with whom Mr. Justice Marshall concurs, dissenting.
Until Linkletter v. Walker, 381 U. S. 618 (1965), the Court traditionally applied new constitutional criminal procedure standards to cases finalized and police practices operative before the promulgation of the new rules.1 Linkletter, however, was the cradle of a new doctrine of nonretroactivity which exempts from relief the earlier victims of unconstitutional police practices. I have disagreed on numerous occasions with applications of various brands of this doctrine and I continue my dissent in this case.2 My own view is that even-handed justice requires either prospectivity only3 or complete retro-*287activity. To me there is something inherently invidious as Mr. Justice Harlan phrased it, in “[s] imply fishing one case from the stream of appellate review, using it as a vehicle for pronouncing new constitutional standards, and then permitting a stream of similar cases subsequently to flow by unaffected by that new rule . . . .” Mackey v. United States, 401 U. S. 667, 679 (1971) (separate opinion). I agree with his critique, id., at 695, that the purported distinction between those rules that are designed to improve the factfinding process and those designed to further other values was “inherently intractable” and to illustrate his point he adverted to the Court’s difficulty in reconciling with its rule such nonretroactivity cases as Johnson v. New Jersey, 384 U. S. 719 (1966); Stovall v. Denno, 388 U. S. 293 (1967), and DeStefano v. Woods, 392 U. S. 631 (1968), all of which held nonretroactive decisions designed, in part, to enhance the integrity of the factfinding process. He also questioned the workability of any rule which requires a guess as to “whether a particular decision has really announced a ‘new’ rule at all or whether it has simply applied a well-established constitutional principle.” Mackey v. United States, supra, at 695; Desist v. United States, 394 U. S. 244, 263 (1969). For example, as I suggest infra, at 293-295, a serious question arises in this case whether Coleman v. Alabama, 399 U. S. 1 (1970), should have been fully anticipated by state judicial authorities.4
*288Additionally, it is curious that the plurality rule is sensitive to “reasonable reliance” on prior standards by law enforcement agencies but is unconcerned about the *289unfairness of arbitrarily granting relief to Coleman but denying it to Adams.
Given my disagreement with the plurality’s rule, I am reluctant even to attempt to apply it, but even by its own *290terms, the balancing approach would appear to require that we hold Coleman retroactive. This conclusion reinforces my fear that the process is too imprecise as a neutral guide for either this Court or the lower courts and will invariably permit retroactivity decisions to turn on predilections, not principles.
I
In applying the rule, I am first troubled by the plurality’s adoption of the finding of the court below that: “On [the] scale of probabilities, we judge that the lack of counsel at a preliminary hearing involves less danger to 'the integrity of the truth-determining process at trial’ than the omission of counsel at the trial itself or on appeal.” Ante, at 282-283. The same might have been said of the right to counsel at sentencing, Mempa *291v. Rhay, 389 U. S. 128 (1967), at certain arraignments, Hamilton v. Alabama, 368 U. S. 52 (1961), or at preliminary hearings where guilty pleas were taken, White v. Maryland, 373 U. S. 59 (1963), all of which have been held retroactive.5
Rather than reaching for these analogies, however, the plurality suggests that the danger to the integrity of the truth-determining process is no greater here than at a pretrial lineup or at an interrogation conducted without counsel. In relying on these analogies, the plurality gives short shrift to the argument that “in practice [the preliminary] hearing may provide the defense with the most valuable discovery technique available to him,” Wheeler v. Flood, 269 F. Supp. 194, 198 (EDNY 1967), an objective which is not so readily achievable at lineups and interrogations at which counsel serves only a protective function. The State’s access to superior investigative resources and its ability to keep its case secret until trial normally puts the defendant at a clear disadvantage.6 *292In light of this disparity, one important service the preliminary hearing performs is to permit counsel to penetrate the evidence offered by the prosecution at the hearing, to test its strengths and weaknesses (without the presence of a jury), to learn the names and addresses of witnesses, to focus upon the key factual issues in the upcoming trial, and to preserve testimony for impeachment purposes. The alternative discovery techniques suggested now by the plurality are puny in comparison. A bill of particulars can usually reach only prosecution witnesses’ names, and it may be cold comfort to defense counsel to learn that he can obtain pretrial statements of prosecution witnesses inasmuch as such statements are often prepared from the State’s viewpoint and have not been subjected to cross-examination. And in many States such statements are not discoverable.
Finally, when read in light of Coleman’s exaltation of the virtues of counseled preliminary hearings, the present language of the plurality may lend itself to a “credibility gap” between it and those involved in the administration of the criminal process. “Plainly,” said the Coleman Court, “the guiding hand of counsel at the preliminary hearing is essential to protect the indigent accused against an erroneous or improper prosecution,” Coleman v. Alabama, supra, at 9, and: “The inability of the indigent accused on his own to realize these advantages of a lawyer’s assistance compels the conclusion that the Alabama preliminary hearing is a ‘critical stage’ of the State’s criminal process at which the accused, is ‘as much entitled to such aid [of counsel] ... as at the trial itself.’ ” Id., at 9-10. It will *293now appear somewhat anomalous that the right to counsel at a preliminary hearing is fundamental enough to be incorporated into the Fourteenth Amendment but not fundamental enough to warrant application to the victims of previous unconstitutional conduct.7
II
I also believe that the plurality’s case for establishing good-faith reliance on “the old standards” by state judicial systems ignores important developments in the right-to-counsel cases prior to Coleman. First of all, no decision of this Court had held that counsel need not be afforded at the preliminary hearing stage. Therefore, to build a case for good-faith reliance the State must wring from our decision the negative implication that uncoun-seled probable-cause hearings were permissible. Such negative implications are found, says the plurality, in Hamilton v. Alabama, 368 U. S. 52 (1961), and White v. Maryland, 373 U. S. 59 (1963), cases reversing convictions obtained through the use at trial of uncounseled guilty pleas entered at preliminary hearings. Neither of those decisions, however, faced the question of whether reversal *294was required on the facts of the instant case. And, though I have studied these two short opinions, I am unable, as is the plurality, to divine any hidden message to law enforcement agencies that we would permit the denial of counsel at preliminary hearings where guilty pleas were not taken. Rather, these cases reinforce, in my mind, the importance of counsel at every stage in the criminal process. In any event, by the time Coleman came down, it was clear, as Mr. Justice Harlan opined, albeit with some regret, that our holding was an inevitable consequence of prior case law:
“If I felt free to consider this case upon a clean slate I would have voted to affirm these convictions. But — in light of the lengths to which the right to appointed counsel has been carried in recent decisions of this Court, see Miranda v. Arizona, 384 U. S. 436 (1966); United States v. Wade, 388 U. S. 218 (1967); Gilbert v. California, 388 U. S. 263 (1967); Mathis v. United States, 391 U. S. 1 (1968); and Orozco v. Texas, 394 U. S. 324 (1969) — I consider that course is not open to me with due regard for the way in which the adjudicatory process of this Court, as I conceive it, should work. . . .
“It would indeed be strange were this Court, having held a suspect or an accused entitled to counsel at such pretrial stages as 'in-custody’ police investigation, whether at the station house (Miranda) or even in the home {Orozco), now to hold that he is left to fend for himself at the first formal confrontation in the courtroom.” Coleman v. Alabama, supra, at 19-20 (separate opinion).8
*295Thus, in the instant case, at the times relevant, the State should have foreseen that the right to counsel attached to the probable-cause hearing.
Ill
I also disagree that “[t]he impact upon the administration of the criminal law of [Coleman retroactivity] needs no elaboration.” Ante, at 284. In the 19 months since Coleman was decided all new prosecutions have presumably followed it and we therefore need only be concerned, for impact purposes, with those state proceedings in which a preliminary hearing was held prior to June 1970. Inasmuch as the median state sentence served by felons when they are first released is about 20.9 months,9 most pre-Coleman sentences would now be served and as a practical matter these former prisoners would not seek judicial review. Moreover, we may exclude from our consideration those 16 or more States that prior to Coleman routinely appointed counsel at or prior to preliminary hearings. See American Bar Association, Project on Standards for Criminal Justice, Providing Defense Services § 5.1 (Approved Draft 1968). Additionally, we may exclude from consideration the possibility of collateral challenges by federal prisoners inasmuch as counsel have routinely been present at preliminary hearings before federal commissioners.10 See Fed. Rule Crim. Proc. 5 (b).
While there are some current prisoners who might challenge their confinements if Coleman were held retro*296spective, many of these attacks would probably fail under the harmless-error rule of Chapman v. California, 386 U. S. 18 (1967). The plurality opinion suggests that conducting such harmless-error proceedings would be onerous. One reason given is that in Illinois, for example, preliminary hearings were not recorded before Coleman. That assertion may not be entirely accurate in light of the fact that this very record contains a transcript of Adams’ preliminary hearing. Perhaps, as the respondent seems to concede,11 transcripts were made available in other Illinois cases. That is the more reasonable assumption in light of our holding in Roberts v. LaVallee, 389 U. S. 40 (1967), that the State must provide a preliminary hearing transcript to an indigent in every circumstance in which the more affluent accused could obtain one.
Even where a transcript was not available, however, a prisoner might be able to show at an evidentiary hearing that he was prejudiced by a particular need for discovery, by the inability to preserve the testimony of either an adverse or favorable witness, or by the inability to secure his release on bail in order to assist in the preparation of his defense.12 Courts are accustomed, of course, to assessing claims of prejudice without the aid of transcripts of previous proceedings, such as is required by Jackson v. Denno, 378 U. S. 368 (1964), or Townsend v. Sain, 372 U. S. 293 (1963). Indeed, in Coleman we remanded for a determination of whether the failure to appoint counsel had been harmless error. 399 U. S., at 11. Not every Coleman claim would warrant an evi-dentiary hearing. Many attacks might be disposed of summarily, such as a challenge to a conviction resulting from a counseled guilty plea entered before any preju*297dice had materialized from an uncounseled preliminary-hearing. See Procunier v. Atchley, 400 U. S. 446 (1971).
Even Stovall v. Denno, 388 U. S., at 299, the analogy frequently invoked by the plurality, held out the possibility of collateral relief in cases where prisoners could show that their lineups had imposed “such unfairness that [they] infringed [their] right to due process of law.” Conducting Coleman harmless-error hearings would not appear to be any more burdensome on the administration of criminal justice than have Stovall “fundamental fairness” post-conviction proceedings.
In any event, whatever litigation might follow a holding of Coleman retrospectivity must be considered part of the price we pay for former failures to provide fair procedures.
E. g., Eskridge v. Washington Prison Board, 357 U. S. 214 (1958) ; Gideon v. Wainwright, 372 U. S. 335 (1963); Jackson v. Denno, 378 U. S. 368 (1964), (see also Desist v. United States, 394 U. S. 244, 250 n. 15 (1969)); Reck v. Pate, 367 U. S. 433 (1961).
Linkletter v. Walker, 381 U. S. 618, 640 (1965); Tehan v. Shott, 382 U. S. 406, 419 (1966); Johnson v. New Jersey, 384 U. S. 719, 736 (1966); Stovall v. Denno, 388 U. S. 293, 302 (1967); DeStefano v. Woods, 392 U. S. 631, 635 (1968); Desist v. United States, 394 U. S. 244, 255 (1969); Holliday v. United States, 394 U. S. 831, 835 (1969); Mackey v. United States, 401 U. S. 667, 713 (1971).
It was suggested in Stovall v. Denno, supra, at 301, that a prospective-only holding would violate the Art. Ill requirement of ease or controversy. But see England v. Louisiana State Board of Medical Examiners, 375 U. S. 411, 422 (1964), where the Court exempted the petitioner from its holding. See also Johnson v. New Jersey, supra, at 733.
While I subscribe to many of the reservations expressed by Mr. Justice Harlan, I nonetheless find his alternative rule of retrospec-tivity unsatisfactory. In Mackey v. United States, 401 U. S. 667, 675 (1971) (separate opinion), he suggested that constitutional decisions be retroactive as to all nonfinal convictions pending at the time of the particular holdings, but that prisoners seeking habeas relief should generally be treated according to the law prevailing at the time of their convictions. It is on this latter score that I am troubled. Surely it would be no more facile a task to unearth the *288state of law of years past than it is to assign, under the plurality’s test, a degree of reasonableness to reliance on older standards by law enforcement agencies. Where the question has arisen in this Court, we have treated habeas petitioners by the modem law, not by older rules. See Reck v. Pate, 367 U. S. 433 (1961) (habeas permitted on basis of current law to release prisoner convicted in 1936). See also Gideon v. Wainwright, 372 U. S. 335 (1963), and Jackson v. Denno, 378 U. S. 368 (1964), announcing new rules in habeas cases. Moreover, as has been concluded by Professor Schwartz, the drawing of a bright line between federal review through habeas and certiorari would be unjustified:
“Where federal review of the constitutionality of state criminal proceedings is concerned, the making of so sharp a distinction between review on certiorari and habeas corpus is unwarranted. There is often no significant difference with respect to age and potential staleness between the two types of cases. Rather than coming years after the conviction is final, habeas corpus is often but a routine step in the criminal defense process — the normal step taken after certiorari has been denied. Sometimes, it actually replaces certiorari, for in Fay v. Noia [372 U. S. 391 (1963)] the Supreme Court advised criminal defendants to skip certiorari and to petition directly to the federal district court for habeas corpus. Even in situations in which a defendant goes through all the direct review steps, it is often nothing more than fortuitous circumstance which determines whether his case is still on direct review or is on collateral attack when the new decision comes down.
“The difference between review on certiorari and habeas corpus seems even less significant when we look to function and actual operation. Although it is sometimes considered the 'normal’ method for obtaining federal review of state convictions, certiorari does not provide, as the Court remarked in Fay v. Noia, ‘a normal appellate channel in any sense comparable to the writ of error,’ for the Court must limit its jurisdiction to questions that have significance beyond the immediate case. Habeas corpus, on the other hand, facilitates the Court’s task in those cases it does take by providing a record focused exclusively on the federal constitutional question. Habeas corpus has thus become the primary vehicle for immediate federal *289review of state convictions. Further, this development has resulted in a gradual shrinking of what were once significant operational differences between review on certiorari and habeas corpus, such as the relationship to the state proceeding, the degree of independent fact-finding authority, and the significance of the defendant’s violation of state procedural rules. From both the functional and the operational standpoints, then, it is justifiable to conclude that ‘the distinctions between habeas corpus proceedings and direct review are largely illusory.’
“In addition, drawing a fine between review [on] certiorari and habeas corpus undercuts the Supreme Court’s bypass suggestion in Fay v. Noia. If a defendant has doubts about the retroactivity of any claim which might both affect him and be subject to Court review in the foreseeable future, he will be well advised always to ignore the Court’s suggestion and to apply for certiorari. Many months may pass before his petition for certiorari is rejected, and so long as it is pending, he will be entitled to receive the benefits of any intervening decisions. As soon as he files his petition for habeas corpus, however, even if he does so only a day after the last state court order is entered, he will have forfeited his right to such benefits. He will thus be put to an election between delayed relief and no relief at all.
“The inequity of drawing a sharp distinction between direct review and habeas corpus is, however, only one aspect of a broader inequity: treating two prisoners deprived of the same fundamental constitutional right differently merely because the Supreme Court did not get around to enunciating a particular right until after the conviction of one of them had become final. Professor Mishkin argues that worry about this point ignores ‘the reasons for barring current convictions and . . . the fact that the new rule in no way undermines the earlier determinations of factual guilt.’ To him, it is as if a guilty person were to complain of his lot because others equally guilty were not prosecuted. And though he recognizes that such claims are sometimes sustained, he concludes that ‘there are certainly rational bases for drawing a line between current convictions and *290those previously final/ citing excerpts from Professors Bator and Amsterdam on finality. Professor Mishkin’s sharp distinction between collateral attack and direct review thus rests ultimately on finality considerations.
“Finality considerations seem especially weak where two cases differ only in the fact that one is still on 'direct’ review whereas the other is not. Where the two cases are far apart in age, finality considerations are admittedly more persuasive. But even there, the mere timing of the Court’s decision to grant federal protection to a fundamental right hardly seems to be a sufficient basis for unequal treatment; after all, in most instances it was not the older prisoner’s fault that the Court did not render its decision earlier. To some extent, of course, the question comes down to a choice between the competing values of equality and repose, and choices of this sort are notoriously immune to reasoned resolution. It will be suggested below, however, that the threat to finality considerations from complete retroactivity appears to have been greatly exaggerated, and if this suggestion is well taken, Professor Mishkin’s rejection of equality is especially untenable.” Schwartz, Retroactivity, Reliability, and Due Process: A Reply to Professor Mishkin, 33 U. Chi. L. Rev. 719, 731-734 (1966).
See McConnell v. Rhay, 393 U. S. 2 (1968) (Mempa retroactive) ; Arsenault v. Massachusetts, 393 U. S. 5 (1968) (White and Hamilton retroactive).
The investigative advantage enjoyed by the State extends beyond the prohibition of the common law against criminal discovery. It also results from the fact that the police are usually first at the scene of the crime, have access to witnesses with fresher recollections, are authorized to confiscate removable evidence,, are positioned to conduct laboratory tests on physical evidence, enjoy a communication channel with a complete undercover world of secret informers, have an air of legitimacy which is conducive to cooperation by witnesses, and have numerous ways to compel testimony even before trial. See generally Norton, Discovery in the Criminal Process, 61 J. Crim. L., C. & P. S. 11, 13-14 (1970); Comment, Criminal Law: Pre-Trial Discovery — The Right of an Indigent's Counsel to Inspect Police Reports, 14 St. Louis U. L. J. 310 (1969); Moore, Criminal Discovery, 19 Hastings L. J. 865 (1968); A State Statute to Liberalize Criminal Discovery, 4 Harv. J. Legis. 105 (1967); Comment, Disclosure and Discovery in Criminal Cases: Where Are We *292Headed?, 6 Duquesne U. L. Rev. 41 (1967); Bibliography: Criminal Discovery, 5 Tulsa L. J. 207 (1968); Symposium: Discovery in Federal Criminal Cases, 33 F. R. D. 53 (1963); Brennan, Criminal Prosecution: Sporting Event or Quest For Truth?, 1963 Wash. U. L. Q. 279.
I am aware that the retroactivity theory presently commanding a Court permits a distinction between rules designed to fortify the reliability of verdicts and rules designed to protect other values. But here, as the plurality suggests, three of the four functions counsel might serve at preliminary hearings would appear to enhance the factfinding process: discovery of the State’s case, preserving of testimony of both hostile and favorable witnesses, and obtaining release on bail. Although the plurality appears to discount the investigative advantage of being free on bail, I believe that this “traditional right to freedom before conviction permits the unhampered preparation of a defense.” Stack v. Boyle, 342 U. S. 1, 4 (1951). See also Kinney v. Lenon, 425 F. 2d 209, 210 (CA9 1970), where the Court of Appeals found that “the appellant is the only person who can effectively prepare his own defense,” because the incarcerated, accused was the only person who could recognize witnesses by sight who might have seen a scuffle.
To this list might have been added Roberts v. LaVallee, 389 U. S. 40 (1967), holding that the State must provide an indigent with a preliminary hearing transcript in every circumstance in which the more affluent accused could obtain one.
Federal Bureau of Prisons, National Prisoner Statistics — Characteristics of State Prisoners, 1960, pp. 26-27 (1965).
In this respect the instant ease further differs from Stovall v. Denno, 388 U. S., at 299, where it was found that: “The law enforcement officials of the Federal Government and of all 50 States have heretofore proceeded on the premise that the Constitution did not require the presence of counsel at pretrial confrontations for identification.”
Brief for Respondent 33.
See n. 7, supra.