Robert Mata appeals from the District Court’s denial of his Petition for Writ of Habeas Corpus. The Petition followed Mata’s murder conviction in California state court and his exhaustion of all available state court remedies. We reverse.
I.
In this opinion we endeavor for the third time to set forth the factual and legal bases for our firm conclusion that impermissibly suggestive pretrial photographic identification procedures violated Mata’s constitutional right to due process of law. The United States Supreme Court vacated and remanded our initial disposition, Mata v. Sumner, 611 F.2d 754 (9th Cir.1979), citing our failure to apply the presumption of correctness mandated by 28 U.S.C. § 2254(d) to the factual determinations made by the California Court of Appeal in this case. Sumner v. Mata, 449 U.S. 539, 547, 101 S.Ct. 764, 769, 66 L.Ed.2d 722 (1981). On remand we considered the state court’s factual determinations in the context of 28 U.S.C. § 2254(d) and concluded that, because we “substantially agree[d]” with the historical or basic facts adduced by the California court, the requirement of explaining by reasoned written references the justification for dispensing with the presumption was inapplicable. Mata v. Sumner, 649 F.2d 713, 715-16, 717 (9th Cir.1981). See Sumner v. Mata, 449 U.S. at 549, 101 S.Ct. at 770. The Supreme Court again vacated and remanded, citing our failure to heed the Court’s earlier conclusion that certain of our factual findings were “considerably at odds” with the findings made by the state appellate court. Sumner v. Mata, 455 U.S. 591, 102 S.Ct. 1303, 1305, 1307, 71 L.Ed.2d 480 (1982) (per curiam).
Having received additional briefing by the parties, we now attempt to review carefully the facts of this case, as reflected in the trial transcript and the findings of the California Court of Appeal, in compliance with 28 U.S.C. § 2254(d) and the most recent mandate of the Supreme Court. In doing so we emphasize that the Court has not reached the merits of our prior conclusion that the pretrial photographic identification procedures were so impermissibly suggestive in this case as to give rise to a very substantial likelihood of irreparable misidentification. See 102 S.Ct. at 1307; 449 U.S. at 552, 101 S.Ct. at 771. Our reevaluation of the facts only strengthens our conviction that the state has denied Mata that fairness required by the Due Process Clause of the Fourteenth Amendment. See Manson v. Brathwaite, 432 U.S. 98, 113, 97 S.Ct. 2243, 2252, 53 L.Ed.2d 140 (1977).
II.
On October 19,1972, inmate Leonard Arias was stabbed to death in a dormitory at the California Correctional Institution at Tehachapi, California. People v. Vargas, 5 Crim. 1735, slip op. at 2 (Cal.Ct.App. Aug. 4, 1975) (unpublished opinion by the Fifth Appellate District on Mata’s direct appeal after trial) [hereinafter cited as Fifth District Opinion]; see Reporter’s Transcript of Trial (“R.T.”) at 365-96, 582-630 (eyewitness description of the incident). Three inmates were charged with the murder: Mata, David Gallegos, and Salvador Vargas. Each denied partaking in the attack and *1247each produced evidence tending to show that he was at another place in the prison at the time of the attack. Fifth District Opinion at 2. The prosecution elicited testimony that the murder was the result of feuding between opposing prison factions— the “Mexican Mafia” and the “Nuestra Family.” See R.T. at 805 — 51 (testimony of member of Mexican Mafia). The victim, Arias, was reputed to be a member of the Nuestra Family who had allegedly previously attacked a member of the Mexican Mafia during his term at San Quentin, another California prison facility. See id. Arias had arrived at Tehachapi the day before he was killed. Fifth District Opinion at 2.
A.
At about 1:30 p.m. on October 19, Arias and fellow inmates Rigoberto Almengor and Jay Allen were drinking coffee and talking near Allen’s bunk in Dormitory Eight. Fifth District Opinion at 2; R.T. at 365-69, 582-84. A call over the institution’s loud speaker ordered all new arrivals to report to the laundry room and, since Arias was a new arrival, Allen and Arias proceeded to the laundry room. Fifth District Opinion at 2; R.T. at 372-73, 584-85. Almengor, after starting to walk to the television room, decided to join Arias and Allen after Allen asked him to come along. R.T. at 372-73, 585.
On the way to the laundry room they passed three men who were standing by the bulletin board. These three then turned and attacked Arias. During the attack Arias was stabbed with a knife several times, causing his death. Fifth District Opinion at 2. Almengor testified that he was walking about even with Arias when the three men at the bulletin board area “turned around real fast” and began attacking Arias. R.T. at 373-74, 376. Almengor thought he should help Arias and approached the assailants. R.T. at 379, 393. The man Almengor later identified as Mata “slooped” away from attacking Arias and came at him with a knife. R.T. at 392, 747. Almengor and this man struggled for a few seconds, R.T. at 960, 747, 381. Almengor threw a kick at him, R.T. at 747, 380, then they both backed off each other. R.T. at 381, 747. The two then stood looking directly at each other, eye-to-eye, about six feet apart, for “several seconds, several minutes.” R.T. at 381-82, 742, 747-49, 960. Almengor, who is nearsighted, was not wearing his glasses at the time, R.T. at 449-50, but he testified that he didn’t need them for the distance involved in the attack, R.T. at 489-91.
Allen testified that he didn’t see who was fighting with Almengor. R.T. at 589, 796-97. Allen stood in front of and was preoccupied by another of the attackers, later identified as Vargas, who carried a large knife. At one point, this assailant lunged at Allen. R.T. at 614-16, 796. Allen backed up and then turned away from the fight, yelling for help. R.T. at 628-29. “It happened fast,” Allen testified. R.T. at 642. When he turned back around the three assailants had turned and were fleeing; he grabbed hold of the fatally wounded Arias and lay him down. R.T. at 629-30, 619-20. Allen later told an investigator for the defendants that he never got a good enough look at any of the assailants to identify them. R.T. at 1110.
B.
Prison investigators interviewed Allen and Almengor on several occasions after the stabbing, each time employing photographic arrays in an attempt to obtain an identification of the attackers:
October 19
Correctional Lieutenant Robert King and Investigator Maurice Higgins interviewed Almengor at about 5:00 p.m. on the day of the incident. Fifth District Opinion at 3; R.T. at 707-10. Almengor asked to see photographs of all the Mexican-American inmates in the facility; Lieutenant King showed him between two and five hundred mug shots of prison inmates. R.T. at 397, 285 — 86, 422, 739. Almengor made a “positive” identification of inmate Pete Nunez as one of the attackers. R.T. at 711. The record is conflicting as to whether Almen*1248gor named Nunez before or after he was shown Nunez’ photograph. See R.T. at 398-400, 420, 707-19, 730-38. Lieutenant King told Almengor that Nunez could not have been involved in the incident because Nunez was not in the prison at the time of the stabbing. R.T. at 713. Almengor made a “possible,” “tentative,” but “never positive” identification of three to seven other inmates pictured in the mug shots. R.T. at 401-02, 432, 437-38, 712, 730-38. One of these was Vargas; also included were inmates Pete Ramirez and Jay Reymundo. Fifth District Opinion at 3; R.T. at 734. Almengor did not identify Mata as a participant. Id. Although Almengor requested to sée the inmates in a line-up, a line-up was never held. R.T. at 781.
Lieutenant King and Officer Higgins also interviewed Allen on this date, but they did not show him any photographs. R.T. at 719-20. Allen said he could make no identification of the participants in the attack. R.T. at 720.
October 27
Almengor was shown a group of 24 photographs of inmates, including photographs of the three defendants. R.T. at 714-15. The defendants’ photographs had been taken three to eight months earlier. R.T. at 917-18. Many of the other photographs in the array were considerably older — some as many as three, four, and even eight years old. R.T. at 918. Outside of Allen’s presence, Almengor “possibly” identified Vargas, but it wasn’t a “positive” identification. R.T. at 715. Again he failed to identify Mata. R.T. at 916. Almengor complained that the photographs were too old and requested up-to-date photographs. R.T. at 716, 916. Also, he repeated his request to see the inmates in a line-up. R.T. at 425, 496-97, 783, 932, 959.
Allen was shown the same group of photographs, but he did not identify anyone. Fifth District Opinion at 3; R.T. at 720-21. Like Almengor, he said the photographs appeared old. R.T. at 721.
October 27 — October 30
Prison authorities shot new photographs of a number of inmates, including the defendants, but they were not useable because of photographer error. R.T. at 716. The photographs of the three defendants were the only ones taken a second time. R.T. at 754. Allen and Almengor, who were housed in segregation together with the defendants and other inmates, saw the three defendants being taken to have their photographs retaken. R.T. at 444r-46, 634, 1118-19. Almengor testified that he could not tell who the three men were but that he knew they were being taken to have their pictures retaken. R.T. at 444 — 46. However, Almengor said he knew that one of the three was defendant Gallegos. R.T. at 452. Allen told an investigator for the defendants that he recognized the three defendants as they were taken out and that, since they were the only ones in segregation to appear in the photographic array on October 30, he assumed “they must be the ones that did it.” R.T. at 1118-19. At trial, however, he testified that he did not remember making such a statement. R.T. at 634. Moreover, while in segregation, both Allen and Almengor had spoken with Gallegos. R.T. at 443, 633.
October 30
Lieutenant King showed Almengor a group of 15 inmates’ photographs, including the new photographs of the three defendants. Fifth District Opinion at 3; R.T. at 716. The defendants’ photographs were dated October 27, 1972; the others were older — some as many as two years old. R.T. at 755-56. Not included in this group were the photographs of Ramirez, Reymundo, and Nunez. R.T. at 717-18. Almengor identified, without qualification, the photographs of Mata, Gallegos, and Vargas as being photographs of the assailants. Fifth District Opinion at 3; R.T. at 718, 741. Allen was then brought into Lieutenant King’s office and, with Almengor at the other side of the room, Allen also identified the three defendants as the assailants. Fifth District Opinion at 3; R.T. at 447 — 48, 732. Allen was not advised that Almengor had selected the same three photographs. R.T. at 722.
*1249C.
During a taped portion of the interview on October 27 Almengor described the three assailants for Lieutenant King and Officer Higgins. R.T. at 969-71. He stated that they wore state-issued jackets, black gloves, and knitted blue Navy watch caps. R.T. at 969. The man he fought with, he said, had a thick mustache, and the man with the large knife had a thicker mustache and was dark complected — “he looked like an Indian, more or less, like an Indian dude.” R.T. at 970. At trial eight months later Almengor repeated his description, adding that the man with the large knife was tall and had a thick, solid black mustache. R.T. at 376, 416, 417, 419-20, 424, 479. However, he could not recall at trial if the man he fought with, whom he had identified as Mata, had a mustache. R.T. at 479. Almengor identified Mata and Gallegos in court as two of the assailants, R.T. at 374-75, but testified that he had been mistaken in his earlier identification of Vargas as the tall assailant with the large knife. R.T. at 501-02. Vargas, Almengor testified, was not as tall as the third attacker he saw. R.T. at 419.
Allen testified at trial that all three assailants wore state-issued clothes, gloves, and blue knit caps and had mustaches. R.T. at 621-22. However, in describing them to the defendants’ private investigator on November 22,1972, Allen stated that only one of the assailants had a mustache. R.T. at 1117-18. Also, he said he was not sure if they were Chicano or Caucasian. R.T. at 1118. Like Almengor, Allen made in-court identifications of Mata and Gallegos and testified that he had been mistaken in his earlier identification of Vargas. R.T. at 587, 623-24.
The 15 photographs shown to Allen and Almengor on October 30 included 10 men with mustaches, six of whom were pictured in state clothes. Three of those six photographs were the updated photographs of the defendants, while the other three photographs were less than a year old. Although taken at different times, most of the 15 photographs in the group were similar in terms of size, color, and pose. All of the inmates pictured were of Mexican descent.
D.
At trial Allen testified that prior to his making an identification, officers brought up with him the subject of his upcoming appearance before the parole board. R.T. at 634-35. He testified that “[t]hey just said they would look at it, you know” and that “[t]hey said it wouldn’t hurt.” R.T. at 635. At the preliminary hearing he testified that prison officials said he could “be shipped” if he didn’t cooperate. R.T. at 636. At trial he explained that this meant that if he didn’t cooperate he might be sent “some other place” where he “[mjight have a better chance of getting killed,” “or ending up getting myself deeper in trouble.” Id. Allen had also told the defendants’ private investigator on November 22 that one of the reasons he made the identification on October 30 was that he felt there was pressure from the Department of Corrections. R.T. at 1119.
III.
At trial the prosecution introduced evidence of the pretrial identifications by Allen and Almengor, as well as their in-court identifications of Mata and Gallegos, as part of its ease-in-chief. See R.T. at 365-506, 582-657, 707-804, 903-71. At the conclusion of the three-week trial the jury found all three defendants guilty of murder in the first degree; each was sentenced to life imprisonment. On direct appeal they raised for the first time the issue of impermissibly suggestive pretrial photographic identification procedures. In addressing this contention on its merits, the Fifth District Court of Appeal, after briefly reviewing the facts, see Fifth District Opinion at 2-3, rejected the argument on the following basis:
Appellants argue that the witnesses Almengor and Allen were housed in the same segregation unit with appellants, that they were aware that appellants were removed from the segregation unit to have their pictures taken and that this *1250makes their identification inadmissible. But they make no showing, and the record supports none, that the witnesses were in fact influenced in their identifications by this action of the investigating officers.
[W]e ... find that the photographs were available for cross-examination purposes at trial. We further find that there is no showing of influence by the investigating officers: that the witnesses had an adequate opportunity to view the crime; and that their descriptions are accurate. The circumstances indicate the inherent fairness of the procedure, and we find no error in the admission of the identification evidence.
Fifth District Opinion at 3-5.
We must accord the factual findings of the Fifth District the presumption of correctness required by 28 U.S.C. § 2254(d).1 Sumner v. Mata, 455 U.S. 591, 102 S.Ct. 1303, 1306-07, 71 L.Ed.2d 480 (1982). Unless one of the eight exceptions listed in section 2254(d) is found, e.g., that a finding is “not fairly supported by the record,” we must defer to the state court’s findings. Id. at 1307.
We have presumed correct the Fifth District’s description of the stabbing incident and the subsequent photographic identification procedures, contained at pages 2-3 of its opinion, and have incorporated those findings into our foregoing recitation of the facts. See part II supra. We conclude, however, that the Fifth District’s rather sketchy account did not adequately develop the material facts concerning the witnesses’ observations of the incident and the photographic procedures later employed by prison authorities. Therefore, our detailed account of the facts in part II supra includes additional facts and testimony, fairly supported by the record, that are material to the issue of photographic identification.
We have also examined the findings contained in the portions of the Fifth District’s opinion quoted above and recognize, in accordance with the opinion of the Supreme Court, 102 S.Ct. at 1305, that certain of the *1251findings therein are considerably at odds with the characterization of the facts in our earlier opinion. See Mata v. Sumner, 611 F.2d 754, 758-59 (9th Cir.1979). In facing up to that conflict, 102 S.Ct. at 1307, we now specifically hold that certain of the Fifth District’s findings are not fairly supported by the record and, therefore, need not be accorded the presumption of correctness mandated by section 2254. See 28 U.S.C. § 2254(d)(8); Sumner, 102 S.Ct. at 1307. In the following section we evaluate these findings in connection with the factors to be considered in determining the constitutionality of the pretrial identification procedures used in this case — a mixed question of law and fact not governed by section 2254. See 102 S.Ct. at 1306.
IV.
Suggestive pretrial photographic identification procedures may taint in-court identifications sufficient to deny the accused due process of law. United States v. Field, 625 F.2d 862, 865 (9th Cir.1980). As well, the admission of testimony concerning the out-of-court identification itself may work to deny due process. See Neil v. Biggers, 409 U.S. 188, 198, 93 S.Ct. 375, 381, 34 L.Ed.2d 401 (1972). In both situations the identification evidence must be excluded “if the photographic identification procedure was so impermissibly suggestive as to give rise to a very substantial likelihood of irreparable misidentification.” Simmons v. United States, 390 U.S. 377, 384, 88 S.Ct. 967, 971, 19 L.Ed.2d 1247 (1968); United States v. Hanigan, 681 F.2d 1127, 1133 (9th Cir.1982); Field, 625 F.2d at 865. See Biggers, 409 U.S. at 198, 93 S.Ct. at 381 (“irreparability” is considered only in context of in-court identifications).
Our review of the trial court’s admission of the in-eourt and out-of-court identification testimony must focus exclusively on the reliability of the identifications in light of the totality of surrounding circumstances. Manson v. Brathwaite, 432 U.S. 98, 113-14, 97 S.Ct. 2243, 2252-53, 53 L.Ed.2d 140 (1977); Biggers, 409 U.S. at 199, 93 S.Ct. at 382; Simmons, 390 U.S. at 382, 88 S.Ct. at 970; Field, 625 F.2d at 865-66. We must weigh the “corrupting effect” of the challenged identification procedure against indicia of the witness’ ability to make an accurate identification. Manson, 432 U.S. at 114, 97 S.Ct. at 2253; Biggers, 409 U.S. at 199-200, 93 S.Ct. at 382; Field, 625 F.2d 867-68. Applying these standards to the facts fairly supported by the record in this case, we conclude that the state employed photographic identification procedures so impermissibly suggestive as to give rise to a very substantial likelihood of misidentification.
A.
In Neil v. Biggers, 409 U.S. 188, 93 S.Ct. 375, 34 L.Ed.2d 401 (1972), the Supreme Court enunciated five factors that reflect a witness’ ability to make an accurate identification: (1) the opportunity of the witness to view the criminal at the time of the crime; (2) the witness’ degree of attention; (3) the accuracy of the witness’ prior description of the criminal; (4) the level of certainty demonstrated by the witness at the confrontation; and (5) the length of time between the crime and the confrontation. Id. at 199-200, 93 S.Ct. at 382. Each of these factors requires a finding of historical fact to which section 2254 applies. Sumner v. Mata, 102 S.Ct. 1303, 1307 n. 10 (1982). We now consider the Biggers factors in light of the trial court record and the Fifth District’s findings.
1. Opportunity to View
The Fifth District found that “the witnesses had an adequate opportunity to view the crime.” Fifth District Opinion at 5. The relevant inquiry, however, is the witness’ opportunity to view the criminal. Biggers, 409 U.S. at 199, 93 S.Ct. at 382. We conclude that, at least with respect to witness Allen, the witness had no opportunity to observe the assailant later identified as Mata. To the extent the Fifth District’s finding can be construed to be contrary, it is not fairly supported by the record. 28 U.S.C. § 2254(d)(8). Allen consistently testified that he did not see the person Almen*1252gor was fighting (allegedly Mata). Allen was preoccupied with the tall man who lunged at him with a large knife (allegedly Vargas). After Allen backed away from this man he turned to yell for help and never again saw the. assailants’ faces. See part 11(A) supra.
Almengor’s opportunity to view the assailant he identified as Mata was somewhat better. After fist-fighting with this man Almengor faced him “eye-to-eye” for “several minutes, several seconds.” However, Almengor was not wearing his glasses at the time, and the entire incident lasted no longer than a few minutes. See id. Cf. United States v. Williams, 626 F.2d 697, 703 (9th Cir.), cert. denied, 449 U.S. 1020, 101 S.Ct. 586, 66 L.Ed.2d 482 (1980) (witness viewed bank robber for about 20 minutes at “close quarters”); United States v. Field, 625 F.2d 862, 865 (9th Cir.1980) (witness viewed bank robber for only one minute from 2-3 feet; identification testimony held tainted). If the Fifth District can be said to have described this as an “adequate” opportunity to view the criminal, that description is supported by the record. “Adequate,” however, is a finding hardly specific enough upon which to rely in reaching a conclusion that must be based on the totality of the circumstances. Manson, 432 U.S. at 113-14, 97 S.Ct. at 2252-53. Therefore, we consider the specific details of Almengor’s encounter with the assailant, see part 11(A) supra, to be fairly supported by the record and to be relevant to his “opportunity ... to view the criminal at the time of the crime.” Biggers, 409 U.S. at 199, 93 S.Ct. at 382.
2. Degree of Attention
Since a witness’ degree of attention bears on his opportunity to view the criminal at the time of the crime, this factor overlaps somewhat with the first Biggers factor. The Fifth District made no specific finding concerning the witness’ degree of attention, although such a finding might be considered implicit in its finding that the witnesses had an “adequate opportunity to view the crime.”
As noted above, Allen focused his attention on the assailant in front of him, not the man identified as Mata. Almengor, however, paid a great degree of attention to the assailant he identified as Mata since the two had fought and then had stood looking directly at each other. See part 11(A) supra.
3. Accuracy of the Prior Description
The Fifth District found that the witnesses’ “descriptions are accurate.” To the extent this finding can be substantiated by reference to the trial court record, it is not fairly supported by the record in several respects. Almengor initially described only two of the assailants as having mustaches; all three defendants, however, had mustaches on October 27, 1972. Allen initially stated that only one of the attackers had a mustache. Both witnesses changed their stories at trial and testified that all three assailants had mustaches. See part 11(C) supra.
A related and more relevant inquiry is the certainty and detail with which the witnesses described the assailants. See Washington v. Cupp, 586 F.2d 134, 137 (9th Cir.1978), cert. denied, 441 U.S. 909, 99 S.Ct. 2003, 60 L.Ed.2d 379 (1979) (“victim provided a specific and detailed description”). Here, although Almengor was able to describe the assailants’ clothing and complexion, his description clearly was not detailed. Cf Biggers, 409 U.S. at 200, 93 S.Ct. at 382 (description included assailant’s approximate age, height, weight, complexion, skin texture, build, and voice); Washington, 586 F.2d at 137 (description included race, height, build, hair style, complexion, age, facial characteristics, and clothing). Allen’s description was even less certain and detailed: he could not remember whether the attackers were Caucasian or Chicano. See part 11(C) supra.
4. Level of Certainty at Confrontation
Again, the Fifth District made no specific finding concerning this factor. Allen and Almengor expressed no certainty about any of the photographs until the photographic *1253array on October 30. See part 11(B) supra. The only person Almengor identified “positively” — Pete Nunez — was not even in the prison facility at the time of the attack. Six other “possible” identifications by Almengor were subsequently ruled out by prison officials. Cf. United States v. Barron, 575 F.2d 752, 754 (9th Cir.1978) (no indication that witnesses had ever selected any other suspect than defendant). Allen was so uncertain that he was unable to make even a “possible” identification until October 30. See part 11(B) supra. The certainty demonstrated by Allen and Almengor on October 30 came only after prison authorities employed the pretrial identification procedures here at issue.
5. Length of Time Between Crime and Confrontation
Although prison investigators presented photographic arrays to Almengor on the day of the incident, and to Almengor and Allen eight days later, neither witness was able to make a positive identification. Not until October 30, eleven days after the stabbing, did Almengor and Allen identify the defendants’ photographs. See part 11(B) supra. While we have held that a lapse of one week between the incident and the confrontation is not enough alone to render an identification inadmissible, see United States v. Hammond, 666 F.2d 435, 440 (9th Cir.1982), the lapse of eleven days in this case is a factor to be considered in assessing the totality of the circumstances.
B.
Against these indicia of the witnesses’ ability to make an accurate identification we are to weigh the corrupting effect of the photographic identification procedures employed by the prison authorities. See Manson v. Brathwaite, 432 U.S. 98, 114, 97 S.Ct. 2243, 2253, 53 L.Ed.2d 140 (1977).
1. Necessity of the Photographic Arrays
In a number of this Circuit’s previous eyewitness identification cases, and in our initial opinion in Mata, we stated that a two-step approach that focuses first on the necessity of the identification procedure employed is the proper inquiry in determining the admissibility of identification evidence. See, e.g., Mata v. Sumner, 611 F.2d 754, 757 (9th Cir.1979), vacated and remanded, 449 U.S. 539, 101 S.Ct. 764, 66 L.Ed.2d 722 (1981); United States v. Peele, 574 F.2d 489, 490 (9th Cir.1978); United States v. Flickinger, 573 F.2d 1349, 1358 (9th Cir.), cert. denied, 439 U.S. 836, 99 S.Ct. 119, 58 L.Ed.2d 132 (1978); United States v. Collins, 559 F.2d 561, 563 (9th Cir.), cert. denied, 434 U.S. 907, 98 S.Ct. 309, 54 L.Ed.2d 195 (1977); United States v. Valdivia, 492 F.2d 199, 210 (9th Cir.1973), cert. denied, 416 U.S. 940, 94 S.Ct. 1945, 40 L.Ed.2d 292 (1974); United States v. Baxter, 492 F.2d 150, 171 (9th Cir.), cert. dismissed, 414 U.S. 801, 94 S.Ct. 16, 38 L.Ed.2d 38 (1973). Cf. United States v. Crawford, 576 F.2d 794, 797-98 (9th Cir.), cert. denied, 439 U.S. 851, 99 S.Ct. 157, 58 L.Ed.2d 155 (1978) (necessity for pretrial photographic identification procedures should be considered but not as a separate first step). The consideration of necessity apparently stemmed from language by the Supreme Court in Simmons v. United States, 390 U.S. 377, 88 S.Ct. 967, 19 L.Ed.2d 1247 (1968), suggesting that the propriety of using a photographic identification procedure was grounded in part on its necessity in the circumstances. 390 U.S. at 384-85, 88 S.Ct. at 971-72. See Baxter, 492 F.2d at 171 (“The Simmons Court suggested two standards by which the permissibility of pretrial photographic identification should be judged. The first is whether the use of photographic identification was necessary in the particular case.”)
The Government’s need to use an inherently less reliable method of identification, e.g., photographic arrays, see Simmons, 390 U.S. at 386 n. 6, 88 S.Ct. at 972 n. 6, would be a relevant factor if the due process interest served by excluding identification evidence focused on the protection of an accused from the deliberate police use of suggestive identification measures. Cf. Manson, 432 U.S. at 111, 97 S.Ct. at 2251 (respondent urges that deterrence of improper identification practice is pre-eminent reason *1254for rule that excludes evidence resulting from unnecessarily suggestive identification procedures without regard to its ultimate reliability); Neil v. Biggers, 409 U.S. 188, 198, 93 S.Ct. 375, 382, 34 L.Ed.2d 401 (1972) (“[Unnecessarily suggestive [confrontations] are condemned for the further reason that the increased chance of misidentification is gratuitous.”) Due process, in this sense, would justify exclusion of identification evidence on grounds analogous to the exclusion of evidence of coerced, but true, confessions: it is reprehensible police conduct — the needless use of suggestive identification measures that focus on the accused in order to secure a conviction — that due process of law prohibits.
[W]hile an involuntary confession is inadmissible in part because such a confession is likely to be unreliable, it is also inadmissible even if it is true, because of the “strongly felt attitude of our society that important human values are sacrificed where an agency of the government, in the course of securing a conviction, wrings a confession out of an accused against his will.”
Watkins v. Sowders, 449 U.S. 341, 347, 101 S.Ct. 654, 658, 66 L.Ed.2d 549 (1981) (quoting Jackson v. Denno, 378 U.S. 368, 386, 84 S.Ct. 1774, 1785, 12 L.Ed.2d 908 (1964).
In Manson, however, the Supreme Court made it clear that in requiring the exclusion of identification evidence that results from suggestive identification procedures, the Due Process Clause protects solely an evidentiary interest, 432 U.S. at 113, 97 S.Ct. at 2252, an interest normally vindicated through the adversarial process of cross-examination, id. at 113-14 n. 14, 97 S.Ct. at 2252-53 n. 14. See Simmons, 390 U.S. at 384, 88 S.Ct. at 971. Due Process protects an accused by prohibiting a jury from hearing eyewitness testimony unless that evidence has aspects of reliability. Manson, 432 U.S. at 112, 97 S.Ct. at 2252. Thus, “reliability is the linchpin in determining the admissibility of identification testimony....” Id. at 114, 97 S.Ct. at 2253. Unlike reliable confessions obtained • through police pressure, eyewitness identification testimony obtained through suggestive and unnecessary procedures is nevertheless admissible if it retains “certain features of reliability.” Id. at 110, 97 S.Ct. at 2251. See Watkins, 449 U.S. at 347-48, 101 5. Ct. at 658-59. “[W]hen improper police conduct ... does not taint the reliability of [the] witness’ identification evidence, the defendant has nothing of which to complain.” United States v. Field, 625 F.2d 862, 868 (9th Cir.1980).
Therefore, we conclude that the rationales of Manson and Watkins dictate that the need, or lack of it, for the identification procedures employed by the prosecution’s officers plays no part in the determination of the admissibility of identification evidence, a determination that focuses solely on reliability. This does not mean, however, that we cannot consider the fact that the officers chose to use a less reliable identification procedure (e.g., photographic array) over a normally more reliable procedure (e.g., corporeal line-up). See Simmons, 390 U.S. at 386 n. 6, 88 S.Ct. at 972 n. 6. The mere fact that a less reliable method was used clearly bears on the extent of the corruptive effect of the identification procedure. We hold only that, because a defendant’s due process interest in this context is only evidentiary, and concerned solely with reliability, the extent to which the officers needed to use a certain identification procedure cannot render eyewitness identification testimony admissible or, on the other hand, mandate its exclusion.
We find that prison authorities in this case chose to use photographic arrays, an identification procedure inherently less reliable than corporeal line-ups. See Simmons, 390 U.S. at 386 n. 6, 88 S.Ct. at 972 n. 6. “Even if the police ... follow the most correct photographic identification procedures and show [the witness] the pictures of a number of individuals without indicating whom they suspect, there is some danger that the witness may make an incorrect identification.” 390 U.S. at 383, 88 S.Ct. at 971. For the reasons discussed above, however, we do not consider relevant whether this procedure was necessary; we find only *1255that its relative unreliability in comparison with line-ups is a factor to be considered in the totality of the circumstances.
2. Suggestive Identification Procedures
In addition to the use of an inherently less reliable identification method, the record is replete with instances of suggestiveness in the identification procedures employed by prison officials. Almengor was shown photographs on not one, but three different occasions; Allen viewed photographs twice. Our statement in United States v. Higginbotham, 539 F.2d 17, 23 (9th Cir.1976) is apposite:
While the repeated showing to a witness of photographic displays for the purpose of identification presents opportunities for abuse and due process problems, when it is shown that the witness was equivocal on the first selection and became firm on a latter showing, that rule should not be applied where the witness has been consistently firm.
In the final photographic array on October 30 prison authorities removed photographs that had been identified earlier by Almengor as “possibly” the assailants and drastically reduced the total number of photographs. The new photographs of the defendants were the only ones updated for the October 30 array, despite the fact that the defendants’ photographs were among the most recent shown in the October 27 array. Allen and Almengor were housed in segregation with the defendants and saw them taken to have their pictures retaken; Allen stated at one point that because of this he assumed the three defendants were the participants in the stabbing. The Fifth District’s finding that the record supports no showing that this action influenced the witnesses’ identification is itself not fairly supported by the record in this case. 28 U.S.C. § 2254(d)(8). Finally, uncontradicted testimony by Allen shows that prison authorities brought pressure to bear upon him by threatening him with getting “shipped” if he didn’t cooperate. The state court’s finding that there was no showing of influence by the investigating officers is likewise not fairly supported by the record. 28 U.S.C. § 2254(d)(8).
C.
Only after these actions by prison investigators did Allen and Almengor identify Mata and his codefendants as the assailants. Keeping in mind that our purpose is to “judge whether the identification was the product of observations at the time of the crime or impressions made during the suggestive pretrial photographic identifications process,” United States v. Field, 625 F.2d 862, 866 (9th Cir.1980), we remain firmly convinced that the circumstances of this case show the latter. Giving due deference to those findings of the Fifth District that are fairly supported by the record, we hold that in the totality of the surrounding circumstances the pretrial photographic procedures employed here were so impermissibly suggestive as to give rise to a very substantial likelihood of irreparable misidentification.
The admission of the in-court identifications and evidence of the pretrial identifications of Mata by Allen and Almengor, therefore, was error. See Field, 625 F.2d at 869-70. Moreover, we cannot conclude that such error was harmless beyond a reasonable doubt. See Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 828, 17 L.Ed.2d 705 (1967); United States v. Stubblefield, 621 F.2d 980, 983 (9th Cir.1980); Green v. Loggins, 614 F.2d 219, 225 (9th Cir.1980). Although the prosecution introduced some circumstantial evidence of Mata’s guilt, its case was based primarily on the pretrial and in-court identifications by Almengor, Allen, and a third witness. We cannot say admission of the identification evidence was harmless when we do not know which witness the jury may have believed. Field, 625 F.2d at 870.
Accordingly, the order of the District Court is reversed. Upon remand, the District Court will hold the Petition in abeyance for a period of ninety days, within which time the state prosecutors may, if they so choose, initiate a new trial against Mata, a trial in which prosecutorial evi*1256dence barred by the Federal Constitution will not be introduced.
REVERSED and REMANDED, with directions.
. Section 2254(d) provides:
(d) In any proceeding instituted in a Federal court by an application for a writ of habeas corpus by a person in custody pursuant to the judgment of a State court, a determination after a hearing on the merits of a factual issue, made by a State court of competent jurisdiction in a proceeding to which the applicant for the writ and the State or an officer or agent thereof were parties, evidenced by a written finding, written opinion, or other reliable and adequate written indicia, shall be presumed to be correct, unless the applicant shall establish or it shall otherwise appear, or the respondent shall admit—
(1) that the merits of the factual dispute were not resolved in the State court hearing;
(2) that the factfinding procedure employed by the State court was not adequate to afford a full and fair hearing;
(3) that the material facts were not adequately developed at the State court hearing;
(4) that the State court lacked jurisdiction of the subject matter or over the person of the applicant in the State court proceeding;
(5) that the applicant was an indigent and the State court, in deprivation of his constitutional right, failed to appoint counsel to represent him in the State court proceeding;
(6) that the applicant did not receive a full, fair, and adequate hearing in the State court proceeding; or
(7) that the applicant was otherwise denied due process of law in the State court proceeding;
(8) or unless that part of the record of the State court proceeding in which the determination of such factual issues was made, pertinent to a determination of the sufficiency of the evidence to support such factual determination, is produced as provided for hereinafter, and the Federal court on a consideration of such part of the record as a whole concludes that such factual determination is not fairly supported by the record:
And in an evidentiary hearing in the proceeding in the Federal court, when due proof of such factual determination has been made, unless the existence of one or more of the circumstances respectively set forth in paragraphs numbered (1) to (7), inclusive, is shown by the applicant, otherwise appears, or is admitted by the respondent, or unless the court concludes pursuant to the provisions of paragraph numbered (8) that the record in the State court proceeding, considered as a whole, does not fairly support such factual determination, the burden shall rest upon the applicant to establish by convincing evidence that the factual determination by the State court was erroneous.
28 U.S.C. § 2254(d).