concurring in part and dissenting in part.
I concur in part and dissent in part. Although I agree with much of the majority opinion, I disagree with Part II.B thereof, in which the majority discusses whether MetLife erred in denying the LaAsmars’ claim for accidental death benefits. As I shall outline in greater detail below, Mark LaAsmar’s death was not the result of an “accident,” and I would thus reverse the decision of the district court and remand with directions to enter summary judgment in favor of defendants.
According to the Summary Plan Description (SPD), the triggering event for payment of “Accidental Injury Benefits” was the occurrence of an “accident.” App. at 82. The SPD did not, however, define the terms “accident” or “accidental.”
Because the Plan in this case “was established under ERISA, federal common law rules of contract interpretation” must be applied in determining the meaning of any undefined Plan term. Santaella v. Metro. Life Ins. Co., 123 F.3d 456, 461 (7th Cir.1997); see Miller v. Monumental Life Ins. Co., 502 F.3d 1245; 1249 (10th Cir. 2007). “[Ajpplying federal common law, ... the proper inquiry is not what [the Plan administrator] intended a term to signify; rather, we consider the ‘common and ordinary meaning as a reasonable person in the position of the [plan] participant ... would have understood the words to mean.’ ” Miller, 502 F.3d at 1249 (quoting Admin. Comm. of Wal-Mart Assocs. Health & Welfare Plan v. Willard, 393 F.3d 1119, 1123 (10th Cir.2004) (internal quotation marks omitted)). Any ambiguities “must be construed against [the plan administrator] in accordance with the doctrine of contra proferentem.” Id. at 1253 (italics in original).
The term “accident” is commonly and generally defined as “[a]nything that happens without foresight or expectation; an unusual event, which proceeds from some unknown cause, or is an unusual effect of a known cause; ... the unforeseen course of events.” Oxford English Dictionary (2d ed.1989). Thus, as suggested by defendants, it is clear that the term, by common definition, necessarily indicates a lack of foreseeability on the part of the person involved in the accident. See Santaella, 123 F.3d at 462 (“[W]e treat the term ‘accidental’ as it is commonly defined, as ‘unexpected or unintentional’ ”). Indeed, in determining “whether a certain result is accidental” in the context of a dispute involving insurance coverage, “ ‘it is customary to look at the casualty [or injury] from the point of view of the insured.’ ” Id. (quoting Appleman, Insurance Law and Practice § 360, at 452-53 (1981)).
Having determined that we must examine the injury from the point of view of the *818insured, the next question is how, precisely, to formulate the foreseeability test. A review of similar ERISA cases reveals three possible approaches. The first, and most narrow, approach to foreseeability asks simply whether the insured subjectively expected to die or be injured. See Todd v. AIG Life Ins. Co., 47 F.3d 1448, 1455-56, n. 8 (5th Cir.1995) (discussing, but not adopting, approach utilized by court in Parker v. Danaher Corp., 851 F.Supp. 1287 (W.D.Ark.1994)). If it is determined that the insured subjectively expected not to die or be injured, then the injury or death is deemed “accidental.” See id. The second and third approaches also begin by examining whether the insured subjectively expected not to die or be injured, but add an additional step, i.e., assessing whether the insured’s subjective expectation was objectively reasonable (or, if the insured’s subjective expectation cannot be determined, how a reasonable person in the shoes of the insured would have viewed the likelihood of injury or death). See id. at 1456 (adopting second approach); Sigler v. Mut. Benefit Life Ins. Co., 663 F.2d 49, 49 (8th Cir.1981) (non-ERISA case adopting third approach). In conducting this additional step, both the second and third approaches analyze objective reasonableness from the perspective of “a reasonable person” “with background and characteristics similar to the insured.” Padfield v. AIG Life Ins. Co., 290 F.3d 1121, 1126 (9th Cir.2002). The second and third approaches differ, however, in precisely how likely an injury or death must be to render it foreseeable, and thus not accidental. Under the second approach, an expectation of survival (or non-injury) is objectively reasonable if death (or injury) was not “substantially certain” or “highly likely” to occur as a result of the insured’s intentional conduct. Todd, 47 F.3d at 1456 (utilizing “substantially certain” test); Wickman v. Nw. Nat’l Ins. Co., 908 F.2d 1077, 1088 (1st Cir.1990) (utilizing “highly likely to occur” test). Under the third approach, death or injury is not considered accidental if a reasonable person in the insured’s position would have recognized that his conduct could result in death or injury. Sigler, 663 F.2d at 49.
In my view, there are a host of reasons favoring adoption of the second, or middle, approach. To begin with, the first approach, though certainly the most favorable to the insured, has not been argued by the LaAsmars in this case, and has, as far as I can determine, only been adopted by a single federal district court. Further, the first approach appears problematic because it is often difficult or impossible to determine the subjective expectations of the insured, and even if the insured’s subjective intent can be determined, the first approach could lead to wildly varying results in cases involving similar policies and circumstances. As for the third approach, not only has it failed to become widely adopted, it is most favorable to the defendants (who have not even argued in favor of the approach), and thus is contrary to the doctrine of contra proferentem. That leaves the second approach, which has been widely adopted in cases involving insurance contracts governed by ERISA, and could fairly be said to be the most rational and reasonable of the three approaches. See Padfield, 290 F.3d at 1127 (concluding “that the ‘substantially certain’ test [wa]s the most appropriate one, for it best allows the objective inquiry to ‘serve [] as a good proxy for actual expectation.’ ”) (quoting Wickman, 908 F.2d at 1088).
That leaves the ultimate question of whether, applying the second approach to the circumstances presented in this case, Mark LaAsmar’s death was “accidental.” “[A]s is usually the case” in circumstances where the insured has died, the record *819“evidence is not sufficient to ascertain with certainty the subjective expectation” of Mark LaAsmar, i.e., whether he expected to be injured or killed by driving while so heavily intoxicated and without wearing a seatbelt. Santaella, 123 F.3d at 462. Thus, we must “ask whether a reasonable person, with background and characteristics similar to the insured, would have viewed [injury or death] as highly likely,” Wickman, 908 F.2d at 1088, or “substantially certain,” Todd, 47 F.3d at 1456, to occur as a result of Mark LaAsmar’s conduct.
The autopsy performed on Mark LaAsmar indicated that his blood alcohol content (BAC) at the time of his death was 0. 227g/100ml, an amount nearly three times greater than Colorado’s legal blood alcohol limit of 0.08g/100ml. Other federal courts have, in similar circumstances, referred to readily available public information, including various on-line resources, regarding the effects of such a BAC level. E. g., Stamp v. Metro. Life Ins. Co., 531 F.3d 84, 90 (1st Cir.2008) (citing Nat’l Hwy. Traffic Safety Admin., U.S. Dep’t of Transp., Setting Limits, Saving Lives: The Case for .08 BAC Laws, DOT HS 809 241, revised Apr. 2001). Consulting such resources in this case, there appears to be no question that a BAC level of 0.227^ 100ml would have resulted in severe impairment of Mark LaAsmar’s ability to drive and, correspondingly, would have substantially increased the likelihood of him crashing his vehicle while driving.1 For example, according to a review of research literature available on the National Highway Traffic Safety Administration’s (NHTSA’s) web site, there is “strong evidence that impairment of some driving-related skills begins with any departure from zero BAC,” and that “[v]irtually all subjects tested in the studies reviewed ... exhibited impairment on some critical driving measure by the time they reached 0.080g/dl.” Nat’l Hwy. Traffic Safety Admin., U.S. Dep’t of Transp., A Review of the Literature on the Effects of Low Doses of Alcohol on Driving-Related Skills, Section 4.1, Apr. 2000 (available at http://www. nhtsa.dot.gov/people/injury/research/pub/ Hs809028/Title.htm). Relatedly, another publicly-available NHTSA publication states that a BAC of .10% results in “[Reduced ability to maintain lane position and brake appropriately,” and that a BAC of .15% results in “[substantial impairment in vehicle control, attention to driving task, and in necessary visual and auditory information processing.” Nat’l Hwy. Traffic Safety Admin., U.S. Dep’t of Transp., The ABCs of BAC, Feb. 2005 (available at http://www.stopimpaireddriving.org/ABCs BACWeb/index.htm). In turn, “[t]he risk of a fatal crash increases rapidly as the blood alcohol concentration of a driver increases.” Robert D. Brewer et al., The Risk of Dying in Alcoholr-Related Automobile Crashes among Habitual Drunk Drivers, 331 The New Eng. J. Med., 513-517 (August 25, 1994). “A driver with a blood alcohol concentration of 100 mg per deciliter (22mmol per liter) or higher is 7 times more likely to be involved in a fatal motor vehicle crash than a driver who has not consumed alcoholic beverages, and a driver with a blood alcohol concentration of 150 mg per deciliter (33 mmol per liter) or more is about 25 times more likely.” Id. “For drivers with [B]AC’s above 0.15% on weekend nights, the likelihood of being *820killed in a single-vehicle crash is more than 380 times higher than it is for non-drinking drivers.” Wis. Dep’t of Transp., Safety & Consumer Protection, Drunk Driving Risk Factors, available at http://www.dot. wisconsin.gov /safety/motorist/drunkdriving/factors.htm.
In light of such widely available and generally well-publicized data, it is clear that an objectively reasonable person in Mark LaAsmar’s position would have viewed driving with a BAC of 0.22%, late at night on a two-lane county road, at sixty miles per hour in a forty mile-per-hour zone, and without a seat belt, as highly or substantially likely to result in serious injury or death.2 Thus, it is likewise clear that Mark LaAsmar’s death cannot be classified as “accidental” for purposes of the Plan at issue. As the Fourth Circuit has noted, “[b]y choosing to drive under circumstances where his vision, motor control, and judgment were likely to be impaired,” Mark LaAsmar “placed himself and fellow motorists in harm’s way,” and “[t]o characterize harm flowing from such behavior as merely ‘accidental’ diminishes the personal responsibility that state laws and the rules of the road require.” Eckelberry v. Reliastar Life Ins. Co., 469 F.3d 340, 346 (4th Cir.2006).
Notably, federal courts performing this same type of analysis “have found with near universal accord that alcohol-related injuries and deaths are not ‘accidental’ under insurance contracts governed by ERISA.” Id. at 344 (citing cases). In doing so, “[tjhese courts have ... reasoned that since the hazards of drinking and driving are widely known and widely publicized the insured should have known that driving while intoxicated was highly likely to result in death or bodily harm.” Id. at 345 (internal quotation marks omitted).
In reaching a different conclusion, the majority opinion states that “[mjost people ... would define accident to include many circumstances where a driver undertakes conduct that makes a crash more likely, such as driving when sleepy or when the weather is bad, talking on the cell phone, reaching for a compact disc, or turning to speak to a child while operating a vehicle.” Maj. Op. 807. In turn, the majority opinion suggests that some “of these volitional acts increase! ] the probability of a wreck ... to an even greater degree than driving drunk.” Id. Finally, the majority opinion asserts that “[sjomewhere in the middle of this spectrum of circumstances falls Mark LaAsmar’s decision to drive home in the early morning darkness on two-lane country roads, with a BAC of .227 and going sixty miles an hour in a forty-mile-per-hour zone.” Id. at 807-08. I strongly disagree with this analysis.
To begin with, the examples of driving-related conduct cited by the majority opinion are quite vague, and it is the precise circumstances of each case that, in the end, determine the foreseeability of the risk undertaken by a driver by engaging in a particular type of conduct. While I do not disagree that at least some of the general categories of causal conduct cited by the majority could reasonably be deemed “accidental,” I submit that each of those categories involve far less of a risk of negative consequences than did the reckless conduct engaged in by Mark LaAsmar immediately prior to the crash that took his life.
Relatedly, I also reject the majority opinion’s suggestion that driving while talking or texting on a cell phone “increases the probability of a wreck ... to an *821even greater degree than” the conduct of Mark LaAsmar in this case. Id. at 807. A careful examination of the two authorities cited by the majority opinion firmly establishes that neither support such a proposition. For example, while the law review article cited by the majority opinion states that “ ‘the performance of drivers who are conversing on cell phones is more impaired than drivers who are intoxicated.,’ ” Douglas R. Richmond, “Drunk in the Serbonian Bog: Intoxicated Drivers’ Deaths as Insurance Accidents,” 32 Seattle U.L.Rev. 83, 86 (Fall 2008) (emphasis in original; quoting Ira H. Lessfield & Richard L. Segal, Driving While on the Cell Phone, BRIEF, Summer 2007, at 58, 59), an examination of the underlying research study that gave rise to this statement involved drivers who were given “a mixture of orange juice and vodka ... calculated to achieve a blood alcohol concentration of 0.08 wt/vol,” David L. Strayer & Frank A. Drews, Multitasking in the Automobile, www.psych.utah.edu/lab/appliedcognition/ pubhcations/multitasking.pdf. In other words, the quoted statement from the law review article was based on a comparison of driving while conversing on a cell phone versus driving with a BAC at the legal limit in most states, including Colorado. Quite obviously, this study says nothing about the relative risks of driving while conversing on a cell phone versus the conduct engaged in by Mark LaAsmar, i.e., driving with a BAC nearly three times the legal limit (and at night, on a two-lane county road, at excessive speed, and without wearing a seat belt). Similarly, the study cited by the Sixth Circuit in Kovach v. Zubrich American Insurance Company, 587 F.3d 323 (6th Cir.2009), refers to “a study of young drivers in England finding] that reaction times of young drivers were reduced by text messaging three times more than by drinking alcohol to the legal limit.” Id. at 335 (italics omitted; emphasis added). Because this study involved participants with substantially lower BACs than Mark LaAsmar, it tells us nothing about the relative risks of driving while text messaging versus driving under the precise conditions that immediately preceded Mark LaAsmar’s fatal wreck.
Finally, I reject the majority opinion’s assertion that “[sjomewhere in the middle of th[e] spectrum of circumstances [it has described] falls Mark LaAsmar’s decision to drive home in the early morning darkness on two-lane country roads, with a BAC of .227 and going sixty miles an hour in a forty-mile-per-hour zone.” Maj. Op. at 807-08. In my view, Mark LaAsmar’s conduct falls at the far end of the spectrum described by the majority. More specifically, whereas most, if not all, of the examples of conduct listed by the majority could be classified as negligent, I believe that Mark LaAsmar’s conduct was reckless or grossly negligent, e.g., Farmer v. Brennan, 511 U.S. 825, 836, 114 S.Ct. 1970, 128 L.Ed.2d 811 (1994) (“The civil law generally calls a person reckless who acts or (if the person has a duty to act) fails to act in the face of an unjustifiably high risk of harm that is either known or so obvious that it should be known.”), and indeed can be fairly compared to the majority opinion’s earlier example of an “insured [who] died as a result of playing Russian roulette,” Maj. Op. at 806.
In sum, I believe the majority opinion is wrong in affirming the district court’s conclusion that Mark LaAsmar’s death was “accidental” within the meaning of the Plan, and in turn affirming the grant of summary judgment in favor of the LaAsmars. I would reverse the judgment of the district court and remand with directions to enter summary judgment in favor of defendants.
. Although the majority opinion likewise refers to some publicly available information regarding the effects of alcohol intoxication, that cited information only appears to focus on the effects of intoxication at a level equivalent to the legal BAC limit in most states, and thus tells us nothing about the effects of the specific BAC level that Mark LaAsmar had at the time of his death.
. I am not persuaded, as suggested by the majority, that “MetLife was applying ... a per se rule based solely upon the degree of intoxication involved.” Maj. Op. at 802. In any event, given our de novo standard of review, it is unnecessary to decide whether MetLife intended such a per se rule.