On this appeal from the denial of a writ of habeas corpus the legal principles are clear; the basic facts, also, are not in dispute. The sole question is the conclusion to be drawn, specifically, whether an initial, show-up, identification of petitioner was so suggestive and unreliable that it must be regarded as tainting the one made in court. The Puerto Rico Supreme Court held that it was not. People v. Casiano Velez, 106 D.P.R. 653 (1977) (4-2); Id., 106 D.P.R. 660 (1978) (3-2); Id., 82 J.T.S. 15 (1982) (3-2), as did the district court. Casiano Velez v. Schmer, Civ. No. 78-625 (D.P.R.1982) (unpublished). We find ourselves obliged to reverse.

On January 12, 1973, at approximately 5:30 p.m., five boys in their late teens were walking down a road in Villalba, Puerto Rico, when they were suddenly confronted by two men who were unknown to them. One of the men held a rifle. He aimed it at the boys, asked what was going on, and then fired a shot in the air. When one of the boys continued walking, the man shot and killed the boy, and then immediately fired upon and wounded another. Both men immediately fled in an automobile. The boys described the car in detail to the police, who, five hours later, found it in the area, being driven by petitioner's cousin. They were unable to stop him, but fifteen minutes afterwards, discovered the car, vacant.1

Petitioner was brought to the police station at 3:00 a.m. He was shown to the boys there with the statement, "This is him, isn't it?" This they answered in the affirmative. Thereafter they identified petitioner in the courtroom. The station identification was not introduced by the Commonwealth, but was advanced, unsuccessfully, by petitioner as a ground for excluding in-court identifications.

Our decision is controlled by the standards set out in Manson v. Brathwaite, 1977, 432 U.S. 98, 113-116, 97 S. Ct. 2243, 2252-2253, 53 L. Ed. 2d 140, and Neil v. Biggers, 1972, 409 U.S. 188, 198, 93 S. Ct. 375, 381, 34 L. Ed. 2d 401; whether, viewing the "totality of the circumstances," the out-of-court identification procedure was so unnecessarily suggestive that it raised a "substantial likelihood of irreparable misidentification." Such a result would violate the principle of fundamental fairness required by the due process clause. See also Foster v. California, 1969, 394 U.S. 440, 89 S. Ct. 1127, 22 L. Ed. 2d 402; Simmons v. United States, 1968, 390 U.S. 377, 88 S. Ct. 967, 19 L. Ed. 2d 1247; Stovall v. Denno, 1967, 388 U.S. 293, 87 S. Ct. 1967, 18 L. Ed. 2d 1199. We must, accordingly, consider the necessity and suggestiveness of the show-up; the reliability of the identification, and the likely carry-over effect. See Simmons v. United States, ante, 390 U.S. at 383-84, 88 S.Ct. at 970-71; cf. Dickerson v. Fogg, 2 Cir., 1982, 692 F.2d 238, 244; Swicegood v. Alabama, 5 Cir., 1978, 577 F.2d 1322, 1325.

The district court's finding the show-up not unnecessarily suggestive was clearly erroneous. A staged show-up is presumptively more suggestive than a line-up, as the police had been well advised by their own court, e.g., People v. Gomez Incera, 97 P.R.R. 243 (1969), and particularly suggestive, as well as totally unnecessary, was the introduction, "This is him, isn't it?" Foster v. California, ante, 394 U.S. at 443, 89 S.Ct. at 1129. Necessity must be measured in terms of the cost of a fairer procedure. Here we are talking only of a short wait for a later morning line-up. Cf. Nassar v. Vinzant, 1 Cir., 1975, 519 F.2d 798, 801, cert. denied, 423 U.S. 898, 96 S. Ct. 202, 46 L. Ed. 2d 132. The district court's litmus, that the police "needed to know right away," so as to abandon, or continue their search, with no specialized need appearing, would excuse any show-up.

Unnecessary suggestiveness does not of itself lead to exclusion; we must still consider reliability. Manson v. Brathwaite, ante, 432 U.S. at 113, 97 S.Ct. at 2252. While suggestiveness reflects generally on reliability, the Court in Manson and in Neil v. Biggers, ante, 409 U.S. at 199-200, 93 S.Ct. at 382, has listed five lines of more specific inquiry.

1. Opportunity to view the criminal at the time of the crime. The witnesses had about a minute. Although nearing sunset, there was light, the weather was clear, and the witnesses were from 20 to 40 feet from the assailant. The Puerto Rico Supreme Court found "the identification ... was reliable since they had plenty of time to see him for a minute or more on a clear day." We, of course, accept that. There was, however, no finding that they made use of this time, or as to their subjective receptivity. It is hard to disagree with the minority justices' opinion that the latter was poor,2 namely,

2. Degree of attention. None of the prior courts' opinions dealt with the degree of attention except to point out, as supportive, the witnesses' ability to describe the automobile. The Court said, "The automobile episode described above by the witnesses supports the reliability of the identification." With respect, we must disagree. Although both the assailant and the automobile could, of course, be seen, a minute is a short time and the more one looked at the one, the less one could look at the other. When one is identifying magnesium wheels, one is not looking at someone's face. Similarly, we note in the superior court's "Resolution" denying a motion for new trial, that the boys testified to the length of the rifle, and the color of the barrel and the butt. Rather than supporting the reliability of the identification, the Puerto Rico courts' recitation of these secondary matters emphasizes the absence of the primary ones.

3. The initial description of the criminal. We have seen no case with such a total absence of description of the individual. All said he wore a white T-shirt; one said his hair was "shaggy," and that was it. The district court disposed of the fact that, when arrested, petitioner wore a colored T-shirt, by saying that he could have changed it. We can agree with that, but we cannot accept its further conclusion that seeing he had a white T-shirt was significant evidence of careful observation. Not one of the witnesses gave the assailant's age, build, height, weight, skin color, other clothing, or other indicia of appearance. Cf. Manson, ante, 432 U.S. at 101, 97 S.Ct. at 2246 (prior description included height, build, hairstyle, skin color, facial characteristics, and clothes). Even shaggy hair is but a broad generalization. The void is both striking, and extraordinary.

4. The certainty demonstrated by the witnesses at the confrontation. The prior courts spoke of the witnesses' certainty. However, certainty must itself be considered in the totality of the circumstances. Here were adolescent boys, doubtless viewing petitioner and commenting, reassuringly, together. Cf. Swicegood v. Alabama, ante, 577 F.2d at 1327 (identification less reliable when witnesses are allowed to discuss among themselves during identification proceedings). Even if that were not so, they were boys who had been kept at the police station some nine hours and who were undoubtedly tired and wanting to go home, and who were told, "This is him, isn't it?" Their ensuing "certainty" is a singular companion to their prior total uncertainty. We must heavily discount it.

5. The length of time between the crime and the confrontation. There is, concededly, an image, or overall impression, that persists in the subconscious beyond the ability to verbalize individual parts. Courts have recognized that identifications, even show-ups, will benefit from this if conducted closely after the crime. Most cases giving substantial weight to this factor, however, involve identifications within an hour of the crime. E.g., Bates v. United States, D.C.Cir., 1968, 405 F.2d 1104 (30 minutes); Harris v. Wyrick, 8 Cir., 1981, 644 F.2d 710 (one hour); Frank v. Blackburn, 1979, 605 F.2d 910, aff'd en banc, 646 F.2d 873, 875, cert. denied, 454 U.S. 840, 102 S. Ct. 148, 70 L. Ed. 2d 123 (30 minutes). Nine and a half hours is over twice as long as any case we have discovered according such weight. Faison v. Zahradnick, 4 Cir., 1977, 563 F.2d 1135, cert. denied, 436 U.S. 908, 98 S. Ct. 2241, 56 L. Ed. 2d 406 (four hours).

In sum, the initial reliability tests, considered collectively, demonstrate a "substantial likelihood of irreparable misidentification." The remaining question is the carry-over, or tainting effect, upon the reliability of the in-court identification. Was a wrongly created impression carried over, so that subsequent identifications likely were merely of the individual so unreliably identified? Were the witnesses imprinted, or, in computer terminology, improperly programmed?3 In considering the effect of a mental imprint one must consider both its force and the field upon which it was impressed. The force here, the suggestiveness, was strong. The field, prior to the show-up, was a negative in the true sense of the word; the original viewing had produced virtually no impression.

We recognize the heavy burden upon petitioner under 28 U.S.C. Sec. 2254(d); Sumner v. Mata, 1981, 449 U.S. 539, 101 S. Ct. 764, 66 L. Ed. 2d 722; s.c. 455 U.S. 591, 102 S. Ct. 1303, 71 L. Ed. 2d 480 (1982); s.c. --- U.S. ----, 104 S. Ct. 386, 78 L. Ed. 2d 332 (1983), but, in this very exceptional case, we hold that he has convincingly demonstrated the Puerto Rico Court's conclusion of reliability cannot be supported by the record. We must agree with the minority justices; the show-up was so paramount that the witnesses' subsequent recollection, in all probability, was simply of the person they saw at the stationhouse, rather than at the scene of the crime. See Simmons, ante.

Finally, there can, on the record, be no claim of harmless error. No other questions need be reached.4 The writ is to be granted unless petitioner is retried within such length of time as may be determined by the district court.