This appeal by the State is from the grant of a writ of habeas corpus by the United States District Court for the Western District of New York, John T. Elfvin, Judge. Concededly, identification testimony admitted against the habeas petitioner, Michael Kampshoff, in his trial for murder and other charges pertaining to the death of his aunt was erroneously admitted into evidence. The evidence was obtained though procedures disapproved as suggestive and unconstitutional in United States v. Wade, 388 U.S. 218, 87 S. Ct. 1926, 18 L. Ed. 2d 1149 (1967), and Simmons v. United States, 390 U.S. 377, 88 S. Ct. 967, 19 L. Ed. 2d 1247 (1968). See also Manson v. Brathwaite, 432 U.S. 98, 97 S. Ct. 2243, 53 L. Ed. 2d 140 (1977); Neil v. Biggers, 409 U.S. 188, 93 S. Ct. 375, 34 L. Ed. 2d 401 (1972). The sole question we have on appeal is whether the constitutional error was harmless or not under Chapman v. California, 386 U.S. 18, 87 S. Ct. 824, 17 L. Ed. 2d 705 (1967), and progeny.

We take our basic statement of the facts from the opinion of the Appellate Division in People v. Kampshoff, 53 A.D.2d 325, 385 N.Y.S.2d 672 (1976):

On February 17, 1973 Mrs. Gladys Kampshoff, age 71, was found by a neighbor dead in the garage attached to the home occupied by her alone at 2797 Tonawanda Creek Road, Amherst, Erie County. Her body was partially frozen. A makeshift blindfold and mouth gag were found taped to her face and the body was in a prone position next to the left rear wheel of her automobile. A board was lying partially on her head and a large monkey wrench was lying over one shoulder. While looking for her the neighbor saw the telephone off the hook and found all doors allowing access to the house secured and locked except a garage car door. The television set was on. Telephone Company records showed that the receiver was taken off the hook somewhere between 3:30 and 4:00 p.m. on February 16. The County Medical Examiner was called to Mrs. Kampshoff's residence February 17 and he confirmed the earlier finding of the location and appearance of the victim's body. The cloth binding around the deceased's head was held by an unusual tape used in surgery for binding the closures in operations. He stated that all four extremities of the victim were frozen. There was evidence of bleeding from the rear of the skull and black and blue marks and reddish marks appeared around her left eye and left temporal region. There was a linear wound across and underneath the angle of her jaw and contusions and abrasions to her forehead. There was a superficial laceration in the neck area and two stab wounds on the left side of the chest. The victim's breast bone was fractured and there were many broken ribs. The victim had a tear in the mesentery approximately four inches in length. The cause of death was given as due to multiple abdominal and chest trauma with hemorrhage, multiple trauma to the head and penetrating stab wounds of the left pleural cavity and lung.

The principal narrative of the events leading up to and surrounding the alleged crimes was given by Warren Strodel, an admitted accomplice of defendants who had been granted immunity. It was his testimony that on February 12 he met with the two defendants to discuss plans for the robbery. Defendant Kampshoff told him that he believed his aunt, Mrs. Kampshoff, had between $50,000 and $75,000 in her home. Kampshoff did not want his aunt to see his face. He wanted her subdued as gently as possible, overpowered, taken to another room and tied, so that Kampshoff could look for the money. In pursuance of the plan, on February 16 Strodel told Kampshoff to pick up defendant Jordan and make sure that he had tape and gloves. Kampshoff left and returned with Jordan, and Strodel asked Kampshoff if he had gotten the tape and Kampshoff said he had. Kampshoff and Jordan got into a beige two-door Cadillac with Strodel. Jordan had an attache case to use as a salesman's prop to assist in gaining access to Mrs. Kampshoff's home and Kampshoff had a gym bag. They also had a pellet gun and darts. Jordan had a bulky light sweater on and Kampshoff a brown coat. Both defendants had shoulder-length hair. They observed children playing hockey across the street from Mrs. Kampshoff's home, and passed the house once. It was then between 2:00 and 2:30 p.m. One of the defendants stated that the presence of the children would make no difference and they approached the house a second time and stopped. Jordan got out of the car, Kampshoff gave him some tape and he and Kampshoff drove ahead a distance. They saw Jordan ring the bell, wait a few moments and then walk around to where the garage doors were located. They then drove around and when they returned they did not see Jordan outside the house. Ten minutes from the time they dropped Jordan off they returned to the house and Strodel let Kampshoff off. Kampshoff took the gym bag with him and he saw him go around the corner of the house where the garage doors were located. He then drove around and passed the house every five minutes until 25 minutes had elapsed. At that time he pulled up in front of the house and, as a prearranged signal discussed several times with both defendants he fired a .22 caliber dart with blue feathers at a large window in the corner of the house. He then drove on, turned around and stopped just past the house. Both defendants, Kampshoff with a brown gym bag and Jordan with an attache case, then got back into the car. Jordan had a knife with him. Kampshoff got in first and said, "I think she's dead. He did a number on her." Strodel asked Jordan why he hit her and Jordan said that she wouldn't be quiet. He cut the victim to scare her. Kampshoff said that his aunt was lying on the floor of the garage and looked in pretty bad shape, and that if she was not dead already she probably would freeze to death. He left the phone off the hook because a busy signal would not be as suspicious as a failure to answer. The money they expected wasn't there and upon returning to Strodel's home they went through the contents of the gym bag and found that contained papers, music sheets, bills, cancelled checks, some government checks and $40 in cash. The money was divided and Strodel suggested that they burn the bag and its contents.

Id. at 327-29, 385 N.Y.S.2d at 675-76.

At trial the eyewitness testimony of Philip Heimiller, age fourteen, was admitted. It is the identification testimony of this witness which, it is conceded, was unconstitutionally obtained. Philip Heimiller testified that he was playing hockey on February 16 on a creek across from Mrs. Kampshoff's house about 130 feet away when at about 2:15 or 2:30 p.m. he saw a car stop and a man with long brown hair, wearing a three-quarter-length brown coat and carrying a brown gym bag exit from the car and walk to Mrs. Kampshoff's driveway. About fifteen to thirty minutes later he saw the same man run around the same car and enter it. He saw the man between fifteen and twenty seconds the first time and about ten seconds the second time. At trial he identified the defendant Kampshoff as the man he saw and a burned gym bag produced by the State as looking like the bag that man was carrying. As Judge Elfvin found,

Philip, who was twelve years old at the time of the Kampshoff homicide, was engaged in play with his companions while the crime was being committed. He saw the man whom he identified as petitioner from a distance of about 130 feet on two occasions lasting between 10 and 20 seconds each and separated by 15 to 30 minutes. W, Vol. I, pp. 111-118; People v. Kampshoff, supra, 385 N.Y.S.2d at 678. On the first occasion he had no clear view of the man's face, and on the second occasion he viewed him frontally for no more than five seconds. T, Vol. II, pp. 1100-1101. His attention to the man was very casual; he did not regard his sighting of him as unusual in any way, or speak of it to anyone until news of the crime committed in the nearby Kampshoff residence surfaced. W, Vol. I, pp. 110-111, 120-122; T, Vol. II, pp. 1098-1099. Cf., Manson v. Brathwaite, supra, 432 U.S. at 115, 97 S. Ct. at 2253. The prior description that Philip gave the police cannot be characterized as highly accurate--a man, about twenty years old, having long brown hair, wearing a medium-brown three-quarter length jacket. W, Vol. I, pp. 125-126. Philip's level of certainty in the identification was quite low. Although he unequivocally stated on direct and cross-examination that petitioner was the man he saw (T, Vol. II, pp. 1103, 1148-1149, 1152), he also testified that he could not say for certain that the man he viewed on the two brief separate occasions was the same person each time (id. at 1143), that he was not certain six months after the incident that that man was petitioner, although he was certain at trial (id. at 1146, 1149-1151), and that in fact he could at most state only that petitioner "looks like" the man he saw, not that he was certain that the two are one and the same person (id. at 1244-1246). As to the final Biggers factor, the lapse of two weeks is ample time for memory of such a casual, brief and distant observation of an individual to fade and be subject to replacement by a plausible substitute suggested by the police. Cf., Manson v. Brathwaite, supra, 432 U.S. at 116 [97 S.Ct. at 2253].

I note further that there was no apparent urgency necessitating use of a suggestive identification procedure, which the cases suggest may sometimes excuse the use of such procedures.

Kampshoff v. Smith, No. Civ.-80-1037E, Memorandum and Order at 6-8 (W.D.N.Y. July 6, 1982) (footnotes omitted).1

On the question of harmless error, the Appellate Division observed that Philip's testimony "certainly tended to corroborate the testimony of the accomplice Strodel as to the identity of defendant Kampshoff." 53 A.D.2d at 335, 385 N.Y.S.2d at 678. The Appellate Division further observed that "[c]learly the accomplice testimony of Warren Strodel was the critical evidence introduced against both Kampshoff and [codefendant] Jordan," id. at 338, 385 N.Y.S.2d at 682, and proceeded to imply that Philip Heimiller's identification testimony was the strongest evidence corroborating Strodel: "Kampshoff was placed at the scene of the crime by Philip Heimiller's in-court identification testimony, which also corroborated the testimony of Strodel." Id.

Judge Elfvin referred to the Chapman harmless error test that "before a federal constitutional error can be held harmless, the court must be able to declare a belief that it was harmless beyond a reasonable doubt." 386 U.S. at 24, 87 S.Ct. at 828. He noted that while uncorroborated accomplice testimony may constitutionally provide the exclusive basis for a criminal conviction, it is better for the courts to caution juries against too much reliance upon such testimony and to require corroborating testimony, Caminetti v. United States, 242 U.S. 470, 495, 37 S. Ct. 192, 198, 61 L. Ed. 442 (1917), which in fact is required under the law of the State of New York. Even without such a rule, a jury would be inclined to look carefully for evidence to corroborate the testimony of Strodel, a borderline psychotic and an abuser of phenothiazine drugs who admitted on the stand to having changed his story a number of times. 53 A.D.2d at 329-30, 385 N.Y.S.2d at 676-77. Judge Elfvin carefully weighed the other corroborating evidence and we cannot improve upon his summary of it, as set out in the margin.2

In the end, however, he found that while a rational trier of fact could have been persuaded of Kampshoff's guilt beyond a reasonable doubt even absent Philip Heimiller's identification testimony, see Jackson v. Virginia, 443 U.S. 307, 320-24, 99 S. Ct. 2781, 2789-2791, 61 L. Ed. 2d 560 (1979), it could not be concluded that this testimony did not contribute to petitioner's conviction and was therefore harmless under Chapman. He added that this is even more clearly seen to be true when the impact of Ronald Heimiller's in-court identification, the admission of which was held harmless error by the Appellate Division, is added to that of Philip's testimony.3

There can be no reasonable doubt that inaccurate eyewitness testimony may be one of the most prejudicial features of a criminal trial. Juries, naturally desirous to punish a vicious crime, may well be unschooled in the effects that the subtle compound of suggestion, anxiety, and forgetfulness in the face of the need to recall4 often has on witnesses. Accordingly, doubts over the strength of the evidence of a defendant's guilt may be resolved on the basis of the eyewitness' seeming certainty when he points to the defendant and exclaims with conviction that veils all doubt, "[T]hat's the man!" United States v. Wade, 388 U.S. at 235-36, 87 S.Ct. at 1936-1937; see also E. Loftus, Eyewitness Testimony, ch. 2 at 9 (1979) (statistical evidence reveals that a jury "seems to find it proof enough when a single person implicates another with a remark such as 'I am certain that's the man!' " and "have been known to accept eyewitness testimony pointing to guilt even when it is far outweighed by evidence of innocence"). But while juries may be led by their experience to believe their eyes, and, by inference, what they hear from those who have seen, the experience of law and psychology has been that eyewitness testimony may sometimes be the least trustworthy means to identify the guilty. See 3 J. Wigmore, Evidence Sec. 786a (1970). England's Royal Committee of Inquiry, established in 1904 to investigate the case of Adolf Beck, twice convicted of crimes he did not commit, adjudged that "evidence as to identity based on personal impressions, however bona fide, is perhaps of all classes of evidence the least to be relied upon, and therefore, unless supported by other facts, an unsafe basis for the verdict of a jury." Parliamentary Papers (Commons) 1905, vol. 62 at 465 (Accounts and Papers) Cmd. 2315 Cd., 1904, Report of Committee of Inquiry into the Case of Mr. Adolf Beck, at vii, cited in F. O'Connor, " 'That's the Man': A Sobering Study of Eyewitness Identification and the Polygraph," 49 St. John's L.Rev. 1, 5 (1974). Even with the support of other facts, Professor Frankfurter would ask:

What is the worth of identification testimony even when uncontradicted? The identification of strangers is proverbially untrustworthy. The hazards of such testimony are established by a formidable number of instances in the records of English and American trials. These instances are recent--not due to the brutalities of ancient criminal procedure.

F. Frankfurter, The Case of Sacco and Vanzetti 30 (1927).

The skepticism of eyewitness identifications is based on both experiment and experience. See 3 Wigmore, supra. Professor Buckhout reports a study in which only 7 of 52 witnesses to a staged purse-snatchingless than 20%--were able correctly to pick the culprit out of a line-up. Buckhout, "Eyewitness Testimony," Sci.Am., Dec. 1974, 23, 30. Forensic science has so thoroughly impeached the reliability of eyewitness identification that studies now focus not on the fact of eyewitness inaccuracy but on the causes of the extraordinary gap between a criminal event and human perception and recall of it. See generally E. Loftus, Eyewitness Testimony, supra; note 4 supra. But too often the doubts cast in the laboratory have not found their way into the courtroom and the judgments of juries. Fifty years ago Professor Borchard's study detailed the cases of sixty-five convictions of entirely innocent defendants. More than half of these convictions were due to wrongful eyewitness identification of the accused even though in eight of these cases the accused and the guilty "bore not the slightest resemblance to each other, whereas in twelve other cases, the resemblance, while fair, was still not at all close. In only two cases can the resemblance be called striking." E. Borchard, Convicting the Innocent xiii (1932). Of course, once the state has focused its investigation and prosecution on an innocent defendant, it is less likely that the real culprit will ever be found; the percentage of defendants accused and convicted, jailed or hanged by unreliable eyewitness identifications, will never fully be known.5 Because the intrinsic unreliability of eyewitness identifications is so often compounded with the distorting effects of the natural suggestion to the witness that the person on trial is the guilty one, it is no exaggeration to venture that "[t]he influence of improper suggestion upon identifying witnesses probably accounts for more miscarriages of justice than any other single factor--perhaps it is responsible for more such errors than all other factors combined." P. Wall, Eye-Witness Identification in Criminal Cases 26 (1965), quoted in United States v. Wade, 388 U.S. at 229, 87 S.Ct. at 1933.

This indictment should, needless to say, hardly be read as a rejection of all eyewitness identification. Often, such testimony is accurate and its exclusion would, by imposing an artificial cover of darkness over all criminal acts, be a broad invitation to crime. But whatever may be said in defense of eyewitness testimony in general applies hardly at all to the testimony of Philip Heimiller, only twelve years old on the day of the crime, who saw a person at a distance of about 130 feet. Philip failed to pick out Kampshoff's photograph from a set of mug shots, but was, by the time of trial, convinced that Kampshoff was the man he saw on the day of the crime. His certainty was, beyond cavil, due to his having seen two television broadcasts in which Kampshoff's picture was televised as that of a suspect in the crime, broadcasts which the police had asked that Philip watch. Documents disclosed after Kampshoff's conviction revealed that the television broadcast was the first time Philip saw Kampshoff's face; he had in fact only seen whomever he did see from behind at the scene of the crime.

We agree with the district court that there was more than a "reasonable possibility" that the eyewitness testimony admitted against Kampshoff influenced the jury: eyewitness testimony is among the most influential, even as it is among the least reliable, forms of proof. Whenever such proof is admitted, there is a genuine likelihood that the jury will base its decision on it. And when such evidence is obtained in violation of the constitutional safeguards against influence and suggestion, the probabilities increase that such evidence will be wrong in identifying the defendant as the person seen at the crime by the eyewitness. Moreover, because eyewitness testimony is such powerful stuff and can decide a case on its own strength, it can blind a jury to other exculpatory evidence or inferences. Once loosed, improperly suggested eyewitness testimony may taint a whole trial. Judge Elfvin was therefore surely correct that where, as here, " 'there is a reasonable possibility that the evidence complained of might have contributed to the conviction' ... [an error in its admission] cannot ... be conceived of as harmless." Chapman v. California, 386 U.S. at 23-24, 87 S.Ct. at 827-828, quoting Fahy v. Connecticut, 375 U.S. 85, 84 S. Ct. 229, 11 L. Ed. 2d 171 (1963).

Having found the error and having found that it is harmful, we may properly follow the course taken in Foster v. California, 394 U.S. 440, 89 S. Ct. 1127, 22 L. Ed. 2d 402 (1969), in which the Court reversed and remanded to the state court a conviction based on erroneously admitted eyewitness testimony and an accomplice's testimony which was required to be corroborated under state law. Cf. Bumper v. North Carolina, 391 U.S. 543, 88 S. Ct. 1788, 20 L. Ed. 2d 797 (1968) (reversal and remand of conviction tainted by admission of evidence found in warrantless search, Court noting that there was doubt that defendant committed crime because eyewitnesses could not identify him out of lineup until after local papers had published his name as "prime suspect" and he was required to state his name during the later lineup. Id. at 550 n. 16, 88 S. Ct. at 1792 n. 16).

Such a course may be proper, however, only if Chapman has not been limited or qualified in relevant part by subsequent Supreme Court decisions the language of which suggests that inadmissible evidence which contributed to a defendant's conviction may be harmless if the other competent evidence of guilt is "overwhelming." In Harrington v. California, 395 U.S. 250, 89 S. Ct. 1726, 23 L. Ed. 2d 284 (1969), six members of the Court thought that "the case against Harrington was so overwhelming that we conclude that [a Bruton violation] was harmless beyond a reasonable doubt." Id. at 254, 89 S.Ct. at 1728. The Court did take note of the admonition in Chapman "against giving too much emphasis to 'overwhelming evidence' of guilt, stating that constitutional errors affecting the substantial rights of the aggrieved party could not be considered to be harmless." 395 U.S. at 254, 89 S. Ct. at 1728, citing Chapman, 386 U.S. at 23, 87 S.Ct. at 827. While three members of the Court thought that Chapman v. California was overruled, the majority opinion in Harrington specifically stated that "[w]e do not depart from Chapman; nor do we dilute it by inference. We reaffirm it." Id. We take the Court, as we must, at its word.

From Schneble v. Florida, 405 U.S. 427, 92 S. Ct. 1056, 31 L. Ed. 2d 340 (1972), one commentator concludes that the Court set "the groundwork ... for a repudiation of the reasonable possibility standard avowed in Chapman." Saltzburg, The Harm of Harmless Error, 59 Va.L.Rev. 988, 1017-18 (1973). The Schneble Court, after weighing the prejudicial effect of an improperly admitted codefendant's admission against the overwhelming nature of properly admitted proof, found the error so "insignificant" as to be harmless beyond a reasonable doubt. 405 U.S. at 429-32, 92 S.Ct. at 1058-1059. The Court emphasized that the allegedly inadmissible statements "at most tended to corroborate certain details of petitioner's comprehensive confession," id. at 431, 92 S.Ct. at 1059, a confession which under the charge could have been held involuntary and inadmissible and without which the State's case was "virtually nonexistent" because "the jurors could on no rational hypothesis have found Schneble guilty without reliance on his confession." Id. The Court then went on to say that, "unless there is a reasonable possibility that the improperly admitted evidence contributed to the conviction, reversal is not required. See Chapman v. California, 386 U.S. 18, 24 [87 S. Ct. 824, 828, 17 L. Ed. 2d 705] (1967)." Id. at 432, 92 S.Ct. at 1059. We note that in the instant case if Strodel's dubious and self-interested testimony were disregarded, it is only the identification testimony of the Heimiller brothers that placed Kampshoff at the scene of the crime at the relevant time. On this basis, within the majority opinion in Schneble there was both "a reasonable possibility that the improperly admitted evidence contributed to the conviction," and merely "sufficient" and not "overwhelming" proof of guilt, so that the district judge properly issued the writ.

In Milton v. Wainwright, 407 U.S. 371, 92 S. Ct. 2174, 33 L. Ed. 2d 1 (1972), admission of a post-indictment confession made to a police officer posing as a fellow prisoner (in violation of Massiah v. United States, 377 U.S. 201, 84 S. Ct. 1199, 12 L. Ed. 2d 246 (1964)) was found harmless beyond a reasonable doubt within Harrington and Chapman, cited by the Court, 407 U.S. at 372, 92 S.Ct. at 2175, in light of three other unchallenged confessions and strong corroborative evidence; in the majority opinion's view the evidence of guilt was "overwhelming." Id. at 373, 377-78, 92 S. Ct. at 2175, 2177-2178. Here, of course, we have no confessions, the circumstantial evidence connecting Kampshoff to the crime is ambiguous, see supra note 2, and, all told, absent the identification testimony, the evidence against Kampshoff is simply not overwhelming. We cannot conclude that there is "no reasonable doubt that the jury ... would have reached the same verdict ..." without hearing the Heimiller testimony. 407 U.S. at 377, 92 S.Ct. at 2177. We do not think Milton is applicable.6

In Donnelly v. DeChristoforo, 416 U.S. 637, 94 S. Ct. 1868, 40 L. Ed. 2d 431 (1974), a first degree murder case, a prosecutor's improper remark7 followed by the trial court's specific disapproving instructions was held not to be "prejudice amounting to a denial of constitutional due process." 416 U.S. at 648 n. 23, 94 S. Ct. at 1873 n. 23. The Court noted that "[c]onflicting inferences have been drawn" from the remark by the courts below. Id. at 643, 94 S.Ct. at 1871. It then went on to say that "it is by no means clear" or "even probable" that the jury would give the remark the meaning attributed to it by the majority of the First Circuit Court of Appeals (and, we would add, two Justices of the Supreme Judicial Court of Massachusetts). Id. at 644, 94 S.Ct. at 1872. Thus while it was possible that the jury did give the remark that meaning, there was no reversible error. We take it that the Court, rather than changing the rule of Chapman, which it did not cite, was speaking in the context of prosecutorial misconduct, and was distinguishing the case from those in which "the State has denied a defendant the benefit of a specific provision of the Bill of Rights ... or in which the prosecutor's remarks so prejudiced a specific right ... as to amount to a denial of that right." Id. at 643, 94 S.Ct. at 1871. DeChristoforo, in other words, does not address the issue in our case.

We conclude as did the district court that the constitutional error in admitting Philip Heimiller's identification testimony was not harmless.

Judgment affirmed.