concurring in part and dissenting in part:
I join Parts I and II of the majority opinion, specifically concurring in the holding in Part II that, for the reasons there given, district courts must have discretion to receive new evidence in appropriate circumstances when conducting de novo review of plan administrators’ denials of ERISA-governed employee benefits, and that the district court did not abuse its discretion in receiving new evidence in this case.
I dissent from the holding in Part III which affirms the district court’s award of insurance proceeds under the ERISA-gov-erned accidental death insurance policy here in issue. I believe the district court failed properly to apply the test of insurance coverage dictated by Adkins v. Reliance Standard Life Insurance Co., 917 F.2d 794 (4th Cir.1990), and that a remand for its proper application should be ordered.
*1035Although I would order a remand whose result conceivably could moot the attorney fee and post-judgment interest issues dealt with in Parts IV and V of the majority opinion, I concur in the analyses and holdings of those Parts as properly establishing the law of the case on those issues.
-o-
As indicated, I disagree with the majority’s conclusion that the district court properly found on its de novo review of the insurer’s denial of the claim for insurance benefits, that on the facts and under the insurance policy’s critical coverage provision, the claimant was entitled to the death proceeds.
The critical coverage provision insures against “loss resulting directly and independently of all other causes from bodily injuries caused by accident.” The specific coverage issue raised by the facts here involves application of this provision to a situation where an insured’s death may have resulted not alone from “injuries caused by accident,” but to some degree from the interaction of those accident-caused injuries with a pre-existing physical condition. As the majority recognizes, we have a rule in this circuit that governs judicial inquiry into the coverage issue raised by such situations. As properly derived by the majority from our decision in Adkins v. Reliance Standard Life Insurance Co., 917 F.2d 794, 797 (4th Cir.1990), it mandates in cases of death a two-step inquiry: first, whether the insured-deceased had a pre-existing disease, predisposition, or susceptibility to injury; and, second, whether that pre-existing disease, predisposition, or susceptibility substantially contributed to the death. Op. 1028.
This inquiry, particularly its second prong, which requires determining the “substantiality” of a pre-existing physical condition as a contributing cause of death, obviously is a difficult one. But it is critical to proper application of the Reliance Standard rule, which was itself based upon the holding in Colonial Life & Accident Insurance Co. v. Weartz, 636 S.W.2d 891 (Ky.Ct.App.1982), which emphasized that
[A] “predisposition” or “susceptibility” to injury ... does not necessarily amount to a substantial contributing cause. A mere “relationship” of undetermined degree is not enough.
Id. at 814.
. Purporting to apply the Reliance Standard test, the district court indicated an understanding of it under which no predisposition or susceptibility could be a substantially contributing cause. In finding coverage on the facts of record, the court said
As the finder of fact, I see no legal significance between the evidence presented by each side as to the cause of Mrs. Quesinberry’s death. Both versions demonstrate that Mrs. Quesinberry had a pre-existing condition which made Reno-grafin 60 toxic to her. It is clear that Renografin 60 is a standard dye used thousands of times for radiographic studies without adverse effects on the patient. Thus, whether the adverse reaction to the dye was due to some unspecified condition which caused a sudden drop in her calcium level or whether it was due to an undiagnosed and rare form of sarcoidosis of the central nervous system is of little moment. Either version would fall within the classification of a “pre-disposition or susceptibility” and, therefore, would not be a substantially contributing factor under the Reliance Standard test.
Quesinberry v. Life Ins. Co. of North America, No. 86-0250-R, slip. op. at 7 (W.D.Va. Oct. 2, 1992) (emphasis added).
In effect, the district court stopped inquiry after finding that Mrs. Quesinberry had a pre-existing susceptibility to injury from the injection of Renografin 60. Proper application of the Reliance Standard test required that it proceed to the second inquiry: whether this pre-existing condition substantially contributed to her death, in which case coverage was properly denied, or was at most a “mere relationship, of undetermined degree,” in which case coverage was improperly denied.
*1036The majority concedes the difficulty posed by the district court’s statement of its apparent understanding of the Reliance Standard test, but reasons that the court couldn’t have meant what it said in view of what it then did. Op. 1028.
With all respect, I don’t think we properly can proceed that way. Obviously, we should simply overlook plain legal slips of the tongue when we can be sure they are that. Our business doesn’t include insuring doctrinal purity in everything said by trial judges acting under pressures we don’t experience, but only insuring — as far as we are able — that principle was properly applied. But when the record at best leaves matters unclear, I think we are obliged in fairness to assume legal misapprehension, and unless it appears harmless, at least to remand to make sure. That is the way I assess the record here. It can best be read in its totality as confirming that what the court said in the quoted portion of its memorandum opinion is a perfect reflection of its understanding of the rule and its proper application.
There is no specific analysis in the court’s opinion of the causation issue other than that reflected in the above-quoted portion. The majority points to the summary conclusion that “Mrs. Quesinberry’s preexisting condition ... was not a substantial cause in bringing about her death.” Op. 1028. But the recitation of this conclusion can’t do the service the majority assigns it. Best indications are that this conclusion, although it tracks the Reliance Standard formulation, was based on a perception that no pre-existing condition could be a substantial cause, and not on any independent inquiry into the substantiality of this particular condition as contributing cause.
Even if it be thought not plain that flat legal misapprehension was at work here, it is at least sufficiently a possibility that I think we should take the precaution of remanding to make sure. As earlier noted, a factual inquiry into substantiality of cause is necessarily a difficult undertaking. But the insurer here is entitled — just as is the ERISA-favored claimant — to have that difficult inquiry properly made and documented according to controlling law. It is of course a possibility that, as the insurer presumably determined, the pre-existing condition here was, under the law, a bar to recovery of these insurance proceeds, however one’s sympathies may lie.
I would remand for reconsideration of the coverage issue on the existing or a reopened record in the district court’s discretion.