Plaintiff Shirley M. Critchlow appeals from a judgment of the United States District Court for the Western District of New York, David G. Larimer, Chief Judge, dismissing her complaint seeking recovery from defendant First UNUM Life Insurance Company of America (“UNUM”) of benefits on an accidental-death-and-dismemberment insurance policy covering her son Daniel Critchlow (“Critchlow”), who died during the practice of autoerotic asphyxiation. The district court granted UNUM’s motion for summary judgment dismissing the complaint, and denied plaintiffs cross-motion for summary judgment in her favor, on the ground that the UNUM policy excluded coverage for losses resulting from intentionally self-inflicted injuries and that autoerotic asphyxiation is an intentionally self-inflicted injury. On appeal, plaintiff contends that the district court erred in concluding that the practice of autoerotic asphyxiation is intended to inflict injury.
On August 12, 2003, a divided panel of this Court, in an amended majority opinion by Judge Van Graafeiland, with B.D. Parker, J., concurring and Kearse, J., dissenting, concluded that the dismissal of the complaint should be affirmed. On August 27, 2003, a judge of this Court requested a poll to have the appeal reheard en banc. The mandate was issued inadvertently on August 28, 2003, and was recalled on June 21, 2004, in light of the pendency of the en banc poll. While the en banc poll was pending, Judge Parker reconsidered his earlier decision and voted to reverse the judgment. Therefore, with the issuance of the present opinion, the earlier decision in this case, reported at 340 F.3d 130 (2003), is vacated. For the reasons that follow, the judgment is reversed, and the matter is remanded for entry of judgment in favor of plaintiff.
I. BACKGROUND
The facts are not disputed. Plaintiff was the named beneficiary of a group accidental-death-and-dismemberment insurance policy covering Critchlow, issued by UNUM (the “Policy” or “UNUM Policy”) to Critchlow’s employer as part of an employee benefit plan covered by the Employee Retirement Income Security Act (“ERISA”), 29 U.S.C. §§ 1001-1461. The Policy’s exclusions section stated, inter alia, that UNUM “will not pay if the loss is caused by: 1) intentionally self-inflicted injuries[, or] ... 5) illness [or] disease.” (UNUM Policy at 2.) The Policy term was December 1, 1998, to December 1, 1999.
A. The Events
In the early morning hours of February 27, 1999, Critchlow, age 32, was found dead in his bedroom. He was unclothed, *250lying on the floor, with ligatures tying various parts of his body. The coroner’s report concluded, and it is undisputed, that his death resulted from his practice of autoerotic asphyxiation, ie., the practice of limiting the flow of oxygen to the brain during masturbation in an effort to heighten sexual pleasure.
Plaintiff applied for death benefits under the accidental death terms of the Policy. She attached copies of the autopsy report on Critchlow and the reports of two members of the County Sheriffs department— Deputy Kevin Kuntz and Investigator R. Hetzke — who had been summoned to the scene. The latter reports stated, in part, as follows:
Daniel had apparently been in the middle of an auto-erotic act. He was tied up in various places by cord, and these cords had evidently been attached to a set of counter weights which were meant to give him an “out” if he started to lose consciousness.
(Report of Dep. Kevin Kuntz dated February 27,1999 (“Kuntz Report”), at 2.).
It appears that the victim was engaged in autoeroticism. It does not appear that he intentionally took his life as he has escape measures built into his binds. Dr. Blasczak responded to the residence for the Coroner’s office. He concurred with writers [sic ] assessment of the cause of death.
The writer retrieved a sales receipt from the victim’s wallet. The deceased made a grocery purchase ... at 6:30PM. From items left in the kitchen, it appears that victim was planning supper. Writer also located a receipt dated 11/20/98 ... for clothes line.
2/28 Writer spoke with Richard Critchlow, the deceased’s father. He had been out of town at the time of his son’s death. He stated that while unnatural he understood the events leading up to Daniel’s death and the cause of his death. Mr. Critchlow ... related an incident that occurred approx. 15 years ago where he had found Daniel after he had bound himself up.
2/28 Writer spoke with Dr. Hannan, who stated that death certificate states that cause of death was accidental asphyxiation.
(Report of Inv. R. Hetzke dated March 1, 1999 (“Hetzke Report”), at 1-2.)
UNUM, in a letter to plaintiff dated July 7, 1999 (“UNUM July 1999 Letter”), denied her application on the grounds that Critchlow’s death was not accidental and was a loss caused by intentionally self-inflicted injuries. It stated, in pertinent part, that
[bjased upon our investigation into the facts and circumstances of the death of the insured and our comprehensive review of prior case law, we have concluded that the death of the insured did not result directly and independently of all other causes from accidental bodily injury.
Additionally, the Policy contains an Exclusion which states:
“We will not pay if the loss is caused by:
(1) Intentionally!] self-inflicted injuries;!”]
Our investigation further reveals that the death of the insured falls within the above Exclusion for intentionally self-inflicted injuries.
For the foregoing reasons, First UNUM denies all coverage under the Policy and declines to pay any benefits thereunder.
(UNUM July 1999 Letter at 1-2.) UNUM stated that its decision was “not a waiver *251of any and all other rights and defenses [it] may have under the provisions of th[e P]olicy.” (Id. at 2.)
In a September 1, 1999 letter to UNUM from plaintiffs attorney Richard A. Cala-brese, plaintiff appealed the denial to UNUM’s appeals committee (“September 1999 Appeal Letter” or “Appeal Letter”). The Appeal Letter described in detail the circumstances of Critchlow’s death, provided medical and technical information on the practice of autoerotic asphyxiation, cited law enforcement officials’ conclusions that Critchlow’s death was in fact unintended and accidental, and argued that Critchlow’s death was covered by the Policy. It also appended, inter alia,
a number of articles from the Journal of Forensic Sciences dealing with autoerotic asphyxiation, which explain the practice, define it and help to explain that the victim of it has not committed suicide, but that death caused by the practice is an unintended result, and hence an accident.
(September 1999 Appeal Letter at 6.) The Appeal Letter also argued that Critchlow’s creation of escape mechanisms showed that he had intended to avoid, not incur, injury and that, therefore, UNUM’s exclusion for ‘“intentionally self-inflicted injury’ ” did not apply. (Id. at 9.)
In a letter to plaintiffs counsel dated December 15, 1999 (“UNUM December 1999 Letter”), the UNUM appeals committee upheld the denial of plaintiffs request for death benefits. UNUM stated that
the basis for the decision of the Appeals Committee are [síc] the reasons given in First UNUM’s initial denial letter dated July 7, 1999, together with the holding and reasoning of the United States District Court for the Northern District of New York in Bennett v. American International Life Assurance Company of New York, 956 F.Supp. 201 (N.D.N.Y.1997). Consistent with the legal standard to be applied in ERISA benefit cases, as stated in the Bennett decision, it is our belief that the evidence supports the conclusion that although the decedent, Daniel Critchlow, may have subjectively expected to survive the activity which caused his death, such expectation was not objectively reasonable.
(UNUM December 1999 Letter at 1.)
B. The Present Action
Plaintiff commenced the present action in April 2000 under ERISA, see 29 U.S.C. § 1132(a)(1)(B), alleging that the Policy was issued pursuant to an employee benefit plan covered by ERISA; that the Policy provides for the payment of a death benefit to the beneficiary of an insured who dies as the result of an accident; that Critchlow’s death was accidental; and that UNUM had, both initially and following an administrative appeal, wrongfully denied plaintiffs claim for the death benefit due her under the Policy.
UNUM filed an answer principally denying that Critchlow’s death was accidental. It moved for summary judgment dismissing the complaint on that ground; it argued alternatively that Critchlow’s death was excluded from coverage because it resulted from an intentionally self-inflicted injury or from illness or disease. (See Memorandum in Support of Defendant’s Motion for Summary Judgment (“UNUM Memorandum”) at 1.) In support of its motion, UNUM submitted two brief affidavits introducing the following documents: (a) the Policy, (b) plaintiffs initial application for benefits, (c) UNUM’s letters denying plaintiffs claim, and (d) two opinions from experts retained by plaintiff, to wit, a January 19, 2001 report by Robert M. Greendyke, M.D. (“Greendyke Opinion”), and a February 18, 2001 report by Ste*252phen J. Hucker, Medical Director, Professor of Psychiatry, and Head of the Division of Forensic Psychiatry in the Forensic Program of McMaster University (“Huck-er Opinion”). UNUM’s memorandum of law stated, inter alia, that
[o]ne of plaintiffs experts, Prof. Stephen J. Hucker, explains ... in his report (Steeker aff., Ex. A. p. 4) that “autoerot-ie asphyxia”
refers to the practice of deliberating [sic] inducing hypoxia (a state of diminished oxygen supply to the brain) with the intention of producing sexual arousal. The Diagnostic and Statistical Manual of the American Psychiatric Association (4th edition), known as DSM-IV, includes the condition of a mental disorder under the general rubric of Sexual Masochism and uses the term “asphyxiophilia.” Others prefer the term “hypoxyphilia.”
As Prof. Hucker goes on to explain, an individual suffering from this disorder may have “no wish to die but rather, to survive and enjoy the experience again.” Id. Unfortunately, however, “[w]hat is not commonly appreciated ... is that unconsciousness occurs very rapidly, i.e. in a matter of seconds.” Id. In short, as plaintiffs other expert, Dr. Robert M. Greendyke, states in his report, Mr. Critchlow died from “self-induced strangulation associated with deviant sexual activity” (id., Ex. B, p. 2).
(Memorandum in Support of Defendant’s Motion for Summary Judgment at 2-3 (footnote omitted)). UNUM proffered no expert opinions other than those of Hucker and Greendyke.
The Hucker Opinion also stated, inter alia, that “[fjortunately for most the experience of hypoxia may be survived with no obvious impairment or injury.” (Hucker Opinion at 4.) The Hucker Opinion stated:
4. Hypoxia, whether or not induced deliberately, as in cases of autoerotic asphyxia, does not, in my opinion, invariably lead to death or injury. Mr. Critchlow, in fact, died of prolonged hypoxia, a condition that he was trying to prevent.
5. In my opinion, therefore, Mr. Critchlow had expected to survive his autoerotic activities.
6. This'expectation would have also been subjectively reasonable, based on his likely experience of having survived similar escapades in the past, and was objectively reasonable based on the knowledge that autoerotic asphyxial episodes do not, inevitably, or even substantially likely, lead to a fatal outcome.
(Hucker Opinion at 5-6.) The Greendyke Opinion stated:
[T]he circumstances conclusively indicate an accidental death resulting from a miscalculation or error in judgment in the course of deviant sexual behavior
... [T]he risk-taking behavior which led to this death cannot be construed as a death wish. People engage in multitudes of possibly dangerous activities on a daily basis, ranging from auto racing to bungee jumping to DWI to alpine mountain climbing, to name a few, and occasionally die if something goes awry, without any claim of self-destructive intent. Simply said, in the heat of sexual excitement, Mr. Critchlow failed to guarantee his own safety. That his precautions had worked previously is quite clear from the evidence that his autoero-ticism was not a first-time event but rather an elaborate, pre-arranged scenario. In other words he had no expectation of death, and that expectation was objectively reasonable, i.e. that death was not substantially likely to result from his conduct.
*253(Greendyke Opinion at 2; see also id. at 3 (“[D]eath was not a normal expected result of the behavior.”).)
Plaintiff cross-moved for judgment in her favor, contending that Critehlow’s death was unintentional and accidental. In support of her motion, plaintiff submitted her own affidavit and affidavits from Critchlow’s father and sister, along with copies of the Hucker and Greendyke Opinions that had been submitted by UNUM in support of its motion, and brief affidavits from the authors of those opinions. The Hucker and Greendyke affidavits stated that “[i]f called to testify as an expert witness in the trial of this matter, my opinion will be that the death of Daniel Critchlow was accidental for the reasons set forth in my report.” (Affidavit of Stephen J. Hucker dated February 19, 2001, ¶4; Affidavit of Robert M. Greendyke, M.D., dated April. 19, 2001 (“Greendyke Aff.”), ¶ 4.). The latter continued as follows:
5. With respect to autoerotic asphyxiation, as practiced by the decedent, Daniel Critchlow, the first injury that occurred to him and the one that eventually lead [sic] to his death, was when there was a sufficient loss of blood supply to his brain, which decreased the availability of oxygen to his brain cells that cause him to lose consciousness. This was never intended by him; he never intended to inflict any injury upon himself, nor was it part of his practice, it being his intent to unbind himself and free himself before unconsciousness occurs. What happened to Daniel Critch-low was accidental and unintended.
6. Up until the time of loss of consciousness, Mr. Critchlow sustained no injury to himself by reason of his practice of autoerotic asphyxiation. If he had not passed out, he would have completed his masturbatory act without any injury to himself, notwithstanding the ■fact that he had constricted his windpipe to create a partial strangulation. Such constriction as practiced in autoerotic asphyxiation would cause no injury, not even a mark on the neck.
7. A change in the blood flow to the head which merely causes light-headedness is a reversible change, such as a change in body temperature or blood pressure, and does not constitute an injury. In no event would such a reversible change in the availability of oxygen in the blood supply to the head constitute a “bodily injury”.
8. As a consequence, it is my opinion that there was no “injury” to the decedent, that was “intentionally self-inflicted” which lead [sic ] to his death.
(Greendyke Aff. ¶¶ 5-8 (emphases in original).)
In a Decision and Order reported at 198 F.Supp.2d 318 (2002), the district court denied plaintiffs cross-motion and granted UNUM’s motion for summary judgment dismissing the complaint. The court reviewed the denial of plaintiffs claim for benefits de novo, “since the policy con-tainted] no language giving UNUM discretion to interpret the policy’s terms or to determine eligibility for benefits.” 198 F.Supp.2d at 321. Noting that “[t]he relevant facts are undisputed,” the court began its analysis by stating that
[i]t is undisputed that decedent died while practicing “autoerotie asphyxiation,” which is described in the coroner’s report as a “dangerous form of sexual mannerism in which arousal is induced by depriving the brain of oxygen in one of several ways: hanging, strangulation, chest compression, covering the mouth and nose with a plastic bag or mask.” Plaintiffs Ex. E [coroner’s report]. Although decedent had constructed a system of ropes and counterweights that *254was apparently intended to incorporate escape mechanisms, or otherwise to ensure that he did not die of asphyxiation, for whatever reason that system failed him on this occasion. It does not appear, however, and UNUM makes no contention, that decedent intended or expected to die that evening, and there was evidence that decedent had engaged in this practice in the past.
198 F.Supp.2d at 320-21 (emphases added). The court held, however, that Critch-low’s death was the result of intentionally self-inflicted injuries as a matter of law:
Plaintiff asserts that the evidence shows that although decedent intentionally constricted his windpipe, that act itself did not cause any injury, since decedent could have performed that act (as he apparently had in the past) without suffering any injury. Since the “injury” here — decedent’s loss of consciousness followed by his death — was not the intended result of decedent’s actions, plaintiff contends that his injuries were not “intentionally self-inflicted.”
Plaintiffs position strains logic and is not persuasive. She states in her memorandum of law that “[cjonstricting the wind pipe, while an intentional act, did not cause Daniel Critchlow any injury.” It certainly did cause him injury, however; it led directly to his death. The cause of death was asphyxiation. Plaintiffs Ex. E. That asphyxiation was caused by decedent’s intentionally constricting his windpipe so as to cut off or reduce the flow of oxygen to his brain. That it is possible to do so for a short period without causing lasting injury, or that injury or death does not immediately occur upon construction of the trachea, does not mean that decedent’s intentional act caused him no injury. Decedent may have thought that he could free himself before he lost consciousness, but he was wrong. His death was nevertheless intentionally self-inflicted, given the serious and obvious risk of death entailed by decedent’s intentional actions.
This result finds support in other cases that have addressed the question of whether death from autoerotic asphyxiation falls within the scope of a provision excluding “intentionally self-inflicted injuries,” or containing similar language. In Sims v. Monumental Gen’l Ins. Co., 960 F.2d 478 (6th Cir.1992), aff'd 778 F.Supp. 325 (E.D.La.1991), the Court of Appeals for the Fifth Circuit held that a death from autoerotic asphyxiation was the result of an “intentionally self-inflicted injury,” and therefore not covered by an insurance policy that, like the one in the case at bar, excluded such injuries. The Sims court reasoned that the fact that the decedent
only intended partial strangulation and did not intentionally kill himself does not avail Mrs. Sims. The policy in this case not only excludes suicide, but also any loss (including death) “resulting directly or indirectly, wholly or partly from ... [an] intentionally self-inflicted injury.” Partial strangulation is an injury in and of itself. His death “resulted directly or indirectly, wholly or partly from” that intentionally self-inflicted injury.
198 F.Supp.2d at 323 (footnote omitted) (emphasis added).
The court stated that it found “no evidence in the record that would support a finding that decedent’s death was not the result of an intentionally self-inflicted injury.” Id. at 327. The court had declined to consider the Hucker and Greendyke opinions (submitted to the court first by UNUM and then by plaintiff) on the grounds that plaintiff had failed to submit those opinions to UNUM in support of her *255application for benefits and that she had not shown good cause for the court to excuse that failure. See id. at 322. However, the court also stated that its decision as to the merits would not be different even if it considered plaintiffs experts’ views. See id. at 327 n. 5.
In support of her position, plaintiff points to evidence that some persons practice autoerotic asphyxiation repeatedly, as well as physical evidence that decedent himself had engaged in this practice on more than one occasion prior to February 26, 1999. This evidence indicates that it is possible to engage in autoerotic asphyxiation without dying, or even without losing consciousness. From that evidence, plaintiff argues that such activity need not cause injury to the person engaged in it. That one may survive autoerotic asphyxiation without suffering any major injury is not dispositive, however. The fact remains that autoerotic asphyxiation, as practiced by decedent, requires not just a slight amount of pressure, or a negligible reduction of the flow of oxygen, but a significant deprivation of oxygen to the brain — in other words, strangulation. Any definition of “injury” that excludes strangulation — whether fatal or not — is simply unreasonable.
.... [B]y constricting the flow of oxygen to his brain, to the point where loss of consciousness and death were certain to occur if the pressure were not released in a relatively short time, the decedent did injure himself. He simply believed (apparently) that he could bring that injury to a halt before the injury became life-threatening. That his belief proved incorrect does not save plaintiffs claim.
Id. at 327 (footnote omitted) (emphasis in original).
Judgment was entered dismissing the complaint. This appeal followed.
II. DISCUSSION
On appeal, plaintiff contends principally that the district court erred in ruling that Critchlow’s death was the result of an intentionally self-inflicted injury and in faffing to grant summary judgment in her favor. She also contends that the district court erred in refusing to consider the Hucker and Greendyke Opinions and affidavits. UNUM contends that the district court correctly ruled that Critehlow died of an intentionally self-inflicted injury. Alternatively, it contends that the dismissal of the complaint should be upheld either because Critchlow’s death was not accidental or because his death fell within a policy exclusion for a loss caused by “illness [or] disease.” Given the record in this case and the principles applicable to employee benefit plans governed by ERISA, we conclude that plaintiff, not UNUM, was entitled to summary judgment.
A. “Intentionally Self-Inflicted Injuñes”
ERISA preempts “any and all State laws insofar as they may ... relate to any employee benefit plan” covered by ERISA. 29 U.S.C. § 1144(a). See, e.g., Krishna v. Colgate Palmolive Co., 7 F.3d 11, 13-14 (2d Cir.1993). “Congress, in adopting ERISA, expected that ‘a federal common law of rights and obligations under ERISA-regulated plans would develop.’ ” Todd v. AIG Life Insurance Co., 47 F.3d 1448, 1451 (5th Cir.1995) (“Todd”) (White, Associate Justice (Ret.), sitting by designation) (quoting Pilot Life Insurance Co. v. Dedeaux, 481 U.S. 41, 56, 107 S.Ct. 1549, 95 L.Ed.2d 39 (1987)). An ERISAregulated plan is thus to be construed in accordance with federal common law. See, e.g., Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101, 110, 109 S.Ct. 948, 103 *256L.Ed.2d 80 (1989). In developing federal common law in an area, the courts may look to state law; but they “may use state common law as the basis of the federal common law only if the state law is consistent with the policies underlying the federal statute in question.” Krishna v. Colgate Palmolive Co., 7 F.3d at 14 (internal quotation marks omitted); see, e.g., Todd, 47 F.3d at 1451 (ERISA questions may be answered by resort to state law only to the extent that state law does not conflict with federal law).
“We interpret ERISA plans in an ordinary and popular sense as would a person of average intelligence and experience.” Todd, 47 F.3d at 1452 n. 1 (internal quotation marks omitted); see, e.g., Fay v. Oxford Health Plan, 287 F.3d 96, 104 (2d Cir.2002) (“Fay ”); Masella v. Blue Cross & Blue Shield of Connecticut, Inc., 936 F.2d 98, 107 (2d Cir.1991) (“Masella”)- If there are ambiguities in the language of an insurance policy that is part of an ERISA plan, they are to be construed against the insurer. See, e.g., Fay, 287 F.3d at 104; Kinstler v. First Reliance Standard Life Insurance Co., 181 F.3d 243, 251-52 (2d Cir.1999); Todd, 47 F.3d at 1451-52. This is a “hornbook rule of contract interpretation” that applies “particularly when [the ambiguity is] found in an exclusionary clause,” Masella, 936 F.2d at 107 (internal quotation marks omitted); and the principle is incorporated into the federal common law governing ERISA-regulated plans in order not to “ ‘afford less protection to employees and their beneficiaries than they enjoyed before ERISA was enacted,’ a result that would be at odds with the congressional purposes of promoting the interests of employees and beneficiaries and protecting contractually defined benefits,” id. (quoting Firestone Tire & Rubber Co. v. Bruch, 489 U.S. at 113-14, 109 S.Ct. 948). Language in a plan “ ‘is ambiguous when it is capable of more than one meaning when viewed objectively by a reasonably intelligent person who has examined the context of the entire ... agreement.’ ” Fay, 287 F.3d at 104 (quoting O’Neil v. Retirement Plan for Salaried Employees of RKO General, Inc., 37 F.3d 55, 59 (2d Cir.1994)).
Similarly, when the issue is the proper interpretation of a clause that excludes a particular condition or occurrence from the coverage provided by the policy, the exclusion clause should be read narrowly rather than expansively. Under traditional principles of state law, exclusions from insurance policy coverage are given strict construction. See, e.g., M.H. Lipiner & Son, Inc. v. Hanover Insurance Co., 869 F.2d 685, 687 (2d Cir.1989) (applying New York law); Seaboard Surety Co. v. Gillette Co., 64 N.Y.2d 304, 311, 486 N.Y.S.2d 873, 876, 476 N.E.2d 272 (1984). In order to be enforced, “exclusions or exceptions from policy coverage must be specific and clear”; they “are not to be extended by interpretation or implication.” Id. at 311, 486 N.Y.S.2d at 876, 476 N.E.2d 272; see also Miller v. Continental Insurance Co., 40 N.Y.2d 675, 678, 389 N.Y.S.2d 565, 567, 358 N.E.2d 258 (1976). Given that ERISA is not to be interpreted to afford employees and their beneficiaries less protection than they enjoyed before ERISA was enacted, see, e.g., Masella, 936 F.2d at 107, these principles governing exclusion clauses are no less applicable to an insurance policy governed by ERISA.
Further, “as a matter of general insurance law, the insured has the burden of proving that a benefit is covered, while the insurer has the burden of proving that an exclusion applies,” Mario v. P & C Food Markets, Inc., 313 F.3d 758, 765 (2d Cir.2002); see, e.g., Technicon Electronics Corp. v. American Home Assurance Co., 74 N.Y.2d 66, 73-74, 544 N.Y.S.2d 531, 533, *257542 N.E.2d 1048 (1989), and these principles too are applicable in ERISA cases, see, e.g., Mario v. P & C Food Markets, Inc., 313 F.3d at 765. Thus, where the insurer’s basis for denying a claim “is set forth in the ‘exclusions’ section of the plan, the burden is usually on the plan sponsor, who must -prove that the exclusion applies.” Id.
Under ERISA, an insurer bears the burden to prove facts supporting an exclusion of coverage.... Thus, as an affirmative defense, an exclusion of coverage requires that the insurer provide proof by a preponderance of the evidence.
Fought v. UNUM Life Insurance Company of America, 357 F.3d 1173, 1185 (10th Cir.2004) (per curiam).
In ruling against plaintiff in the present case, the district court relied in large part on the Fifth and Eighth Circuits’ decisions in Sims v. Monumental General Insurance Co., 960 F.2d 478, 480 (5th Cir.1992) (“Sims ”), and Sigler v. Mutual Benefit Life Insurance Co., 663 F.2d 49, 50 (8th Cir.1981) (“Sigler”), which ruled that deaths caused by autoerotic asphyxiation were caused by self-inflicted injuries. Sims and Sigler, however, were not ERISA cases but rather were diversity actions controlled by state law. After Sims, the Fifth Circuit itself in Todd, dealing with autoerotic asphyxiation in the context of an ERISA-regulated insurance policy, declined to follow Sims and Sigler, instead applying federal common law principles developed for claims asserted under ERISA. See Todd, 47 F.3d at 1453 & n. 4.
Todd dealt with the question of whether, under an ERISA-regulated policy, death resulting from autoerotic asphyxiation can be considered “accidental,” rather than with whether such a death falls within the policy’s exclusion for self-inflicted injuries. See id. at 1452. So far as we are aware, the latter question has been resolved in only one federal appellate case, Padfield v. AIG Life Insurance Co., 290 F.3d 1121, 1127-28 (9th Cir.) (“Padfield”), cert. denied, 537 U.S. 1067, 123 S.Ct. 602, 154 L.Ed.2d 556 (2002), which was decided several weeks after the district court decision in the present case. The Padfield court stated that “voluntary risky acts resulting in injury are not necessarily acts that result in ‘intentionally self-inflicted injury,’ ” id. at 1129, and that the answer to the question of whether autoerotic asphyxiation results in an intentionally self-inflicted injury
hinges on whether the physical consequences that [the deceased] intended were injuries. If they were injuries, and if they led to his death, the exclusion applies, and [the insurer] correctly denied benefits,
id.
Citing the principles that provisions of an ERISA-governed policy are to be read in their ordinary and popular sense and in accordance with the federal common law of rights and obligations developed for ERISA-regulated plans, see id. at 1130, the Padfield court noted that courts in other ERISA cases confronted with similar questions had used an analysis that has both subjective and objective components, see id. at 1126. Under the subjective/objective analysis,
[t]he court first asks whether the insured subjectively lacked an expectation of death or injury. See Wickman v. Northwestern Nat'l Ins. Co., 908 F.2d 1077, 1088 (1st Cir.1990) (“Requiring an analysis from the perspective of the reasonable person in the shoes of the insured fulfills the axiom that accident should be judged from the perspective of the insured.”). If so, the court asks whether the suppositions that underlay the insured’s expectation were reason*258able, from the perspective of the insured, allowing the insured a great deal of latitude and talcing into account the insured’s personal characteristics and experiences. See id. If the subjective expectation of the insured cannot be ascertained, the court asks whether a reasonable person, with background and characteristics similar to the insured, would have viewed the resulting injury or death as substantially certain to result from the insured’s conduct.
Padfield, 290 F.3d at 1126 (emphases added).
In applying that analysis to the practice of autoerotic asphyxiation in general, the Padfield court noted that
[ajutoerotic asphyxiation is a repetitive pattern of behavior that individuals engage in over a period of years, and generally the intent of the individuals performing this act is not death .... When performed successfully, the act results only in a temporary decrease in oxygen levels that causes light-headedness, and usually does not leave visible marks on the neck.
Id. (internal quotation marks omitted) (emphases added). However,
[bjecause of equipment malfunction, errors in the placement of the noose or ligature, or other mistakes, accidental deaths sometimes occur. Data from the United States, England, Australia, and Canada indicate that one to two [autoer-otic asphyxiationj-caused deaths per million population are detected and reported each year.... But the use of asphyxia to heighten sexual arousal more often than not [has] a non fatal outcome ....
Id. at 1125-26 (internal quotation marks omitted) (emphases added); see also id. at 1127 (“death by autoerotic asphyxiation is statistically rare”).
Reviewing the record before it, the Pad-field court ruled that Padfield plainly had no subjective intent to injure himself.
All of the evidence indicates that if the events ... had gone as Mr. Padfield intended, he would have experienced a temporary deprivation of oxygen, a euphoric light-headedness ... and an intensified sexual experience. His oxygen level would then have been restored, his euphoric state would have subsided, and he would have returned home uninjured. None of these consequences is an “injury” as that term is defined in the ordinary and popular sense [by] persons of average intelligence and experience.
290 F.3d at 1129 (internal quotation marks omitted). Padfield’s death ensued because the events did not go as he had intended and expected, although his expectation was reasonable:
Autoerotic asphyxiation practitioners expect to survive the experience, and there is nothing to suggest that Mr. Padfield subjectively expected otherwise. Though the record is limited, it appears that Mr. Padfield had a history of engaging in this autoerotic behavior and surviving it.... Because death by au-toerotic asphyxiation is statistically rare, his expectation of survival certainly was reasonable.
Even if we could not determine Mr. Padfield’s subjective expectations, the same conclusion would be warranted under a purely objective analysis because death is not the “substantially certain” result of autoerotic asphyxiation.... Given the uniform medical and behavioral science evidence indicating that autoerotic activity ordinarily has a nonfatal outcome, the likelihood of death from autoerotic activity falls far short of what would be required to negate coverage under an accidental death policy.
*259Id. at 1127 (other internal quotation marks omitted) (emphases added). In concluding that the proposition that death was unlikely was uniformly supported by the medical and scientific evidence, the Padfield court cited, inter alia, Todd, 47 F.3d at 1457, which in turn cited R. Hazelwood, P. Dietz & A. Burgess, Autoerotic Fatalities 49 (1983). The Padfield court concluded that
[although] Mr. Padfield voluntarily engaged in actions that led to a fatal injury, ... his reasonable expectation was that this behavior would not have resulted in “injury ” as that word is commonly defined. Given both the usual pattern of autoerotic asphyxiation and the statements by Mrs. Padfield, the undisputed facts in this case show that Mr. Padfield, having performed the act in the past without inflicting any injury, had a reasonable expectation that he would be able to do so again. Thus, ... Mr. Padfield did not die from an intentionally self-inflicted injury .... Rather, he made a “fatal mistake.” .... Generally, insureds purchase accident insurance for the very purpose of obtaining protection from their own miscalculations and misjudgments.
Id. at 1130 (other internal quotation marks omitted) (emphasis added).
We conclude, as did Padfield, that this subjective/ objective analysis reflects the developing federal common law used in ERISA cases to determine whether a death, including a death during the practice of autoerotic asphyxiation, was, within the meaning of an ERISA-regulated insurance policy, either accidental or the result of an intentionally self-inflicted injury. Thus, in the present case, we ask, first, whether Critchlow subjectively lacked an expectation of death or injury, and second, if so, whether the suppositions that underlay that expectation were reasonable from Critchlow’s perspective, taking into account, inter alia, his own personal characteristics and experiences.
As to Critchlow’s subjective intent, it has never been disputed that his death was subjectively unexpected and unintended. In denying plaintiffs administrative appeal from the denial of benefits, UNUM, while disputing the objective reasonableness of Critchlow’s expectation of survival, stated that “it is our belief that the evidence supports the conclusion that ... the decedent, Daniel Critchlow, may have subjectively expected to survive the activity which caused his death .... ” (UNUM December 1999 Letter at 1.) UNUM did not take a contrary position in the litigation. The district court noted that “UNUM makes no contention!] that decedent intended or expected to die that evening ....” 198 F.Supp.2d at 321. Further, the court’s own interpretation of the evidence was that “[i]t d[id] not appear ... that [Critchlow] intended or expected to die that evening,” id.; the court described “his death” as an “unintended injury,” id. at 324 (emphasis added).
The conclusions that Critchlow’s death was subjectively unintended and that he subjectively expected to survive were amply supported by the police reports that Critchlow “ha[d] escape measures built into his binds” (Hetzke Report at 1); that “these cords had evidently been attached to a set of counter weights which were meant to give him an ‘out’ if he started to lose consciousness” (Kuntz Report at 2); and that just hours before his death Critchlow had shopped at a grocery store and left items on the kitchen counter indicating that he was “planning supper” (Hetzke Report at 2). Given this record, no rational factfinder could fail to conclude that Critchlow intended and expected to survive his February 27, 1999 episode of autoerotic asphyxiation.
*260Nor do we see any evidence in the record from which a rational factfinder could find that Critchlow’s subjective intent to survive was objectively unreasonable. Parenthetically, we note that the district court’s statement that there was “no evidence in the record that would support a finding that decedent’s death was not the result of an intentionally self-inflicted injury,” 198 F.Supp.2d at 327 (emphases added), appears to have misallocated the burden of proof. As discussed above, an insurer that wishes to defend its denial of benefits by relying on an exclusion clause has the burden of proving that that clause, strictly construed, is applicable. The UNUM Policy at issue here covered accidental losses, and as discussed in Part II.B.l. below, the loss here was accidental. It was thus incumbent on UNUM to show affirmatively that Critchlow’s accidental death was the result of an intentionally self-inflicted injury; the burden was not on plaintiff to show that it was not.
As to the objective reasonableness of Critchlow’s expectation, although the district court stated the conclusion that “deliberately strangling ” oneself constitutes an intentional self-infliction of injury, id. at 326 (emphasis added), that statement tends to merge the concepts of intent and result. To the extent that the court was referring to total strangulation, the court focused on the result of Critchlow’s activities, not on his intent. The above evidence entirely refutes any suggestion that that result was what Critchlow intended. Moreover, the district court itself acknowledged that Critchlow had not intended total strangulation, as it noted that he had set up a complicated escape mechanism “to ensure that he did not die of asphyxiation,” id. at 320.
To the extent that the court instead meant to endorse the proposition that “ ‘[pjartial strangulation is an injury in and of itself,’ ” 198 F.Supp.2d at 323 (quoting Sims, 960 F.2d at 480 (emphasis ours)), the court ignored, inter alia, the fact that Critchlow’s death was not caused by “partial” strangulation but by the total loss of oxygen for a sustained period. To the extent that the district court here adopted the view that partial strangulation is an injury in and of itself, it ignored the apparently well-accepted medical and scientific views that the physiological effects of partial strangulation without loss of consciousness — absent an accident — are a temporary lightheadedness and euphoria with no serious or lasting adverse impact on one’s health, see Padfield, 290 F.3d at 1126, and that autoerotic asphyxiation is not likely to result in death, id. at 1127 (“uniform medical and behavioral science evidence indi-eat[es] that autoerotic activity ordinarily has a nonfatal outcome”).
In addition, Critchlow’s own experiences with autoerotic asphyxiation indicated that his subjective belief that he would survive was objectively reasonable. Critchlow’s father’s affidavit indicated that Critchlow had practiced autoerotic asphyxiation at least as early as the age of 12 or 13. By age 32, when he died, therefore, Critchlow apparently had engaged in that activity for some two decades. UNUM, in the district court, conceded “Critchlow’s long experience as a practitioner of autoerotic asphyxiation.” (UNUM Reply Memorandum in Further Support of Defendant’s Motion for Summary Judgment and in Opposition to Plaintiffs Cross-Motion for Summary Judgment at 3.) Thus, Critchlow had survived the practice of autoerotic asphyxiation for some 20 years, with no evidence of any injurious effects; and on the night of his death he had, by all accounts, set up an elaborate escape mechanism designed to save him if he began to lose consciousness. His subjective belief that he would survive what turned out to be his final episode cannot be considered objectively unreason*261able. The district court stated that Critchlow “may have thought he could free himself before he lost consciousness, but he was wrong.” 198 F.Supp.2d at 323. Wrong he was; but his being “wrong” in being confident of his safety forecloses the conclusion that his injury was intentional.
Finally, we note that the district court refused to consider the Hucker and Greendyke Opinions and affidavits on the ground that plaintiff had not submitted them to UNUM during the administrative proceedings. See 198 F.Supp.2d at 322. Although we have noted that a district court ordinarily ought not, without a showing of good cause, accept additional evidence that was not in the administrative record, Zervos v. Verizon New York, Inc., 277 F.3d 635, 646 (2d Cir.2002), “[w]e also have implied that the district court should not accept information from the insurer that was not part of the original record unless the insured is instrumental in causing the information to be added to the record,” id. (citing Zuckerbrod v. Phoenix Mutual Life Insurance Co., 78 F.3d 46, 50 n. 2 (2d Cir.1996)) (emphases added). Further, we have held that additional evidence may be considered by the district court on de novo review of an issue of plan interpretation, see Masella, 936 F.2d at 103-05, and have pointed out that, while “a claimant might be obligated to present to plan administrators all evidence regarding the claimant’s condition,” he “would be unfairly burdened by an obligation to present expert testimony to plan administrators on an issue of plan interpretation,” DeFelice v. American International Life Assurance Co., 112 F.3d 61, 65 (2d Cir.1997) (discussing Masella, 936 F.2d at 105). In the present case, we would have expected the district court to follow the Zuckerbrod and Masella principles, given that the Opinions (a) were prepared for plaintiff, (b) were first submitted to the court by UNUM, not by plaintiff, and (c) were thereafter submitted by plaintiff as well. Moreover, there was no question as to the facts and circumstances of Critchlow’s death; the Opinions dealt with whether the UNUM Policy could properly be interpreted to exclude that death. No other medical or scientific evidence was formally proffered by either side, and it seems odd that the district court elected to deprive itself of all expert assistance directed specifically at the present Policy, especially when the identical expert reports were proffered by both sides.
Plaintiff contends that the district court’s exclusion of the Opinions constituted an abuse of discretion; UNUM, having first offered the Opinions but having prevailed without them, now endorses their rejection. We conclude that we need not rule on this question, however, for even without admission of the Opinions themselves, the record in this case was sufficient both to carry plaintiffs burden of showing that Critchlow’s death was accidental (see Part II.B.l. below) under the standard expressly invoked by UNUM itself and to show that UNUM failed to carry its burden of showing that Critch-low’s death fell within the exclusion for intentionally self-inflicted injuries.
Although the district court excluded the reports of Hucker and Greendyke, a significant part of Hucker’s views was already, in essence, part of the administrative record by reason of UNUM’s invocation of Bennett v. American International Life Assurance Co. of New York, 956 F.Supp. 201, 211 (N.D.N.Y.1997) (“Bennett ”), as a basis for its rejection of plaintiffs administrative appeal. In denying that appeal, UNUM stated that it relied on “the holding and reasoning of the United States District Court for the Northern District of New York in Bennett v. American International Life Assurance Company of New *262York, 956 F.Supp. 201 (N.D.N.Y.1997),” and on “the legal standard to be applied in ERISA benefit cases, as stated in the Bennett decision.” (UNUM December 1999 Letter at 1.)
The Bennett opinion reveals that in that case, like this one, Hueker gave an expert opinion on, inter alia, the general likelihood of death as a result of autoerotic asphyxiation. In the Hueker Opinion excluded by the district court here, Hueker described the nature of autoerotic asphyxiation and stated, inter alia, that “[t]he individual has no wish to die but rather, to survive and enjoy the experience again. Fortunately for most the experience of hypoxia may be survived with no obvious impairment or injury.” (Hueker Opinion at 4; see also id. at 6 (“autoerotic asphyxial episodes do not, inevitably, or even substantially likely, lead to a fatal outcome”)). In the opinion given by Hueker in the Bennett case, he stated, inter alia, that “ ‘most practitioners likely survive the experience and expect to repeat it again.’ (Hueker Aff., Exh. 2),” Bennett, 956 F.Supp. at 211. Thus, a key aspect of the Hueker Opinion, touching on both the unlikelihood of death and the practitioner’s normal expectation of survival, was part of the Bennett opinion invoked by UNUM and hence was, in effect, part of the administrative record.
The Hueker expert opinion as quoted in Bennett is entirely consistent with the “uniform medical and behavioral science evidence” referred to in Padfield, “indicating that autoerotic activity ordinarily has a nonfatal outcome,” 290 F.3d at 1127. That scientific evidence, the evidence that Critchlow had had a long history of autoer-otie asphyxiation without any apparent ill effects, and the evidence that he had created escape mechanisms on the night in question, compel the conclusion that his expectation of survival on that night was objectively reasonable as a matter of law.
In sum, no scientific evidence before the court indicated that autoerotic asphyxiation, if practiced without accident, constitutes an injury rather than simply producing a temporary lightheadedness that the practitioner believes will increase his sexual gratification; no evidence indicated that one engaging in that practice expects to die, rather than to survive the experience and repeat it again. Such non-serious, temporary changes in condition are not what persons of reasonable or average intelligence and experience would ordinarily understand to be meant by “injuries” in the phrase “loss ... caused by ... intentionally self-inflicted injuries.”
Although the district court emphasized that the practice of autoerotic asphyxiation is “risk[y],” 198 F.Supp.2d at 323, 325, 327 n. 6, and UNUM argues that “it was unreasonable for Mr. Critchlow to believe that autoerotic asphyxiation did not present a risk of serious injury or death” (UNUM brief on appeal at 27 (emphasis added)), the language of the Policy provision relied on by UNUM does not exclude activities that are simply hazardous. Such activities as cliff rappelling, rock climbing, and sky-diving are dangerous activities; and sometimes the precautions taken prove to be insufficient, and the result is death. UNUM appears to concede that death resulting from such “extreme-sport” activities would not be excluded under the present Policy, but it attempts to distinguish those activities by arguing that they involve “controlled risks” (id. at 17 (emphasis in original)) and that “[s]kydivers and rockclimbers do not set out to injure themselves” (id. at 18 (internal quotation marks omitted)). These arguments provide no meaningful distinction. As discussed above, practitioners of autoerotic asphyxiation, wishing to survive the expe*263rience and to repeat the process, create escape mechanisms precisely to control the risk; the effect sought by the practice of autoerotic asphyxiation is a temporary effect that, absent an accident, is not injurious; and injury resulting from that practice is unintended. We thus see no differences between engaging in autoerotic asphyxiation and engaging in hazardous extreme-sport activities insofar as the coverage provided by the UNUM Policy is concerned. Given the language of the exclusion at issue here, it is not sufficient merely that a serious risk was willingly undertaken, so long as injury was not intended and, objectively, was not likely to occur.
We conclude that the UNUM Policy provision excluding loss resulting from “intentionally self-inflicted injuries” does not unambiguously apply to a death resulting from autoerotic asphyxiation. And to the extent that that phrase, or the term “injuries” within it, is ambiguous, it must, in accordance with ERISA principles, be construed against UNUM. The district court should have rejected as a matter of law UNUM’s contention that it properly denied plaintiff benefits on the basis of the exclusion for “intentionally self-inflicted injuries.”
B. UNUM’s Alternative Contentions
UNUM also contends that the dismissal of the complaint may be upheld on either of two grounds not reached by the district court, to wit, (1) that Critehlow’s death was not “accidental” or (2) that it fell within a policy exclusion for a loss caused by “illness or disease.” (UNUM brief on appeal at 28.) We reject both contentions.
1. The Contention that Critchlow’s Death Was Not “Accidental”
In denying plaintiffs initial application and her appeal, UNUM asserted the position that Critchlow’s death was not the result of an accident. As indicated in Parts I.A. and II.A. above, UNUM’s letter denying plaintiffs appeal stated that UNUM relied on the holding, reasoning, and legal standard adopted in Bennett, 956 F.Supp. 201 (See UNUM December 1999 Letter at 1).
In Bennett, the plaintiffs husband had died during autoerotic asphyxiation, and the insurer refused to pay benefits under an ERISA-regulated accidental-death insurance policy, contending that the insured’s death was not “accidental.” 956 F.Supp. at 202. The Bennett court reviewed numerous cases, including those decided under state law, such as Sigler and Sims, and those decided under ERISA, such as Todd (dealing with autoerotic asphyxiation) and Wickman v. Northwestern National Insurance Co., 908 F.2d 1077, 1088 (1st Cir.1990) (dealing with a decedent who climbed over a bridge guard rail, held on with only one hand, and fell, later dying from his injuries). The Bennett court concluded that the following legal standard, an amalgam of the principles stated by Todd, see 47 F.3d. at 1456, and Wickman, see 908 F.2d at 1088, was appropriate for ERISA cases:
For death under an accidental death policy to be deemed an accident, it must be determined (1) that the deceased had a subjective expectation of survival, and (2) that such expectation was objectively reasonable, which it is if death is not substantially likely to result from the insured’s conduct.
Bennett, 956 F.Supp. at 211 (emphasis in original). The court also noted that the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, commonly referred to as “DSM-IV,” indicates that deaths from the practice of autoerotic asphyxiation are ordinarily accidental:
*264[T]he DSM-IV states that “[b]eeause of equipment malfunction, errors in the placement of the noose or ligature, or other mistakes, accidental deaths sometimes occur.” DSM-IV § 302.83, at 529 (emphasis added). Thus, according to the DSM-IV death sometimes does not occur, and ivhen it does it is considered accidental.
Bennett, 956 F.Supp. at 211 (second emphasis ours).
In Bennett, the court ruled “as a matter of law, that Mr. Bennett had a subjective expectation of survival.” Id. However, it noted that at the scene of his death “no ‘mechanism’ for escape was apparent,” id.; and the defendant insurer produced an expert witness who opined that Mr. Bennett’s death was not accidental, see id. at 212. The district court thus concluded that the objective reasonableness of Mr. Bennett’s expectation of survival could not be decided as a matter of law. See id.
In the present case, in contrast, it is undisputed that Critchlow had in fact set up an elaborate escape mechanism. And unlike the insurer in Bennett, UNUM produced no expert opinion stating that Critchlow’s death was anything other than unintended and accidental.
We need not decide whether the precise formulation used in Bennett is to be applied in all ERISA cases. We conclude only that it is not inconsistent with federal common law, and that, UNUM having expressly adopted it in denying plaintiff benefits, it is the appropriate standard to apply to UNUM’s contention that Critchlow’s death was not accidental. And having determined in Part II.A. above that Critch-low had a subjective expectation of survival and that his expectation was objectively reasonable because death was not likely to result, we must conclude that Critchlow’s death was accidental under the Bennett standard.
Accordingly, given the present record and the Bennett standard invoked by UNUM in denying benefits, we conclude that UNUM’s denial of benefits on the basis that Critchlow’s death was not within the scope of the Policy because it was not “accidental” was a wrongful denial as a matter of law.
2. The Contention that Critchlow’s Death Resulted from “Illness” or “Disease”
Finally, we reject UNUM’s contention that Critchlow’s death by autoerotic asphyxiation fell within the Policy’s exclusion for loss caused by “illness” or “disease” (Policy at 2). The only authorities cited by UNUM for this proposition— and dealing with this practice — are DSM-IV, which, UNUM submits, classifies autoerotic asphyxia as a “mental disorder,” and Beare v. Prudential Insurance Co., 66 A.D.2d 936, 411 N.Y.S.2d 442 (3d Dep’t 1978), which affirmed a jury verdict that an autoerotie asphyxiation death resulted from “mental infirmity or disease.” We find neither source pertinent here.
Beare is a case decided under New York law, not under ERISA. Moreover, it allowed exclusion for, inter alia, a “mental infirmity.” As discussed in Part II.A. above, the language of exclusionary causes is to be strictly construed. To equate the terms “illness” or “disease,” which appear in UNUM’s exclusion clause, with “infirmity,” which may mean simply “feebleness,” “frailty,” or “an imperfection or weakness,” Webster’s New International Dictionary 1275 (2d ed.1957), would expand the meaning of the exclusionary clause, in contravention of controlling ERISA principles. We thus cannot conclude that Beare provides any support for UNUM’s present attempt to invoke the exclusion for loss caused by “illness” or “disease.”
*265Nor do we see that the mere presence of autoerotic asphyxiation as an entry in DSM-IVs catalogue of disorders means that one who engages in that practice suffers from an “illness” or “disease.” UNUM has pointed to nothing in that work that either equates the term “disorder” with “illness” or “disease” or states that illness and disease encompass all disorders. And although some disorders may properly be classified as illnesses or diseases, DSM-IV plainly includes some disorders that cannot, such as “Mathematics Disorder,” see DSM-IV at 50 (“mathematical ability (as measured by individually administered standardized tests ...) that falls substantially below that expected for the individual's chronological age”); and “Disorder of Written Expression,” see id. at 51, 411 N.Y.S.2d 442 (similarly assessed worse-than-expected writing ability); and “Caffeine-Induced Sleep Disorder,” see id. at 212, 411 N.Y.S.2d 442. UNUM has pointed us to nothing that shows that one who practices autoerotic asphyxiation suffers from an “illness” or a “disease,” rather than simply a “disorder,” and the Policy may not, consistent with ERISA principles, be expanded to encompass mere disorders.
CONCLUSION
We have considered all of UNUM’s contentions and find them to be without merit. UNUM has not argued on this appeal that there is any genuine issue of material fact to be tried. Accordingly, for the reasons stated above, we reverse the grant of summary judgment in favor of UNUM, as well as the denial of plaintiffs cross-motion for summary judgment, and we remand for entry of judgment in favor of plaintiff.