Justice Souter, with whom Justice Stevens, Justice Ginsburg, and Justice Breyer join, dissenting.

A defendant’s right to representation on appeal is limited by the prohibition against frivolous litigation, and I realize that when a lawyer’s corresponding obligations are at odds with each other, there is no perfect place to draw the line between them. But because I believe the procedure adopted in People v. Wende, 25 Cal. 3d 436, 600 P. 2d 1071 (1979), fails to assure representation by counsel with the adversarial character demanded by the Constitution, I respectfully dissent.

Although the Sixth Amendment guarantees trial counsel to a felony defendant, see Gideon v. Wainwright, 372 U. S. 335 (1963), the Constitution contains no similarly freestanding, unconditional right to counsel on appeal, there being no obligation to provide appellate review at all, see Ross v. Moffitt, 417 U. S. 600, 606 (1974). When a State elects to provide appellate review, however, the terms on which it does so are subject to constitutional notice. See, e.g., Griffin v. Illinois, 351 U. S. 12, 18 (1956); Rinaldi v. Yeager, 384 U. S. 305, 310 (1966); Evitts v. Lucey, 469 U. S. 387, 393 (1985).

In a line of cases beginning with Griffin, this Court examined appellate procedural schemes under the principle that justice may not be conditioned on ability to pay, see generally Ross, supra, at 605–609. Even though “[a]bsolute equality is not required,” Douglas v. California, 372 U. S. 353, 357 (1963), we held in Douglas that when state criminal defendants are free to retain counsel for a first appeal as of right, the Fourteenth Amendment requires that indigent appellants be placed on a substantially equal footing through the appointment of counsel at the State’s expense. See McCoy v. Court of Appeals of Wis., Dist. 1, 486 U. S. 429, 438 (1988) (referring to “principle of substantial equality”).

Two services of appellate counsel are on point here. Appellate counsel examines the trial record with an advocate’s eye, identifying and weighing potential issues for appeal. This is review not by a dispassionate legal mind but by a committed representative, pledged to his client’s interests, primed to attack the conviction on any ground the record may reveal. If counsel’s review reveals arguable trial error, he prepares and submits a brief on the merits and argues the appeal.

The right to the first of these services, a partisan scrutiny of the record and assessment of potential issues, goes to the irreducible core of the lawyer’s obligation to a litigant in an adversary system, and we have consistently held it essential to substantial equality of representation by assigned counsel. “The paramount importance of vigorous representation follows from the nature of our adversarial system of justice.” Penson v. Ohio, 488 U. S. 75, 84 (1988). See, e.g., Ellis v. United States, 356 U. S. 674, 675 (1958) (per curiam); Douglas, supra, at 357–358; McCoy, supra, at 438. The right is unqualified when a defendant has retained counsel, and I can imagine no reason that it should not be so when counsel has been appointed.

Because the right to the second service, merits briefing, is not similarly unqualified, however, the issue we address today arises. The limitation on the right to a merits brief is that no one has a right to a wholly frivolous appeal, see Anders v. California, 386 U. S. 738, 742 (1967), against which the judicial system’s first line of defense is its lawyers. Being officers of the court, members of the bar are bound “not to clog the courts with frivolous motions or appeals,” Polk County v. Dodson, 454 U. S. 312, 323 (1981); see also McCoy, supra, at 436, and this is of course true regardless of a lawyer’s retained or appointed status in a given case. The problem to which Anders responds arises when counsel views his client’s appeal as frivolous, leaving him duty barred from pressing it upon a court.

The rub is that although counsel may properly refuse to brief a frivolous issue and a court may just as properly deny leave to take a frivolous appeal, there needs to be some reasonable assurance that the lawyer has not relaxed his partisan instinct prior to refusing, in which case the court’s review could never compensate for the lawyer’s failure of advocacy. A simple statement by counsel that an appeal has no merit, coupled with an appellate court’s endorsement of counsel’s conclusion, gives no affirmative indication that anyone has sought out the appellant’s best arguments or championed his cause to the degree contemplated by the adversary system. Nor do such conclusions acquire any implicit persuasiveness through exposure to an interested opponent’s readiness to mount a challenge. The government is unlikely to dispute or even test counsel’s evaluation; one does not berate an opponent for giving up. To guard against the possibility, then, that counsel has not done the advocate’s work of looking hard for potential issues, there must be some prod to find any reclusive merit in an ostensibly unpromising case and some process to assess the lawyer’s efforts after the fact. A judicial process that renders constitutional error invisible is, after all, itself an affront to the Constitution. See Penson, supra, at 81–82.

In Anders, we devised such a mechanism to ensure respect for an appellant’s rights. See Penson, supra, at 80. A lawyer’s request to withdraw on the ground that an appeal is frivolous “must … be accompanied by a brief referring to anything in the record that might arguably support the appeal.” Anders, 386 U. S., at 744. This simply means that counsel must do his partisan best, short of calling black white, to flag the points that come closest to being appealable; the lawyer’s job is to state the issues that give the defendant his best chances to prevail, even if the best comes up short under the rule against trifling with the court. “[T]he court—not counsel—,” we continued, “then proceeds, after a full examination of all the proceedings, to decide whether the case is wholly frivolous.” Ibid.

Anders thus contemplates two reviews of the record, each of a markedly different character. First comes review by the advocate, the defendant’s interested representative. His job is to identify the best issues the partisan eye can spot. Then comes judicial review from a disinterested judge, who asks two questions: whether the lawyer really did function as a committed advocate, and whether he misjudged the legitimate appealability of any issue. In reviewing the advocate’s work, the court is responsible for assuring that counsel has gone as far as advocacy will take him with the best issues undiscounted. We have repeatedly described the task of an appellate court in terms of this dual responsibility. “ ‘First, [the court] must satisfy itself that the attorney has provided the client with a diligent and thorough search of the record for any arguable claim that might support the client’s appeal. Second, it must determine whether counsel has correctly concluded that the appeal is frivolous.’ ” Penson, 488 U. S., at 83 (quoting McCoy, 486 U. S., at 442).

Griffin and Anders thus require significantly more than the abstract evaluation of the merits of conceivably appealable points. Without the assurance that assigned counsel has done his best as a partisan, his substantial equality to a lawyer retained at a defendant’s expense cannot be assumed. And without the benefit of the lawyer’s statement of strongest claims, the appellate panel cannot act as a reviewing court, but is relegated to an inquisitorial role.

It is owing to the importance of assuring that an adversarial, not an inquisitorial, system is at work that I disagree with the Court’s statement today that our cases approve of any state procedure that “reasonably ensures that an indigent’s appeal will be resolved in a way that is related to the merit of that appeal.” Ante, at 14. A purely inquisitorial system could satisfy that criterion, and so could one that appoints counsel only if the appellate court deems it useful. But we have rejected the former and have explicitly held the latter unconstitutional, see Douglas, 372 U. S., at 355, the reason in each case being that the Constitution looks to the means as well as to the ends. See Singer v. United States, 380 U. S. 24, 36 (1965) (“The Constitution recognizes an adversary system as the proper method of determining guilt …”). See also, e.g., Penson, supra, at 87 (“A criminal appellant is entitled to a single-minded advocacy …”); Strickland v. Washington, 466 U. S. 668, 685 (1984) (“The Sixth Amendment recognizes the right to the assistance of counsel because it envisions counsel’s playing a role that is critical to the ability of the adversarial system to reach just results”); United States v. Cronic, 466 U. S. 648, 656 (1984) (“Thus, the adversarial process protected by the Sixth Amendment requires that the accused have ‘counsel acting in the role of an advocate.’ ”) (quoting Anders, supra, at 743).

We have not held the details of Anders to be exclusive, but it does make sense to read the case as exemplifying what substantial equality requires on behalf of indigent appellants entitled to an advocate’s review and to reasonable certainty that arguable issues will be briefed on their merits. With Anders thus as a benchmark, California’s Wende procedure fails to measure up. Its primary failing is in permitting counsel to refrain as a matter of course from mentioning possibly arguable issues in a no-merit brief; its second deficiency is a correlative of the first, in obliging an appellate court to search the record for arguable issues without benefit of an issue-spotting, no-merit brief to review. See 25 Cal. 3d, at 440–442, 600 P. 2d, at 1074–1075.

Although Wende assumes that counsel will act as an advocate, see id., at 441–442, 600 P. 2d, at 1075, it fails to assure, or even promote, the partisan attention that the Constitution requires. While the lawyer must summarize the procedural and factual history of the case with citations to the record, nothing in the Wende scheme requires counsel to show affirmatively, subject to evaluation, that he has made the committed search for issues and the advocate’s assessment of their merits that go to the heart of appellate representation in our adversary system. It begs the question to say that “[c]ounsel’s inability to find any arguable issues may readily be inferred from his failure to raise any,” id., at 442, 600 p. 2d, at 1075 and it misses the point to argue that the indigent appellant is adequately protected because the lawyer assigned to a case under California’s assigned counsel scheme may not file a Wende brief without the approval of a supervisor. The point is the need for some affirmative and express indicator that an advocate has been at work, in the form of a product that an appellate court can specifically review. Thus Anders requires counsel to flag the best issues for the sake of keeping counsel on his toes and giving focus to judicial review of his judgment. Wende on the other hand requires no indication of conceivable issues and hence nothing specifically reviewable by a court bound to preserve the system’s adversary character. Wende does no more to protect the indigent’s right to advocacy than the no-merit letter condemned in Anders, or the conclusory statement disapproved in Penson.

On like reasoning, Wende is deficient in relying on a judge’s nonpartisan review to assure that a defendant suffers no prejudice at the hands of a lawyer who has failed to document his best effort at partisan review. Exactly because our system assumes that a lawyer committed to a client is the most dependable guardian of the client’s interest, see supra, at 6, we have consistently rejected procedures leaving the determination of frivolousness to the court in the first instance, see Douglas, supra, at 355–356, or to the court following a conclusory declaration by counsel, see Penson, 488 U. S., at 81–82, or to the court assisted by counsel in the role of amicus curiae, see Ellis, 356 U. S., at 675. The defect in these procedures is their entire reliance on review by a detached magistrate who does not apply the partisan scrutiny in the first instance that defendants with paid lawyers get as a matter of course.

It goes without saying, too, that Wende’s reliance on judges to start from scratch in seeking arguable issues adds substantially to the burden on the judicial shoulders. While I have no need to decide whether this drawback of the Wende scheme is of constitutional significance, it raises questions that certainly underscore the constitutional failing of relying on judicial scrutiny uninformed by counsel’s partisan analysis. In an amicus brief filed in this case, 13 retired Justices of the Supreme Court or Courts of Appeal of California have pointed out the “risk that the review of the cold record [under the Wende scheme] will be more perfunctory without the issue-spotting guidance, and associated record citations, of counsel.” Brief for Retired Justice Armand Arabian et al. as Amici Curiae 5. The amici have candidly represented that “[w]hen a California appellate court receives a Wende brief, it assigns the case to a staff attorney who prepares a memorandum analyzing all possible legal issues in the case. Typically, the staff attorney then makes an oral presentation to the appellate panel … .” Brief for Retired Justice Arabian, supra, at 6. When the responsibility of counsel is thrown onto the court, the court gives way to a staff attorney; it could not be clearer that Wende is seriously at odds with the respective obligations of counsel and the courts as contemplated by the Constitution.

III

Unlike the Court, I reach the question of appropriate relief. With respect to respondent’s Anders claim, the Court of Appeals premised its disposition on finding that two potentially meritorious issues showed that Robbins had been prejudiced by the failure of the Wende scheme to result in their litigation. I think it unnecessary to invoke such findings, however, and would hold for Robbins simply because of the failure to provide an advocate’s analysis of issues as a predicate of court review. Without more, I would, in effect, require the state courts to reinstate the appeal for treatment consistent with the Anders application of Griffin.

It is true of course, that before relief is normally granted for want of adequate assistance of trial counsel, a defendant must show not only his lawyer’s failure to represent him with reasonable competence (demonstrated here by the failure to file an advocate’s issue-spotting brief), but also a “reasonable probability” that competent representation would have produced a different result in his case, see Strickland, 466 U. S., at 694. But the assumption behind Strickland’s prejudice requirement is that the defendant had a lawyer who was representing him as his advocate at least at some level, whereas that premise cannot be assumed when a defendant receives the benefit of nothing more than a Wende brief. In a Wende situation, nominal counsel is functioning merely as a friend of the court, helping the judge to grasp the structure of the record but not even purporting to highlight the record’s nearest approach to supporting his client’s hope to appeal. Counsel under Wende is doing less than the judge’s law clerk (or a staff attorney) might do, and he is doing nothing at all in the way of advocacy. When a lawyer abandons the role of advocate and adopts that of amicus curiae, he is no longer functioning as counsel or rendering assistance within the meaning of the Sixth Amendment. See Cronic, 466 U. S., at 654–655. Since the apparently missing ingredient of the advocate’s analysis goes to the very essence of the right to counsel, a lawyer who does nothing more than file a Wende brief is closer to being no counsel at all than to being subpar counsel under Strickland.

This, I think, is the answer to any suggestion that a specific assessment of prejudice need be shown in order to get relief from Wende. A complete absence of counsel is a reversible violation of the constitutional right to representation, even when there is no question that at the end of the day the smartest lawyer in the world would have watched his client being led off to prison. See Cronic, supra, at 658–659; cf. Rodriquez v. United States, 395 U. S. 327 (1969). We do not ask how the defendant would have fared if he had been given counsel, and we should not look to what sort of appeal might have ensued if an appellant’s lawyer had flagged the points that came closest to appealable issues. Such a result is equally consistent with our cases holding a violation of due process to be complete when a defendant is denied a right to the appeal he is otherwise entitled to pursue. See Peguero v. United States, 526 U. S. 23, 30–31 (1999) (O’Connor, J., concurring); Rodriquez, supra, at 330.

This conclusion was anticipated in Penson, in which we dealt with the violation of Anders standards when counsel was allowed to withdraw without supplying the court with his best effort to identify appealable weaknesses, and prior to any judicial determination that counsel had missed nothing in finding no arguable appellate issues in the record. The appellate court in Penson subsequently identified arguable issues but thought the appointment of new counsel unnecessary after finding that any legitimately appealable issues would be losers. This Court recognized a presumption of prejudice without more, for purposes of both Strickland and Chapman v. California, 386 U. S. 18 (1967). See Penson, 488 U. S., at 85–86. Although the state court’s failure to appoint counsel after identifying issues made Penson an egregious case, id., at 83, the failure of advocacy and consequent constructive absence of counsel was clear even at the point at which the lawyer withdrew, id., at 82, and the presumption of prejudice applicable then is applicable in this case now.

There is practical sense as well as good theory behind this presumption of prejudice, for any requirement to demonstrate prejudice specifically would often place federal judges on habeas in highly precarious positions calling for judgments that state judges are generally better qualified to make. Since there will have been no advocate’s help in analyzing the record on the direct state appeal, and since counsel may well have been absent formally as well as constructively in any state post-conviction proceedings, the federal judge would be looking for (among other things) previously unidentified state law issues not previously waived. One could not ask for a more certain guarantee of inefficient and time consuming judicial effort.

What remains is only to say a word about the State’s argument that relief in this case is barred under Teague v. Lane, 489 U. S. 288 (1989), as requiring application of a new rule of law not clearly entailed by our prior holdings. The argument seems to be that California has relied on Wende for so long that any disapproval from a federal court at this juncture is some sort of novelty (resulting from the failure of other state defendants to reach the federal courts earlier with Wende objections). The obvious answer is that the application of Douglas and Griffin standards to meritless appeals has been subject to repeated explanation starting with Anders and echoed in McCoy and Penson. Once general rules are announced they do not become “new” again with every particular violation that may subsequently occur. See Saffle v. Parks, 494 U. S. 484, 491–492 (1990) (discussing application of the rule of Jurek v. Texas, 428 U. S. 262 (1976), in Penry v. Lynaugh, 492 U. S. 302 (1989)). The same point, of course, would answer any objection under the AEDPA that an Anders petitioner was seeking to go beyond “clearly established Federal law, as determined by the Supreme Court of the United States,” 28 U. S. C. §2254(d)(1) (1994 ed. Supp. III).

* * *

The Wende procedure does not assure even the most minimal assistance of counsel in an adversarial role. The Constitution demands such assurances, and I would hold Robbins entitled to an appeal that provides them.