wanting, we can find no better method to reconcile Firestone's dual commands than to apply the arbitrary and capricious standard, and integrate conflicts as factors in applying that standard, approximately calibrating the intensity of our review to the intensity of the conflict. While the approach of Professor Langbein, see supra note 2, and of the Eleventh Circuit would seem more compatible with the basic principles of trust law, and hence a better "fit", only the Supreme Court can undo the legacy of Firestone. In sum, we adopt the sliding scale approach, and, accordingly, will expect district courts to consider the nature and degree of apparent conflicts with a view to shaping their arbitrary and capricious review of the benefits determinations of discretionary decisionmakers.

Were we to apply extremely deferential arbitrary and capricious review, we would likely affirm the judgment of the district court, because there is some credible evidence which an administrator could have relied upon to conclude that Maria Pinto was not totally disabled. Two doctors, one of whom is a specialist in cardiology, stated that they did not believe that she was totally disabled. Therefore, Reliance Standard's decision was not "without reason, unsupported by substantial evidence or erroneous as a matter of law." Abnathya, 2 F.3d at 45 (quoting Adamo v. Anchor Hocking Corp., 720 F. Supp. 491, 500 (W.D.Pa. 1989)). On the other hand, were we to apply de novo review, we would probably conclude that Reliance Standard made the incorrect determination, because Pinto presented credible evidence from her long-time treating cardiologist that she was totally disabled for cardiological reasons, evidence which was affirmed by another cardiologist, and only one other cardiologist, who had much less opportunity to perform tests and examine her than her own doctor, concluded that she was not. According deference to neither side, Pinto's case seems stronger.

However, applying a heightened arbitrary and capricious review, we are deferential, but not absolutely deferential. Like the Fifth Circuit, "[t]he greater the evidence of conflict on the part of the administrator, the less deferential our abuse of discretion standard." Vega, 188 F.3d at 297. Therefore, we look not only at the result--whether it is supported by reason--but at the process by which the result was achieved. In so doing, we note several problems. First, Reliance Standard reversed its own initial determination that Pinto was totally disabled without receiving any additional medical information. The only thing that changed between Reliance Standard's initial acceptance and its subsequent denial was that Pinto notified Reliance Standard that her SSA application had been rejected (a determination that SSA subsequently abandoned). The SSA's rejection of Pinto appears to have triggered Reliance Standard's investigation into Pinto's disability, suggesting that Reliance Standard places significant trust in the SSA process, yet the SSA's subsequent reversal had no such effect. Inconsistent treatment of the same facts was viewed with suspicion by the Brown court. See 898 F.2d at 1569 ("That [the insurance company] would reach opposing conclusions on the basis of the same evidence seriously challenges the assumptions upon which deference is accorded to[its] interpretation of the plan.").

Second, looking at the final decision, we see a selectivity that appears self-serving in the administrator's use of Dr. Bahler's expertise. Reliance Standard used some of Dr. Bahler's specific limitations to explain its rejection, but it did not accept (or satisfactorily explain its rejection of) his conclusion that she was totally disabled. The Fifth Circuit addressed a similar circumstance, where the administrator credited one part of the advice of a treating doctor, but not his other advice. That court held that this was unacceptable in the context. See Salley v. E.I. DuPont de Nemours & Co., 966 F.2d 1011, 1015 (5th Cir. 1992). This inconsistent treatment of the same authority in two separate instances (the SSA, Dr. Bahler) raises the likelihood of self-dealing. Applying the sliding scale to this case, our review is ratcheted upward by these suspicious events.

Finally, when a staff worker reviewing the files recommended that Pinto be reestablished pending further testing, her suggestion was rejected, and Reliance Standard decided to do the opposite: suspend the resumption of benefits. Although this in itself does not prove bias, it lends further support to the view that whenever it was at a crossroads, Reliance Standard chose the decision disfavorable to Pinto. The default position was that benefits were not granted.

Taking all of these procedural anomalies into account, we find ourselves on the far end of the arbitrary and capricious "range," and we examine the facts before the administrator with a high degree of skepticism.8

Reliance Standard relies heavily on the "two-to-two" argument, arguing that because there are two doctors on either side of the Pinto disability debate, a decision to credit either side cannot be arbitrary and capricious. However, neither of the doctors retained by Reliance Standard had the same contact with Pinto that Dr. Bahler did. Dr. Rosenthal read Dr. Bahler's reports, examined Pinto, and talked with her, but this examination, however professional, does not compare with the eighteen years of interaction between Dr. Bahler and Pinto. The essence of Dr. Bahler's conclusion was that Pinto's condition was "labile"; that is, her condition could severely worsen under stress or activity ("high stress situations . . . could precipitate her cardiac asthma."). Although she might be able to persist in an occupation for some time, and she had basic motor skills, the risk of work was too great. Reliance Standard gave no explanation for its rejection of this aspect of Dr. Bahler's assessment.

Moreover, while Reliance Standard relies on Dr. Capone, Dr. Capone is a pulmonologist; he could only, and did only, assess whether she had pulmonary problems. The pulmonary examination was at Dr. Rosenthal's suggestion, but neither Pinto nor Dr. Bahler have suggested that the source of her disability was pulmonary. That Dr. Capone concluded that she was not disabled by respiratory disease should not discredit Dr. Bahler. Moreover, Dr. Capone originally suggested that in order to decide whether Pinto was disabled, she would first need to undergo therapy to see if it made a difference. She never underwent therapy, but despite his initial reluctance--and after being pressed-he concluded that she was not totally disabled. If Reliance Standard was as accommodating with Dr. Bahler's conclusions as it was with Dr. Capone's, it would likely have granted Pinto benefits.

For these reasons, a factfinder could conclude that Reliance Standard's decision to credit its doctors over Drs. Bahler and Goodman was the result of self-dealing instead of the result of a trustee carefully exercising its fiduciary duties to grant Pinto the benefits due her under the insurance plan. Summary judgment was therefore inappropriate, for there is a genuine issue of material fact as to whether Reliance Standard acted arbitrarily and capriciously. The judgment of the District Court will be reversed, and the case remanded for further proceedings consistent with this opinion. There is sufficient evidence at this stage to merit a penetrating review of the decision under the heightened standard. The decision was close enough that such a review may result in a determination that it was arbitrary and capricious. On remand, the District Court may take evidence regarding the conflict of interest, and ways in which the conflict may have influenced the decision, and then determine whether, considering the conflict, the decision was "arbitrary and capricious" in the sense described in Section IV.