THORNBERRY, Circuit Judge:
Appellant Paul Swicegood, a state prisoner, was convicted of robbery in 1976. His conviction was affirmed by the Alabama Court of Criminal Appeals, Swicegood v. State, 343 So.2d 806 (1977), and the Alabama Supreme Court declined to hear the case. Ex Parte Swicegood, 343 So.2d 810 (1977). Appellant then brought this habeas corpus action in federal district court. 28 U.S.C. § 2254.
The district court denied the writ, concluding that although there had been an illegal arrest, federal review was precluded by Stone v. Powell, 428 U.S. 465, 96 S.Ct. 3037, 49 L.Ed.2d 1067 (1976). The court also rejected Swicegood’s argument that lineup procedures were constitutionally deficient but admitted that the question was “very close.” For the reasons stated below, we affirm the district court’s application of Stone but remand for further proceedings in connection with the lineup issue.
I. FACTUAL BACKGROUND
Pauline and Claxton Carter operated a Birmingham night club called the New Spot Lounge. In the early morning hours of May 4, 1975, Mrs. Carter returned home from the club and was seized by two men wearing ski masks, carried to her bedroom, bound, gagged, and blindfolded. Mr. Carter arrived an hour and a half later and, upon entering the house, was struck in the face by the intruders, who then tied and blindfolded him and locked him in a closet. The men escaped with approximately $7,000 in proceeds from the club and various personal property belonging to the Carters.1
Both robbers wore ski masks that covered their faces, and Mr. Carter saw neither without his mask. Mrs. Carter, however, was able to see one man for no more than two seconds when he unblindfolded her for purposes of identifying the keys to her car. The man, who had pushed his mask off his face, saw her looking at him, struck her, and blindfolded her again. Both of the Carters heard the men talk, although Mrs. Carter’s blindfold covered her ears at least part of the time and Mr. Carter spent most of the ordeal locked in a closet. The incident spanned approximately two hours, from 4 to 6 a. m.
A Birmingham police detective arrested Swicegood at his trailer house on May 22, 1975, without a warrant. The detective subsequently testified that the basis for the arrest was information from a confidential informant, who had been reliable for several years, that Swicegood had participated in the robbery and was in possession of a handgun. On the day of the arrest, the detective and other officers went to the trailer and identified themselves as lawmen. Swicegood came to the door naked and then went into another room as a young boy let the officers in. The police found a .25 caliber pistol in Swicegood’s bedroom and arrested him for illegal possession of a handgun.2 Swicegood was also informed that he was a robbery suspect.
On the day after the arrest, Swicegood, who at that point had not been charged with the handgun offense, was required to participate in a lineup. Neither of the Carters was able to identify him when they first viewed the lineup with the participants wearing masks; rather, both tenta[1324]*1324tively identified other individuals. After a second viewing — this time with the men unmasked — both “positively” identified Swicegood, and both subsequently made in-court identifications of him and testified that they had picked him out of a police lineup. The circumstances surrounding the lineup are discussed in more detail in Section III, infra.
II. FOURTH AMENDMENT CLAIM
Appellant contends that he was arrested without probable cause in contravention of the fourth amendment and that the product of that arrest — the lineup identification— was improperly admitted into evidence. '
In Stone v. Powell, supra, the Supreme Court held that “where the State has provided an opportunity for full and fair litigation of a Fourth Amendment claim, the Constitution does not require that a state prisoner be granted federal habeas corpus relief on the ground that evidence obtained in an unconstitutional search or seizure was introduced at his trial.” 428 U.S. at 482, 96 S.Ct. at 3046. As this Court explained in O’Berry v. Wainwright, 546 F.2d 1204, 1209-10 (5 Cir.), cert. denied, 433 U.S. 911, 97 S.Ct. 2981, 53 L.Ed.2d 1096 (1977), our initial inquiry is whether the state provided an opportunity for a full and fair hearing on the fourth amendment issue. If it did, our review is precluded by Stone.
Appellant argues that he did not receive full consideration because the state courts dealt with the illegal arrest question only “in passing” and “without any reasoned consideration” and that he was denied fair consideration because the state courts misapplied settled principles of constitutional law. If we were to accept these arguments, Stone would be rendered meaningless.
In O’Berry v. Wainwright, supra, this court held that there had been full and fair consideration where the state appellate court merely said that each contention had been examined and found without merit, that there had been no timely objection to preserve the error, and that there was no plain error. 546 F.2d at 1216. In the instant case, appellant presented the question to the trial court, the Alabama Court of Criminal Appeals, and the Supreme Court of Alabama, and all rejected his fourth amendment argument. Further, the Court of Criminal Appeals wrote in its opinion:
We have carefully examined the record and determined from the testimony of [the arresting officer] that there was sufficient information possessed by him upon which to arrest the appellant. He was aware that a felony had been committed and that the appellant was a participant therein. He was also aware of the appellant’s past criminal record.
Swicegood v. State, supra, 343 So.2d at 809. Moreover, on petition for rehearing, one judge on the Court of Criminal Appeals dissented, arguing that there was no probable cause for arrest and that the fruit of that arrest, i. e., the in-custody identification, was thus inadmissible, 343 So.2d at 809-810. In these circumstances, appellant cannot successfully argue that he did not receive full consideration of his claim.
Appellant’s second point — that the state denied him a fair hearing by misapplying federal constitutional law — must also fail. If this argument were correct, the federal courts would consider the merits of fourth amendment habeas cases whenever the state courts erred in their fourth amendment analysis and would refuse to consider the merits of those cases in which the state courts were correct. That, of course, is federal habeas review of state court fourth amendment decisions, precisely what Stone forbids. If the term “fair hearing” means that the state courts must correctly apply federal constitutional law, Stone becomes a nullity.
Moreover, the Court in Stone rejected the notion that federal judges are institutionally more receptive to federal constitutional norms that are their state brethren. 428 U.S. at 493-94 n. 35, 96 S.Ct. 3037. Underlying Stone is the conviction that state courts are as competent as the federal courts in insuring that federal constitutional rights are protected. O’Berry v. Wainwright, supra, 546 F.2d at 1215. That premise is obviously destroyed if a federal [1325]*1325court reaches the merits in a fourth amendment habeas case simply because it believes the state courts decided the issue incorrectly-
Having found that appellant received an opportunity for full and fair consideration of his fourth amendment claim, we now turn to a second argument designed to rid his path of the Stone obstacle: that the rule of that decision is inapplicable where, as here, the fruit of the illegal arrest is not physical evidence but rather a lineup identification.
This argument can be viewed in two ways. At its broadest, it is a contention that Stone does not preclude consideration of the legality of an arrest that leads to non-physical evidence, such as identification at a lineup. Thus, we would determine whether the lineup identification, a fruit of the allegedly unlawful arrest, was properly admitted against appellant. However, our prior decisions make clear that Stone applies with equal force to lineup evidence and physical evidence, and we necessarily reject this contention. Of course, we need not ponder the effect of Stone on cases involving other types of nonphysical evidence, since the situation before us presents only a lineup identification.
In Caver v. Alabama, 537 F.2d 1333 (5 Cir. 1976), cert. denied, 430 U.S. 910, 97 S.Ct. 1183, 51 L.Ed.2d 587 (1977), and White v. Alabama, 541 F.2d 1092 (5 Cir. 1976), cert. denied, 430 U.S. 910, 97 S.Ct. 1183, 51 L.Ed.2d 587 (1977), the appellants contended that their arrests were merely pretextual and that subsequent voice identifications obtained from a lineup held while they were in police custody were inadmissible. In both cases we concluded that Stone would bar this type of collateral attack unless the appellants could show that the state failed to provide them an opportunity for full and fair litigation of the fourth amendment claim. 537 F.2d at 1336; 541 F.2d at 1093. In the instant ease, we have determined that appellant received such an opportunity, and, accordingly, we hold that Stone precludes consideration of the arrest question.3
A second reading of appellant’s argument is also possible: Stone does not apply when an appellant has urged that the lineup procedures violated his due process rights. So read, the assertion is undoubtedly correct, for the claim relates not to the validity of the arrest but rather to the suggestiveness of the lineup, and Stone is no bar to consideration of this alleged violation of due process. Caver v. Alabama, supra; White v. Alabama, supra. We now proceed with the due process inquiry.
III. SUGGESTIVE LINEUP
In determining whether police identification procedures are “unnecessarily suggestive and conducive to irreparable mistaken identification” and thus violative of due process, we are required to examine the “totality of the circumstances.” Stovall v. Denno, 388 U.S. 293, 302, 87 S.Ct. 1967, 1972, 18 L.Ed.2d 1199 (1967); Allen v. Estelle, 568 F.2d 1108,1111 (5 Cir. 1978). Our initial inquiry is whether the identification procedure was unnecessarily suggestive; if so, the court must then determine whether the procedure created a substantial risk of misidentifieation. Allen v. Estelle, supra at 1112; United States v. Smith, 546 F.2d 1275 (5 Cir. 1977); United States v. Gidley, 527 F.2d 1345 (5 Cir. 1976). Only when the eyewitness identification is “unreliable because of suggestive line-up procedures should a court rule that the evidence is constitutionally inadmissible as a matter of law.” Caver v. Alabama, supra at 1335.
The district court concluded that the suggestiveness question was “very close,” but did “not feel justified” in upsetting the judgment of the Alabama courts that had “carefully considered this point and decided adversely to the Petitioner.” Neither court directly addressed the reliability question, having determined that the lineup was not impermissibly suggestive.
Our suggestiveness inquiry requires a detailed examination of the facts surrounding the lineup. Unfortunately, neither the district court nor the Alabama Court of Crimi[1326]*1326nal Appeals addressed these facts at length. The district court summarized the relevant events as follows:
Petitioner was held in custody without being presented to a committing magistrate and was required to participate in a lineup. Neither of the victims of the robbery identified Petitioner when they first viewed the lineup. The victims indicated at the first viewing that individuals in the lineup other than Petitioner appeared to be one of the robbers. On a second viewing of the lineup, Mrs. Carter identified Petitioner and subsequently made an in-court identification of Petitioner.
The Alabama Court of Criminal Appeals went into somewhat more detail, see 343 So.2d at 808, and our own independent review of the record has uncovered additional facts that we deem significant. '
Six white males, including appellant Swi-cegood, participated in the lineup, which was held some three weeks after the robbery. Although all were then being detained at the jail, none wore a jail uniform; rather, all were attired in their own “street” clothes. The men ranged in height from 5'7" to 5'11" and in weight from 125 to 150 pounds. Some had moustaches and some did not, and their hair color varied.4 In talking with police immediately after the robbery, Mrs. Carter had described the intruder who had briefly removed his mask as approximately 5'8" and 160 pounds, although she later testified that he had light-colored hair and was clean shaven. All of the men in the lineup were 26 years of age or younger, except Swicegood, who was 35 at the time.5
The Carters viewed the lineup separately, each having two opportunities to do so. At the first viewing, the men all wore ski masks as the robbers had done during the crime. Neither Mr. nor Mrs. Carter could make an identification at that point, although all six men, using a microphone-intercom system, spoke phrases used by the intruders. Mrs. Carter indicated that Swice-good’s voice sounded familiar but could not be sure. She tentatively identified another man as the robber. Mr. Carter thought he recognized the voice of a third man but also could not be certain. After they had viewed the lineup, Mr. and Mrs. Carter went to a waiting room where other witnesses6 — but no police personnel — were present. Each of them viewed the lineup a second time, with the men having removed the masks. Mrs. Carter identified Swice-good, as did her husband. Mr. Carter had not seen the robber’s face and said his identification was made by voice alone. However, he admitted having previously seen Swicegood in his night club.
After the lineup, a police officer apparently told the Carters that Swicegood, whom they had identified, was “the suspect that we had.” The officer indicated that he did not remember having made the statement but said he could have made it. Mrs. Carter’s testimony on the point was equivocal. Counsel for Swicegood was present at the lineup, although another attorney represented him at trial.7 During trial, Mr. and Mrs. Carter each identified Swicegood in court and testified that they had picked him out of a lineup. Defense counsel had made a pretrial motion to suppress the line[1327]*1327up identification and renewed this objection at trial.
Persons in a lineup can be required to speak words or phrases supposedly uttered by the culprit. United States v. Wade, 388 U.S. 218, 222-23, 87 S.Ct. 1926, 18 L.Ed.2d 1149 (1967); see also Crume v. Beto, 383 F.2d 36 (5 Cir. 1967). Moreover, the disparate physical appearances of the lineup participants is not alone sufficient to warrant a finding of suggestiveness. Caver v. Alabama, supra; United States v. Reid, 517 F.2d 953, 965-66 n.15 (2 Cir. 1975); United States ex rel. Pella v. Reid, 527 F.2d 380, 384 (2 Cir. 1975); United States v. Jackson, 166 U.S.App.D.C. 166, 172, 509 F.2d 499, 505 (1974). “Police stations are not theatrical casting offices; a reasonable effort to harmonize the lineup is normally all that is required.” United States v. Lewis, 547 F.2d 1030, 1035 (8 Cir. 1976), cert. denied, 429 U.S. 1111, 97 S.Ct. 1149, 51 L.Ed.2d 566 (1977).
Accordingly, if only these problems were involved in the instant case, we would have no difficulty holding that the lineup was not impermissibly suggestive.8 However, other factors come into play in our analysis of the “totality of the circumstances.” First, there was a considerable difference in the ages of the lineup participants, with Swicegood being nine years older than the next-oldest man and 15 years older than the youngest. Although age generally manifests itself in appearance-related characteristics and courts often treat age as a physical attribute, e. g., United States v. Hines, 147 U.S.App.D.C. 249, 455 F.2d 1317, cert. denied, 406 U.S. 969, 975, 92 S.Ct. 2427, 32 L.Ed.2d 669, 675 (1972), we think it should be treated as an independent element in this case. A person’s age, for example, often has a decided effect on his vernacular, speech patterns, or voice characteristics, and both Mr. and Mrs. Carter identified Swicegood on the basis of his voice. Cf. United States v. Kopacsi, 488 F.2d 900, 903-04 (5 Cir. 1973), cert. denied, 416 U.S. 987, 94 S.Ct. 2393, 40 L.Ed.2d 765 (1974) (dissimilarity of accents among line-' up participants). Further, a significant disparity in the ages of the lineup participants could influence a witness who believed that the culprit was an “older” person. Thus, a lineup with only one “older” person could impermissibly point the finger at that individual, albeit in a much more subtle fashion than in other situations. See United States v. Wade, supra, 388 U.S. at 232, 87 S.Ct. 1926.
Second, the Carters had the opportunity to discuss the lineup during the period between the first and second viewings, although Mr. Carter testified that they did not do so. However, upon the second viewing both Mr. and Mrs. Carter identified Swicegood, although neither had picked him on the first viewing and each had tentatively identified different persons.9 Moreover, the police did not take relatively simple precautionary measures to guard against communication between witnesses, for example, in United States v. Hines, supra, the police changed the positions of the lineup participants between the first and second viewings. 147 U.S.App.D.C. at 262, 455 F.2d at 1330. We have previously indicated that one of the factors on which the fairness of pretrial lineup procedures depends is whether witnesses were allowed to “discuss among themselves their conclusions.” Pearson v. United, States, 389 F.2d 684, 688 (5 Cir. 1968).
Third, Mr. Carter was allowed to examine the men without masks, although he had seen the robbers only with their masks on. He was unable to identify Swicegood after the first viewing — when the lineup partici[1328]*1328pants wore masks — but made a “positive” identification after seeing them again, this time without masks. Moreover, Mr. Carter testified that he had seen Swicegood on previous occasions in the Carters’ club, a fact that suggests he may have selected Swicegood because he was the only member of the lineup who looked familiar.
Finally, the police, in putting together the lineup, were guided only by a very general description given by Mrs. Carter on the night of the crime — white male, 5'8", 160 pounds. Neither of the Carters was asked to provide a more precise physical description prior to the lineup, nor were they requested to give a description of the robbers’ voices. In addition, the police record of the lineup was complete as to all of the mep except Swicegood, whose clothing was not described.
After carefully analyzing the circumstances surrounding the lineup, we agree with the district court that the suggestiveness question is “very close.” Although the lineup may not have been overwhelmingly suggestive, we think it was “so close to the mark” that due process considerations compel us to examine the reliability of the identification. See Allen v. Estelle, supra at 1113; United States v. Lewis, supra at 1035; United States v. Jackson, supra, 166 U.S.App.D.C. at 172, 509 F.2d at 505.
As the Supreme Court recently emphasized, “reliability is the linchpin in determining the admissibility of identification testimony . . . .” Manson v. Brath-waite, 432 U.S. 98, 114, 97 S.Ct. 2243, 2253, 53 L.Ed.2d 140 (1977). Analysis of the reliability issue requires examination of several factors: the opportunity of the witness to view the criminal at the time of the crime; the witness’ degree of attention; the accuracy of the witness’ prior description of the criminal; the level of certainty demonstrated at the confrontation; and the time between the crime and the confrontation. “Against these factors is to be weighed the corrupting effect of the suggestive identification itself.” Id. This procedure requires careful analysis of the facts, and we note in this regard that the district court did not specifically undertake such an examination.
I- The opportunity to view or listen. Mrs. Carter had only a fleeting opportunity to view the robber, and her husband had no chance whatsoever. Although they had ample opportunity to hear the men talk, listening conditions were far from optimum — Mr. Carter was locked in a closet, and Mrs. Carter’s blindfold covered her ears at least part of the time. Moreover, at trial Mrs. Carter admitted that the robber had only an “ordinary” voice with no distinguishing characteristics, and Mr. Carter, although maintaining that “all of us have distinctive voices,” could not point to anything unusual about the voice. Finally, common experience teaches that voice identification is not as reliable as visual identification, although in the proper case voice identification can be sufficient. See Caver v. Alabama, supra; United States v. Moore, 571 F.2d 76, 91 (2 Cir. 1978). Cf. United States v. Ladd, 527 F.2d 1341 (5 Cir. 1976); United States v. Turner, 528 F.2d 143 (9 Cir. 1975).
2. The decree of attention. We have little doubt that the Carters, who were physically mistreated and threatened with death, paid rapt attention to their captors. This is certainly not a situation in which the witnesses were casual bystanders. Cf. Neil v. Biggers, 409 U.S. 188, 93 S.Ct. 375, 34 L.Ed.2d 401 (1972) (eyewitness testimony of rape victim); Moore v. Illinois, 434 U.S. 220, 98 S.Ct. 458, 468, 54 L.Ed.2d 424 (Blackmun, J., concurring). However, the Court emphasized in Manson that the eyewitness was a “trained police officer on duty” whose powers of observation were likely to be considerably greater than those of the ordinary individual. Although neither of the Carters could be characterized as such an expert, both testified that they attempted to listen carefully to the voices of the robbers.
3. The accuracy of the description. The only description given police prior to the lineup came from Mrs. Carter: white male, 5'8'', 160 pounds. Apparently neither Mr. nor Mrs. Carter made an effort to describe the robbers’ voices in advance of the lineup. Therefore, we are left with no measuring stick except Mrs. Carter’s generalized state[1329]*1329ment to police immediately after the crime. Swicegood is a white male, 5'7", 150 pounds, according to the police record of the lineup and thus comes very close to Mrs. Carter’s description; however, so do the other five men who took part in the lineup.
4. Witness’ level of certainty. After first viewing the lineup, neither of the Carters could make a positive identification. In fact, each selected different individuals as the likely robber. See footnote 9, supra. After the second viewing, however, both positively identified Swicegood, and at trial both testified that they were positive Swi-cegood was one of the robbers. This in-court certainty could have been tainted, however, by a police officer’s remark after the lineup to the effect that the Carters had chosen the “right man.” 10
5. The time between crime and confrontation. There was a considerable delay between the robbery and the lineup. The crime took place on May 4, 1975, and the lineup was conducted on May 23, nearly three weeks later. By way of contrast, the witness in Manson identified the defendant’s photograph only two days after the crime, and the Court noted that “[w]e do not have here the passage of weeks or months between the crime and the viewing of the photograph.” 432 U.S. at 116, 97 S.Ct. at 2253. In Allen v. Estelle, supra, a police “showup” took place only an hour and a half after the crime.
Manson requires that we assess the reliability of the identification by weighing against these “indicators” the “corrupting effect” of the lineup procedure. We cannot say that the likelihood of misidentifieation was slight or that the suggestiveness of the lineup was an uncorrupting influence. Unlike McGuff v. Alabama, 566 F.2d 939 (5 Cir. 1978), the instant case is not one in which “[o]nly the great reliability of these identifications saves the procedures utilized from constitutional infirmity.” Id. at 941. There is no such “great reliability” here, and we conclude that under all the circumstances of this case, the suggestiveness of the lineup and the unreliability of the identifications coalesce to produce a due process violation.
Because the error was not harmless beyond a reasonable doubt, we reverse the district court’s denial of Swicegood’s application for writ of habeas corpus and direct that the writ be granted, subject to the state’s right to retry him within a time established by the district court after consultation with the parties.
REVERSED and REMANDED with directions.