(dissenting).
In my view, the erroneous admission in this case of evidence derived from the EXIS computer database violated the defendant’s Sixth Amendment right to confront witnesses against him. Contrary to my brethren, I do not believe that this error was harmless beyond a reasonable doubt. I therefore dissent.
I.
Trenkler admitted to building a device that exploded in Quincy in 1986. The government’s central strategy at trial27 was to prove that the Quincy device was so similar to the Roslindale bomb that they had to have been built by the same person. Stephen Scheid, an Intelligence Research Specialist with the Bureau of Alcohol, Tobacco and Firearms (“ATF”), testified that he conducted a computer query on the ATF’s EXIS database28 to identify bomb incidents which shared certain characteristics with the Ros-lindale incident. Based on this analysis, Scheid told the jury that, out of the 14,252 bombings and attempted bombings reported in EXIS, only the Roslindale and the Quincy incidents shared all the queried characteristics.
For a jury reviewing otherwise weak circumstantial evidence of defendant’s guilt (see infra), this is powerful stuff—tangible, “scientific” evidence which seems to conclusively establish that the same person who made the Quincy device in 1986 made the Roslindale bomb in 1991. Unfortunately, as the majority concedes, the reports from which the EXIS information is derived are utterly un*63reliable, thus rendering its conclusion equally unreliable, and, as will be shown, completely misleading. For three related reasons, I disagree with the majority’s conclusion that admission of the EXIS-derived evidence was “harmless beyond a reasonable doubt.” First, the EXIS-derived evidence plainly influenced the district court’s decision to allow the government’s motion to admit evidence of the Quincy incident, under Fed.R.Evid. 404(b), to show that the same person must have built the Roslindale bomb. Second, the EXIS-derived evidence was very powerful and very misleading. Third, the other evidence against Trenkler was not “overwhelming,” as is required under our precedent.
II.
The majority assumes, without deciding, that Trenkler’s Sixth Amendment right to confront witnesses against him was violated by introduction of the EXIS-derived evidence. Supra n. 22. As the majority recognizes, constitutional cases are governed by a stringent harmless error analysis — a conviction cannot stand unless the effect of the evidence is “harmless beyond a reasonable doubt.” Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 828, 17 L.Ed.2d 705 (1967) (emphasis added); United States v. de Jesús-Ríos, 990 F.2d 672, 678 (1st Cir.1993).29 To comprehend why admission of the EXIS-derived evidence was not harmless beyond a reasonable doubt, one must understand the nature and extent of the constitutional violation. Because the majority barely acknowledges, much less discusses, the constitutional right at stake in this case, its result appears both analytically sound and benign. It is neither. I will therefore begin by explaining why, and to what extent, Trenkler’s Sixth Amendment right to confront witnesses against him was violated. I will then endeav- or to show why this error cannot be considered harmless.
III.
The Confrontation Clause of the Sixth Amendment provides that, “[i]n all criminal prosecutions, the accused shall enjoy the right ... to be confronted with the witnesses against him.” The Supreme Court has explained that “[t]he central concern of the Confrontation Clause is to ensure the reliability of the evidence against a criminal defendant by subjecting it to rigorous testing in the context of an adversary proceeding before the trier of fact.” Maryland v. Craig, 497 U.S. 836, 845, 110 S.Ct. 3157, 3163, 111 L.Ed.2d 666 (1990); United States v. Zannino, 895 F.2d 1, 5 (1st Cir.1990) (“the mission of the Confrontation Clause is to advance a practical concern for the accuracy of the truth-determining process in criminal trials by assuring that the trier of fact has a satisfactory basis for evaluating the truth of the prior statement”) (quoting Dutton v. Evans, 400 U.S. 74, 89, 91 S.Ct. 210, 219-220, 27 L.Ed.2d 213 (1970) (plurality opinion)).
Hearsay evidence from an unavailable de-clarant30 may only be admitted against a defendant in a criminal case if the government can demonstrate that the proffered evidence “bears adequate indicia of reliability.” Ohio v. Roberts, 448 U.S. 56, 66, 100 S.Ct. 2531, 2539, 65 L.Ed.2d 597 (1980) (internal quotation marks omitted).31 The govern*64ment may satisfy this burden by establishing either that the evidence “falls within a firmly rooted hearsay exception” or by showing that the evidence possesses “particularized guarantees of trustworthiness.” Id.; accord Idaho v. Wright, 497 U.S. 805, 816-17, 110 S.Ct. 3139, 3147, 111 L.Ed.2d 638 (1990) (collecting cases); Manocchio, 919 F.2d at 773. The majority properly holds that the EXIS-de-rived evidence satisfies neither of these criteria, but neglects to fully explain why.
The critical inquiry for determining “particularized guarantees of trustworthiness” is whether “the test of cross-examination would be of marginal utility.” Wright, 497 U.S. at 820, 110 S.Ct. at 3149-50.32 The government in this ease, through Scheid, was permitted to introduce the statement that, out of 14,252 bombing and attempted bombing incidents in the EXIS database, only the Roslindale and Quincy incidents share the queried characteristics. The individuals who reported those bomb incidents were witnesses against Trenkler, each of them testifying, in effect: “This bomb incident had the following characteristics .... ” Despite the importance of their “testimony,” neither Trenkler nor the jury ever saw any of these witnesses. Trenkler’s attorney was unable to cross-examine these witnesses with respect to their credibility and reliability. Because they were not subject to cross-examination, neither we nor the jury will ever know, for example, the answers to the following questions. Were the authors of these reports bomb experts? Were they even police officers? Did they follow certain procedures in compiling evidence? In filing their reports? What criteria did they use for determining that the device in question was a quote “bomb”? Did they even have first hand knowledge of the devices, or was the information provided to them second-hand from lay witnesses? Scheid did not know the answers to these questions, nor did he have first hand knowledge of the incidents themselves, supra pp. 58-59, thus making it impossible for Trenkler’s attorney to effectively cross-examine him. Moreover, Scheid admitted that the bomb reports need not be signed, and that nothing required the author of a submitted report to have personal knowledge of its contents.33
The majority also alludes to a potentially more pernicious problem concerning the EXIS-derived evidence. The majority notes that the database entry for the Roslindale incident lists approximately twenty-two characteristics describing that incident, but Scheid, inexplicably, chose only to query ten of those characteristics.34 Supra n. 21. The majority notes that there is nothing to suggest that these ten characteristics are more important to a bomb-signature analysis than any of the other characteristics not chosen. Scheid offers no reason why he chose to query only certain generic characteristics instead of the more specific characteristics of the Roslindale bomb, which would be more evincing of a “signature.” For example, the Quincy device would not have been a match if Scheid had queried any of the following characteristics of the Roslindale bombing: Futa-ba antenna, Rockstar detonator, use of dynamite, nails, glue, 6-volt battery, slide switch, paint, magazine page, or black electrical tape. The majority leaves the implication unspo*65ken. I will not be so discreet. The obvious implication is that Scheid chose the particular characteristics in an attempt to find a match with the Quincy device. This implication is enforced by the fact that, according to Scheid’s own testimony, the Quincy incident was not entered into the database until after the Roslindale incident. That is, government agents brought the Quincy bombing to Scheid’s attention when they asked him to investigate the Roslindale bombing.
The majority thinks these concerns go more to the weight of the evidence than to its admissibility; to the contrary, they go directly to the question of whether the evidence has particularized guarantees of trustworthiness under the Confrontation Clause. They demonstrate that it does not. Because the reports upon which the EXIS evidence is based are inherently and utterly unreliable, the EXIS evidence itself is inherently and utterly unreliable, and Trenkler’s Sixth Amendment right to confront the witnesses against him was violated. See Wright, 497 U.S. at 805, 110 S.Ct. at 3139. The question then becomes whether this error was harmless beyond a reasonable doubt.35
IV.
Under the harmless beyond a reasonable doubt standard, we must vacate the conviction if there is “some reasonable possibility that error of constitutional dimension influenced the jury in reaching [its] verdict.” United States v. Maraj, 947 F.2d 520, 526 n. 8 (1st Cir.1991) (emphasis added) (quoting United States v. Argentine, 814 F.2d 783, 789 (1st Cir.1987)). See also United States v. Flores, 968 F.2d 1366, 1372 (1st Cir.1992). Under this standard, we will only find harmless error when the untainted evidence, standing alone, provides “overwhelming evidence” of the defendant’s guilt. Clark v. Moran, 942 F.2d 24, 27 (1st Cir.1991). In conducting this inquiry, we “must consider the evidence as a whole, weighing the effect of the tainted evidence against the effect of that evidence which was properly admitted.” Id. (citing Lacy v. Gardino, 791 F.2d 980, 986 (1st Cir.), cert. denied, 479 U.S. 888, 107 S.Ct. 284, 93 L.Ed.2d 259 (1986)). Thus, the relative strength of the tainted evidence— i.e., its potential effect on the jury — is a highly significant consideration.
As I see it, there are three related reasons why admission of the EXIS evidence cannot be considered harmless beyond a reasonable doubt. First, it is clear to me that the district court relied on the improper EXIS evidence in its decision to allow the government to present evidence of the Quincy incident to the jury to prove identity under Rule 404(b).
At the hearing on its motion in limine to admit evidence of the Quincy incident under Fed.R.Evid. 404(b), the government presented the testimony of Scheid, regarding the EXIS computer analysis, and the testimony of the government’s bomb expert, Waskom, who testified that, in his opinion, the Quincy and Roslindale devices were so similar that they must have been built by the same person. In turn, Trenkler presented expert testimony that the devices were too different for anyone to be able to determine if they were built by the same person. After hearing this *66evidence, the district court concluded that “the similarities [between the two incidents] are sufficient to admit the evidence under the rules established ... by the First Circuit.”
The majority states that, based upon its review of the record, it is convinced that the EXIS-based evidence “was not a critical factor in the district court’s decision to admit the Quincy bomb evidence for purposes of identity. The EXIS-derived evidence was merely cumulative, corroborating the testimony of the government’s explosives expert.” Swpra pp. 60-61. Yet the record demonstrates that the district court judge thought otherwise when she decided to admit evidence of the 1986 Quincy incident. In her oral opinion on the government’s motion, the district court judge began by summarizing the testimony of Waskom, and then stated: “Adding to this evidence, the statistical evidence from the EXIS system, I am persuaded that the two devices are sufficiently similar to prove that the same person built them, and thus relevant to the issues in this case.” (emphasis added). The district court judge did not say that the EXIS evidence “corroborated” Waskom’s testimony. She stated that, when she adds the EXIS evidence to Waskom’s testimony, she becomes convinced that the two devices are sufficiently similar. It is plain that the district court judge relied on the EXIS evidence to form the critical final link between the two devices. Indeed, in arguing its motion, the government chose to first present the EXIS evidence and then to present the Waskom testimony, suggesting that it intended the latter to corroborate the former. The district court’s erroneous determination that the EXIS evidence was admissible led not only to the jury hearing that evidence, but also to the jury hearing Waskom’s testimony with respect to the two incidents. I cannot agree, therefore, that admission of this evidence was harmless beyond a reasonable doubt.
The second reason that admission of the EXIS evidence cannot be considered harmless is that this type of “scientific” evidence is too misleading, too powerful, and has too great a potential impact on lay jurors, to be disregarded as harmless.
The EXIS-derived evidence was, in the best case scenario, unintentionally misleading, and, in the worst case scenario, deliberately skewed. Scheid testified that, in entering information about the Quincy incident into the EXIS database, he relied solely on a laboratory report prepared in 1986 by investigators from the Massachusetts Department of Public Safety. This report does not state that the Quincy device was attached to the underside of the Capeway truck. Rather, it refers only to an “[e]xplosion on truck.” Somebody must have given Scheid further information about the Quincy explosion because he entered “under vehicle” as a characteristic of the Quincy incident. The majority acknowledges these facts but, inexplicably, makes no comment. See supra at 50 n. 8. These facts are important for three reasons. First, they illustrate the fallibility of the underlying reports. How many of the other 14,232 reports had similar defects? Second, they illustrate how easily one wrong or incomplete entry can affect a query result. If Scheid had actually followed the report, the Quincy incident would not have matched the Roslindale bombing because Scheid’s query entry was for a bomb “under vehicle.”36 Finally, these facts indicate that the EXIS test was skewed (whether intentionally or unintentionally) to find a match between the Quincy and Roslindale incidents.37
The EXIS-derived evidence is also misleading because it focuses the jury’s attention on the trees instead of the forest. By focusing on similar minor aspects between the two devices — e.g., duet tape, magnets and soldering — the majority completely brushes aside *67the fact that the central and most important ingredient in the two devices is fundamentally different. The central ingredient in a bomb, one would think, is the explosive content (in much the same way that the central ingredient in a high-performance car is the engine). The Roslindale bomb used two to three sticks of dynamite — a very powerful explosive. The Quincy device used an M-21 Hoffman artillery simulator, which is a device used by the military to simulate, in a safe fashion, the flash and noise of artillery. The simulator is, in effect, a firecracker-like device; it has no where near the strength of dynamite. In stark contrast to dynamite, a simulator is not designed to cause physical or property damage. Indeed, while the Roslin-dale device created an explosion large enough to kill, the Quincy device caused no visible damage to the truck it was placed under. Equating the two devices is like equating a BB gun with a high caliber rifle.38
The misleading nature of the EXIS-de-rived statement is compounded by the nature of its source, and the way in which it was presented to the jury. Not only is it rank hearsay evidence, it is hearsay evidence wrapped in a shroud of “scientific” authenticity. This is not a paid government expert testifying that, in his opinion, the two devices were built by the same person; this is a computer declaring that the two devices were built by the same person. Computers deal in facts,39 not opinions. Computers are not paid by one side to testify. Computers do not have prejudices. And computers are not subject to cross-examination. Moreover, the chart of the EXIS queries performed by Scheid, and the printouts of the results of those queries, were introduced into evidence and presented as exhibits to the jury. Consequently, the jury had this misleading, physical evidence with them in the jury room during deliberations.40 Does it not stand to reason that the lay juror will accord greater weight to a computer’s written findings than to the testimony of a government expert witness? The common-sense answer is, of course.41
The majority decision in this case not only defies common sense, it is also contrary to our precedent. In de Jesús-Ríos, 990 F.2d 672, we held that the defendant’s due process rights were violated when the district court admitted certain identification testimony by a witness. Significantly, we concluded that the error was not harmless beyond a reasonable *68doubt, even though another witness testified at trial that he also had identified the defendant. Rather than concluding, as the majority does here, that the one erroneously admitted identification was “merely cumulative” of the other, the court reasoned:
[Tjhere is no way for us to discern the role that Rivera’s identification played in the jury’s deliberations. We are concerned that the jury may have been persuaded to convict by the very fact that there were two witnesses who identified [the defendant]. It is also possible that the jury relied solely upon the testimony of Rivera in reaching its conclusion. Thus, we find reasonable doubt exists as to whether the jury would have convicted [the defendant] based solely upon Mejias’s identification testimony.
Id. at 678 (emphasis added). Is it not equally plausible that the jury in this case “may have been persuaded to convict” by the very fact that two “witnesses” — Waskom and the EXIS-derived evidence — identified the builder of the Quincy device as the builder of the Roslindale bomb? Is it not also equally plausible that the jury relied solely upon the EXIS-derived evidence in reaching its conclusion? Because the EXIS-derived statement came from a computer, and was presented in tangible, exhibit form, it is more powerful and seemingly credible evidence to a lay jury than the testimony of a human being. The jury may well have relied on the EXIS-derived evidence to break the tie between the competing experts. This is particularly so since, as the trial judge noted, defendant’s expert witness had “considerably more experience in making ... signature comparisons.” Since the EXIS-derived evidence could well have been “the clincher” for the jury, it cannot be considered harmless beyond a reasonable doubt. See Coppola v. Powell, 878 F.2d 1562 (1st Cir.1989).
The third reason that admission of the EXIS evidence is not harmless beyond a reasonable doubt is that the other evidence against Trenkler was not “overwhelming.” See Clark, 942 F.2d at 27. The majority points to a conglomeration of other testimony in support of its conclusion that there was “substantial evidence” of Trenkler’s guilt, independent of the Quincy incident. The test, of course, is not whether there is “substantial evidence” of Trenkler’s guilt but whether there is “overwhelming evidence” of Trenk-ler’s guilt. The two standards are qualitatively and quantitatively different. In any case, I will begin by addressing Trenkler’s “statements” to government agents.
ATF Agent D’Ambrosio testified that he asked Trenkler to draw a sketch of the Quincy device, which Trenkler did. D’Ambrosio then told Trenkler that the Roslindale bomb also used remote control, but that, rather than a firecracker type device, it used dynamite. D’Ambrosio asked Trenkler how, in light of these facts, the wiring diagram he had just drawn for the Quincy deviee would have been different for the Roslindale bomb. D’Ambrosio testified that Trenkler then drew a diagram which showed two blasting caps inserted into two sticks of dynamite. The majority considers this significant evidence of Trenkler’s guilt because the fact that the Roslindale bomb used blasting caps had not been publicly disclosed. The majority fails to note, however, that D’Ambrosio actually testified that at least two blasting caps were used in the Roslindale bombing. Thus, Trenkler’s drawing of only two blasting caps was not an exact match. Moreover, the jury heard evidence that Trenkler had extensive knowledge of both electronics and explosives, so it is not necessarily significant that Trenk-ler was able to reconstruct an aspect of the Roslindale bomb, particularly considering the information concerning the bomb provided to Trenkler by D’Ambrosio. Trenkler merely identified that blasting caps were a likely way in which a bomb of this size and power would be constructed. In the absence of any testimony that the use of blasting caps is unusual or unique (a proposition which is highly unlikely), the jury could only speculate as to the significance of the drawing.
The majority also finds significance in ATF Agent Leahy’s testimony that Trenkler said to him: “If we did it, then only we know about it ... how will you ever find out ... if neither one of us talk[]?” The majority paints this statement in a confessional light. This testimony may or may not have been of some circumstantial relevance to the jury *69(although standing alone, of course, it would not be sufficient to sustain a conviction). But, upon review, when the court is looking for “overwhelming evidence of guilt,” one would think the court would not have to resort to this sort of an ambiguous, taunting statement.42 Similarly, the court notes that there was evidence that Trenkler and Shay knew each other, and that Trenkler had knowledge of both electronics and explosives. While the jury might consider this type of circumstantial evidence relevant, it can hardly be said that it does much in the way of providing “overwhelming evidence” of defendant’s guilt. Cf. United States v. Innamorati, 996 F.2d 456, 476 (1st Cir.1993) (holding that the erroneous admission of inculpatory grand jury testimony was harmless beyond a reasonable doubt when seven people testified at trial that defendant was engaged in marijuana and cocaine dealing, and drugs and money were found in defendant’s constructive possession).
The majority relies most heavily on the testimony of David Lindholm, who testified that Trenkler confessed to budding the Ros-lindale bomb. But Lindholm had some serious credibility problems which make his testimony “shaky,” to say the least. Lindholm testified that he met Trenkler while Lin-dholm was serving a 97-month sentence for conspiracy to distribute marijuana and tax evasion. He further testified that he was in the marijuana business from approximately 1969 through 1988, and that he did not pay any income taxes during that time. Lin-dholm also testified that, in order to secure bank loans to purchase property during that period, he showed several banks false income tax returns. On the basis of Lindholm’s shady past alone, the jury might have completely disregarded his testimony.
But Lindholm also had some less obvious credibility problems. The circumstances of his meeting Trenkler strike me as a little too coincidental. On December 17, 1992, after a year and a half incarceration in Texas, Lin-dholm is brought back to Boston concerning certain unspecified charges related to his conviction. He is then placed in the orientation unit at the Plymouth House of Correction where he meets Alfred Trenkler, who is being held in connection with the Roslindale bombing. The two subsequently discover that they have an extraordinary amount in common. First, they are both from the town of Milton, Massachusetts. Second, Trenkler attended Thayer Academy and Milton Academy, and Lindholm’s father also attended Thayer Academy and Milton Academy. Third, they both lived for a time — overlapping by one year — on White Lawn Avenue in Milton. Based on these commonalities, and Lindholm’s generosity in sharing his knowledge of the criminal justice system with Trenkler, they form a friendship. Trenkler then, allegedly, confesses to Lin-dholm that he built the bomb.
In my view, a reasonable juror might question whether Lindholm was placed in the orientation unit by the government for the purpose of obtaining a confession from Trenkler. If so, that juror would likely wonder what Lindholm got in return. Not surprisingly, Lindholm testified that he had no agreements with the government and that he did not receive any promises or inducements for his testimony.43 He did testify on cross-examination, however, that he knew, when he provided the information about Trenkler to the government, that the only way his 97-month sentence could be reduced was if he supplied new information to the government.44
*70We do not know how much weight the jury gave Lindhobn’s testimony, but we do know that, at least on paper — for we did not observe his demeanor at trial — Lindholm had some significant credibility problems. Consequently, I cannot conclude beyond a reasonable doubt that the jury would have believed his testimony; particularly in a case such as this where there is absolutely no physical evidence tying Trenkler to the bombing. Cf. Coppola, 878 F.2d at 1571 (discounting inculpatory testimony of three jail inmates because it “raises serious questions of credibility” and noting the absence of any conclusive physical evidence tying the defendant to the crime). The only evidence coming near that level of reliability was the improperly admitted EXIS evidence.
Absent the EXIS-derived evidence, the government’s case against Trenkler consists of a smorgasbord of inconclusive circumstantial evidence and an inherently unreliable alleged jailhouse confession. Faced with this sort of evidence, a reasonable jury would probably look for some sort of tangible evidence upon which to hang its hat. The EXIS-derived evidence was just that. Because it was the only ostensibly conclusive evidence tying Trenkler to the crime, it may have been the clincher for the jury. See Coppola, 878 F.2d at 1571. It was therefore not harmless beyond a reasonable doubt.
V.
A horrible crime was committed in which one police officer was killed and another seriously injured. Society rightfully demands that the guilty be apprehended, tried, and punished. But the distinguishing feature of our legal system is that even those charged with grotesque crimes are guaranteed certain constitutional rights intended to ensure that they receive a fair trial. Unfortunately, and with all due respect to my brethren, I believe the defendant’s right to a fair trial was violated when the government was permitted to introduce the highly prejudicial evidence derived from the EXIS computer database. Because this error so severely violated defendant’s Sixth Amendment right to confront the witnesses against him, and because the remainder of the evidence against him was not “overwhelming,” I dissent.
. In support of its motion in limine to admit evidence of the 1986 incident, the government described this evidence as "the centerpiece of the Government's case in chief."
. For a description of the EXIS database, see supra pp. 49-50.
. Under the standard for analyzing harmless error in a nonconstitutional case, the court will uphold a conviction provided it can be said "that the judgement was not substantially swayed by the error.” United States v. Flores, 968 F.2d 1366, 1372 n. 7 (1st Cir.1992) (quoting Kotteakos v. United States, 328 U.S. 750, 765, 66 S.Ct. 1239, 1248, 90 L.Ed. 1557 (1946)).
. For practical purposes, the authors of the over 14,000 underlying EXIS reports were effectively "unavailable” in this case. See United States v. Inadi, 475 U.S. 387, 394, 106 S.Ct. 1121, 1125-26, 89 L.Ed.2d 390 (1986) (absolute unavailability not constitutionally required in all cases); Manocchio v. Moran, 919 F.2d 770, 774—76 (1st Cir.1990) (same).
.The majority properly holds that the EXIS-derived statement — that out of more than 14,000 bombings and attempted bombings in the EXIS database only the Roslindale and Quincy incidents shared the specific queried characteristics — is inadmissible totem pole hearsay. That is, it was based on a host of out-of-court statements (the 14,252 underlying reports submitted by unknown authors) offered in court for the truth of the matters asserted therein (the characteristics of those bombings). See Fed.R.Evid. 801. Because we know neither who submitted those underlying reports, nor under what conditions, the majority properly holds that the statements do not satisfy any of the hearsay exceptions listed in Fed.R.Evid. 803(l)-(24).
. The residual hearsay exception contained in Fed.R.Evid. 803(24), under which the EXIS evidence was admitted, is not a “firmly rooted hearsay exception.” See Idaho v. Wright, 497 U.S. 805, 817, 110 S.Ct. 3139, 3147, 111 L.Ed.2d 638 (1990); Government of Virgin Islands v. Joseph, 964 F.2d 1380, 1387 (3d Cir.1992).
. Even the majority questions the validity of the EXIS conclusion that only the Roslindale and Quincy devices share the same characteristics. As the majority points out, because we know absolutely nothing about how the underlying EXIS reports were generated, there is no way to know what the absence of an item at a bomb site means. Both Scheid and the government's explosives expert admitted as much. Supra n. 21.
.The queried characteristics were 1) bombings and attempted bombings; 2) involving cars or trucks; 3) with bomb placed under the car or truck; 4) using remote-control; and 5) magnets. EXIS listed seven incidents which included these characteristics. Scheid testified that he then performed a manual query of the seven incidents using other characteristics of the Roslindale bombing. He checked the other incidents to see if they involved 1) duct tape; 2) soldering; 3) AA batteries; 4) a toggle switch; and 5) round magnets. Scheid did not check all 14,252 bombings and attempted bombings for these latter characteristics, only the seven.
. This Circuit has demonstrated that it is not shy about applying the harmless error rule to sustain a criminal conviction, but rather, shows a persistent inclination to so rule. See, e.g., United States v. Romero-Carrión, 54 F.3d 15 (1st Cir.1995); United States v. Cotal-Crespo, 47 F.3d 1 (1st Cir.1995); United States v. Smith, 46 F.3d 1223 (1st Cir.1995); United States v. Lewis, 40 F.3d 1325 (1st Cir.1994); United States v. Tuesta-Toro, 29 F.3d 771 (1st Cir.1994); Singleton v. United States, 26 F.3d 233 (1st Cir.1994); United States v. Isaacs, 14 F.3d 106 (1st Cir.1994); United States v. Welch; 15 F.3d 1202 (1st Cir.1993); United States v. Sepúlveda, 15 F.3d 1161 (1st Cir.1993); United States v. Innamorati, 996 F.2d 456 (1st Cir.1993); United States v. Williams, 985 F.2d 634 (1st Cir.1993); United States v. Spinosa, 982 F.2d 620 (1st Cir.1992); United States v. Figueroa, 976 F.2d 1446 (1st Cir.1992); United States v. Tejeda, 974 F.2d 210 (1st Cir.1992); United States v. Parent, 954 F.2d 23 (1st Cir.1992); United States v. Karas, 950 F.2d 31 (1st Cir.1991); United States v. Minnick, 949 F.2d 8 (1st Cir.1991); United States v. Maraj, 947 F.2d 520 (1st Cir.1991); Clark v. Moran, 942 F.2d 24 (1st Cir.1991); United States v. McMahon, 938 F.2d 1501 (1991); United States v. Brown, 938 F.2d 1482 (1st Cir.1991); United States v. Ellis, 935 F.2d 385 (1st Cir.1991); United States v. Sutherland, 929 F.2d 765 (1st Cir.1991); United States v. Wood, 924 F.2d 399 (1st Cir.1991); United States v. Paiva, 892 F.2d 148 (1st Cir.1989).
. The majority acknowledges that "[t]he statement that out of more than 14,000 bombing and attempted bombing incidents in the EXIS database only the Roslindale and Quincy incidents share the eight specific queried characteristics (bombings and attempted bombings, attached under car or truck, remote-control, round magnets, duct tape, solder, AA batteries, toggle switches) is a fairly powerful statement, but perhaps a somewhat misleading one." Supra n. 21 (emphasis added).
. As discussed previously, there is other evidence (i.e., the suspect nature of Scheid's query choices) which tends to show that the EXES query may have been skewed to reach a predictable result. See supra pp. 64-65.
. Federal authorities apparently did not deem the Quincy incident serious enough to warrant bringing charges against Trenkler pursuant to 18 U.S.C. § 844(i) (malicious destruction of property by means of an explosive), one of the statutes at issue in this case. State charges stemming from the Quincy incident were dismissed.
. Of course, the facts generated by the computer are only as accurate and reliable as the facts fed into it by its operator. As the majority recognizes, in this case the facts fed into the computer were, unbeknownst to the jury, manifestly unreliable. Thus, its conclusion based on those facts is similarly unreliable.
. Common sense tells us that lay jurors often will lend more weight to tangible evidence than to oral testimony. See generally 22 C. Wright & K. Graham, Federal Practice and Procedure, § 5173 (1978) ("It is often asserted that the psychological impact of the concrete has a capacity to suggest matters not proved, to lead the jury to draw unconscious inferences that would not be drawn if the object was the subject of testimony rather than being produced in court.”) (internal citations omitted). See also People v. Moore, 71 N.Y.2d 684, 529 N.Y.S.2d 739, 525 N.E.2d 460, 463 (1988) (Kaye, J., dissenting) ("No point in a trial can be more critical than jury deliberations. Materials taken into the jury room at those crucial moments may well influence the verdict.”).
.As one commentator has noted:
Scientific evidence impresses lay jurors. They tend to assume it is more accurate and objective than lay testimony. A juror who thinks of scientific evidence visualizes instruments capable of amazingly precise measurement, of findings arrived at by dispassionate scientific tests. In short, in the mind of the typical lay juror, a scientific witness has a special aura of credibility-
Imwinkelried, Evidence Law and Tactics for the Proponents of Scientific Evidence, In Scientific and Expert Evidence 33, 37 (E. Imwinkelried ed. 1981). See also Giannelli, The Admissibility of Novel Scientific Evidence: Frye v. United States, a Half-Century Later, 80 Colum.L.Rev. 1197, 1237 (1980) (“The major danger of scientific evidence is its potential to mislead the jury; an aura of scientific infallibility may shroud the evidence and thus lead the jury to accept it without critical scrutiny.”); 22 C. Wright & K. Graham, supra note 41, § 5217 ("Scientific ... evidence has great potential for misleading the jury. The low probative worth can often be concealed in the jargon of some expert ... ”).
. In Coppola, for example, we lent little weight to defendant’s statement to another inmate— "What did I have to lose?" — in response to a question whether he had committed the rape. See 878 F.2d at 1569-70.
. If the government makes an explicit promise to a witness, of course, this will come out at trial and likely decrease the witness's credibility in the eyes of the jury. But if the government lawyers explain to the witness why they do not want to make any explicit promises, leaving the inference that one good deed begets another, the witness can testify that he has no agreement. I note, in this regard, that this court has previously questioned the validity of these "no agreement” statements by criminal defendants. See, e.g., Coppola, 878 F.2d at 1569-70.
.When asked on direct examination why he testified, Lindholm stated:
Since I have been incarcerated, I have come to realize that the sole function of prison is not just punishment. I think rehabilitation is im*70portant for an individual. And I think, when I talk about rehabilitation, I mean rehabilitation of a person's values in terms of how they live one’s life and the decisions they make, knowing the decisions they make, knowing the difference between what’s wrong and what’s right, what's illegal and legal.