OPINION OF THE COURT
BECKER, Circuit Judge.
This appeal by defendant Richard Stevens from the judgment of sentence of the district court for the District of New Jersey presents a number of substantial issues. Stevens, having been convicted of aggravated sexual assault and robbery, both within the special territorial jurisdiction of the United States in violation of 18 U.S.C. §§ 2241 and 2111 respectively, mounts a multi-pronged challenge to his conviction. Two of these challenges are bottomed on the fifth amendment's due process clause. To begin with, he objects under Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), to the government's "destruction" of the one item of physical evidence that might have exonerated him — a saliva/semen sample taken from the mouth of one of the victims. Stevens also complains of the procedure used to identify him, claiming that it failed to pass muster under Simmons v. United States, 390 U.S. 377, 88 S.Ct. 967, 19 L.Ed.2d 1247 (1968).
Stevens advances several evidentiary challenges as well. He objects to the introduction of the testimony of his Pretrial Services Officer concerning a defense witness's alleged prior inconsistent statement on the ground that pretrial services information is rendered confidential by 18 U.S.C. § 3153(c)(1) (1982). Stevens also protests the extent to which the district court, applying our opinion in United States v. Downing, 753 F.2d 1224 (3d Cir.1985), limited the testimony of his expert witness regarding the reliability of eyewitness identification. He further remonstrates about the district court's refusal to admit the testimony of one Tyrone Mitchell under Fed.R.Evid. 404(b), or, more accurately, under a seldomly used subspecies of Rule 404(b) known as "reverse 404(b)." Stevens proffered Mitchell's testimony that he (Mitchell) was the victim of a crime which was so similar to the instant crime that the investigating officers believed that the same individual had committed both. Mitchell also would have testified that he, unlike the victims here, did not identify Stevens as his assailant. According to Stevens, Mitchell's testimony would have tended to show that some unknown third person had perpetrated both crimes, and that the victims had misidentified him as their attacker.
We will affirm the district court's Brady and Simmons determinations, its admission of the testimony of Stevens's Pretrial Services Officer, and the majority of its Downing rulings. Stevens's Brady argument fails because there is insufficient evidence of governmental bad faith to satisfy the standard for relief laid down in Arizona v. Youngblood, 488 U.S. 51, 109 S.Ct. 333, 102 L.Ed.2d 281 (1988). Nor has Stevens shown that the "identification procedure was so impermissibly suggestive as to give rise to a very substantial likelihood of ... misidentification," Simmons, 390 U.S. at 384, 88 S.Ct. at 971, especially since the victims' identifications were unusually reliable under Neil v. Biggers, 409 U.S. 188, 93 S.Ct. 375, 34 L.Ed.2d 401 (1972). We also think that the admission of the testimony
We believe, however, that Downing requires that Stevens's expert be permitted to testify concerning the lack of a correlation between confidence and accuracy in eyewitness identifications. Both of the victims expressed a great deal of confidence in their identifications of Stevens. To counteract this highly damaging testimony, Stevens offered expert testimony that, contrary to popular belief, scientific studies have shown "a fairly weak relationship" between confidence and accuracy. We conclude that the district court erred in holding that there was no "fit" between this testimony and the facts at bar. We also are satisfied that such testimony would have proven helpful to the jury in assessing the victims' identifications.
We also think that Stevens should have been allowed to call Tyrone Mitchell as a witness, and to introduce other evidence concerning the parallels between the Mitchell crime and the crime sub judice. When a defendant proffers "other crimes" evidence under Rule 404(b), there is no possibility of prejudice to the defendant; therefore, the other crime need not be a "signature" crime. Instead, it only need be sufficiently similar to the crime at bar so that it is relevant under Fed.R.Evid. 401 and 402, and that its probative value is not substantially outweighed by Fed.R.Evid. 403 considerations. Applying this standard to the instant case, we are satisfied that the Mitchell crime clears the relatively low relevancy hurdle. We find it significant that the investigating authorities thought that the same individual had committed both crimes, and that the fruits of both crimes, which occurred within a few hundred yards of one another at Fort Dix, New Jersey, ended up at the same time in Fort Meade, Maryland. We also are confident that the probative value of Mitchell's testimony is not trumped by Rule 403 factors such as undue waste of time or confusion of the issues.
Given the closeness of the case against Stevens — his first trial ended in a mistrial after the jury could not agree upon a verdict — we think that these errors cannot be deemed harmless. We therefore will reverse Stevens's conviction and remand for a new trial, at which Stevens's expert will be allowed to testify about the lack of a correlation between confidence and accuracy in eyewitness identifications and evidence will be admitted about the similarities between the Mitchell crime and the crime at issue here.
I. FACTS AND PROCEDURAL HISTORY
At about 9:30 p.m. on April 15, 1989, a damp and chilly Saturday evening, two white Air Force police officers, Jane Smith
A few minutes later, a black male, wearing a wool cap and a tan nylon jogging suit, entered the shelter, paused for a moment about ten feet away from the officers, and then asked them who they were. Both officers stood, but neither responded to the inquiry. The man approached the officers and, standing just a few feet away, reached his right hand into the small of his back and drew a small, silver handgun from his pants. Pointing the gun at McCormack's chest, the man demanded McCormack's wallet. McCormack handed over his wallet, which contained an unsigned $100 money order. To assure himself that McCormack was not withholding any cash, the
The man then turned his attention, and his gun, to Smith. He asked for her money, but she had none. After patting down Smith's pockets, the man ordered her to drop her pants; totally helpless, Smith complied. As he pulled down his own pants, the man leveled his gun at McCormack's head and demanded that Smith kneel before him. He told Smith that unless she performed fellatio on him, he would "blow [McCormack's] brains out." Not wanting to endanger her friend's life, Smith again complied. Three or four minutes later, the man sat down on the bench and directed McCormack to sit beside him. Placing the gun on McCormack's left temple, he insisted that Smith complete the act. She did.
Soon afterward, a car drove by with its headlights on, and all three stood up. The man told McCormack to leave the bus shelter and run across an adjacent field, which McCormack did. Seconds later, the man turned and faced Smith, instructing her to run across the field as well. Smith followed after her friend. The two ran to the nearest building, the non-commissioned officers' club, where they telephoned the military police. Within five minutes, two military police officers arrived and escorted Smith and McCormack to the Fort Dix military police station.
At the station, Smith and McCormack met Christine Amos, a military police investigator. Smith advised Amos that she needed to go to the hospital, because she believed that traces of her attacker's semen remained in her mouth and on the sleeve of her jacket. Before leaving to arrange for transportation to the hospital, Amos invited Smith and McCormack to take a look at the wanted board on the wall and to see if anyone resembled their assailant. This wanted board consisted of eight posters, containing mostly composite sketches but also some photographs. McCormack rose, approached the wanted board, and almost immediately focused upon a photograph of the defendant, Richard Stevens. McCormack said: "[T]his is him. This is the man." Smith agreed that the photograph resembled their attacker, but thought that it made him appear a bit heavier. When Amos returned, Smith and McCormack informed her that they had identified a photograph of their assailant. Amos removed the poster of Stevens from the wall and then accompanied Smith to Walson Army Community Hospital.
Upon arriving at the hospital, Smith was taken into an examination room where a doctor administered a rape crisis kit. Among other tests performed, the doctor attempted to secure samples of the assailant's semen from Smith's saliva and clothing. Once these tests were completed, Smith was met by Agent Timothy Jackson of the United States Army Criminal Investigation Division ("CID"). Smith and Agent Jackson returned to the Fort Dix military police station where Smith recounted to Agent Jackson that evening's traumatic events. Agent Jackson also showed Smith an array of six photographs from which she identified Stevens as her attacker. While Smith was at the hospital, another CID agent interviewed McCormack and showed him the same photographic spread. Like Smith, he identified Stevens as the assailant.
Five days later, Smith and McCormack returned to the police station to view a lineup. The lineup, which consisted of seven individuals, was prepared and conducted by Agent James Maxwell of the Federal Bureau of Investigation. Smith and McCormack viewed the lineup separately and, without consulting with one another, each identified Stevens as their assailant. Both stated that they were positive that Stevens was the man who had robbed and assaulted them at the roadside bus stop.
Based on the victims' identifications, a federal grand jury sitting in the District of New Jersey returned a two-count indictment charging Stevens with aggravated sexual assault and robbery in the first degree, both within the special territorial jurisdiction of the United States, in violation
Stevens filed a number of pretrial motions. He moved to dismiss the indictment on the ground that the government had violated Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), by destroying, in contravention of a consent order, a semen sample taken from Smith's mouth. Had the government adhered to the terms of the consent order, Stevens contended that he could have performed DNA testing on the sample, which might have proven conclusively his innocence. A hearing was held on this motion, but it was denied. Stevens also moved to suppress evidence of Smith's and McCormack's identification of him from the wanted board, claiming that the identification procedure used was impermissibly suggestive and conducive to mistaken identification. After conducting a Wade hearing,
In January of 1990, Stevens was tried before a jury, which ultimately became deadlocked, so a mistrial was declared. Stevens was retried in March of 1990 on the same charges. Following a four-day trial, this jury convicted Stevens on both counts. Stevens thereafter filed a motion for judgment of acquittal, Fed.R.Crim.P. 29(c), which was denied by the district court. The court then sentenced Stevens to concurrent terms of 168 months of incarceration, together with concurrent terms of three years supervised release and special assessments totalling $100. This appeal followed, over which we have jurisdiction pursuant to 28 U.S.C. § 1291.
II. THE BRADY ISSUE
Following the assault, Smith was taken to a hospital where saliva samples were extracted from her mouth. Three glass slides were prepared from these samples, which, along with other materials, were forwarded to an FBI laboratory for forensic testing. At Stevens's first detention hearing, Agent Morris Austin, the FBI agent in charge of the Stevens investigation, testified that the FBI was performing DNA tests on the slides to determine whether or not Stevens was the perpetrator, but that the results of these tests would not be available for another six to eight weeks. The three glass slides were discussed again at Stevens's second detention hearing, at the conclusion of which Stevens's counsel made the following request: "I would request that the samples that the FBI is now testing for D & A [sic], I would like them preserved, whatever they can preserve so that we may later test
Meanwhile, the FBI laboratory was attempting to perform tests on the three glass slides containing Smith's saliva samples. Microscopic analysis of the slides revealed that: (1) slide # 1 contained no semen at all; (2) slide # 2 contained traces of semen, but too little for forensic testing of any kind; and (3) slide # 3 contained sufficient semen to attempt serological testing, but not enough to permit DNA testing. Because DNA testing appeared out of the question, FBI Agent Mark Babyak decided to attempt to perform a standard serological analysis on slide # 3 to ascertain blood type.
Stevens contends that the district court should have dismissed the indictment in view of the government's bad-faith destruction of potentially exculpatory evidence (i.e, the saliva/semen sample on slide # 3) in violation of the consent order. He asserts, in particular, that the magistrate judge's comments at the conclusion of the second detention hearing and the consent order entered into shortly thereafter required the government to allow the defense to conduct DNA testing on the sample before the FBI performed any serological testing. The government's putative failure to adhere to the terms of the consent order, Stevens argues, constitutes "per se bad faith" in violation of the fifth amendment's due process clause, see Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963). Resolution of this claim requires us to apply the standard announced by the Supreme Court in Arizona v. Youngblood, 488 U.S. 51, 109 S.Ct. 333, 102 L.Ed.2d 281 (1988). There, the Court held that "unless a criminal defendant can show bad faith on the part of the police, failure to preserve potentially useful evidence does not constitute a denial of due process of law." Id., 109 S.Ct. at 337.
At the conclusion of the hearing on Stevens's Brady motion, the district court applied this standard and found that the government did not destroy the saliva/semen sample in bad faith:
We may not disturb this finding on appeal unless it is "clearly erroneous." See United States v. Galvan-Garcia, 872 F.2d 638, 641 (5th Cir.), cert. denied, ___ U.S. ___, 110 S.Ct. 164, 107 L.Ed.2d 122 (1989). After reviewing the record and Stevens's allegations, we hold that the district court's conclusion that the government acted in good faith when it elected to perform serological testing on Smith's saliva sample is not clearly erroneous.
As the Supreme Court remarked in Youngblood, "[t]he presence or absence of bad faith by the [government] for purposes of the Due Process Clause must necessarily turn on the [government's] knowledge of the exculpatory value of the evidence at the time it was lost or destroyed." 109 S.Ct. at 337 n. **. Nothing in the record suggests that the government suspected that the saliva/semen sample on slide # 3 could form a basis for exonerating Stevens. The FBI agents knew nothing about what slide # 3 might reveal; they knew only that the slide contained too little semen to permit DNA testing. They therefore elected to pursue another type of forensic test in an effort to discover the assailant's blood type. This test either might have inculpated or exculpated Stevens; no one knew. We agree with the district court that the performance of this alternative test bespoke of no bad faith.
Also, we do not think that the performance of serological testing on the sample contained on slide # 3 contravened the terms of the consent order. This order, which was drafted by the parties and signed by the magistrate judge, instructed the FBI "to preserve, if possible, after the FBI has conducted its own tests, bodily samples extracted from [Smith], ... so that defendant can later conduct, if necessary, his own tests for DNA content." (Emphasis added).
Stevens construes this order as requiring the FBI to disclose the samples to him after the FBI has conducted its own DNA tests. He contends that neither the order nor the parties' prior discussions permitted the FBI to perform serological testing before he was afforded an opportunity to test the samples for DNA content. We disagree. The terms of the consent order required the FBI to forward to Stevens whatever remained of the samples after the FBI had tested them to their satisfaction. Nothing in the order expressly limited the FBI to testing for DNA content. The order instead employed broad language, permitting the FBI to conduct its "own tests." Based on this plain language, we think that the government complied fully. After performing a standard serological analysis, which (unfortunately) consumed all of the materials on slide # 3, the government made available to Stevens for testing and inspection all remaining samples taken from Smith's body.
In sum, we conclude that the district court's finding that the government did not destroy in bad faith the sample on slide # 3 is not clearly erroneous, and that, a fortiori, the court did not err in refusing to dismiss the indictment against Stevens based on the alleged Brady violation.
III. THE WANTED BOARD IDENTIFICATION
Stevens next contends that the wanted board at the Fort Dix military police station was impermissibly suggestive and that it created a substantial risk of misidentification.
In Simmons v. United States, 390 U.S. 377, 383, 88 S.Ct. 967, 971, 19 L.Ed.2d 1247 (1968), the Supreme Court acknowledged that the "employment of photographs by police may sometimes cause witnesses to err in identifying criminals." This danger, the Court noted, is acute when the witness is shown "pictures of several persons among which the photograph of a single such individual recurs or is in some way emphasized." Id. (emphases added). The Supreme Court nonetheless declined to erect a per se bar against initial identification by photograph. Id. at 384, 88 S.Ct. at 971. The Court instead held that evidence of a pretrial photographic identification is inadmissible "only if the photographic identification procedure was so impermissibly suggestive as to give rise to a very substantial likelihood of irreparable misidentification."
The Simmons/Stovall inquiry is essentially two-pronged. See United States ex rel. Phipps v. Follette, 428 F.2d 912, 914-15 (2d Cir.), cert. denied, 400 U.S. 908, 91 S.Ct. 151, 27 L.Ed.2d 146 (1970). The first question is whether the initial identification procedure was "unnecessarily" or "impermissibly" suggestive. This inquiry actually contains two component parts: "that concerning the suggestiveness of the identification, and that concerning whether there was some good reason for the failure to resort to less suggestive procedures." 1 W. LaFave & J. Israel, Criminal Procedure § 7.4(b), at 581 (1984) (emphasis added).
After conducting a Wade hearing, the district court concluded that, "looking at the totality of the circumstances," the wanted board identification was not unnecessarily
It cannot seriously be controverted that the wanted board had several suggestive attributes. Whereas the posters of Stevens each contained photographs, most of the other posters had only composite sketches. Stevens's picture, moreover, was the only one that appeared twice on the wanted board. Both of these features, quite possibly, could have drawn the victims' attention to Stevens. See, e.g., Heiman v. State, 511 N.E.2d 458, 459-60 (Ind.1987) (scrutinizing photographic array that contained two photographs of the suspect). Even more bothersome, in our view, are that Stevens's photographs were significantly larger than the others, and that Stevens was the only suspect portrayed in color. See, e.g., Passman v. Blackburn, 652 F.2d 559, 570 (5th Cir.1981) ("The use of a single color photograph ... with eleven other black and white mug shots ... carried an inherent suggestivity."), cert. denied, 455 U.S. 1022, 102 S.Ct. 1722, 72 L.Ed.2d 141 (1982); O'Brien v. Wainwright, 738 F.2d 1139, 1140-41 (11th Cir.1984) (same), cert. denied, 469 U.S. 1162, 105 S.Ct. 918, 83 L.Ed.2d 931 (1985). Given all this, we believe that, had the military police employed the wanted board as a conventional photographic array, it would have been unnecessarily suggestive.
The wanted board, however, was not a traditional investigative photographic array that the police had arranged with this particular crime in mind. It was, rather, a collection of random sketches and photographs that had been assembled in order to facilitate chance identifications like the one that occurred here.
More importantly, even if the identification procedure in this case was unnecessarily suggestive, "a degree of suggestiveness
Id. at 199, 93 S.Ct. at 382. Applying these criteria to this case, we are confident that the victims' identifications of Stevens from the wanted board were reliable, despite the suggestiveness of the wanted board.
1. The opportunity to view the assailant. McCormack and Smith, both of whom are military police officers trained in observation, had an unusually lengthy and vivid opportunity to view the perpetrator. The attack lasted at least ten minutes, during which time the assailant was in close proximity and made no effort to conceal his face. McCormack and Smith testified that the attacker directly faced them on several occasions, and that there was sufficient light in the bus shelter to see his face. Smith testified that there was ample light in the shelter to read a book or the face of a watch.
2. The degree of attention. Neither McCormack nor Smith was a casual or passing observer; they were, instead, victims of a terrifying crime that riveted their attention to the assailant. Both emphasized in their trial testimony that they paid close attention throughout the incident. Smith stated:
In addition, as the Supreme Court noted in Manson, "specially trained" officers, like Smith and McCormack, can "be expected to pay scrupulous attention to detail." 432 U.S. at 115, 97 S.Ct. at 2253.
3. The witnesses' level of certainty. When McCormack and Smith first spotted Stevens's photographs on the wanted board, they expressed a high degree of certainty that he was the perpetrator. Smith said, "[I] looked down at him and the first thing that hit me was, yes, it was. It's the man. It was his eyes." She did, however, admit that the assailant appeared to be thinner in the wanted board photograph: "[I]n real life his face looked slightly heavier." McCormack's identification, in contrast, was not qualified in any respect. He testified that he recognized Stevens's picture "as soon as [he] came to it":
4. The time between the crime and the identification. That the victims identified Stevens almost immediately after the crime adds considerable weight to their emphatic assurances of certainty. "We do not have here the passage of weeks or months between the crime and the viewing of the photograph," Manson, 432 U.S. at 116, 97 S.Ct. at 2254. McCormack and Smith recognized Stevens on the wanted board within about an hour of the attack. It is reasonable
The above factors are by no means exhaustive, and, admittedly, several countervailing considerations exist that detract from the reliability of the victims' identifications. First, both victims were white, whereas the assailant was black. Scholarly literature attacking the trustworthiness of cross-racial identification is now legion.
Despite these countervailing considerations and the latent suggestivity of the wanted board, we are satisfied that the Neil factors indicate a reliable basis for the victims' identifications. About an hour after intensely observing their attacker at close range for an extended period of time in a well-lit bus shelter, McCormack and Smith, two specially trained military police officers, confidently identified Stevens from a police-station wanted board. Because the suggestivity of the wanted board did not, in our view, create "a very substantial likelihood of ... misidentification," Simmons, 390 U.S. at 384, 88 S.Ct. at 971, we conclude that the evidence of McCormack's and Smith's identifications from the wanted board was properly admitted.
IV. THE TESTIMONY OF PRETRIAL SERVICES OFFICER WELHAM
Both McCormack and Smith testified that their assailant held a gun in his right hand throughout the incident. In response to this testimony, Stevens called his father, Johnny Richardson, to testify that Stevens is left-handed. The district court admitted Richardson's testimony as lay opinion under Fed.R.Evid. 701. The government, in its cross-examination of Richardson, attempted to undermine the natural assumption that Richardson knew whether his son was left- or right-handed, by establishing that Richardson only had known Stevens for about a year and that he had been with Stevens only fifteen times during that year. In pursuing this line of cross-examination, the government asked Richardson how much time he had spent with Stevens in mid-April of 1989. After Richardson responded that Stevens had stayed with him for three or four days, the government asked Richardson if he had told Pretrial Services Officer ("PSO") Kurt Welham that Stevens had only spent one night at his house in mid-April. Richardson answered that he could not recall having a conversation with a PSO.
The district court then permitted the government to call Welham as a rebuttal witness. Welham testified that Richardson had informed him that Stevens's mid-April visit had lasted only one night. This testimony constituted extrinsic evidence of a prior inconsistent statement, satisfying the requisites of Fed.R.Evid. 613(b):
See also Sisto, 534 F.2d at 622 ("`[I]f on cross-examination the witness has denied making the [prior inconsistent] statement, or has failed to remember it, the making of the statement may be proved by another witness.'" (citation omitted)).
Stevens challenges Welham's testimony on the ground that Welham contacted Richardson in his official capacity as a PSO for the sole purpose of securing information pertaining to Stevens's bail application. Section 3153(c)(1) of Title 18 of the United States Code provides that "information
The government responds by directing us to subsection (c)(3) of section 3153, which provides that:
18 U.S.C. § 3153(c)(3) (emphasis added). The government reads this subsection as prohibiting the admission of PSO testimony only "on the issue of guilt." Because the district court admitted Welham's testimony on the issue of Richardson's credibility, not Stevens's guilt, the government maintains that section 3153(c)(1) is not implicated.
The government also refers us to the history of the Pretrial Services Act. This history, the government argues, reveals that Congress rendered pretrial services information confidential in order to promote candor and truthfulness in bail interviews by the defendant:
H.R.Conf.Rep. No. 97-792, 97th Cong., 2d Sess. 8, reprinted in 1982 U.S.Code Cong. & Admin.News 2377, 2394 (emphasis added). Welham, the government points out, testified about a conversation with Richardson, not Stevens. Given that Welham did not betray Stevens's confidence, the government argues that the district court's decision to admit Welham's testimony did not run afoul of Congress's primary concern, i.e., ensuring the free flow of information between the defendant and his PSO.
As usual, the legislative history provides much that supports the other side's position as well. At oral argument, Stevens identified a portion of this statute's history that he contends belies the government's assertion that PSO testimony is admissible for impeachment purposes. The Senate's version of section 3153(c)(3), Stevens notes, provided that pretrial services information "may be used ... for the purpose of impeachment in any subsequent proceeding." S.Rep. No. 97-77, 97th Cong., 2d Sess. 12, reprinted in 1982 U.S.Code Cong. & Admin.News 2377, 2388. This exception was deleted, however, from the final version of the bill. Stevens reasons that this deliberate omission indicates that the House/Senate Conference Committee rejected the selfsame argument advanced by the government here.
Few cases address the reach of the confidentiality requirement of section 3153(c)(1). Two of them frame the issue, but neither resolves it in a manner helpful to this appeal. In United States v. Hammond, 666 F.2d 435 (9th Cir.1982), the Ninth Circuit examined the confidentiality provision of 18 U.S.C. § 3154(1), the predecessor to sections 3153(c)(1) and (3).
Id. at 438.
Similarly uninstructive is the Eighth Circuit's decision in United States v. McLaughlin, 777 F.2d 388 (8th Cir.1985). Although the defendant did not raise the issue, the court was disturbed by the government's use of pretrial services information to impeach the defendant. The court stated:
Id. at 392.
The Eighth Circuit recently revisited this question in United States v. Wilson, 930 F.2d 616 (8th Cir.1991). The defendant in Wilson asserted that the district court had violated section 3153(c)(1) by allowing the government to introduce his pretrial services statements to impeach him on cross-examination. The Eighth Circuit rejected this challenge in language we find significant for present purposes:
Id. At 619 (emphasis added).
Like the Eighth Circuit, we find section 3153(c)(3) to be facially unambiguous. Information disclosed during the course of a pretrial services interview "is not admissible on the issue of guilt." (Emphasis added). There is, however, no clear statutory bar to using pretrial services statements for impeachment purposes. If, as here, "the statutory language is clear and plain, a court must give it effect unless the legislative history is such that a literal reading `will produce a result demonstrably at odds with the intention of [the] drafters,' or in other words, `would thwart the obvious purposes of the ... [statute].'" Smith v. Fidelity Consumer Discount Co., 898 F.2d 907, 910 (3d Cir.1990) (citations omitted). We do not find Stevens's legislative history argument to be sufficiently compelling to convince us to depart from what, in our view, is the statute's plain meaning.
Our interpretation of section 3153(c)(3) is, moreover, consistent with the established distinction between impeachment and substantive guilt evidence, which has been emphasized by the Supreme Court in numerous constitutional contexts. See Michigan v. Harvey, 494 U.S. 344, 110 S.Ct. 1176, 1180, 108 L.Ed.2d 293 (1990) (evidence secured during a police-initiated conversation occurring after the defendant has invoked his sixth amendment rights is inadmissible as substantive evidence in the government's case-in-chief, but is admissible to impeach the defendant's inconsistent trial testimony); United States v. Havens, 446 U.S. 620, 626-28, 100 S.Ct. 1912, 1916-17, 64 L.Ed.2d 559 (1980) (evidence suppressed
This dichotomy is also recognized in other areas of criminal law. Plea agreements, for example, commonly contain a provision stating that proffer information that is disclosed during the course of plea negotiations is inadmissible as substantive evidence of guilt, but is admissible for purposes of impeachment. The standard plea agreement provides:
United States v. Nemetz, No. 87-196-C, slip op. at 2-3, 1987 WL 17543 (D. Mass. Sept. 21, 1987) (quoting a standard plea agreement) (holding that the defendant, by signing the plea agreement, waives the protection of Fed.R.Crim.P. 11(e)(6) for the purposes of impeachment). Similarly, grand jury testimony, which is rendered confidential by Fed.R.Crim.P. 6(e), may be used to contradict the witness's inconsistent trial testimony. See, e.g., United States v. Stockton, 788 F.2d 210, 219-20 (4th Cir.) (government permitted to impeach witness's trial testimony by introducing into evidence her prior inconsistent statements before the grand jury), cert. denied, 479 U.S. 840, 107 S.Ct. 147, 93 L.Ed.2d 89 (1986).
We also think that Stevens exaggerates a bit Congress's confidentiality concerns. For example, section 3153(c)(2)(C) creates an exception to the confidentiality requirement that requires a PSO to disclose information gathered during the pretrial services process to the defendant's probation officer. 18 U.S.C. § 3153(c)(2)(C). This information then is used by the probation officer in compiling the defendant's presentence report and in recommending a sentence to the district court. Thus, the probation officer, by virtue of section 3153(c)(2)(C), has access to the defendant's pretrial services statements on such subjects as employment, prior criminal record, and drug use, all of which can affect significantly the length of the sentence imposed by the district court under the Federal Sentencing Guidelines. This exception, we believe, further underscores that Congress did not impose a blanket prohibition on the use of pretrial services information.
We note, finally, that the question presented by this appeal is much narrower (and easier) than the question answered by the Eighth Circuit in Wilson. Whereas the Eighth Circuit held that the defendant may be impeached by his own inconsistent pretrial services statements, at issue here is the impeachment of a third party, the defendant's father. We are dealing then with a situation that does not impinge directly on Congress's core concern — the sanctity of the defendant/PSO relationship. We do not suggest that it would be immaterial to a defendant that his family and friends may be interviewed by a PSO with a diminished obligation of confidentiality. Rather, we submit that, to the extent that Congress's decision to strike the impeachment exception evinces an intent to expand the breadth of the confidentiality requirement, Congress was concerned primarily about information relayed by the defendant to the PSO.
V. THE DOWNING ISSUE
Stevens's defense at trial was that Smith and McCormack misidentified him as their assailant. The cornerstone of this defense was the expert testimony of Dr. Steven Penrod, a psychologist, regarding the fallibility of eyewitness identification. Prior to trial, Stevens moved in limine for a ruling on the admissibility of Dr. Penrod's testimony in six specific areas: (1) the accuracy of cross-racial identifications; (2) the effect of weapon focus on identifications; (3) the effect of stress on identifications; (4) the suggestiveness of the wanted board; (5) the relation back of subsequent identifications to the initial identification; and (6) the lack of correlation between confidence and accuracy in eyewitness identifications.
Rule 702 of the Federal Rules of Evidence "authorizes the admission of expert testimony so long as it is rendered by a qualified expert and is helpful to the trier of fact." DeLuca v. Merrell Dow Pharmaceuticals, Inc., 911 F.2d 941, 954 (3d Cir.1990).
Id. In addition, "admission depends upon the `fit,' i.e., a specific proffer that the testimony will focus on particular characteristics of the eyewitness identification at issue and discuss how those characteristics call into question the reliability of the identification." United States v. Sebetich, 776 F.2d 412, 419 (3d Cir.1985), cert. denied, 484 U.S. 1017, 108 S.Ct. 725, 98 L.Ed.2d 673 (1988). More specifically, as we held in Downing,
753 F.2d at 1242.
After a preliminary Downing hearing, the district court allowed Dr. Penrod to testify concerning cross-racial identification, weapon focus, and stress; but barred testimony concerning the suggestiveness of the wanted board, the "relation-back" question, and the correlation between confidence and accuracy. The court based its decision on the last Downing criterion — the "fit" between the expert testimony and the facts of the case.
The district court permitted testimony about cross-racial identification, weapon focus and stress, because Dr. Penrod's proffered testimony was sufficiently tied to the facts of this case: the victims were white and the assailant was black; the assailant brandished a gun; and both victims were
A. The Suggestiveness of the Wanted Board
At the in limine hearing, Dr. Penrod testified that the wanted board was a "eureka type" array, meaning that the wanted board, in contrast to the typical photographic array, was a random arrangement of photographs and sketches that was not constructed for the express purpose of allowing these witnesses (Smith and McCormack) to identify this suspect (Stevens). Dr. Penrod further testified that all of the existing scientific research in the area of suggestiveness has been based upon "non-eureka type" arrays — i.e., arrays that were assembled by a psychologist for the express purpose of allowing a subject to identify a particular individual. Based on this distinction, the district court held that testimony on the suggestiveness of the wanted board would be unconnected to the facts of this case. We disagree.
Smith and McCormack were directed by Amos to inspect the wanted board "to see if anyone on [the board] resembled the perpetrator." At trial, Stevens alleged that the wanted board contained several suggestive features that subliminally attracted the victims' attention to him. To bolster this theory, Stevens hoped to have Dr. Penrod testify that, if there are features in an array that draw the witness's attention to a particular person, the witness is more likely to identify that person as the perpetrator. Given these facts, there was, we think, a sufficient "fit" between Dr. Penrod's tendered testimony and the victims' identifications of Stevens from the assertedly suggestive wanted board.
That Dr. Penrod's proposed testimony derived from studies involving "non-eureka type" arrays does not undermine the "fit." The "eureka"/"non-eureka" distinction is, in our view, a red herring: a suggestive feature may induce a witness to misidentify a particular individual regardless of whether the array was constructed for the express purpose of allowing that witness to identify that individual. We instead think that this distinction simply reflects a limitation inherent in this type of scientific research. Dr. Penrod testified that "the only way a psychologist [can] study [the] suggestiveness of an array is to assemble an array[,] so ... it's implicit in our methodology that we can't study eureka type arrays." This strikes us as a matter of common sense. A psychologist cannot analyze suggestive arrays without first arranging an array with some subtly suggestive features. We therefore are satisfied that there is an ample connection between Dr. Penrod's tentative testimony and the facts of this case. Had the district court permitted him to do so, Dr. Penrod would have testified that the wanted board's suggestive attributes could have induced Smith and McCormack to scrutinize Stevens's photographs more carefully and thus could have resulted in a mistaken identification.
We believe, nonetheless, that the district court's exclusion of Dr. Penrod's testimony on this point did not amount to an abuse of discretion. "The touchstone of Rule 702 ... is the helpfulness of the expert testimony, i.e., whether it `will assist the trier of fact to understand the evidence or to determine a fact in issue.'" Downing, 753 F.2d at 1235 (quoting Fed.R.Evid. 702). As we noted supra at 1390, the wanted board was potentially suggestive for several obvious reasons. Stevens was the only person on the display whose picture appeared twice and was the only person whose photograph was in color. Also, most of the other pictures on the wanted board were sketches, not photographs; and those that were photographs were significantly smaller than
In sum, we are persuaded that the probative value of Dr. Penrod's proposed testimony was substantially outweighed by considerations of "undue delay, waste of time, or needless presentation of cumulative evidence," Fed.R.Evid. 403. The district court, although purportedly barring Dr. Penrod's testimony on suggestiveness for lack of "fit," recognized as much: "The fact that there are two pictures ... of the same person and one [is] in color, I think[,] ... is certainly sufficiently apparent [to] the jury ... for the jury to make its own determination."
B. The Relation Back of Subsequent Identifications to the Initial Identification of Stevens From the Wanted Board
Stevens also sought to elicit from Dr. Penrod testimony concerning the pitfalls of multiple identifications. At the preliminary Downing hearing, Dr. Penrod explained that, once a witness makes an identification, he or she will tend to stick with that initial choice at subsequent photographic arrays or lineups, even if it was erroneous. The reason for this phenomenon, Dr. Penrod submits, is that "information acquired at an initial identification [often] influence[s] identifications made later on." That is, witnesses sometimes base subsequent identifications on their vague recollection of a face viewed in a prior array or lineup, not on their memory of the crime itself.
Once again, we do not think that there is a "fit" problem with this aspect of Dr. Penrod's testimony. Stevens was the only individual from the wanted board who also appeared in either the photographic array or the lineup. According to Dr. Penrod, this factor, together with the alleged suggestiveness of the wanted board, could have brought about successive misidentifications. If the victims erroneously identified Stevens from the wanted board, the scientific studies cited by Dr. Penrod suggest that the victims would tend to remain faithful to that choice at later identifications, because they would recognize Stevens's face from the wanted board. There is, in short, a nexus between Dr. Penrod's tendered testimony and the facts of this case.
But, as we noted supra, Downing demands more than just a "fit." Stevens asserts that the introduction of Dr. Penrod's testimony on the "relation-back" issue would have prompted the jury to discount the corroborative value of the victims' identifications of him from the photographic array and lineup. We think, however, that this point, like Dr. Penrod's comments on the suggestiveness of the wanted board, is rather pedestrian. It is, we believe, susceptible of elucidation without specialized scientific knowledge and thus could have been fleshed out adequately by counsel through probing cross-examination and arguments pitched to the common
C. The Lack of Correlation Between Confidence and Accuracy in Eyewitness Identifications
Dr. Penrod also testified at the in limine hearing about scientific studies that seek to measure the relationship between the degree of confidence a witness purports to have in his or her identification and the accuracy of that identification. In these studies, subjects either are exposed to a live staging of some highly unusual event or are shown a videotape of a reenacted crime. The subjects then are requested to make an identification and to rate how confident they are in that identification. According to Dr. Penrod, these studies have revealed "a fairly weak relationship" between confidence and accuracy.
At the conclusion of the in limine hearing, the district court prohibited Dr. Penrod's testimony on these confidence/accuracy studies, finding no "fit" between the proffered testimony and the facts of this case. The court based this conclusion on its belief that witnesses function differently in "real life situations" than in tests performed in a controlled environment. The district court stated that, unlike the studies that Dr. Penrod outlined, the crime in this case transpired over several minutes and placed the assailant in close proximity to the witnesses. The court further noted that Smith and McCormack, both of whom were trained in surveillance techniques, observed the assailant knowing that an offense was occurring and that they later would be called upon to identify him. In contrast, the subjects in the aforedescribed studies were unaware at the time of the staged events that they eventually would be asked to make an identification.
We think that the district court misapprehended Downing's "fit" requirement. Both Smith and McCormack expressed high confidence in their identifications of Stevens as the perpetrator. To rebut the natural assumption that such a strong expression of confidence indicates an unusually reliable identification, Stevens sought to admit Dr. Penrod's testimony that there is a low correlation between confidence and accuracy. We believe that Dr. Penrod's proposed testimony "is sufficiently tied to the facts of the case that it will aid the jury in resolving a factual dispute." Downing, 753 F.2d at 1242.
The factors listed by the district court as destroying the "fit" are characteristic of all studies in the field of eyewitness identifications. Scientists cannot replicate real-life violent crimes; therefore, they are forced to conduct their testing in a simulated, yet somewhat artificial, environment. This limitation, we suspect, also applies to studies concerning cross-racial identification, weapon focus, and stress; yet the district court readily admitted Dr. Penrod's testimony on these subjects. The fact that the subjects in the classroom and videotape studies observed the "assailant" under much different circumstances than did Smith and McCormack obviously constitutes a fertile ground for cross-examination. But we fail to see how these differences undo the "fit" in this case: both Smith and McCormack proclaimed that they were exceedingly confident in their identifications of Stevens, and Dr. Penrod offered to testify that such declarations do not necessarily mean that the victims' identifications were accurate.
Moreover, in contradistinction to the proffered testimony about the suggestiveness of the wanted board and the "relation-back" issue, Dr. Penrod's explication of the confidence/accuracy studies could prove helpful to the jury in assessing the reliability of Smith's and McCormack's identifications. That witnesses ofttimes profess considerable confidence in erroneous identifications is fairly counterintuitive. See id. at 1230 n. 6 ("To the extent that a mistaken witness may retain great confidence in an inaccurate identification, cross-examination can hardly be seen as an effective way to reveal the weakness in a witness' recollection
VI. THE "REVERSE 404(b)" ISSUE
To shore up his theory that Smith and McCormack misidentified him as the perpetrator, Stevens sought to introduce under Fed.R.Evid. 404(b) the testimony of Tyrone Mitchell, the victim of a similar crime at Fort Dix. Three days after Smith and McCormack were assaulted, Mitchell, a black man, was robbed at gunpoint by another black man who, according to Mitchell's description, resembled Smith's and McCormack's attacker. Unlike Smith and McCormack, however, Mitchell stated that Stevens was not his assailant. Stevens reasons that Mitchell's failure to identify him tends to establish that he did not assault Smith and McCormack. The syllogism goes as follows. In view of the many parallels between the two crimes, one person very likely committed both; and because Stevens was exonerated by Mitchell, a black man whose identification (or lack thereof) is arguably more reliable than that of the two white victims, Stevens was not that person. The critical question is, of course, one of degree of similarity.
The similarities between the Mitchell robbery and the Smith and McCormack robbery/sexual assault are significant. Both crimes: (1) took place within a few hundred yards of one another; (2) were armed robberies; (3) involved a handgun; (4) occurred between 9:30 p.m. and 10:30 p.m.; (5) were perpetrated on military personnel; and (6) involved a black assailant who was described similarly by his victims. Indeed, based on these similarities, the United States Army Criminal Investigation Division came to believe, initially, that the same person had committed both crimes. Agent Jackson, the CID officer assigned to investigate the instant crime, reported that:
He further remarked:
Stevens, as a result, quickly became the CID's leading suspect in the Mitchell robbery, as well as the Smith and McCormack robbery/sexual assault. CID Agent Bronisz stated: "Dix CID is certain that STEVENS robbed MITCHELL of his ID." The FBI, cognizant of this fact, thus had Mitchell view Stevens in the same lineup and on the same day that Smith and McCormack viewed him. Mitchell, however, did not identify Stevens as his assailant.
An additional, and even more striking, parallel subsequently developed which made the similarities between the two crimes more difficult to dismiss as mere coincidence. Mitchell was robbed of various items, including his military identification card. This card later was used to cash two stolen checks at the Fort Meade exchange in Maryland. Significantly, McCormack's stolen money order, like Mitchell's identification, also ended up near Fort Meade: it was cashed by someone other than Stevens at the Odenton Pharmacy located across the street from Fort Meade. That the fruits of the Mitchell and McCormack robberies, which occurred within days of one another at Fort Dix, New Jersey, both surfaced near Fort Meade, Maryland, is undoubtedly probative that the same individual committed both offenses.
The analytical basis for Stevens's proffer of Mitchell's testimony was a rarely-used
2 Wigmore, Wigmore on Evidence § 304, at 252 (J. Chadbourn rev. ed. 1979) (emphases in original).
In Commonwealth v. Murphy, 282 Mass. 593, 185 N.E. 486 (1933), there was no evidence connecting the defendant to the offenses except for the victims' identifications. To offset these identifications, the defendant offered evidence of three similar crimes, one of which occurred while he was in custody. He also sought to introduce evidence that two of the victims of the other crimes had stated that he looked somewhat like (but was not) the assailant, and that the third victim had admitted to misidentifying him as the assailant. On appeal, the Supreme Judicial Court of Massachusetts reversed the trial court's exclusion of this evidence, emphasizing that the jury was entitled to consider it in evaluating the victims' identifications: "No one, we think, will deny that if the evidence offered is the truth it well might shake confidence in the identifications upon which alone this conviction rests." Id. at 596-597, 185 N.E. at 487.
In State v. Bock, 229 Minn. 449, 39 N.W.2d 887 (1949), the defendant, who was convicted of uttering a forged check, appealed from the trial court's refusal to admit evidence that checks identical to that which he allegedly passed were negotiated on the same day and in a like manner by someone else. The Supreme Court of Minnesota reversed, holding that:
Id. at 458, 39 N.W.2d at 892.
Likewise, the defendant in Holt v. United States, 342 F.2d 163 (5th Cir.1965), proffered "reverse 404(b)" evidence to bolster his defense of misidentification. More specifically, he attempted to introduce evidence that he had been misidentified twice before as the perpetrator of virtually identical offenses. Id. at 164-65. The district court excluded this evidence on relevancy grounds, but the Fifth Circuit reversed, noting that the proffered evidence tended to support defendant's assertions that another man who resembled him had committed the identical crime, and that he had been misidentified. Id. at 166. "In these circumstances," the court stated, "the evidence was relevant, and of substantial probative value. The jury should have been
Two New Jersey cases are also helpful. In State v. Garfole, 76 N.J. 445, 388 A.2d 587 (1978), the defendant, who was charged with sexually molesting two children, sought to introduce evidence that he had been indicted on charges arising out of five similar episodes; that the charges arising out of four of those episodes had been dismissed; and that he had an alibi for all but two of the episodes. Id. at 448, 388 A.2d at 589. The trial court rejected this proffer, stating that it was irrelevant to the specific charge at issue.
The New Jersey Supreme Court disagreed, noting that the evidence clearly had some relevance to the case. Id. at 451; 388 A.2d at 590. The Court then turned to the question whether the other crimes were sufficiently relevant (i.e., similar) to outweigh countervailing considerations such as undue consumption of time and confusion of the issues. The Court observed, in this regard, that the trial court had imposed on the defendant the same standard of similarity that would have applied if the State had offered the evidence, in essence requiring that the other crimes be so distinctive as to constitute "signature" crimes. This, the Court held, was error:
Id. at 452-53, 388 A.2d at 591 (emphasis added) (footnote and citation omitted).
The Superior Court of New Jersey explored this same issue in State v. Williams, 214 N.J.Super. 12, 518 A.2d 234 (1986), wherein the defendant appealed his conviction for attempted murder on the ground that the trial court improperly had excluded his "reverse 404(b)" proffer. The victim, who ultimately identified the defendant (her ex-boyfriend) as the perpetrator, was grabbed from behind, told not to scream, and (after a brief struggle) stabbed nine times. She also stated that her attacker had "pull[ed] on her clothes," prompting the investigating officer to report the attack as a possible rape attempt. Id. at 16, 518 A.2d at 236. At trial, the defendant sought to adduce evidence of two similar crimes committed at the same time of day and at the same location by an individual who lived nearby. In both of the other crimes, the perpetrator grabbed his victim from behind, threatened her with a knife, and raped her; one of the victims was stabbed eight times following the rape.
In refusing to admit this evidence, the trial court stated that it would unduly confuse the jury, and that the other crimes were not sufficiently similar to outweigh the potential confusion. On appeal, the Superior Court reversed, concluding that the parallels between the three crimes satisfied the diminished standard of similarity articulated in Garfole:
Id. at 21, 518 A.2d at 239. Had the trial court admitted this evidence, the Superior Court noted, it would have shown that someone else had committed similar crimes in the same area and at about the same time, which, in turn, would have undercut the persuasiveness of the victim's identification.
The precise rationale for the district court's exclusion of Stevens's "reverse 404(b)" evidence is difficult to glean from the record. It appears that the court barred the evidence on relevancy grounds, holding that the Mitchell robbery was a "separate offense ... not directly related or tied ... to this defendant." There is, however, language in the district court's ruling suggesting that the court thought the existence of a prior misidentification to be a prerequisite to the admission of "reverse 404(b)" evidence:
The government defends the district court's exclusion by asserting that a defendant may avail himself of Rule 404(b) in only three "constrained" circumstances, none of which apply to the instant case. First, the government submits, the defense may introduce evidence that the government induced others to commit crimes in order to show that the defendant was induced to commit the charged offense. See United States v. Rodriguez, 917 F.2d 1286, 1289-91 (11th Cir.1990). This usage obviously has no bearing on Stevens's case. Second, the government argues, Rule 404(b) permits a defendant to introduce evidence that another person committed a similar crime, and that he (the defendant) was misidentified as the perpetrator of that similar crime. Holt and Murphy, the government contends, are of examples of this line of cases, which, the government asserts, is unavailable to Stevens because Mitchell never misidentified him as the assailant. Finally, the government states, defendants may invoke Rule 404(b) to admit evidence of other crimes "where those other crimes were sufficiently numerous and similar in their execution as to form a clear pattern." Appellee's Br. at 28. This is the so-called "signature" crime exception. The government maintains that this use, which supposedly is typified by Bock and Garfole, is also unavailable to Stevens, because Stevens has proffered evidence of only one other crime, and because that crime was dissimilar in that there was no sexual assault.
Based on our survey of the case law, we believe that the district court imposed too stringent a standard of similarity on Stevens, and that the government has unnecessarily compartmentalized the permissible uses of "reverse 404(b)" evidence. "It is well established that a defendant may use similar `other crimes' evidence defensively if in reason it tends, alone or with other evidence, to negate his guilt of the crime charged against him." Williams, 214 N.J.Super. at 20, 518 A.2d at 238. In our view, the most persuasive treatment of "reverse 404(b)" evidence is found in Garfole, see supra at 1402-03, wherein the New Jersey Supreme Court observed that a lower standard of similarity should govern "reverse 404(b)" evidence because prejudice to the defendant is not a factor. We agree with the reasoning of Garfole and with its holding that the admissibility of "reverse 404(b)" evidence depends on a straightforward balancing of the evidence's
Given the flexible contours of the above equation, we reject the government's attempt to impose hard and fast preconditions on the admission of "reverse 404(b)" evidence. More specifically, the defendant, in order to introduce other crimes evidence, need not show that there has been more than one similar crime, that he has been misidentified as the assailant in a similar crime, or that the other crime was sufficiently similar to be called a "signature" crime. These criteria, although relevant to measuring the probative value of the defendant's proffer, should not be erected as absolute barriers to its admission. Rather, a defendant must demonstrate that the "reverse 404(b)" evidence has a tendency to negate his guilt, and that it passes the Rule 403 balancing test.
Here, Mitchell's testimony, in conjunction with Dr. Penrod's comments about the suspect nature of cross-racial identifications, easily satisfies the Rule 401 threshold. The government's case against Stevens rested entirely on two white victims' identifications of him as the assailant. To rebut the natural force of these identifications, Stevens attempted to mount a two-pronged defense. He began by offering Dr. Penrod's expert testimony that cross-racial identifications are notoriously unreliable. Then, to ground more concretely Dr. Penrod's opinion in the facts of this case, Stevens sought to introduce evidence that Mitchell, a black victim of a similar crime, had failed to identify him as the assailant.
The similarities between the Mitchell robbery and the Smith and McCormack robbery/sexual assault are demonstrated amply, in our view, by the facts that the CID was convinced that the same person had committed both crimes, and that the fruits of both robberies (i.e., McCormack's money order and Mitchell's military identification) ended up in Fort Meade.
Under the circumstances of this case, the potential for waste of time or for misleading the jury, if the "reverse 404(b)" evidence were admitted, was minimal. Although the government objected to Mitchell's testimony, it agreed to stipulate to all of the essential facts regarding the Mitchell robbery in the event that his testimony were admitted. Stevens maintains that, as a result of this stipulation, Mitchell would have been a "fifteen-minute" witness, who simply would have testified regarding the details of the robbery and his viewing of the lineup. In addition, Stevens probably would have asked CID Agent Jackson and FBI Agent Maxwell some questions on cross-examination about the similarities between the Mitchell robbery and the Smith and McCormack robbery/sexual assault
Thus, even if Stevens's estimate of "fifteen minutes" was overly optimistic, there was no appreciable risk that Stevens's presentation of "reverse 404(b)" evidence would have degenerated into a mini-trial about whether or not Stevens had robbed Mitchell. Nor was there any real danger that Stevens's portrayal of the parallels between the two crimes would have obstructed the orderly progress of the trial or would have distracted the jurors' attention from the real issues in this case. The government, in fact, does not dispute seriously these points. The thrust of its challenge to Stevens's "reverse 404(b)" evidence was that it did not pass muster under Rule 401, not Rule 403. We thus conclude that the probative value of the Mitchell robbery was not substantially outweighed by the prospect of undue delay or confusion of the issues.
Our resolution of this issue is informed by our general belief that a criminal defendant should be able to advance any evidence that, first, rationally tends to disprove his guilt, and second, passes the Rule 403 balancing test. To garner an acquittal, the defendant need only plant in the jury's mind a reasonable doubt. Had Stevens been allowed to adduce at trial evidence of the similarities between the Mitchell robbery and the Smith and McCormack robbery/sexual assault — including the Fort Meade connection — the jury might have determined that it was possible that another person had committed both crimes, thereby giving rise to a reasonable doubt. We do not know, of course, how a jury would weigh this evidence, but we do think that, at the very least, Stevens was entitled to have the jury consider the evidence and draw its own conclusions.
VII. HARMLESS ERROR
The government contends that we should affirm Stevens's conviction because the district court's errors were harmless under Fed.R.Crim.P. 52(a). If the evidence against Stevens were overwhelming, we could adopt such an approach, but that is not the case here. As mentioned above, Stevens's first trial ended in a mistrial because the jury was unable to reach a verdict. Given the apparent closeness of the evidence, we think it possible that the admission of evidence about the Mitchell robbery might have swayed the jury toward acquittal. We note, in addition, that both of the district court's errors involved evidence that detracted from the reliability of the victims' identifications — the sole predicate for Stevens's convictions. Had the jurors learned that confidence is a poor indicator of the accuracy of an identification, and that Mitchell (a black man who perhaps was robbed by the same assailant) failed to identify Stevens as the perpetrator, the outcome of their deliberations could have been different. Simply put, we
We conclude that the district court erred by barring Dr. Penrod's testimony about the lack of a correlation between confidence and accuracy in eyewitness identifications and by excluding evidence about the similarities between the Mitchell robbery and the Smith and McCormack robbery/sexual assault. Because these errors were not harmless, the judgment of sentence will be reversed and the case remanded for a new trial.
In our view, the district court's decision not to redact the wanted board reflects a sound exercise of its discretion. After considering both parties' arguments, the district court concluded that the probative value of the jury's seeing the board as a whole outweighed any potential for prejudice. The court attached considerable importance to enabling the jury to see the wanted board as McCormack and Smith did when they identified Stevens. Allowing the jury to inspect the board in its unaltered condition, the court thought, would permit the jurors to determine what, if anything, drew the victims' attention to Stevens. The district court emphasized that Stevens's counsel could explain to the jury that Stevens was not wanted at the time of the identification. We decline to disturb this balancing.
Stevens's first argument — that McCormack and Smith "were reflecting back to the impermissibly suggestive and unreliable `Wanted Board,' and not the crime, when they identified appellant at the later identifications," Appellant's Reply Br. at 10 — is based on a flawed premise. An inquiry into "taint" is appropriate only if a prior identification procedure was so impermissibly suggestive that the prior identification lacks reliability. Because we have determined that the victims' identifications of Stevens from the wanted board bore sufficient indicia of reliability, Stevens's claim that the later identifications were "tainted" fails a fortiori.
Stevens's second argument — that the photographic array and lineup were unnecessarily suggestive because he was the only individual on the wanted board who also appeared in either the array or lineup — suffers from two infirmities. To begin with, Stevens raised this argument for the first time on appeal. At the Wade hearing, Stevens's counsel made clear that "the identification I'm concerned with is the first one. I would argue [only] that the second and third ones were tainted by the first identification." Alternatively, we think that the victims' identifications of Stevens from the photographic array and lineup were admissible, because — as we explained supra at 1390-92 — even if these later identification procedures were "unnecessarily suggestive, the Neil v. Biggers factors indicate that the victims' identifications were nonetheless sufficiently reliable.
Here, two military police officers, trained in observation, were subjected to a prolonged attack in a well-illuminated bus shelter by a man who made no effort to mask his face and who remained in close proximity to them. Shortly afterward, they both identified Stevens as their assailant from the wanted board and from a photographic array. And a few days later, the victims identified Stevens from a line-up. At trial, the jury obviously chose to credit the victims' identifications of Stevens. Such "[c]redibility determinations are for the jury." United States v. Jannotti, 673 F.2d 578, 598 (3d Cir.) (en banc), cert. denied, 457 U.S. 1106, 102 S.Ct. 2906, 73 L.Ed.2d 1315 (1982). After combing the record, we conclude that there was sufficient evidence for the jury to have found Stevens guilty beyond a reasonable doubt, and that therefore the district court correctly denied Stevens's Rule 29(c) motion. Accordingly, we will not remand the case with instructions to enter an order of acquittal.