ON PETITION FOR REHEARING
(Opinion Sept. 4, 1987, 5th Cir., 1987, 826 F.2d 375).
Rosario Joseph Dispensa was convicted of burglary of a habitation with intent to commit rape and sentenced to fifteen years imprisonment. The district court granted his writ of habeas corpus on the basis that the police’s use of an overly suggestive out-of-court identification procedure made the rape victim’s subsequent in-court identification unreliable. The State of Texas appealed, and, in a prior opinion, we concluded that because Dispensa had introduced new evidence at the federal habeas corpus hearing, which made his claims stronger than they had been when presented to the state court in a motion to suppress prior to his original trial and in the evidence at the trial, he had not exhausted his state remedies. For that reason, we reversed the judgment granting the writ.1
Dispensa then sought reconsideration of our decision. Having considered his motion and supporting brief as well as the State’s brief filed in opposition to it, we are persuaded that we erred in holding that he had not exhausted state remedies. He had sought habeas corpus in state court and presented to it the very claim he now presents, basing his state court petition on essentially the same factual evidence and exactly the same legal theory on which he now relies in his federal habeas petition. Dispensa thus gave the State an opportunity to correct the error before petitioning the federal court for relief. Because we are also persuaded that the State convicted Dispensa by means of a suggestive identification in violation of his constitutional right to due process of law, we affirm the judgment of the district court granting him the writ he seeks unless the State should elect to retry him within the time fixed by the district court.
I.
In the early morning hours of July 16, 1981, Theresa Ellen Barthel was alone in her apartment in the City of Houston, Texas. After returning from work the previous evening, she had cleaned the apartment thoroughly, then gone to bed at about 10 p.m., leaving a hall light and the lights on a patio adjoining her bedroom illuminated as a security measure since her husband was away. At about 4 a.m., she awoke to find a man standing by her bed with his hand over her mouth. The lights she had left on when she went to bed illuminated the area well, giving her a clear view of her assailant. After a brief struggle, he began to rape her. During the rape, there was a pillow over Barthel’s face. When she pulled the pillow away after the rape, she saw the assailant, who was wearing only *214his pants, grab his shirt and the rest of his belongings and leave the apartment. After the rapist had gone, Barthel found that both her apartment door and a window were open. She had complained to the management of the apartment complex many times about faulty locks on the windows; apparently the rapist had entered the apartment through one of these windows.
Immediately after the rape, Barthel telephoned her husband, who was attending a training school in Florida, then reported the crime to the police. She described the rapist as being 5'8" to 5'9" in height, weighing 160 to 170 pounds, and having brown wavy or curly hair. She said she didn’t know whether or not the rapist had a mous-tache. She did not know Rosario Dispensa, who had recently moved to Houston from another city and who lived in her apartment complex. Dispensa is approximately six feet tall; weighed 180 to 185 pounds; wore glasses; and had straight hair, prominent tattoos on his arms and shoulders, a large amount of body hair, and a.prominent moustache. In addition, he offered the testimony of a female companion that he has a visibly deformed penis and that the deformity is perceptible during intercourse. As a result of having been mugged shortly before the date of the rape, he had cuts on his hip and scabs and rough places on the palms of his hands.
A police officer, J.L. Pratt, who had been at the apartment complex on another matter, testified that he had seen Dispensa at 3:00 a.m., wearing only his trousers and carrying his shoes and what appeared to be a shirt, heading towards his apartment, which was a few doors away from the Barthels’. Dispensa testified that frequently upon arriving at the apartment complex after his long day of work, he went to the swimming pool and soaked his feet before proceeding to his apartment.
Following the rape, the police found seven sets of fingerprints in the apartment, none of which were identified as Dispen-sa’s. They found a cigarette butt, but it was not from the brand Dispensa smoked. Dispensa’s friend, Sandra Fay Wilson, testified that he was asleep in bed with her at 4:00 a.m. on July 16. The police examined the sheets on Barthel’s bed. Although Dispensa is hirsute, the police found only two hairs that were not Barthel’s or those of her dog or cat, plus a few other fragments too short to identify. A toxicologist testified that these two hairs were pubic hairs and that one had microscopic characteristics similar to those of a pubic hair furnished by Dispensa pursuant to a court order.
Dispensa was the manager of Papa Joe’s Seafood Restaurant. His duties required him to remain after the restaurant closed at 11:00 p.m., checking receipts and preparing for the next day’s operations. He usually returned home about 2:00 a.m.
Houston police detectives L.W. Henning and Ralph Yarborough decided to take Bar-thel to Papa Joe’s to see if she could recognize Dispensa as her assailant. They arrived at the mall in which the restaurant was located at about 11:00 a.m. and learned that the restaurant would not open until 11:30. They therefore strolled through the mall for about half an hour. During their walk, the police asked Barthel to observe the crowd and let them know if she recognized anyone. Then, they testified, without ever having explicitly told her the purpose of their trip, they took her to eat lunch at Papa Joe’s. After he had finished eating, Yarborough excused himself from the table and asked the cashier if he could speak to Dispensa. He was directed to Dispensa, who was in a back office in the kitchen area at the rear of the restaurant. Yarbor-ough there informed Dispensa that he was a suspect in a sex crime and told him that he could either walk through the restaurant so that the victim could see him or be subject to arrest and a police line-up. Dis-pensa chose to walk through the restaurant.
None of the witnesses to the identification procedure agreed precisely on how it was conducted. At the trial, Detective Yarborough testified that he remained in the office while Dispensa walked through the restaurant. He said he then asked Dispensa to wait while he checked with *215Henning to see if Barthel had reacted as Dispensa walked by. In a whispered conversation at the table, Henning told him that nothing had occurred. Consequently, Yarborough returned to Dispensa’s office and asked him to walk through the restaurant again. This time, however, Yarbor-ough stated, he preceded Dispensa from the office and waited for him at the cashier’s desk. He said that Dispensa then walked through the restaurant a second time, taking a different path that placed him directly in front of Barthel, and finally walked up to Yarborough. Yarborough asserts that he then looked over Dispensa’s shoulder to see if there had been any reaction and saw Henning struggling to keep Barthel seated.
Detective Henning testified at the trial that he did not know what Dispensa looked like when he and Yarborough took Barthel to the restaurant and that he did not signal to her to look at Dispensa as he approached their table. Instead, he related, he merely told her to be generally observant of the people in the restaurant. He testified at the suppression hearing that, while Yarbor-ough was away from the table, Barthel told him that a man had just passed by who looked like the rapist but that Henning himself had only gotten a glimpse of the man to whom she referred. He stated that he next saw Dispensa when Dispensa was approaching the cashier’s desk. Although he said that Yarborough was also in that general area, he denied seeing the two men walk to the register together and could not remember how close together they were standing when Barthel turned around on her own volition, saw Dispensa clearly, exclaimed that he was the rapist, and burst into tears.
Barthel’s testimony at the suppression hearing contradicted portions of both Yar-borough’s and Henning’s accounts. She asserted that Yarborough did not return to the table after he excused himself. More importantly, she testified that Henning “told me to talk to him like I was his brother [sic], just start talking to him, and he said when I tell you to look, I want you to look. So when Mr. Henning told me to look, walked past me [sic], and I didn’t get a look at his face, but then he walked by me again and that’s when I ran out” of the restaurant. She testified that she did not know the purpose of the trip to the restaurant but she was pressed by counsel for Dispensa with the question: “Now, didn’t you suspect why he [Henning] was asking you to act in a peculiar fashion?” Barthel answered, “At that point I did.” Counsel persisted, “So at that point then you knew what you were supposed to see when you looked around?” Barthel responded, “Well, at that point when somebody tells me to look behind me — I’m really getting mad,” and she asked for a recess. She stated that the first time Henning told her to look, Dispensa walked past before she could see his face, but that when Henning told her to look a second time, she saw Dispensa well and recognized him immediately. She also asserted that she did not see Yarborough with Dispensa in the restaurant and that she fled the dining room area after making the identification.
At both the suppression hearing and the trial, Dispensa described making two trips through the restaurant. He stated that the first time he walked by every table, checking to see that the place settings and arrangements of condiments were in order, and then returned to his office, where Yar-borough was waiting. Yarborough then left the office “to check on something” and, upon returning, told Dispensa that someone had not gotten a good look at him. He therefore asked Dispensa to walk through again, and Dispensa reluctantly complied. Dispensa testified, however, that on the second trip Yarborough followed behind him, and that rather than meandering through the tables as he had done on the first pass he proceeded directly to the receptionist’s desk at the front of the restaurant. There, he said, he and Yarbor-ough talked a few minutes before he returned to his office. Shortly thereafter, Yarborough entered the office with a uniformed policeman to make the arrest. Dis-pensa stated further that he did not notice any commotion indicating that an identification had been made while he was in the restaurant.
*216Arguing in state court that the identification procedure was impermissibly suggestive, Dispensa’s trial counsel moved to suppress any in-court identification of Dispen-sa and any reference to the pretrial identification. After a hearing on this motion, the trial court ruled that the identification procedure was fair and reasonable under the circumstances and that the procedure was not so impermissibly suggestive as to violate Dispensa’s due process rights. The court stated that the reliability of the identification was an issue for the jury.
Following his conviction and an unsuccessful appeal, Dispensa sought relief in state court by writ of habeas corpus, contending that the identification procedure was impermissibly suggestive. The state district court denied the petition without granting an evidentiary hearing, and this judgment was affirmed on appeal. Dispen-sa then sought relief in federal court, which granted an evidentiary hearing before a magistrate.
Dispensa’s testimony at the federal hearing differed from his prior account of the identification procedure. Instead of describing a single full pass through the restaurant and a second abbreviated trip to the receptionist’s desk and back, he asserted that he was asked to make two full passes through the restaurant before he and Yarborough walked together from his office to the receptionist’s desk and talked a few minutes. Dispensa testified that Yarborough left him in the office between the first and second and the second and third passes while he checked with Hen-ning in the restaurant to see if Barthel had made an identification. He explained that he failed to describe the second full trip through the restaurant during his testimony in the trial court because his trial counsel had told him that the court would be less likely to believe his account if he testified to more trips than Yarborough and Henning admitted. At his federal hearing, Dispensa also testified for the first time that, during his first walk through the restaurant, he greeted the people at each occupied table as he checked their table settings. Because he did not know where Barthel and Henning were seated, he testified, he could not tell when the identification actually was made.
In addition to offering these modifications to his account of the identification procedure at the federal habeas hearing, Dispensa offered expert psychological testimony about the significance of Barthel’s choice of words and her apparent emotional state at the suppression hearing. Dr. Fred Fason explained that Barthel’s repeated assertion that Henning told her when to look and her stated anger during cross examination about that testimony suggest that she unconsciously relied on the police to identify the rapist for her and that she felt guilty and angry with herself for not being able to make the identification alone. By analyzing the transcript of her testimony, Dr. Fason also concluded that Barthel’s assertions that she did not know why the police were taking her to the mall, despite the fact that there was only one obvious reason for such a trip, suggest that she had suspended her critical thinking and therefore was especially susceptible to suggestion. In response to questions posed by Dispensa’s habeas counsel about discrepancies between the description of her assailant that Barthel gave to the police and Dispensa’s actual appearance, Dr. Fason testified that she most likely would have noticed if her attacker was especially hairy, that she would also have been likely to notice prominent tattoos, and that poor descriptions such as the one she offered the police after the attack are consistent with a determination that she had repressed her memory of the attacker and could no longer accurately recall his appearance. Finally, he offered general testimony about the poor performance of victims in witness-identification experiments and about the low correlation between witnesses’ expressed confidence in the identifications they have made and the accuracy of their choices.
The State called no witnesses at the federal evidentiary hearing. Examining the transcripts of the testimony in state court, the magistrate concluded that Henning had testified falsely when he said he had not directed Barthel to look at Dispensa. He *217found that Barthel had identified Dispensa as the rapist “only because she saw him standing with officer [sic] Yarborough and was prompted to look at petitioner by officer [sic] Henning” and that her later in-court identification of Dispensa at the trial “was the direct result of the identification by her at Papa Joe’s Restaurant on July 23, 1981.” He concluded that, therefore, the procedure was “highly and impermissibly suggestive and was intended more to confirm their belief in petitioner’s guilt than in seeking [sic] a truthful resolution of the matter.” Referring to Dr. Fason’s testimony, the magistrate stated,
Although Dr. Fason’s testimony may explain why she [Barthel] relied upon Hen-ning and Yarborough to make the selection for her, it is obvious from the record that she did so and thus would have been obvious without Dr. Fason’s testimony.... [I]t is clear from the record that complainant relied upon Henning and Yarborough to point out the petitioner to her.
After making his findings of fact, the magistrate concluded: “Petitioner has exhausted sufficiently his state remedies and the testimony of Dr. Fason and the testimony of petitioner as to the additional pass-by do not significantly alter either the legal theory or factual basis presented to the state courts.” He recommended that the writ be granted. He did not suggest that he had relied on Dr. Fason’s testimony anywhere in his findings.
After the State had filed objections to the report, the district court denied the State’s motion to dismiss for failure to exhaust state remedies and granted the application for a writ, directing that Dis-pensa be released from custody unless the State elected to retry him within 60 days. After the State filed its notice of appeal, the district court directed that Dispensa, who had by then been in state custody more than four years, be released on posting of $25,000 personal recognizance bond.
II.
The Supreme Court observed in Engle v. Isaac 2
If a defendant perceives a constitutional claim and believes it may find favor in the federal courts, he may not bypass the state courts simply because he thinks they will be unsympathetic to the claim.
This requirement, that a petitioner who seeks federal redress must first seek relief in state courts and thus exhaust his state remedies, is not a jurisdictional prerequisite, but a prudential policy based on concerns for federalism.3 The exhaustion requirement is now codified in the federal habeas corpus statute, which forbids a federal court to grant an application for a writ of habeas corpus unless “it appears that the applicant has exhausted the remedies available in the courts of the State, or that there is either an absence of available State corrective process or the existence of circumstances rendering such process ineffective to protect the rights of the prisoner.”4
This court has consistently held that a federal habeas petitioner has failed to exhaust his state remedies when he relies on a different legal theory than he did in state court5 or when he makes the same legal claim to a federal court but supports the claim with factual allegations that he did not make to the state courts.6 As the State points out, the exhaustion requirement is not satisfied merely because the state court denied Dispensa an evidentiary hearing on his habeas petition. If the petitioner did not fairly present the substance of his claims to the state courts7 or presents new legal theories or entirely new *218factual claims to the federal court,8 the petition must be dismissed even in the absence of a state court hearing,9 so that the state court may have a fair opportunity to determine whether an evidentiary hearing is needed.
In discussing the State’s motion to dismiss for failure to exhaust state remedies, the magistrate reported that he found no material difference in Dispensa’s testimony before him and the earlier testimony in state court. Indeed, the magistrate found the differences in the factual allegations so slight as to make the State’s motion frivolous. As our prior opinion shows, we found that there were differences and that the differences were substantial. By reevaluating the evidence in this way, however, we overstepped the limits of our role as an appellate court, which require that we review the findings of the district court under a standard that affirms all but clear error as to factual determinations.10 The district court decided that Dispensa had not introduced new factual claims at his federal habeas hearing. On this basis, it rejected the State's contention that Dispensa had not satisfied the exhaustion requirement.
In neither the pleadings in federal court nor the evidence at the federal evidentiary hearing did Dispensa present any claim upon which the state courts had not passed. He presented to the federal court the identical legal theory and what the district court found to be substantially the same facts that he had presented to the state courts, which have therefore had an opportunity to consider the case fully. The State has not presented evidence or even a contention that Dispensa has “attempted to expedite federal review by deliberately withholding essential facts from the state
courts.”11 The district court decided which facts presented to it were significant and which were not. This weighing of the evidence for importance was the kind of decision reserved for the trier of fact. Because the district court’s determination that the same facts were presented to state and federal courts is supported by the record, we are bound by it, and we affirm the legal conclusion that results from it: Dispensa has exhausted his state remedies.
III.
The identification of a defendant in a manner that suggests whom the witness should identify is a denial of the defendant’s right to due process of law.12 As the Supreme Court said in United States v. Wade: “The vagaries of eyewitness identification are well-known; the annals of criminal law are rife with instances of mistaken identification.”13 Procedures that induce a witness to identify a particular individual magnify this potential for error. As the Court recognized in Neil v. Biggers, “[s]uggestive confrontations are disapproved because they increase the likelihood of misidentification, and unnecessarily suggestive ones are condemned for the further reason that the increased chance of misidentification is gratuitous.”14
Unnecessarily suggestive out-of-court identifications are not per se subject to exclusion, however; they are admissible if, under the totality of the circumstances, they are sufficiently reliable.15 In reviewing the constitutionality of the admission of out-of-court identifications, courts are to assess their reliability by analyzing such factors as the witness’s opportunity to view the criminal at the time of the crime, the *219witness’s degree of attention, the accuracy of the witness’s prior description of the criminal, the level of certainty demonstrated at the confrontation, and the time between the crime and the confrontation.16 Against these factors, the court ruling on admissibility must weigh the corrupting effect of the suggestive identification itself.17
The State contends that the federal court should not have itself weighed the suggestiveness of Dispensa’s identification but was bound by a state court determination of that issue. In so contending, it relies on 28 U.S.C. § 2254(d), which provides that, in suits by state prisoners seeking federal habeas corpus, “a determination after a hearing on the merits of a factual issue, ... evidenced by a written finding, written opinion, or other reliable and accurate written indicia, shall be presumed to be correct ...”18 unless one of eight enumerated exceptions applies.
The State’s reliance on § 2254(d) is misplaced because the state court record is devoid of written or even clearly inferred findings of fact. After a hearing on Dis-pensa’s motion to suppress, the state trial court declared, “I do not feel that the manner in which the identification was obtained at the restaurant there was unfairly done, and this is your due process.... I don’t think the method employed there was unreasonable under the circumstances and I don’t believe that that would taint her in court identification.” The court stated that misidentification and suggestiveness of the identification procedure were questions for the jury. It announced that it would file a statement of facts and conclusions of law with reference to the identification issue, but it never did. Nor did it submit any question of the identification procedure to the jury or instruct the jury on the law of the reliability of identifications.
From this state record, the most we can deduce is that the trial court found the identification not so unreliable as to be inadmissible as a matter of law. We cannot, however, as the district court could not, work backwards from this conclusion to divine the factual findings on which it is premised.
Factual findings by a state appellate court are also entitled to a conclusive presumption of correctness under § 2254(d).19 On direct appeal, the Texas Court of Appeals for the First Supreme Judicial District affirmed Dispensa’s conviction, finding that
although there is some conflict in the testimony regarding the manner in which the identification procedure was conducted, the totality of the circumstances does not demonstrate that the procedure utilized was so suggestive as to present a substantial likelihood of misidentification. When the five factors set down in Manson v. Brathwaite, and Neal [sic] v. Biggers, supra, are applied in the instant case, it appears that the witness’s identification of the appellant was reliable and did not violate due process.
This statement is also too conclusory to enable us to identify discrete factual findings that we must presume to be correct. The court alluded to a conflict in the testimony about the identification procedure, but failed to record any factual findings about how it resolved that conflict. It mentioned the five factors relevant to a determination of reliability under Neil v. Biggers,20 but did not discuss the application of these factors to the facts of this case. The court’s conclusion that the identification was reliable does not give us sufficient guidance to let us reconstruct its underlying factual bases. Because we cannot identify either state court’s factual basis for its general and conclusory findings, there is nothing to which we can accord the statutorily mandated presumption of correctness.
*220? federal magistrate, having concluded, as we do, that no presumption of correctness applied, entered explicit and detailed findings of fact in which he resolved many contested issues in Dispensa’s favor. He expressly found Dispensa to be “a credible, truthful witness,” and stated, “I accept his testimony given in the [federal] evidentiary hearing as truthful.” He found that Dispensa spoke to all parties in the dining room on the first walk through the restaurant, that Barthel did not identify Dispensa on the first walk-through, that Yarborough conferred with Henning at the table where Barthel was sitting between each pair of walks through the restaurant, that Dispensa made a total of three walks through the restaurant, that Yarborough accompanied Dispensa on the third walk-through, that Henning told Barthel to look at Dispensa while Dispensa and Yarbor-ough stood together talking at the cashier’s desk, and that Barthel thereupon made the identification. The magistrate further found that Henning testified untruthfully when he denied having told Barthel to look in Dispensa’s direction and that Barthel testified untruthfully when she stated that Yarborough did not return to the table to confer with Henning. The magistrate found that Barthel identified Dispensa at the restaurant “only because she saw him standing with officer [sic] Yarborough and was prompted to look ... by officer [sic] Henning.” The district court adopted these findings.
The State asserts that the district court erred in appraising Dispensa's credibility when it had before it only his testimony and the transcripts of the state court hearings. The State, however, had the opportunity to call the policemen and any other witnesses it chose at the federal court hearing. It cannot now object on a basis that asserts, in essence, its failure to do so.
Based on its adoption of the magistrate’s findings of fact, the district court held that the identification procedure was unfair and unnecessarily suggestive. We affirm these findings as not clearly erroneous. Bound by these findings, we reach the same legal conclusion the district court did. Barthel expected to be presented with a suspect to identify; Dispensa was presented to her three times, as Henning told her to look, with Yarborough and Henning conferring in whispers to see whether she had made an identification yet; Yarborough walked with and stood by Dispensa in Barthel’s field of view, thus essentially telegraphing to Barthel, who had until that time failed to perform on cue, “This is the one.”
Our conclusion that the procedure was unfair and unnecessarily suggestive does not mandate the suppression of testimony about the identification if that identification is sufficiently reliable.21 We must therefore test the reliability of Barthel’s identification by the measures the Supreme Court set forth in Neil v. Biggers.22 Of these, by far the most significant in this case is the lack of accuracy of the witness’s prior description of the criminal. Barthel described an assailant shorter and lighter than Dispensa. These discrepancies, considered alone, are not great enough to undermine the reliability of the identification conclusively. Overwhelmingly more significant are those features that Barthel failed to include in her description of the assailant, particularly his moustache, his general hirsuteness, and the striking tattoos. This pronounced incongruity between the assailant described and the suspect identified lead us to conclude that the identification was unreliable and should have been suppressed.
Barthel’s identification of Dispensa was the primary, if not the sole, basis for his arrest. Her out-of-court identification made possible her later identification of him at the hearing on the motion to suppress and at the trial.
If both identifications had been suppressed, the remaining evidence would not have been sufficient to support Dispensa’s conviction. The only other evidence to link *221Dispensa with the assault was the pubic hair sample and Officer Pratt’s testimony. Even the expert who testified about the similarities between the hair samples could not categorically state that one of the two hairs removed from Bartel’s sheets came from Dispensa. Given Dispensa’s hirsuteness and the testimony of his companion that he shed hairs liberally on the sheets at his home, it is noteworthy that only one hair arguably Dispensa’s could be found on the victim’s sheets if the assailant were Dispensa. None of the seven fingerprints found in the complainant’s apartment matched Dispensa’s. Neither the State nor Dispensa’s trial attorney ever established that semen samples taken from Barthel and the sheets on her bed matched Dispen-sa’s blood type. The cigarette found in the ashtray in her apartment was not the brand smoked by either Dispensa or Bar-thel. While Officer Pratt said he saw Dis-pensa at 3:00 a.m. going to his apartment, Dispensa’s friend Wilson testified that Dis-pensa was asleep between 3:00 and 5:00 a.m. on the morning the complainant was raped. The record shows no basis for concluding that even Pratt’s identification was not a result of suggestion.
The record therefore could not support a conviction of Dispensa without Barthel’s in-court identification, which was buttressed by the identification she made at Papa Joe’s Seafood Restaurant. The out-of-court identification was inadmissible and the in-court identification could not stand without it. The district court correctly found that the procedure followed in suggesting Dispensa as the suspect “was likely to and did result in a substantial likelihood of irreparable misidentification.”
For these reasons, the motion for rehearing is granted, our prior opinion is vacated, and the judgment of the district court is AFFIRMED.