dissenting.
I believe the decision on the merits to be very close. However, as a judge on a reviewing court, I am unable to say that the district court misapplied the law to the generally uncontroverted facts. Accordingly, I join in Judge McMillian’s view that we should affirm the district court.
McMILLIAN, Circuit Judge, dissenting, with whom LAY, Chief Judge, and HEA-NEY and BRIGHT, Circuit Judges, join.
I do not agree with the majority opinion’s decision on the merits and therefore dissent. I believe the district court properly weighed the “corrupting effect” of the impermissi-bly suggestive identification procedures against the indicia of the witness’ ability to make a reliable identification.
Graham’s first and strongest argument is that the victim’s in-court identification of him as one of the rapists violated his constitutional right to due process. The South Dakota Supreme Court held that even though the victim’s in-court identification was unnecessarily suggestive, it was admissible because it had been purged of any taint arising from the illegal out-of-court identification procedures. State v. Reiman, 284 N.W.2d 860, 871 (S.D.1979). The South Dakota Supreme Court also ruled that even if the taint had not been purged, the in-court identification still would be admissible because its origin was independent of the improper out-of-court identification. The state supreme court found that the in-court identification “was based upon a mental image indelibly fixed in [the victim’s] mind during the extended assault. As she stated at trial, ‘I’m positive it’s these four. I can never forget their faces.’ ” Id. at 872. The state argues that this court should adopt *1549this three-tiered approach and hold the in-court identification to be admissible, even if constitutionally lacking, because it was of independent origin. In my view this argument misperceives the current standard of due process review for eyewitness identifications.
In Manson v. Brathwaite, 432 U.S. 98, 113-14, 97 S.Ct. 2243, 2252-53, 53 L.Ed.2d 140 (1977) (Manson), the Supreme Court set forth the dual elements of a successful due process challenge to an in-court eyewitness identification. First, a defendant must prove that the in-court identification was the result of unnecessarily suggestive out-of-court identification procedures which by their nature engender a “very substantial likelihood of irreparable misidentification.” Neil v. Biggers, 409 U.S. 188, 198, 93 S.Ct. 375, 381, 35 L.Ed.2d 401 (1972) (Biggers). Second, the defendant must prove that the identification was unreliable. Manson, 432 U.S. at 114, 97 S.Ct. at 2253. Biggers listed several criteria for determining reliability, including: the witness’ opportunity to view the criminal during the crime, the witness’ degree of attention, the accuracy of the witness’ prior descriptions, the level of certainty demonstrated by the witness at the confrontation, and the length of time between the crime and the confrontation. 409 U.S. at 199-200, 93 S.Ct. at 382-383. In the final analysis, the court must balance the corrupting effect of the suggestive procedures against the reliability of the identification to determine if the use of the in-court identification violated the defendant’s due process rights. Manson, 432 U.S. at 114, 97 S.Ct. at 2253; Biggers, 409 U.S. at 199-200, 93 S.Ct. at 382-383.
Thus, concepts of “purged taint” and “independent origin” have been blended into, and superseded by, the two-step process of weighing reliability against suggestiveness articulated in Biggers, 409 U.S. at 199-200, 93 S.Ct. at 382-383. See Solomon v. Smith, 645 F.2d 1179, 1185, 1188 & n. 9 (2d Cir. 1981). Today, “reliability is the linchpin in determining the admissibility of identification testimony.” Manson, 432 U.S. at 114, 97 S.Ct. at 2253. Nonetheless, the South Dakota Supreme Court did rely on the Biggers criteria of reliability in determining that the in-court identification was purged of taint and was of independent origin. Therefore, I will assume that the state supreme court concluded that the identification was reliable and will review the court’s findings accordingly.
I. SUGGESTIVENESS
The South Dakota Supreme Court did not dispute that the one-man lineup identification of Graham was unnecessarily suggestive. State v. Reiman, 284 N.W.2d at 871.1 The federal district court also held that the entire procedure used by the police to assist the victim’s identification of Graham was *1550impermissibly suggestive and posed a very substantial likelihood of irreparable mis-identification. The police focused their investigation on Graham even before the victim had a chance to view or identify him. Tr. at 1413. The police followed Graham throughout the day and allowed the victim to view Graham five times before presenting him to the victim in a one-man lineup. It was only after viewing Graham for the fifth time in a single day, including once through binoculars, that the victim became “ninety percent sure” Graham was one of her rapists. In my opinion the district court correctly determined that:
By the use of these procedures I am sure that the investigating officers planted the seeds of an identification in the mind of [the victim]. The seeds of the identification were nurtured by each successive showing of the petitioner to [the victim]. By the time of trial what had been mere seeds blossomed into a complete and unequivocal identification.
Graham v. Solem, No. CIV80-4174, mem. op. at 9 (D.S.D. Feb. 16, 1982).
II. RELIABILITY
In applying the Biggers reliability criteria to the present case, the South Dakota Supreme Court concluded that: (a) the victim had an unusual opportunity to view the defendant over the course of several hours; (b) her degree of attention was high due to the nature of the crime; (c) although none of the victim’s prior descriptions of her rapists matched Graham’s appearance, Graham could have been the fifth rapist for whom no description was given; (d) the victim was very certain of her in-court identification; and (e) only ten days separated the confrontation from the crime. State v. Reiman, 284 N.W.2d at 871-72.
Under Sumner v. Mata, 449 U.S. 539, 551, 101 S.Ct. 764, 771, 66 L.Ed.2d 722 (1981), federal courts must give a presumption of correctness to these factual determinations unless one of the eight exceptions contained in 28 U.S.C. § 2254(d) is present.2 The failure of the record in the state court proceeding, considered as a whole, to fairly support the state court’s factual determination is one such exception. Id. § 2254(d)(8). I agree with the federal district court that Graham has borne his burden of proving that the state trial and supreme courts’ factual determinations in applying the Big-gers criteria of reliability are not supported by the record. I come to this conclusion cognizant of the “high measure of deference” to be accorded state factual findings, as well as the virtually absolute deference to be paid the credibility determinations of a state court judge who has heard the witness’ live testimony. See Marshall v. Lon-berger, 459 U.S. 422, 103 S.Ct. 843, 850-51, 74 L.Ed.2d 646 (1983). In this case, however, there have been no credibility determinations, explicit or implied, by a state court judge who has heard the witness’ live testimony because no live testimony was presented to the state courts. The state trial court decided Graham’s motion to suppress the identification based solely on the cold record of the relevant witnesses’ written depositions, including the deposition of the victim. See note 1 supra. I am convinced, and convinced mainly by the testimony of the victim herself, that several of the state trial and supreme courts’ factual findings lack fair support in the record.
A. Opportunity to View
The South Dakota Supreme Court failed to mention in its Biggers analysis that there was extensive testimony that the victim was under the influence of both alcohol and marijuana at the time of the rape. See, e.g., Tr. at 800, 888,1301-03,1311-12,1903-04.3 It also failed to note that the victim *1551was not wearing her prescription glasses at the time of the rape. The victim testified that she is nearsighted and especially needs her glasses to see at night. Tr. at 896-97. The rape occurred at night. The record shows that the victim’s opportunity to view her assailants was limited by impaired perception.
B. Attentiveness
I agree that, because of the nature of the crime, the victim’s attention was high. Her ability to perceive and recollect, however, may have been impaired.
C. Prior Descriptions
The victim testified that Graham was the passenger of the van in which she was accosted and that the passenger was one of the two assailants she remembered most. Tr. at 893. Yet the victim testified that none of the descriptions she gave the police matched Graham’s appearance. Tr. at 891. The South Dakota Supreme Court stated that any discrepancies between the descriptions and Graham’s appearance were not substantial because the victim gave only four descriptions and there were at least five assailants. Graham, the court concluded, could have been the fifth undescribed assailant. State v. Reiman, 284 N.W.2d at 872.
This factual conclusion is not supported by the record. First, it is difficult to believe that the victim would not have described one of the persons she remembered most, especially when his face was “indelibly fixed in her mind.” Id. Second, the victim admitted on the stand that she had described the four assailants which she remembered most and that none of those descriptions matched Graham’s appearance. Tr. at 893. Third, and most importantly, the victim testified that Graham was the passenger in the front of the van in which she was abducted. In her deposition the victim gave a description of that passenger to the police before being subjected to the impermissibly suggestive identification procedures. The description of the van’s passenger given by the victim does not in any way match Graham’s appearance. This shows that Graham could not have been the fifth (or possibly sixth) undescribed assailant. The victim herself testified that she did describe the passenger in the front of the van and that Graham was that passenger. The majority opinion’s conclusion to the contrary is untenable in the face of the victim’s own testimony. The only conclusion which can be drawn from the record is that the victim did give a description of the assailant she claims is Graham and that description does not match Graham’s appearance. Tr. at 893-94. This criterion suggests that the victim’s in-court identification was unreliable.
D. Level of Certainty
At trial and at the one-man lineup, the victim said she was absolutely sure that Graham was one of her rapists. But the trial and the one-man lineup were not the victim’s only confrontations with Graham. Earlier confrontations evoked less certainty which, through the course of impermissibly suggestive identification procedures, was gradually increased to 80-90% certainty at the fifth confrontation, just prior to the lineup. Tr. at 855. The focus of the certainty criterion must be on the eyewitness’ level of certainty before being exposed to legally impermissible confrontations. As the Second Circuit has recently emphasized:
Certainty entails confidence in one’s identification of a suspected perpetrator. Hesitancy, the inability to be positively sure about a suspect, and the extent to which an affirmative identification is the product of prodding by others, are signals which undermine the certainty of the witness’s identification of the suspect at the pre-trial confrontation.
Dickerson v. Fogg, 692 F.2d 238, 246 (2d Cir.1982). See United States ex rel. Phipps v. Follette, 428 F.2d 912, 915 (2d Cir.), cert. denied, 400 U.S. 908, 91 S.Ct. 151, 27 *1552L.Ed.2d 146 (1970). Here, the victim displayed a great deal of uncertainty during the initial stages of the impermissibly suggestive procedures used by the police to help her identify Graham. This uncertainty stands in marked contrast to the instant certainty the victim exhibited in identifying the other three defendants. Tr. at 847-51. This contrast is highlighted further by the victim’s testimony that she remembered the rapist she now claims is Graham better than the other three defendants whom she identified upon seeing for the first time. The record reveals a discernible progression in the victim’s level of certainty — from her initial descriptions to the police when Graham’s physical features had not yet appeared in her memory, Tr. at 892, to the trial where she testified that she would never forget Graham’s face. This rising level of certitude reinforces the conclusion that the victim’s certainty in identifying Graham at trial is more the product of the police’s prodding than it is an indication of reliability.
The South Dakota Supreme Court gave special weight to the victim’s testimony that she was able to positively identify Graham at the one-man lineup because she recognized the tattoo on his arm. State v. Reiman, 284 N.W.2d at 872. Yet she never described this tattoo to the police prior to the lineup and could not describe it at trial without prompting from the prosecuting attorney. Tr. at 855. Recognition of a tattoo is even less significant when one notes that both defendants Reiman and Elliott had tattoos and that the victim testified she saw “a lot of tattoos throughout the night.” Tr. at 855 — 66.
The most telling indicia of unreliability of the victim’s identification of Graham is the instant and absolute certainty that the victim displayed when she identified the other three co-defendants in a crowded room without prompting from the police. This factor sets Graham apart from his co-defendants. The victim gave no indication that she had a greater or lesser opportunity to view any particular assailant. Indeed, she testified that she remembered the particular assailant, whom she now says is Graham, the most. When she was taken to identify individuals as possible assailants, she instantly and without hesitation pointed out Graham’s co-defendants. Yet when her attention was specifically directed toward Graham, Tr. at 1060, 1062, she was unable to positively identify Graham as one of her assailants. See Tr. at 1056-66. This uncertainty continued until the impermissibly suggestive identification procedures indelibly fixed Graham’s features into the victim’s memory. The majority opinion speaks of the practical, common sense workings of memory and notes that memory ripens over time. This is indeed true, but Biggers teaches us that if an eyewitness’ memory is nurtured by successive showings of a defendant to the victim, the victim’s identification must be accompanied by other indicia of reliability. See generally E. Loftus, Eyewitness Testimony (1979); Rahaim & Brod-sky, Empirical Evidence Versus Common Sense: Juror & Lawyer Knowledge of Eyewitness Accuracy, 7 Law & Psych.Rev. 1 (1982); Note, Eyewitness Identification Testimony & the Need for Cautionary Jury Instructions in Criminal Cases, 60 Wash U.L.Q. 1387,1387-02 (1983). One such indication is the level of certainty before the impermissibly suggestive identification procedures take hold. Here, the victim displayed a great deal of uncertainty about Graham, while displaying no uncertainty about identifying the other defendants. This difference in the level of the victim’s uncertainty at the initial stages of what turned into an impermissibly suggestive identification procedure is a pivotal ingredient in my decision. I must also emphasize that no single factor convinced me that the victim’s identification of Graham is unreliable. Rather, taken as a whole, the surrounding circumstances show that the few indications of reliability do not outweigh the highly suggestive nature of the identification procedures used in Graham’s case, and Graham’s case alone.
E. Time Lapse
Only about ten days lapsed between the rape and the lineup. This was a relatively short time, short enough for the victim to *1553have any memories of the rape still fresh in her mind.
F. Other Indicia of Reliability
The state urges us to consider as additional indicia of reliability the fact that all four defendants admitted they were at or near the Outasite Paint Shop during the hours the rape was committed. The victim had pinpointed the Outasite Paint Shop as the place where the rape occurred by identifying a mirror which came from the Outa-site Paint Shop. The victim testified that during the course of the rape she broke free and ran into a bathroom where she saw a mirror with a red design on it. When the police showed her a mirror which came from the Outasite Paint Shop, she recognized it as the one she saw the night of the rape.
First, I note that Graham did not admit to being in the Outasite Paint Shop during the time the rape occurred. He did admit that he momentarily stopped by the Outasite Paint Shop but did not go in. He stated that he merely sat outside the area of the Outasite Paint Shop while talking on a CB radio and then left. Second, other evidence of guilt which does not relate to the witness’ ability to perceive or remember should not enter into the process of determining an identification’s reliability. Green v. Loggins, 614 F.2d 219, 225 (9th Cir.1980). Compare Manson, 432 U.S. at 116, 97 S.Ct. at 2253 (fact that defendant was found at the scene of the crime and admitted being there on several occasions played no part in the reliability analysis), with Cronnon v. Alabama, 587 F.2d 246, 250 (5th Cir.) (reliability supported by witness testimony that murderer left the store with a bag of peanuts because later-identified defendant was reported to have arrived home with a bag of nuts that he got “at the store”), cert. denied, 440 U.S. 974, 99 S.Ct. 1542, 59 L.Ed.2d 792 (1979). But even assuming that Graham’s admission to being outside of the Outasite Paint Shop the night of the rape can be used to show reliability, I find that the victim’s identification of the Outa-site Paint Shop as the scene of the crime was itself of questionable reliability.
Shortly after she was raped, the victim went to the home of two of her friends. While there she told her friends that she had been raped in a vacant room with just a sink and a mattress in it. Tr. at 1806. She did not mention the mirror4 or that many cars cluttered up the room to her friends at that time. Tr. at 1806. When the victim told her roommate that she had been raped, her roommate suggested that Bud Reiman was probably the culprit. The roommate then described Reiman to the victim as having “long hair, a beard and tattoos.” Tr. at 772. The victim responded that this description was accurate. The victim also described the room in which she was raped to her roommate. According to the roommate, the victim described the room as containing a van and several bikes. The roommate then suggested to the victim that the rape probably occurred in the Outasite Paint Shop, even though the roommate had never been inside the Outasite Paint Shop. Tr. at 787-88.
After her friends told her that Bud Rei-man was probably one of the rapists and that he probably raped her in his garage — • the Outasite Paint Shop — the police took the victim along with them when they executed a search warrant at Reiman’s garage. It was only after these highly suggestive events that the victim’s description of the place of her rape includes a reference to a room filled with cars in various states of repair — a description which matches Rei-man’s garage. Compare Tr. at 820 with Tr. at 1806.
This analysis does not suggest that the victim’s identification of the Outasite Paint *1554Shop is inadmissible. The degree of reliability of the paint shop identification as the location of the rape is for the jury to consider. The victim’s identification of the paint shop as the location of the crime has only been reviewed in order to determine whether the victim, in general, could or could not perceive and recollect the events of that night accurately. The Outasite Paint Shop identification shows that the victim’s ability to recollect accurately ripened only after the police subjected the victim to the impermissibly suggestive identification procedures.
Another factor which militates against a finding of reliability is the victim’s descriptions of the van in which she was kidnapped. Initially, she described the van as being orange. Tr. at 864. Later, the victim testified that “[f]or a while I thought it was a green van.” Tr. at 865. Finally, she described the van to the police as being white with gold trim and possibly bearing a large “happy face” insignia on its exterior. Tr. at 990. One of the defendants did own a van, but it was a two-tone gray color with a Maltese Cross painted on its exterior. Tr. at 1373-74.
One last indication of the victim’s ability to accurately perceive and recollect events on the night she was raped is her recollection of the details of her kidnapping. Graham’s counsel elicited the following testimony from the victim concerning Michael Peck, an acquaintance of the victim whom she saw at the bar just prior to her abduction, and the events which transpired while the victim was being forcibly restrained in the van outside the bar.
Q. Previously, you have testified, have you not, that Mike Peck came to the window of this van when you were inside held captive?
A. I said I wasn’t for sure, but I thought he did. If someone did, it was Mike, but I can’t swear that he did come to the window.
Q. Perhaps you can explain it for me. You said if someone did, it was Mike.
A. I remember someone coming and talking to the passenger.
Q. Okay then, you remember it. Then it must have been Mike.
A. Someone did come and talk to the passenger.
Q. Okay. We are back to that now. And you have previously just told me that if someone did, it was Mike. Are we sure now, someone did and it as Mike Peck.
A. I can’t swear that it was Mike, but I know someone was there.
Q. Well, I don’t want to argue with you, and I want to be fair, but you just told me if someone did, it was Mike. Now, are you retreating on that and saying it could have been someone other than him?
A. No, what I’m trying to say is I thought I saw Mike at the van door.
Q. Have you also testified previously that Mike looked in and saw you and then looked away?
A. Like he didn’t know I was in there.
Q. But you did testify that you thought he saw you?
A. Yes.
A. He was there, and then he was gone. That’s why I can’t say for sure if it was even him. Sometimes I think I just imagined it.
Q. You think you may have imagined it?
A. No, I think — I believe — I think — I believe he was there.
Tr. at 886-88. Mike Peck testified that he did not look in any van window on that night. Tr. 1304. It is interesting to note that Graham was also at the bar at the same time Mike Peck and the victim were there.
III. WEIGHING RELIABILITY AGAINST SUGGESTIVENESS
As the United States Supreme Court said in Manson, a “witness must testify about an encounter with a total stranger under circumstances of emergency or emotional stresses. The witness’ recollection of the stranger can be distorted easily by the circumstances or by later actions of the po*1555lice.” 432 U.S. at 112, 97 S.Ct. at 2252. I believe that is what occurred in this case.
The case of Foster v. California, 394 U.S. 440, 89 S.Ct. 1127, 22 L.Ed.2d 402 (1969), is closely analogous to the present case. In Foster, the witness viewed the defendant on several occasions before the witness could make a positive identification. The witness first viewed the defendant in a lineup, but failed to identify the defendant. Then the police arranged the one-to-one confrontation between the defendant and the witness. When the witness could only make a tentative identification even after these confrontations, the police placed the defendant in another lineup a few days later. The defendant was the only person who was in both the first and second lineups. This third confrontation “produced a definitive identification.” Id. at 443, 89 S.Ct. at 1129. The Supreme Court held that, as a matter of law, the identification procedures were so defective that the identification was constitutionally inadmissible. Id. n. 2. On far less egregious facts, the Second Circuit reached a similar conclusion. See Dickerson v. Fogg, 692 F.2d at 244-47.
In the present case, the few facts supporting the reliability of the victim’s identification of Graham do not outweigh the suggestiveness of the identification procedures used. This leads to the conclusion that there was a very substantial likelihood of misidentification which violated Graham’s due process rights. Because I would hold the admission of the victim’s identification of Graham as one of her rapists was not harmless error, I would affirm the judgment of the district court granting the petition for writ of habeas corpus.
. The South Dakota trial court, however, did state in denying Graham’s pre-trial suppression motion that the identification procedures were not “unduly suggestive.” State v. Reiman, No. 77-30, letter opinion at 2 (S.D. 6th Cir.Ct. Dec. 9, 1977) (Jones, J.) (located in Designated Record at 296-98, State v. Reiman, 284 N.W.2d 860 (S.D.1979)). Yet Judge Jones also stated:
Such identification procedures are condemned by both United States v. Wade, 1967 [388 U.S. 218], 87 S.Ct. 1926 [18 L.Ed.2d 1149], and State v. Keeling [89] SD [436] 1975, 233 N.W.2d 586. Keeling establishes our rule that the burden is on the State after such a showing to establish by clear and convincing evidence that the pretrial identification was not unnecessarily suggestive. In deciding this motion, I have carefully read the depositions of all witnesses with particular reference to the identification procedures following with respect to Mr. Graham prior to this one-man lineup. Considering the facts of the rape, that there were no substantial discrepancies between [the victim’s] pre-lineup description of Mr. Graham and his actual description, that she had previously identified Mr. Graham in a positive manner while he was standing near Nelson’s Cycle Shop, I conclude that the one-man lineup of Mr. Graham did not taint possible future identifications of him so as to require that this court prohibit [the victim] from identifying Mr. Graham at the trial, if she is able to do so.
Id.
Translating these findings into the terms of a Neil v. Biggers analysis, it appears Judge Jones reasoned that the reliability of the identification, judged by the Biggers criteria, outweighed the suggestiveness of the “condemned” identification procedures. Thus, according to this analysis, the identification procedures, although they are to be condemned as impermis-sibly suggestive, were not “unduly” suggestive.
. Findings relating to the Biggers criteria are factual determinations and are entitled to the presumption of correctness. The conclusion that there was or was not a substantial likelihood of misidentification, which is drawn from weighing the Biggers criteria against the suggestiveness of the identification procedures, is a question of law and is not entitled to the presumption of correctness. See Neil v. Big-gers, 409 U.S. 188, 193 n. 3, 93 S.Ct. 375, 379 n. 3, 34 L.Ed.2d 401 (1972); Dickerson v. Fogg, 692 F.2d 238, 242-43 (2d Cir.1982).
. The majority opinion points out that the friends who had accompanied the victim to the bar testified that the victim was not intoxicated. However, these friends had parted company with the victim shortly after arriving at the *1551bar. Testimony from two witnesses who saw the victim shortly before the victim was abducted indicates that the victim was probably under the influence of an intoxicant at the time of her abduction. Tr. at 702, 705, 707, 709, 712-15, 732, 735.
. I have tried to recite only the victim’s testimony or other relatively uncontested testimony and did not originally explore the evidence in the record that tends to show that the mirror in the Outsasite Paint Shop that the victim said she remembered from the night of the rape was actually purchased by Reiman after the night of the rape. Tr. at 1871-84. There are other indicia of unreliability in the record beyond those elicited from the victim’s own testimony. Most of the transcript citations in the dissenting opinion, however, are citations to the victim’s testimony.