Opinion for the Court filed by SILBERMAN, Circuit Judge, in which BUCKLEY, STEPHEN F. WILLIAMS, D.H. GINSBURG, SENTELLE, KAREN LeCRAFT HENDERSON, and RANDOLPH, Circuit Judges, concur.
Dissenting opinion filed by HARRY T. EDWARDS, Circuit Judge, with whom MIKVA, Chief Judge, WALD and RUTH BADER GINSBURG, Circuit Judges, join.
SILBERMAN, Circuit Judge, in which BUCKLEY, STEPHEN F. WILLIAMS, D.H. GINSBURG, SENTELLE, KAREN LeCRAFT HENDERSON, and RANDOLPH, Circuit Judges, concur:
This case concerns the circumstances in which a court making an award of reasonable attorney’s fees under federal fee-shifting statutes may augment the lodestar with a contingency enhancement designed to compensate the prevailing party’s attorney for the risk of losing the case. The panel opinion in this case, King v. Palmer, 906 F.2d 762 (D.C.Cir.1990), reviewed a district court award of attorney’s fees and costs made to the plaintiff, Mabel King, pursuant to the fee-shifting provisions of Title VII. See 42 U.S.C. §§ 2000e-5(k), 2000e-16(d).1 The panel rejected the District of Columbia’s contention that no enhancement for the risk of nonpayment was proper but set aside the district court’s award of an enhancement of 50% of attorney’s fees subject to contingency, holding instead that Ms. King was entitled to a full 100% enhancement of those fees, relying on this court’s decision in McKenzie v. Kennickell, 875 F.2d 330 (D.C.Cir.1989). On September 12, 1990, we granted the District of Columbia’s petition suggesting rehearing en banc to reconsider the holding on contingency enhancements in McKenzie. Having reviewed the issue en banc, we overrule McKenzie and reverse the award of a contingency enhancement to Ms. King.
I.
Mabel King brought a gender discrimination claim against her employer, the District of Columbia, and ultimately received an award of back pay and retroactive promotion. See King v. Palmer, 778 F.2d 878, 882 n. 7 (D.C.Cir.1985), on remand, 641 F.Supp. 186, 187-89 (D.D.C.1986). The history of the substantive litigation underlying the dispute over attorney’s fees is summarized in the panel opinion. See King, 906 F.2d at 764.
Ms. King experienced no difficulty in securing an attorney. She was represented throughout the litigation by the first attorney she contacted, Robert Adler, who took the case on a partial contingency basis. Ms. King contacted Mr. Adler as a result of his successful representation of a colleague of hers in another Title VII case, for which he had received a 10% contingency enhancement. Ms. King and Mr. Adler agreed that she would be responsible for litigation costs and expenses, as well as for fees of up to $5000, and that she would receive any award of damages, while Mr. Adler would receive any statutory attor*774ney’s fees that might be awarded, should Ms. King prevail. See id.
Mr. Adler averred that he took the case expecting that a contingency enhancement would be available. In his applications for attorney’s fees following Ms. King’s success on the merits, Mr. Adler twice requested a 35% fee bonus to compensate him for the risk of nonpayment he had borne during the litigation, but the district court held this request in abeyance pending the Supreme Court’s decision concerning the availability of contingency enhancements in Pennsylvania v. Delaware Valley Citizens’ Council for Clean Air, 483 U.S. 711, 107 S.Ct. 3078, 97 L.Ed.2d 585 (1987) (Delaware Valley II). In the interim, the district court awarded a lodestar fee totaling $232,707.62, which comprised the reasonable number of hours Mr. Adler spent on the case multiplied by a reasonable hourly rate, and noted that a “15 [percent] bonus for the risk of not prevailing would ... be appropriate in the event that such an award is authorized by the Supreme Court.” King v. Palmer, Civ. No. 83-1980, Rev.Mem. at 13, 1987 WL 12794 (D.D.C. June 10, 1987) {Mem.Op.I).
After the Court issued its fragmented decision in Delaware Valley II, Mr. Adler reapplied for a contingency enhancement, increasing his request to 100%. Reading Justice O’Connor’s concurrence in Delaware Valley II as controlling the availability and degree of contingency enhancement, the district court held that the plaintiff must establish how the market compensates for contingent cases on a class-wide basis and then show that without such enhancement she would have had substantial difficulty attracting competent counsel to her case. See King v. Palmer, Civ. No. 83-1980-LFO, Mem. at 2, 1988 WL 104970 (D.D.C. Sept. 20, 1988) (Mem.Op.II). In the district court’s view, Ms. King met these requirements by introducing affidavits from a number of local attorneys experienced in Title VII work asserting that they would not accept fee-shifting cases where fees were available only if the case was won, absent the prospect of contingency enhancements. See id. at 3. Since Ms. King had agreed to pay all costs and expenses and the first $5000 of fees, however, the district court found that Mr. Adler’s representation of her was only partially contingent and awarded a 50% enhancement instead of the 100% requested, on the authority of an earlier district court opinion in Palmer v. Shultz, 679 F.Supp. 68 (D.D.C.1988), appeal dismissed, No. 88-5108 (D.C.Cir.1988). See Mem.Op.II at 3-4. Both parties appealed.
The panel, following our previous opinion in McKenzie v. Kennickell, 875 F.2d 330 (D.C.Cir.1989), affirmed the district court’s award of a contingency enhancement but increased it from 50% of the lodestar to 100%. McKenzie established a regime in which contingency enhancements would be routinely available in statutory fee-shifting cases. In reaching this result, the McKenzie panel treated Justice O'Connor’s concurring opinion in Delaware Valley II as controlling and explicitly applied her admonition that “no enhancement for risk is appropriate unless the applicant can establish that without an adjustment for risk the prevailing party ‘would have faced substantial difficulties in finding counsel in the local or other relevant market.’ ” Delaware Valley II, 483 U.S. at 733, 107 S.Ct. at 3091 (O'Connor, J., concurring in part and concurring in the judgment) (quoting plurality opinion at 731, 107 S.Ct. at 3089). The McKenzie majority described the inquiry to be conducted under this test as “counterfactual,” meaning that plaintiffs “need not show that [they] actually experienced difficulty in obtaining representation,” but merely that, “absent a contingency enhancement, plaintiffs would have encountered substantial difficulties in finding counsel ... [at the time] they commenced their lawsuit.” McKenzie, 875 F.2d at 337 (first emphasis added, second in original). Thus, the majority concluded that a prevailing plaintiff under the typical fee shifting-statute could gain a contingency enhancement by producing affidavits from lawyers in the District of Columbia stating that those lawyers would not normally take a case on contingency unless they were paid more than their normal hourly fees if they won. Indeed, according to the majority, it *775was entirely irrelevant whether counsel in the case had been “attracted by the possibility of a contingency enhancement”; the panel dismissed as beside the point the fact that one of the lawyers who took the case, the head of the pro bono section of a major Washington law firm, candidly stated that his firm would have taken the ease even without the prospect of a contingency enhancement. Id. at 338. The majority reasoned that were it to deny a contingency enhancement on an “actual difficulty” basis, it would simply encourage a “charade” in which “public interest lawyers would accept a case only after announcing loudly that they were doing so on the assumption of a contingency enhancement.” Id. at 337-38. In short, under the McKenzie holding, even a plaintiff who easily found counsel and whose counsel presumably expected no contingency bonus could satisfy the “substantial difficulties” test.
Judge Buckley, dissenting on this issue, thought the McKenzie majority misread Delaware Valley II. He pointed out that “Justice O’Connor joined the plurality in requiring proof that the prevailing party ‘would have faced substantial difficulties’ in obtaining competent counsel ... absent an upward fee adjustment for contingency risks.” Id. at 340-41 (Buckley, J., concurring in part and dissenting in part) (quoting Delaware Valley II, 483 U.S. at 733, 107 S.Ct. at 3090) (emphasis in original). That meant, according to Judge Buckley, that we were required to pursue an “individualized approach” in which evidence of actual difficulties would be extremely important. Id. at 341.
Because the record establishes that McKenzie and his fellow plaintiffs had in fact located qualified lawyers willing to represent them in Washington, D.C. in the early 1970’s, I conclude that under Delaware Valley II these fee applicants have failed to prove that the prevailing party “would have faced substantial difficulties” in securing competent attorneys absent the incentive of an enhanced fee.
Id. (emphasis added).
In accordance with McKenzie, the panel in this case thought the failure of the district court to make a specific finding that Ms. King would have faced substantial difficulties in obtaining counsel without a risk enhancement was of no real significance. See King, 906 F.2d at 768. The district court had instead relied on Ms. King’s attorney affidavits and on a case in which a different district judge had made a blanket finding “that attorneys in the District would not accept contingent cases without some risk enhancement.” Id. (citing Mem.Op.IIat 2, 4 (citing Palmer v. Shultz, 679 F.Supp. 68 (D.D.C.1988), appeal dismissed, No. 88-5108 (D.C.Cir.1988))). The panel held that the cross-citation to the other district judge’s finding in conjunction with the affidavits filed in the case was sufficient to satisfy McKenzie’s reading of Justice O’Connor’s opinion.
We decided to rehear the case en banc in order to reconsider the interpretation of Delaware Valley II that the panel in McKenzie adopted. It is our view that the approach followed by the majority in McKenzie is, as Judge Buckley argued, a misreading of Justice O’Connor’s concurring opinion. Moreover, we do not think that Justice O’Connor’s concurring opinion in Delaware Valley II controls the issue of the circumstances under which contingency enhancements are permitted. We conclude that the fragmented decision in Delaware Valley II provides no test for determining the availability, much less the calculation, of contingency enhancements under fee-shifting statutes and that we are therefore obliged to continue the search for the most sensible rule to govern contingency enhancements. The rule we adopt is that of the Delaware Valley II plurality: a reasonable lodestar fee awarded under federal fee-shifting statutes may not be enhanced to compensate a prevailing party for his initial risk of loss.
II.
A.
Delaware Valley II has given rise to a spate of circuit court opinions that attempt — with varying degrees of confidence — to interpret the Supreme Court’s *776position. See, e.g., Rode v. Dellarciprete, 892 F.2d 1177, 1184-85 (3d Cir.1990); Student Pub. Interest Research Group v. AT & T Bell Laboratories, 842 F.2d 1436, 1451 (3d Cir.1988); Blum v. Witco Chem. Corp., 829 F.2d 367, 379-82 (3d Cir.1987); Craig v. Secretary, Dep’t of Health and Human Servs., 864 F.2d 324, 327-28 (4th Cir.1989); Spell v. McDaniel, 824 F.2d 1380, 1403-05 (4th Cir.1987), cert. denied, 484 U.S. 1027, 108 S.Ct. 752, 98 L.Ed.2d 765 (1988); Leroy v. City of Houston, 831 F.2d 576, 583-84 (5th Cir.1987), cert. denied, 486 U.S. 1008, 108 S.Ct. 1735, 100 L.Ed.2d 199 (1988); Skelton v. General Motors Corp., 860 F.2d 250, 254 (7th Cir.1988); Hendrickson v. Branstad, 934 F.2d 158, 162-63 (8th Cir.1991); D’Emanuele v. Montgomery Ward & Co., 904 F.2d 1379, 1384 (9th Cir.1990); Fadhl v. City of San Francisco, 859 F.2d 649, 650-51 (9th Cir.1988) (per curiam); Smith v. Freeman, 921 F.2d 1120, 1122-23 (10th Cir.1990); Wulf v. City of Wichita, 883 F.2d 842, 876 (10th Cir.1989); Norman v. Housing Auth., 836 F.2d 1292, 1302 (11th Cir.1988). We, like our sister courts of appeals, have struggled to take from the case a rule of law that defines the circumstances in which contingency enhancements may be awarded to the lawyers who represent prevailing plaintiffs under the myriad of federal fee-shifting statutes. See McKenzie v. Kennickell, 875 F.2d 330, 332-38 (D.C.Cir.1989); id. at 340-43 (Buckley, J., dissenting); Weisberg v. U.S. Dep’t of Justice, 848 F.2d 1265, 1272-73 (D.C.Cir.1988); Thompson v. Kennickell, 836 F.2d 616, 621 (D.C.Cir.1988); Save Our Cumberland Mountains, Inc. v. Hodel, 826 F.2d 43, 53 n. 6 (D.C.Cir.1987), vacated on other grounds, 857 F.2d 1516 (D.C.Cir.1988) (en banc).
The Supreme Court’s decisions on attorney’s fees prior to Delaware Valley II had established the lodestar — a measure of fees defined by the number of hours reasonably expended on a case multiplied by a reasonable market rate per hour — as the presumptively reasonable award, steadily subsuming most other factors into that single calculation. See Hensley v. Eckerhart, 461 U.S. 424, 433, 103 S.Ct. 1933, 1939, 76 L.Ed.2d 40 (1983); Blum v. Stenson, 465 U.S. 886, 897-902, 104 S.Ct. 1541, 1548-1551, 79 L.Ed.2d 891 (1984); Pennsylvania v. Delaware Valley Citizens’ Council for Clean Air, 478 U.S. 546, 564-66, 106 S.Ct. 3088, 3097-99, 92 L.Ed.2d 439 (1986) (Delaware Valley I). Twice, however, the Supreme Court had specifically reserved the question whether the lodestar could ever be enhanced to reflect the risk of nonpayment assumed by an attorney accepting a case under a statute that authorized fees only to the prevailing party. See Blum, 465 U.S. at 901 n. 17, 104 S.Ct. at 1550 n. 17; Delaware Valley I, 478 U.S. at 568, 106 S.Ct. at 3199-3200. Delaware Valley II attempted to resolve this issue.2
The judgment in Delaware Valley II reversed an award of a 100% contingency enhancement.3 Justice White, writing for a plurality of four Justices, concluded in Part IV of his opinion that contingency enhancements under fee-shifting statutes are simply “impermissible.” Delaware Valley II, 483 U.S. at 727, 107 S.Ct. at 3087-88 (plurality opinion). Nevertheless, the plurality went on in Part V to suggest that if contingency bonuses were to be made available at all, they “should be re*777served for exceptional cases.” Id. at 728, 107 S.Ct. at 3088. Four Justices in dissent would have allowed a contingency enhancement in any case in which “an attorney and client have been unable to mitigate the risk of nonpayment,” id. at 749, 107 S.Ct. at 3099 (Blackmun, J., dissenting), as well as “additional enhancement” in those cases posing great “ ‘legal’ risks.” Id. at 751, 107 S.Ct. at 3100. Under the dissent’s test, contingency enhancements would be “appropriate in most circumstances.” Id. at 741, 107 S.Ct. at 3095.
Justice O’Connor concurred in part and concurred in the judgment reversing the award. She agreed with the dissenters that “Congress did not intend to foreclose consideration of contingency in setting a reasonable fee.” Id. at 731, 107 S.Ct. at 3089 (O’Connor, J., concurring in part and concurring in the judgment). But she joined in the plurality’s judgment that the record before the Court did not justify a contingency enhancement. See id. at 734, 107 S.Ct. at 3091. She also agreed with the plurality that no enhancement could be awarded for the “legal” risks peculiar to the specific case. See id. at 731, 734, 107 S.Ct. at 3089-90, 3091. Finally, Justice O’Connor agreed with the statement in Part V of the plurality opinion “that no enhancement for risk is appropriate unless the applicant can establish that without an adjustment for risk the prevailing party ‘would have faced substantial difficulties in finding counsel in the local or other relevant market.’ ” Id. at 733, 107 S.Ct. at 3091 (quoting plurality opinion at 731, 107 S.Ct. at 3089).
B.
In our prior opinions interpreting Delaware Valley II, we, like other circuit courts, have assumed that Justice O’Con-nor’s concurrence controls. See McKenzie v. Kennickell, 875 F.2d 330, 332-38 (D.C.Cir.1989); id. at 340-43 (Buckley, J., dissenting); Weisberg v. U.S. Dep’t of Justice, 848 F.2d 1265, 1272-73 (D.C.Cir.1988); Thompson v. Kennickell, 836 F.2d 616, 621 (D.C.Cir.1988); Save Our Cumberland Mountains, Inc. v. Hodel, 826 F.2d 43, 53 n. 6 (D.C.Cir.1987), vacated on other grounds, 857 F.2d 1516 (D.C.Cir.1988) (en banc). But we have not focused on the fact that there are two analytically distinct questions involved in awarding a contingency enhancement. First, a court must decide whether an enhancement is available at all. Then, assuming an enhancement is warranted, the court must calculate its amount. Virtually all of Justice O’Con-nor’s relatively brief opinion deals with the second question. But the question of availability of enhancements logically precedes the question of their calculation.
To ascertain when contingency enhancements should be made available under Delaware Valley II, we have looked for some common ground between Justice O’Con-nor’s concurrence and the plurality opinion. We have had little difficulty placing a label on that common ground, since Justice O’Connor expressly joined the plurality’s statement in Part V that enhancements should be available only when a plaintiff would have faced “substantial difficulties” in attracting counsel to his case without the prospect of an enhancement. However, we and the other courts of appeals have had considerable trouble determining the content of that “substantial difficulties” label — that is, determining just how “substantial” the “difficulties” in attracting counsel have to be, and how they must be proven.
In this search for content, several of our sister circuits have read Justice O’Connor’s concurrence as implicitly agreeing with the plurality’s statement, Delaware Valley II, 483 U.S. at 727, 107 S.Ct. at 3087-88 (plurality opinion), that contingency bonuses should be available only in “exceptional cases.” See, e.g., Student Pub. Interest Research Group v. AT & T Bell Laboratories, 842 F.2d 1436, 1451-52 (3d Cir.1988) (“[Cjontingency multipliers should be granted only rarely.”); Norman v. Housing Auth., 836 F.2d 1292, 1302 (11th Cir.1988) (“[I]n the rare case enhancement may be appropriate_”). Appellant presses this position upon us here. And in Thompson v. Kennickell, we made a similar suggestion, describing Delaware Valley II in Gilbert and Sullivan terms: “What, never? *778No, never!” for the plurality, and “What, never? Hardly ever!” for Justice O’Con-nor. Thompson, 836 F.2d at 621 (emphasis in original).
To be sure, Justice O’Connor does not say at any point that she disagrees with the plurality’s “exceptional cases” position. And she does endorse the plurality’s view that the lodestar is a presumptively adequate fee. See Delaware Valley II, 483 U.S. at 733-34, 107 S.Ct. at 3090-91 (O’Con-nor, J., concurring in part and concurring in the judgment). Moreover, Justice O’Connor joined the reversal of the award of a contingency enhancement without a remand, notwithstanding the dissenters’ powerful argument that the applicant should be given an opportunity to develop the record to meet the Supreme Court’s standard. See id. at 754-55, 107 S.Ct. at 3101-02 (BJackmun, J., dissenting); cf. Thompson, 836 F.2d at 621 (remanding for application of Delaware Valley II). This at least suggests that Justice O’Connor believed that contingency enhancements should be available only in those presumably rare situations in which the need was readily apparent. Still, she did not join Part V of the plurality opinion, so we cannot be sure that she accepted the “exceptional cases” limitation.
There is only one point concerning the availability of contingency enhancements that a fair reading of Justice O’Connor’s concurrence clearly supports. This point is that evidence of actual difficulties is highly probative of the “substantial difficulties” Delaware Valley II describes. In adopting the plurality’s “substantial difficulties” test, Justice O’Connor quotes from Part V of the plurality opinion. The passage she quotes concludes with a footnote that we presume Justice O’Connor adopted along with the textual language she cited. The footnote states: “ ‘an attorney’s fee award should be only as large as necessary to attract competent counsel,’ and ‘one relevant factor bearing on high-risk is whether other counsel had declined to take the case because there was little or no prospect of earning a fee.' ” Id. at 731 n. 12, 107 S.Ct. at 3089 n. 12 (plurality opinion) (quoting Lewis v. Coughlin, 801 F.2d 570, 576 (2d Cir.1986)) (emphasis added). And, as Judge Buckley noted in McKenzie, “Justice O’Connor joined the plurality in requiring proof that the prevailing party ‘would have faced substantial difficulties’ obtaining competent counsel ‘in the relevant market,’ absent an upward fee adjustment for contingency risks.” McKenzie, 875 F.2d at 340-41 (Buckley, J., concurring in part and dissenting in part) (quoting Delaware Valley II, 483 U.S. at 733, 107 S.Ct. at 3090 (O’Con-nor, J., concurring in part and concurring in the judgment)) (emphasis in original). We thus believe that five Justices envisioned a particularized factual inquiry into the plaintiff’s actual difficulties in retaining counsel — the kind of inquiry that Judge Buckley thought necessary but the majority in McKenzie eschewed. But see Morris v. American Nat'l Can Corp., 941 F.2d 710, 715 (8th Cir.1991) (stating no actual difficulties need be shown) (citing McKenzie, 875 F.2d at 337).
This is, we recognize, a hard standard to meet. Indeed, if evidence that other counsel actually refused the case is only “one relevant factor” in determining whether the plaintiff would have had “substantial difficulties” in obtaining counsel without a risk enhancement — a factor insufficient by itself to justify awarding an enhancement — the plaintiff’s burden in producing sufficient evidence to meet the test must be quite daunting. And it also follows that a plaintiff’s failure to put on any evidence of actual difficulties in attracting counsel without extra compensation would severely undermine a claim for a contingency enhancement.
The district court here made no finding that Ms. King would have faced substantial difficulties in attracting counsel without a contingency bonus. Nor was there any evidence that Ms. King faced actual difficulties in securing representation. As it happened, Robert Adler was the first attorney the plaintiff contacted, and, although he later stated that he would not have accepted representation without the “definite possibility” of a contingency enhancement, his fee award in a previous case had *779been enhanced by only 10%. Given the uncertain state of the law at the time he took this case (which is not to say that it is particularly clear today) and his previous experience, the “definite possibility” to which he referred does not seem very weighty. In his engagement letter to Ms. King, Mr. Adler referred only to charging his “hourly rates” and stated that he would “seek an award of attorneys’ fees from the defendants with respect to those amounts, should we be the prevailing party.” Joint Appendix (J.A.) at 73a (emphasis added). It does not seem to us that Mr. Adler’s testimony goes very far to meet the plaintiffs burden under the “substantial difficulties” test.
Before the district court and again before us, Ms. King has also sought to rely on the affidavits of attorneys who were not approached by Ms. King and were never involved in the case.4 These affidavits— some from Title YII practitioners, some from practitioners from other areas — contend that lawyers would not take cases on a non-fee-paying basis without contingency enhancements. We do not think that we can accept such evidence as meeting the substantial difficulties test. Without in any way denigrating the bona fides of these lawyers, we cannot blink the fact that they are obviously self-interested. We think it is indisputable that if such evidence were treated as determinative, or even weighty, the substantial difficulties test would be met so easily as to become a mere formality. The Supreme Court has itself recently disparaged such anecdotal evidence from attorneys unconnected with the case in the context of attorney’s fees disputes. See United States Dep’t of Labor v. Triplett, 494 U.S. 715, 110 S.Ct. 1428, 1433-34, 108 L.Ed.2d 701 (1990) (holding such evidence to be “blatantly insufficient” to raise a constitutional doubt about federal limits on attorney’s fees, “even if entirely unrebutted”).
Nor do we believe the few affidavits presented that expressed a view as to whether the affiant lawyer would or would not have taken Ms. King’s case add much to her claim for an enhancement. Insofar as they seek to hypothesize whether the affiants would have taken her case, they focus (inevitably it seems to us) on the strength or weakness of her claim. See supra note 3. But in Delaware Valley II, it will be recalled, both the plurality and Justice O’Connor regarded that factor as inappropriate. See Delaware Valley II, 483 U.S. at 726, 107 S.Ct. at 3087 (plurality opinion); id. at 734, 107 S.Ct. at 3091 (O’Connor, J., concurring in part and concurring in the judgment).
In sum, even if we were to apply Justice O’Connor’s concurrence as the holding of Delaware Valley II, we think Ms. King’s evidence does not paint a picture of a situation where a contingency enhancement is necessary to “mak[e] it possible for poor clients with good claims to secure competent help.” Id. at 730-31,107 S.Ct. at 3089.
C.
Although we have determined that Ms. King failed to carry her burden under Justice O’Connor’s opinion in Delaware Valley II, candor obliges us to concede that we are unable to set forth a conceptual framework that would govern further litigation on the subject of contingency enhancements. We have certainly suggested, in accordance with our understanding of the substantial difficulties test, that actual evidence that attorneys did refuse a case is of greater probative value *780than the hypothetical testimony of non-involved and self-interested lawyers. But we are sorely troubled by, and indeed we have no answer to, the McKenzie majority’s argument that focusing on actual difficulties will encourage “a charade in which clients seeking representation under fee shifting statutes would be steered to several attorneys whose pre-arranged role it would be to ‘refuse’ the case, knowing that such refusals were necessary to permit the eventual award of fees.” McKenzie, 875 F.2d at 337. We think the McKenzie majority was also correct in suggesting that emphasizing the actual difficulties a plaintiff had in obtaining counsel will create perverse incentives by discouraging those very “reference services ... that make it easier for litigants to find legal representation.” Id.
To add to our quandary, even if we did have evidence that several lawyers had declined Ms. King’s case, we think it would be impossible to separate out from their decision not to represent her the strength or weakness of her claim as it appeared to them at the time. After all, this is surely the principal reason a lawyer will turn down a case under a fee-shifting statute. Delaware Valley II, however, tells us unequivocally that the risk of loss in a particular case is not a factor that courts may look at in determining whether a contingency enhancement is appropriate. “[A] court should not award any enhancement based on ‘legal’ risks or risks peculiar to the case.” Delaware Valley II, 483 U.S. at 734, 107 S.Ct. at 3091 (O’Connor, J., concurring in part and concurring in the judgment); see also id. at 726-27, 107 S.Ct. at 3087-88 (plurality opinion).5 If the courts cannot do so directly, how can it be appropriate to do so vicariously through the eyes of lawyers who declined the case?
The more we struggle with this problem, the more we are convinced that it is virtually impossible to determine whether a given plaintiff would have had “substantial difficulties” in obtaining counsel without a contingency enhancement. The inquiry is quite artificial, because, by definition, the plaintiff stands before the court with counsel. And since counsel could not possibly know whether a risk enhancement was in the offing until a court decides the question years later, our inquiry is circular. As Judge Williams noted in his concurrence in the panel opinion in this case, whether a plaintiff would have faced substantial difficulties absent the possibility of a contingency enhancement is essentially unknowable when the most critical assumption necessary to make such a counterfac-tual judgment is itself the issue before the court. See King, 906 F.2d at 770 (Williams, J., concurring) (“I view causation as running in the opposite direction from that supposed by the controlling precedents; I see the judicial judgment as defining the market, not vice versa.”).
As Judge Buckley correctly observed in McKenzie, if Justice O’Connor’s opinion controls the holding of Delaware Valley II, we would be obliged to apply the “substantial difficulties” test, notwithstanding these analytical difficulties, “like it or not.” McKenzie, 875 F.2d at 342 (Buckley, J., concurring in part and dissenting in part). But the difficulties we have described have prompted us to think harder about what the controlling principles of Delaware Valley II really are; specifically, we have reconsidered whether we have been correct in assuming that Justice O’Connor’s concurring opinion governs the subject of contingency enhancements. We have regarded her concurrence as controlling largely in reliance on the Supreme Court’s admonition in Marks v. United States, 430 U.S. 188, 97 S.Ct. 990, 51 L.Ed.2d 260 (1977), that when the Court issues fragmented opinions, the opinion of the Justices concurring in the judgment on the “ ‘narrowest grounds' ” should be regarded as the Court’s holding. Id. at 193, 97 S.Ct. at 993-94 (quoting *781Gregg v. Georgia, 428 U.S. 153, 169 n. 15, 96 S.Ct. 2909, 2923 n. 15, 49 L.Ed.2d 859 (1976) (opinion of Stewart, Powell, and Stevens, JJ.)). But Marks is workable — one opinion can be meaningfully regarded as “narrower” than another — only when one opinion is a logical subset of other, broader opinions. In essence, the narrowest opinion must represent a common denominator of the Court’s reasoning; it must embody a position implicitly approved by at least five Justices who support the judgment.
In Gregg v. Georgia, 428 U.S. 153, 96 S.Ct. 2909, 49 L.Ed.2d 859 (1976), for example, the Court interpreted its earlier nine-way split in Furman v. Georgia, 408 U.S. 238, 92 S.Ct. 2726, 33 L.Ed.2d 346 (1972) (per curiam). In Furman, the five Justices who supported the judgment that Georgia’s death penalty statute was unconstitutional produced five separate opinions. Two Justices concluded that the death penalty was unconstitutional in all circumstances. See id. at 305-06, 92 S.Ct. at 2760-61 (Brennan, J., concurring); id. at 370-71, 92 S.Ct. at 2793-94 (Marshall, J., concurring). Three others found specific defects in the Georgia statute but declined to decide whether capital punishment might be constitutional under other circumstances. Of these, Justices Stewart and White felt that the Georgia death penalty was unconstitutional because it was applied in an arbitrary and capricious manner, see id. at 309-10, 92 S.Ct. at 2762-63 (Stewart, J., concurring); id. at 313, 92 S.Ct. at 2764 (White, J., concurring), while Justice Douglas stated that it was unconstitutional because it was “pregnant with discrimination,” falling more harshly on minorities and the poor both because its application was discretionary rather than mandatory and because wealthy defendants could afford better counsel. See id. at 255-57, 92 S.Ct. at 2734-36 (Douglas, J., concurring).
In Gregg, the Court — in another fragmented opinion — treated the opinions of Justices Stewart and White as controlling. See Gregg, 428 U.S. at 169 n. 15, 96 S.Ct. at 2923 n. 15 (plurality opinion). Justices Marshall and Brennan, who believed the death penalty unconstitutional in all circumstances, surely agreed with Justices Stewart and White that it was unconstitutional when administered in an arbitrary and capricious manner. By the same token, Justice Douglas, who insisted that any discretion in the judge or jury to decide when to impose capital punishment rendered the arrangement unconstitutional, would certainly have subscribed to Justice Stewart’s notion that the death penalty could not be administered constitutionally to “a capriciously selected random handful” of criminals. Furman, 408 U.S. at 309-10, 92 S.Ct. at 2762-63 (Stewart, J., concurring). Selecting the opinions of Justices Stewart and White as the holding of Furman in Gregg was thus unproblematic.
Similarly, in Marks itself, the Court adopted as the governing definition of obscenity the position of the plurality from the earlier case of A Book Named “John Cleland’s Memoirs of a Woman of Pleasure” v. Attorney General of Massachusetts, 383 U.S. 413, 86 S.Ct. 975, 16 L.Ed.2d 1 (1966) (Fanny Hill). In Fanny Hill, three separate views supported the judgment that the book was not obscene: the view expressed in the plurality opinion, which said that a book had to be “utterly without redeeming social value” to be considered obscene, see id. at 419, 86 S.Ct. at 977-78 (opinion of Brennan and Fortas, JJ., and Warren, C.J.) (emphasis omitted); the view of Justice Stewart that only “hardcore” pornography could be banned as obscene, see id. at 421, 86 S.Ct. at 978-79 (opinion of Stewart, J.); and the view of Justices Black and Douglas, who believed that obscenity could never be banned. See id. at 421, 86 S.Ct. at 978-79 (opinion of Black, J.); id. at 433, 86 S.Ct. at 985 (opinion of Douglas, J.). Because Justices Black and Douglas had to agree, as a logical consequence of their own position, with the plurality’s view that anything with redeeming social value is not obscene, the plurality of three in effect spoke for five Justices: Marks’ “narrowest grounds” approach yielded a logical result.6
*782When, however, one opinion supporting the judgment does not fit entirely within a broader circle drawn by the others, Marks is problematic. If applied in situations where the various opinions supporting the judgment are mutually exclusive, Marks will turn a single opinion that lacks majority support into national law. When eight of nine Justices do not subscribe to a given approach to a legal question, it surely cannot be proper to endow that approach with controlling force, no matter how persuasive it may be.
The Court itself appears not to apply Marks in cases of this type. To take one example, in Coolidge v. New Hampshire, 403 U.S. 443, 91 S.Ct. 2022, 29 L.Ed.2d 564 (1971), a plurality of four Justices held that only when evidence was discovered “inadvertently” could it be seized pursuant to the plain view exception to the Fourth Amendment’s warrant requirement. See id. at 469, 91 S.Ct. at 2040 (plurality opinion). Four other Justices believed that inadvertence was not necessary for a valid seizure of evidence in plain view. See id. at 492, 91 S.Ct. at 2051 (Burger, C.J., concurring in part and dissenting in part); id. at 506, 91 S.Ct. at' 2058 (Black, J., concurring in part and dissenting in part); id. at 510, 91 S.Ct. at 2060 (Blackmun, J., concurring in part and dissenting in part); id. at 516, 91 S.Ct. at 2063 (White, J., concurring in part and dissenting in part). Justice Harlan concurred in the judgment that the search in question was unconstitutional but provided no reasoning by which one could discern his position on the inadvertence requirement. See id. at 490, 91 S.Ct. at 2050 (Harlan, J., concurring in the judgment). The Court subsequently stated that the inadvertence requirement was “not a binding precedent” and was merely “the considered opinion of four Members of this Court” that should be “the point of reference for further discussion of the issue.” Texas v. Brown, 460 U.S. 730, 737, 103 S.Ct. 1535, 1540-41, 75 L.Ed.2d 502 (1983) (plurality opinion). The Court eventually disavowed the inadvertence requirement entirely. See Horton v. California, 496 U.S. 128, 110 S.Ct. 2301, 2308-10, 110 L.Ed.2d 112 (1990).
It seems to us that Delaware Valley II is one of the fragmented opinion cases that cannot be resolved satisfactorily by Marks. Unlike Furman or Fanny Hill, Delaware Valley II is not a case in which the concurrence posits a narrow test to which the plurality must necessarily agree as a logical consequence of its own, broader position. In other words, it is not a case in which there is an implicit majority of the Court. Rather, Delaware Valley II involves three distinct approaches to the issue of contingency enhancements in fee-shifting statutes, none of which enjoys the support of five Justices.
Superficially, to be sure, there is a common link between the plurality opinion and Justice O’Connor’s concurrence; both Part V of the plurality opinion and Justice O’Connor seem to endorse the substantial difficulties test we sought to apply earlier in this opinion. But the plurality quite clearly indicated in Part IV that it did not believe that contingency enhancements were ever available. See Delaware Valley II, 483 U.S. at 727, 107 S.Ct. at 3087-88 (plurality opinion). Therefore, Part V appears to have been composed not as an alternative holding but rather as a fall-back position, an invitation, as it were, to Justice O’Connor to reach common ground. Since Justice O’Connor did not accept the plurality’s invitation, explicitly declining to join Part V, the plurality’s true position remains that expressed in Part IV.
Justice O’Connor does appear to accept the bare concept that contingency enhancements should not be awarded unless the plaintiff shows substantial difficulties, but it is far from clear, as we noted earlier, what content she would give to the “substantial difficulties” test the plurality articulates in Part V. Her concurrence does not contain enough independent reasoning on the question of availability to allow us to compare her position analytically to that of the plurality. In that sense, her opinion *783on that issue approaches Justice Harlan’s in Coolidge.
Even if it were possible to determine from Justice O’Connor’s opinion when to apply a contingency enhancement and to conclude that her views on that subject were somehow narrower than the plurality’s, it is quite clear that one could not say in the Marks sense that her carefully explained view of how the contingency enhancement should be calculated is narrower than the plurality’s answer to that question. The plurality in Part V stated that if an upward adjustment for contingency risk were applied, the amount should be based on the “real risk-of-not-prevailing” in the case — but as a general rule should be “no more than one-third of the lodestar.” Id. at 730, 107 S.Ct. at 3089. Justice O’Con-nor, on the other hand, takes a different approach altogether, one that does not vary with the riskiness of the individual case but rather is based on a class determination of the amount of contingency enhancement usually paid in the relevant market, with no explicit ceiling. See id. at 731-34, 107 S.Ct. at 3089-91 (O’Connor, J. concurring in part and concurring in the judgment). We do not see how either approach can be thought “narrower” than the other; they are simply different.
To apply Marks to Delaware Valley II, we would have to conclude that Justice O’Connor’s answers to both the “when” and the “how” questions were “narrower” than the plurality’s in Part V. Without implicit agreement on both, it is simply impossible to regard the substantial difficulties test as controlling. Of course, as we have recognized, how one calculates a contingency enhancement could be thought to be a separate analytic issue from the question whether and under what circumstances a contingency enhancement is available. But as the panel opinion and other cases demonstrate, the two questions tend to run together. See King, 906 F.2d at 765-68; McKenzie, 875 F.2d at 334-37; Student Pub. Interest Research Group v. AT & T Bell Laboratories, 842 F.2d 1436, 1451 (3d Cir.1988). It is very difficult to consider the circumstances under which a contingency enhancement is “necessary” to attract counsel without contemplating the amount of the enhancement; each part of the inquiry has inevitable ramifications for the other. This may be the reason Justice O’Connor’s opinion focuses so heavily on the “how” question. Because her answer to that question is so clearly at odds with that of the plurality, however, we are left without a controlling opinion or a governing test for awarding contingency enhancements under Delaware Valley II.
The Third Circuit, taking a different approach, has reasoned that Justice O’Con-nor’s opinion can be regarded as a subset of the dissent if not the plurality. As such, Justice O’Connor’s concurrence would speak for a majority of the Court. See id. (“Because the four dissenters would allow contingency multipliers in all cases in which Justice O’Connor would allow them, her position commands a majority of the Court.”). The Third Circuit appears to apply the Marks methodology to reach this result, but it does not explicitly rely on Marks. See id. at 1451 n. 16 (citing Marks as a “see also” in a footnote appended to the citation of a circuit opinion). This is understandable, because Marks has never been so applied by the Supreme Court, and we do not think we are free to combine a dissent with a concurrence to form a Marks majority. As the Court said in Marks itself, “When a fragmented Court decides a case and no single rationale explaining the result enjoys the assent of five Justices, ‘the holding of the Court may be viewed as that position taken by those members who concurred in the judgments on the narrowest grounds.’ ” Marks, 430 U.S. at 193, 97 S.Ct. at 993 (quoting Gregg, 428 U.S. at 169 n. 15, 96 S.Ct. at 2923 n. 15 (opinion of Stewart, Powell, and Stevens, JJ.)).
To be sure, in Vasquez v. Hillery, 474 U.S. 254, 106 S.Ct. 617, 88 L.Ed.2d 598 (1986), the Supreme Court, in interpreting Rose v. Mitchell, 443 U.S. 545, 99 S.Ct. 2993, 61 L.Ed.2d 739 (1979), emphasized that an opinion that combines shifting majorities in various portions is no less binding than would be an opinion in which the same Justices formed the majority for all the sections. See Vasquez, 474 U.S. at *784261-62 n. 4, 106 S.Ct. at 622 n. 4; see also Arizona v. Fulminante, — U.S. -, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991) (employing two distinct majorities to arrive at a judgment, both of which therefore constitute binding law). That, however, is quite a different situation than the one that the Marks methodology addresses, where there is no explicit majority agreement on all the analytically necessary portions of a Supreme Court opinion. Under these latter circumstances, if the application of Marks will not yield a majority holding, nothing will.7
To say that Delaware Valley II provides no controlling legal holding is not to say that it has no binding impact on us. Because the Court’s result was to deny a contingency enhancement without even a remand, we think we could not authorize the routine awarding of contingency enhancements of whatever size. Cf. National Mut. Ins. Co. v. Tidewater Transfer Co., 337 U.S. 582, 655, 69 S.Ct. 1173, 1199-1200, 93 L.Ed. 1556 (1949) (Frankfurter, J., dissenting) (noting that the result is binding even when the Court fails to agree on reasoning). We furthermore believe, as this opinion and the dissent make clear, that there simply is no practical middle ground between providing enhancements routinely and not providing them at all. Keeping in mind that a majority of the Supreme Court clearly agrees that the question of attorney’s fees must not turn into major litigation in itself, see Delaware Valley II, 483 U.S. at 722, 107 S.Ct. at 3085, we think the appropriate course is to hold that contingency enhancements will not be available in this Circuit.8 We note that although other circuit courts have set forth various tests for awarding contingency enhancements under Delaware Valley II, most of the tests appear to be difficult, if not impossible, to meet in practice. See, e.g., Student Pub. Interest Research Group v. AT & T Bell Laboratories, 842 F.2d 1436, 1451-52 (3d Cir.1988) (“Contingency multipliers should be granted only rarely.”); Craig v. Secretary, Dep’t of Health and Human Servs., 864 F.2d 324, 327 (4th Cir.1989) (no contingency enhancement available “in the absence of exceptional circumstances”); Leroy v. City of Houston, 831 F.2d 576, 583 (5th Cir.1987) (contingency enhancements should be “reserved for ‘exceptional cases’ ” (citation omitted)); Skelton v. General Motors Corp., 860 F.2d 250, 254 (7th Cir.1988) (contingency enhancements available only if plaintiffs meet “stringent requirements”); Hendrickson v. Branstad, 934 F.2d 158, 162 (8th Cir.1991) (“[Ejnhancement is reserved for ‘rare’ and ‘exceptional’ cases_”); Smith v. Freeman, 921 F.2d 1120, 1123 (10th Cir.1990) (quoting plurality’s view that “enhancement for the risk of nonpayment should be reserved for exceptional cases”); Norman v. Housing Auth., 836 F.2d 1292, 1302 (11th Cir.1988) (“[I]n the rare case enhancement may be appropriate ”). But cf. D’Emanuele v. Montgomery Ward & Co., 904 F.2d 1379, 1383 (9th Cir.1990) (implying that routine *785contingency enhancements might be justified).
We have done our best to apply Delaware Valley II but have been unable to derive a governing rule from the opinion. Considering our struggle to understand and apply Delaware Valley II as well as the difficulties our sister circuits have experienced, we urge the Supreme Court to clarify its position.
* * * * * *
For the foregoing reasons, we overrule McKenzie v. Kennickell, 875 F.2d 330 (D.C.Cir.1989), and those portions of our other previous opinions inconsistent with our current disposition, and reverse the contingency enhancement portion of the attorney’s fees allowed to appellant.
It is so ordered.
. 42 U.S.C. § 2000e-5(k) provides in pertinent part:
In any action or proceeding under this sub-chapter the court, in its discretion, may allow the prevailing party ... a reasonable attorney’s fee as part of the costs_
Section 2000e-16(d) extends the provisions of § 2000e-5(k) to actions by employees of the District of Columbia. Congress has made clear that it intends the courts to resolve the policy questions inherent in determining what is "reasonable.” See H.R.Rep. No. 1558, 94th Cong., 2d Sess. 8 (1976).
. We do not understand how, as the dissent suggests, the discrete question presented in this case — whether a contingency enhancement is properly included within an award of attorney’s fees — can possibly be thought a matter for the discretion of the trial judge. If the overall reasonableness of a statutory attorney’s fee award were always a matter for the trial judge’s discretion, unguided by a legal structure, the Supreme Court certainly has wasted a good deal of time and effort attempting to develop uniform rules. See Blanchard v. Bergeron, 489 U.S. 87, 109 S.Ct. 939, 103 L.Ed.2d 67 (1989); Delaware Valley II, 483 U.S. 711, 107 S.Ct. 3078, 97 L.Ed.2d 585 (1987); Delaware Valley I, 478 U.S. 546, 106 S.Ct. 3088, 92 L.Ed.2d 439 (1986); Blum, 465 U.S. 886, 104 S.Ct. 1541, 79 L.Ed.2d 891 (1984); Hensley, 461 U.S. 424, 103 S.Ct. 1933, 76 L.Ed.2d 40 (1983).
. Delaware Valley II interpreted the fee-shifting provision of the Clean Air Act, 42 U.S.C. § 7604(d). However, the Court has said that its standards for determining "reasonable” fees apply to all federal statutes awarding "reasonable" attorney’s fees to a "prevailing party,” including Title VII. See Hensley, 461 U.S. at 433 n. 7, 103 S.Ct. at 1939 n. 7.
. Of these numerous affidavits, only five addressed the facts of Ms. King’s particular case or expressed any opinion at all about whether Ms. King herself would have faced substantial difficulties in attracting competent counsel absent the availability of a contingency enhancement. Two of these described the weakness and difficulty of her case as the principal reason the affiants would have been unwilling to assume representation. See J.A. at 112d (Cashdan); 173b (Fitzpatrick). Another attorney allowed that his firm would possibly have represented Ms. King if she had an "exceptionally strong claim.” J.A. at 124b (Chuzi). The fourth admitted there was at least a possibility, although "remote,” of finding pro bono representation for Ms. King. See J.A. at 214 (Lapidus). Only a single affidavit stated flatly that the affiant would not take Ms. King’s case because of the unavailability of contingency enhancements. See J.A. at 177 (Gottfried).
. The dissent, nevertheless, would calculate the amount of contingency enhancement based on the degree of risk faced by the plaintiffs lawyer in each case. As the plurality in Delaware Valley II noted, that approach would provide incentives to bring the weakest cases to court, and it would put the district judge who had to make that determination in a difficult psychological posture. See Delaware Valley II, 483 U.S. at 722, 107 S.Ct. at 3085; id. at 725, 107 S.Ct. at 3086-87 (plurality opinion).
. Justice Stewart’s “hardcore” test, though it was also a logical subset of Justice Black and Doug*782las’ opinion, would only have spoken for three Justices and could therefore not have been the controlling rationale.
. In our view, even applying the Third Circuit’s reasoning, we do not think Justice O’Connor’s concurrence constitutes a controlling opinion in Delaware Valley II. For similar reasons to those we outlined in our discussion as to whether her opinion could be thought narrower than the plurality opinion, Justice O’Connor’s thoughtful answer to the question of how to calculate a contingency enhancement should it be available cannot possibly be thought a subset of the dissent’s approach to the same issue. She herself recognizes this. See Delaware Valley II, 483 U.S. at 732, 107 S.Ct. at 3090 (O'Connor, J., concurring in part and concurring in the judgment).
. The dissent’s reliance on the legislative history of 42 U.S.C. § 1988 — particularly citations to Johnson v. Georgia Highway Express, Inc., 488 F.2d 714, 717-19 (5th Cir.1974)—comes too late. The Supreme Court has, on several occasions, indicated that it does not regard factors listed separately in Johnson as appropriate enhancements to the lodestar. See Blum v. Stenson, 465 U.S. 886, 898-99, 104 S.Ct. 1541, 1548-49, 79 L.Ed.2d 891 (1984); Delaware Valley I, 478 U.S. 546, 564, 566, 106 S.Ct. 3088, 3097-98, 3098-99, 3098, 92 L.Ed.2d 439 (1986). And the dissent’s citation of Blanchard v. Bergeron, 489 U.S. 87, 109 S.Ct. 939, 103 L.Ed.2d 67 (1989), is also misplaced. There the Supreme Court was dealing with an entirely different issue — the question whether an actual, private contingency fee arrangement limited a statutory award, not whether a contingency factor should be added to the lodestar.