David Wayne Spence v. Gary L. Johnson, Director, Texas Department of Criminal Justice, Institutional Division

80 F.3d 989

• Court: Court of Appeals for the Fifth Circuit
• Decided: March 29, 1996
• Docket: 94-20212, 94-20213
• Status: Published

Opinion

EDITH H. JONES, Circuit Judge:

In 1982, David Wayne Spence tortured and stabbed to death three teenagers in Waco, Texas. He has been convicted and sentenced to die for two of the murders in Texas courts and has been denied habeas corpus relief by the federal district court. Spence now seeks relief from this court. We find no reversible error and affirm.

I. FACTS

On the morning of July 18,1982, Jill Montgomery and Raylene Rice drove to Waco to pick up and cash Jill’s paycheck from the Fort Fisher Ranger Museum and to meet their friend Kenneth Franks. Later that day, the three drove to Koehne Park, located on the banks of Lake Waco. They were never seen alive again by their loved ones. Their bodies were found in a wooded area of Speegleville Park, across the lake from Koehne Park, a day later. Their last moments are recorded in the record of the two trials.

After arriving at the park, Kenneth, Jill, and Raylene encountered Spence and his cohorts, Anthony (Tony) and Gilbert Melendez. The six “hung out” together for some time drinking beer and smoking marijuana joints. After a few hours, in the evening, Spence persuaded the group to go to a convenience store and buy more beer. En route in his ear, Spence attempted to grab Jill’s breast. When she resisted, they argued heatedly. Spence then warned Kenneth and Jill he was going to “get even” with them “for some dope that he [Kenneth] had burned me for.” Kenneth denied he had “burned” Spence for anything.

Instead of driving to the convenience store, Spence turned and drove back into a wooded area in Koehne Park. After everyone got out of the car, he produced a knife and in vile terms ordered Jill and Raylene to undress. The girls immediately complied. Spence then forced Jill to walk with him to another part of the park. Gilbert ordered Raylene to get into the car. Gilbert then raped her.

Spence told Anthony to bring Kenneth, forcibly, to where he was with Jill, so that Kenneth could watch him rape Jill. Spence forced Jill to the ground, sat on her legs, and rubbed her breasts with his knife. He then raped her while Kenneth and Anthony watched. After Spence finished, Anthony traded places with him and raped Jill. Spence marched Kenneth back to the car where he raped Raylene.

After this rape, Spence returned to Jill. He cut her breasts and repeatedly stabbed her. At some point, Spence bit off one of Jill’s nipples. He then handed his knife to Anthony, telling him to stab Jill. When Spence believed that Anthony was not properly stabbing her, Spence took back his knife and finally inflicted the lethal wounds.

Spence then returned to the car and stabbed Kenneth to death. After this second murder, Spence grabbed Raylene and repeatedly stabbed her. He also ordered Tony to stab her. He then bit her body several times and rammed a piece of wood — which he referred to as his “lovestick” — into her vagina.

While Spence remained with the bodies, Anthony and Gilbert drove to Spence’s mother’s home and exchanged the car for Gilbert’s pick-up truck. During their absence, Spence bound the bodies. When the Melendezes returned, the trio threw the bodies into the back of the truck and, at Spence’s direction, drove to Speegleville Park. They dumped the dead teenagers’ bodies apart from each other in an off-the-road area. In placing Kenneth’s body, Spence boasted to Tony that the police “were going to freak out when they find this boy because he will be sitting up.” Spence and the Melendez brothers divided up several hundred dollars Spence had taken from Jill’s wallet.

The tortured bodies of Jill, Raylene, and Kenneth were found the next day.

From the federal habeas record, it transpires that intensive investigation of the highly publicized murders yielded no'clear suspects, leading the Waco Police Department to reduce the case to inactive status in September. That is to say, the department no longer committed extraordinary resources to the case. One dogged police officer, Truman Simons, volunteered to pursue the investigation and continued to work on it even after he changed jobs and became a jailer employed by the McLennan County Sheriffs Office. After January 1993, a newly elected district attorney, Vic Feazell, encouraged the investigation. By late March 1993, Simons persuaded Gilbert Melendez to confess his involvement and that of Spence in the crimes. Gilbert and Anthony Melendez, Spence and Muneer Deeb were indicted for the capital murders in November 1983. Spence was tried first in McLennan County for killing Jill, while his second capital murder prosecution was moved to Brazos County. The evidence in the two trials differs somewhat, as will be described below.

II. JUDICIAL PROCEEDINGS

In July 1984, Spence was convicted in the 54th Judicial District Court of McLennan County for the capital murder of Jill Montgomery. Following a separate punishment hearing, the jury affirmatively answered the court’s special issues required by the former Texas Code of Criminal Procedure Article 37.071. Accordingly, the court sentenced Spence to death. The Texas Court of Criminal Appeals affirmed Spence’s conviction and sentence, Spence v. Texas, 795 S.W.2d 743 (Tex.Crim.App.1990) (en banc), cert. denied, 499 U.S. 932, 111 S.Ct. 1339, 113 L.Ed.2d 271 (1991). Spence then filed an application for a writ of habeas corpus with the convicting court. The court adopted the State’s response as its findings and recommended relief be denied. The Texas Court of Criminal Appeals denied relief based on the trial court’s findings, Ex parte Spence, Application No. 15,346-03 (Dec. 12, 1991).

In September 1985, Spence was convicted after a change of venue in the 85th Judicial District Court of Brazos County for the capital murder of Kenneth Franks. Again, after a separate punishment hearing, Spence was sentenced to death. The Texas Court of Criminal Appeals affirmed his conviction and sentence, Spence v. Texas, No. 69,554 (Tex.Crim.App.1990), cer t. denied, 501 U.S. 1239, 111 S.Ct. 2875, 115 L.Ed.2d 1038 (1991). Spence’s application for a writ of habeas corpus was denied by the Texas Court of Criminal Appeals based on the trial court’s recommendation. Ex parte Spence, Application No. 15,346-03 (Dec. 12, 1991).

In December 1991, Spence filed federal petitions for writ of habeas corpus regarding each conviction and sentence. The petitions were assigned to one judge. After reviewing his allegations in each case, the court denied the petitions in separate opinions and orders dated April 29,1992.

Spence filed a voluminous motion to alter or amend the judgments. The district court consolidated its proceedings and ordered an evidentiary hearing on Spence’s allegations that the State had suppressed material exculpatory evidence. The parties agreed the best way to develop the evidentiary record was through depositions and affidavits, followed by massive briefing and oral argument. Numerous depositions were taken and numerous affidavits submitted. After the hearing, the district court re-entered, with very brief opinions, its earlier judgments denying habeas corpus relief.

This court granted certificates of probable cause to appeal, heard consolidated oral argument, and treats both cases in this opinion. For convenience, the claims raised in the Kenneth Franks ease are discussed first, followed by the Jill Montgomery ease issues.

III. DISCUSSION

This court reviews the district court’s legal conclusions de novo, applying the same standard as the district court, and its factual conclusions for clear error. Amos v. Scott, 61 F.3d 333, 337-38 (5th Cir.), cert. denied, - U.S. -, 116 S.Ct. 557, 133 L.Ed.2d 458 (1995) (citations omitted). “A finding of fact is clearly erroneous when, although there is evidence to support it, the reviewing court on the entire evidence is left with a firm and definite conviction that a mistake has been committed.” In re Henderson, 18 F.3d 1305, 1307 (5th Cir.), cert. denied, — U.S.-, 115 S.Ct. 573, 130 L.Ed.2d 490 (1994) (internal quotations and citations omitted).

A. Spence’s Claims Regarding Kenneth Franks’s Murder

Critical to evaluating Spence’s claims is a general understanding of the evidence offered at the trial for capital murder of Kenneth Franks.

By the time of the Kenneth Franks trial, both Gilbert and Tony Melendez had agreed to testify for the state, each having pled guilty to two life sentences while the third capital murder indictment hung over their heads. The State introduced expert evidence identifying Spence as the source of Jill’s bite injuries. The State also offered incriminating statements Spence had made to Regina Rosenbaum and inmate David Puryear. Spence’s defense was predicated on the lack of conventional demonstrative evidence connecting him to the crime; the internal and mutual inconsistencies in the testimony and statements of Gilbert and Tony Melendez; the asserted lack of corroborating evidence for the Melendez brothers’ version of events; and the unreliability of the State’s forensic odontology evidence. The defense also sought to show that Gilbert’s truck was inoperable on the night of the crimes and that people present in Koehne Park that evening did not recall seeing Spence’s group with the victims.

Challenging the constitutionality of his conviction, Spence contends that (1) the State failed to disclose some benefits Gilbert Melendez received for his testimony; (2) the State knowingly presented perjured testimony from Gilbert and David Puryear; (3) the State suppressed Waco Police Department reports from citizens who believed they possessed evidence linking other persons to the murders; (4) the federal district court erred in not granting a hearing on the reliability of the State’s odontology evidence; and (5) the district court failed to evaluate the cumulative effect of the above violations.

1. Gilbert Melendez’s Testimony

Spence argues that the State violated Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), by failing to disclose certain privileges and benefits Gilbert Melendez received in exchange for his testimony. The standards for a Brady violation are well settled. In Brady, the Supreme Court held that “the suppression by the prosecution of evidence favorable to an accused upon request violates due process where the evidence is material either to guilt or punishment, irrespective of the good faith or bad faith of the prosecution.” Id. at 87, 83 S.Ct. at 1196-97. The Court further established in Giglio v. U.S., 405 U.S. 150, 154-55, 92 S.Ct. 763, 766, 31 L.Ed.2d 104 (1972), that:

“when the reliability of a given witness may well be determinative of guilt or innocence, nondisclosure of evidence affecting credibility falls within the general rule of Brady.... Here, the Government’s case depended almost entirely on [one witness’s] testimony. [The person’s] credibility as a witness was therefore an important issue in the case, and evidence of any understanding or agreement as to a future prosecution would be relevant to his credibility and the jury was entitled to know of it.”

Id. at 154-55, 92 S.Ct. at 766.

To establish a Brady claim, a habeas petitioner must demonstrate that (1) the prosecution suppressed evidence, (2) the evidence was favorable to the petitioner, and (3) the evidence was material. U.S. v. Ellender, 947 F.2d 748, 756 (5th Cir.1991) (citations omitted). In assessing the materiality of suppressed evidence, the Supreme Court explained that “evidence is material only if there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” U.S. v. Bagley, 473 U.S. 667, 682, 105 S.Ct. 3375, 3383, 87 L.Ed.2d 481 (1985). “A ‘reasonable probability’ is a probability sufficient to undermine confidence in the outcome.” Id., at 682, 105 S.Ct. at 3383. Recently, the Court further observed that a “reasonable probability” of a different result is shown when the non-disclosure “could reasonably be taken to put the whole ease in such a different light as to undermine confidence in the jury verdict.” Kyles v. Whitley, -U.S.-,-, 115 S.Ct. 1555, 1566, 131 L.Ed.2d 490 (1995) (footnote omitted). “[A] showing of materiality does not require demonstration by a preponderance that disclosure of the suppressed evidence would have resulted ultimately in the defendant’s acquittal.” Id. at-, 115 S.Ct. at 1566. Finally, the materiality inquiry is applied to “the suppressed evidence collectively, not item-by-item.” Id. at-, 115 S.Ct. at 1567.

Fifth Circuit decisions have expanded upon these statements, holding that “[t]he materiality of Brady evidence depends almost entirely on the value of the evidence relative to the other evidence mustered by the State.” Smith v. Black, 904 F.2d 950, 967 (5th Cir.1990), judgment vacated on other grounds, 503 U.S. 930, 112 S.Ct. 1463, 117 L.Ed.2d 609 (1992). “[W]hen the testimony of a witness who might have been impeached is strongly corroborated by additional evidence supporting a guilty verdict, the undisclosed evidence is generally not found to be material.” Wilson v. Whitley, 28 F.3d 433, 439 (5th Cir.), cert. denied, — U.S. -, 115 S.Ct. 754, 130 L.Ed.2d 653 (1994). Similarly, when the undisclosed evidence is merely cumulative of other evidence, no Brady violation occurs. See Allridge v. Scott, 41 F.3d 213, 218 (5th Cir.1994), cert. denied, — U.S.-, 115 S.Ct. 1959, 131 L.Ed.2d 851 (1995).

Regarding Gilbert Melendez, Spence complains that the State failed to disclose that (a) Gilbert’s initial cooperation was induced by the prosecution’s insinuation that he could be granted immunity from prosecution for truthfully assisting the state; (b) the then-district attorney Feazell and investigator Simons orally promised not to oppose Gilbert’s release on parole; and (c) Gilbert received special privileges with his girlfriend while in the McLennan county jail leading up to his trial testimony. Because the State does not deny the nondisclosure of this information, the issue before this court is whether the evidence is material. Like the district court, we conclude it is not.

Before addressing Spence’s specific allegations, it is important to set forth several overriding considerations against finding the undisclosed evidence to be material. First, Gilbert’s testimony was supported by other evidence presented at Spence’s trial. Gilbert’s eyewitness-testimony was closely corroborated by his brother Tony, whose testimony has not been attacked at all by Spence. It was also consistent with the State’s autopsy reports, 1 the testimony of Regina Ann Rosenbaum and David Puryear, 2 and the odontological evidence. 3

Second, the undisclosed evidence is cumulative of other evidence impeaching Gilbert. Gilbert admitted at trial his prior convictions for assault with intent to murder and aggravated sexual assault, both of which resulted in prison terms. Gilbert further testified that he had been charged with three counts of capital murder for the deaths of Jill, Ray-lene, and Kenneth. He stated that he had received two life sentences in prison in exchange for his testimony against Spence, and that he had accepted the plea bargain to avoid facing the death penalty. Also at trial, Spence’s lawyer questioned Gilbert extensively about inconsistencies between his testimony and statements made to investigators about the crimes in March and April, 1983. Gilbert conceded that portions of his previous statements relating details of the murders had not been true, in part because he was initially concealing his brother’s role. Thus, the jury was amply informed about Gilbert’s criminal record, his inconsistent statements, and his motives for testifying. Defense counsel emphasized Gilbert’s untrustworthiness in closing arguments.

Third, the State presented other evidence of Spence’s guilt. Most prominent was the testimony of Tony Melendez to essentially the same gruesome facts of the murders. Consistent with the Melendez brothers’ testimony of the manner of the victims’ deaths, the Dallas County medical examiner testified that Jill, Raylene, and Kenneth had sustained multiple stab wounds and that the bite marks on Jill’s and Raylene’s bodies had been inflicted near the time of their deaths. Also, the State presented evidence that Spence had made general admissions of guilt soon after the murders. A few weeks after the murders, he told Regina Ann Rosenbaum and others present in his apartment that he had taken some girls to Lake Waco and raped them; no other similar crime was reported to police. Spence told David Puryear that he had committed the Lake Waco murders and was glad he had done so. Further, the State’s forensic odontological expert concluded that the bite marks on Jill’s and Ray-lene’s bodies were inflicted by Spence. Even Spence’s rebuttal expert in this field could not rule out the possibility that Spence’s teeth caused the wounds, although he believed there was too little evidence to support a firm conclusion.

Keeping in mind the general state of the evidence, we address Spence’s specific Brady arguments regarding Gilbert Melendez. 4

(i)Spence contends that the State violated Brady by not disclosing its initial “overtures” of immunity, which induced Gilbert to testify against him. This argument is unpersuasive. As the federal habeas testimony shows, the State’s initial immunity offer to Gilbert was based on his telling the complete truth about the murders. Gilbert admits that he did not tell the complete truth in his early statements and thus foreclosed any chance of immunity. Having misled the police in his early statements, Gilbert was left with two choices: he could refuse to cooperate and face prosecutions for capital murder, or he could plea bargain and avoid the death penalty. Gilbert chose the latter course and explained his plea bargain to the jury. Even if the state’s overture toward immunity might have influenced Gilbert’s decision to make his initial statements, that effect could hardly be considered material to the jury’s consideration of his truthfulness after they had learned of his plea bargain. Any hint that Gilbert might have originally falsely incriminated himself had vanished.

(ii) Spence next argues the State violated Brady by not disclosing that Truman Simons, then a deputy sheriff, and Vic Feazell, then the district attorney of McLennan County, orally promised they would not oppose Gilbert’s release on parole from his two concurrent life sentences in prison. The state now concedes that such oral promises were made and have been complied with by those two individuals. Nevertheless, the fact that such promises were made is not material on the record before us. No reasonable jury would have believed Gilbert fabricated his testimony — incriminating himself and his brother in kidnapping, rape, and murder — and pleaded guilty to two life sentences in prison based on oral promises that two individuals would not oppose his parole efforts some time in the distant future. Such promises were hardly a guarantee that Melendez would be paroled, nor did they bind future McLennan County officials.

(iii) Spence’s last assertion, concerning the State’s failure to disclose that Gilbert received unsupervised visits with his girlfriend while in the McLennan County jail, is also without merit. Even if the allegations are true, no reasonable jury would have believed that Gilbert fabricated his testimony and statements given over the course of two and a half years, from March 1983 to trial in September 1985, just to receive a few conjugal visits. This evidence is not material.

2. Perjury Allegations

Spence contends the State violated Napue v. Illinois, 360 U.S. 264, 79 S.Ct. 1173, 3 L.Ed.2d 1217 (1959), by failing to correct perjured testimony given by Gilbert and David Puryear. In Napue, the Supreme Court held that a prosecutor’s knowing use of, or deliberate failure to correct, perjured testimony, violates a defendant’s Fourteenth Amendment rights. 360 U.S. at 271, 79 S.Ct. at 1178-79. However, the Supreme Court also established that a defendant will only receive a new trial if there is a reasonable likelihood that evidence of the perjury would have affected the guilty verdict. U.S. v. Agurs, 427 U.S. 97, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976). We conclude the district court did not clearly err in finding that neither Gilbert’s nor Puryear’s trial testimony was perjured. Further, we conclude that Spence’s claims, even if proven, would have not have affected the jury’s verdict.

a. Gilbert’s Testimony

Spence argues Gilbert’s trial testimony was perjured because he told an extravagant tale in his 1993 federal habeas deposition concerning how then-deputy sheriff Simons helped him fabricate his statements, changing the details of the murders allegedly to match revelations in the ongoing investigation. The district court was unpersuaded that perjury had been promoted, and so are we. First, as has been repeated, Gilbert’s trial testimony incriminated himself and his brother for multiple murders, kidnappings, and rapes, for which they each received two life sentences in prison. The third capital murder indictment remains outstanding. It seems highly unlikely that Gilbert, having accepted such consequences as these, would retain any incentive to fabricate his testimony about the murders.

Second, that Gilbert’s 1993 deposition was given under oath is not dispositive. His trial testimony was also sworn. Given these conflicting sworn statements, the district court did not clearly err in finding the trial testimony more convincing. The Fifth Circuit has often noted' that “[rjecanting affidavits and witnesses are viewed with extreme suspicion by the courts.” May v. Collins, 955 F.2d 299, 314 (5th Cir.), cert. denied, 504 U.S. 901, 112 S.Ct. 1925, 118 L.Ed.2d 533 (1992) (internal citations and quotations omitted).

Third, the differences between Gilbert’s and Tony’s trial testimony only concern details of their involvement in the murders, leading even Spence, in his appellate brief, to concede their similarity in relating the basic course of events. Spence’s hideous cruelty permeates the testimony of each man in graphic and startlingly parallel particulars, e.g. Spence’s systematic torture of Jill, the tag team rapes of both girls, Spence’s insistence that Kenneth watch as he raped Jill, and Spence’s gleefully sitting Kenneth’s body upright at Speegleville Park. From the discrepancies in Gilbert’s and Tony’s testimony, one could more reasonably conclude that there was no fabrication. If Simons or Fea-zell had masterminded a conspiracy falsely to convict Spence, surely their efforts would have been designed to harmonize the brothers’ versions of the murders as closely as possible.

Fourth, as Gilbert’s testimony was corroborated by other evidence presented at Spence’s trial, the probability that the state knowingly presented false testimony further diminishes.

b. David Puryear’s Testimony

Spence contends Puryear’s trial testimony that Spence told him to draw on a bandana a picture of a knife and two girls— one blonde like Raylene and one brunette like Jill — was perjured. 5 We conclude the district court’s factual finding that Puryear did not commit perjury is not clearly erroneous. Spence bases his assertion on an affidavit from Puryear’s former brother-in-law, Steve Moore, stating that Puryear told Moore he had lied at Spence’s trial. 6 Not only does this affidavit relate inadmissible hearsay, but the record shows Puryear wrote an unsolicited letter to prosecutor Ned Butler stating his willingness to testify against Spence. Puryear’s letter asserted he was not seeking any “deal” for his testimony.

Spence’s argument that Puryear committed perjury because he did not tell the Waco police about the bandana when initially interviewed is also not dispositive. The omission does not prove he lied at trial, but may only demonstrate he was reluctant to “get involved” when first approached. See Smith v. Black, 904 F.2d at 961.

3. Police Reports

Spence contends that the State violated Brady by not disclosing Waco Poliee Department reports that might have implicated other persons in the murders, and that the district court applied erroneous standards in evaluating this argument. We reject these contentions.

a. Standard of Materiality

Before considering materiality, we address the district court’s legal approach. The district court concluded that the undisclosed evidence was not material because under Texas law it would not have been admissible at trial. The Fifth Circuit has expressly found otherwise in Sellers v. Estelle:

In addressing the issue of materiality, the Magistrate found that [statements in police reports] would have been inadmissible, hence these reports were immaterial. Such a conclusion is unwarranted. First, by enabling the defense to examine these reports, [the petitioner] may have been able to produce witnesses whose testimony or written statements may have been admissible. Second, the evidence here suppressed was material to the preparation of the petitioner’s defense, regardless of whether it intended to be admitted into evidence or not.

651 F.2d 1074, 1077 n. 6 (5th Cir.1981), cert. denied, 455 U.S. 927, 102 S.Ct. 1292, 71 L.Ed.2d 472 (1982) (citations omitted). Nonetheless, the court’s error was harmless, as will be seen.

Second, Spence contends the district court erred in interpi-eting Brady’s materiality standard as a result-oriented inquiry. The Supreme Court recently clarified that Brady is not a sufficiency of evidence test:

Bagley’s touchstone of materiality is a “reasonable probability” of a different result, and the adjective is important. The question is not whether the defendant would more likely than not have received a different verdict with the evidence, but whether in its absence he received a fair trial, understood as a trial resulting in a verdict worthy of confidence. A “reasonable probability” of a different result is accordingly shown when the Government’s evidentiary suppression “undermines confidence in the outcome of the trial.”

Kyles, — U.S. at-, 115 S.Ct. at 1566 (internal quotations and citations omitted). Taken as a whole, the district court’s opinion did not narrow the Brady standard. The court’s analysis begins, at page 2, by quoting Bagley to require a petitioner “to show that the suppressed evidence is material in that ‘there is a reasonable probability that had the evidence been disclosed to the defense, the result of the proceeding would have been different.’ (emphasis added and citation omitted).” We have no reason to believe the court deviated from this standard simply because he repeated it in a shortened version later in the opinion, doubtless, it seemed needlessly repetitious to include the entire Brady materiality formula in discussing each of Spence’s claims.

Spence focuses on undisclosed police reports that implicated Terry (“Tab”) Harper as a suspect, and suggested that a drug deal motivated Kenneth’s murder. 7

b. Reports on Terry “Tab” Harper

Spence contends the State violated Brady by suppressing Waco Police Department reports indicating Terry “Tab” Harper was a suspect. We reject this claim; information in the poliee reports, even if admissible evidence, does not undermine confidence in the jury’s verdict. While Spence’s argument accepts as true all of the reports that tried to link Harper to the murders, it ignores contradictory reports. The investigat ing officers, whose depositions are in the federal habeas record, unanimously concluded that the murders were not drug-related, that they were not consistent with Harper’s prior criminal behavior, and/or that he had an alibi. Harper was well-known as a bully, and the officers testified that whenever a high-profile crime occurred locally, young people would call the police station and associate Harper with the incident. Harper was brought to the police station and questioned about the Lake Waco murders but then released for lack of evidence. Spence highlights reports made to the police by two witnesses to whom Harper allegedly bragged about killing someone, but the reports furnish no other “evidence” of Harper’s involvement. In short, had Spence been given these police records and presented his theory regarding Harper at his trial, the State could have countered with other facts exonerating Harper. Thus, nondisclosure of the Tab Harper reports does not undermine confidence in the jury’s verdict; the information is not material.

c. Kenneth Franks’s Alleged Drug Use

Spence next argues the State violated Brady by failing to disclose police reports suggesting that Kenneth Franks was in debt over drugs and was a “known drug associate” of Harper. These reports, inadmissible by themselves, are not material and would not have supported a defense theory that the murders reflect a failed drug deal.

Spence’s argument again ignores reports and evidence that contradicted his drug deal theory. The autopsy report indicated no evidence of drugs in Kenneth’s body, and the police could not find any persons who would verify that Kenneth was a drug dealer or a customer or supplier of Tab Harper. The police unanimously concluded the murders were not drug-related, because of their viciousness and the fact that the bodies were not hidden or disposed of. Spence’s assertions that Koehne Park was a location known for teenage drug use and that Jill had cashed her paycheck before going to the park hardly raise an inference that Kenneth was murdered because of drugs.

d. Cumulative Effects of Suppressed Information

Kyles reminds us that the determination of the materiality of withheld evidence must be made “collectively, not item-by-item.” Kyles, — U.S. at-, 115 S.Ct. at 1567. So that no misunderstanding arises, we have considered cumulatively the significance of the withheld evidence — Gilbert Melendez’s benefits for testifying, police reports on Tab Harper, reports on Kenneth Franks’s status as a drug user and on Koehne Park witnesses — against the state’s total case on Spence. In stark distinction to the result of such a comparison in Kyles, the defense does not profit here. The strength of Tony Melendez’s testimony is unchallenged, while that of Gilbert is only somewhat weakened. Spence’s statements against penal interest to Rosenbaum and Puryear remain unassailed, while the status of hotly contested bite injury testimony is not changed. Without credible support, the Kenneth Franks-as-drug-dealer scenario could easily have backfired on the defense as a shoddy tactic to blacken the victim’s reputation. The police set no store by the reports of Tab Harper’s involvement after they investigated it; there is no reason to suspect defense counsel could have challenged their work based only on innuendo from the police reports. The Koehne Park witnesses would have repeated testimony that came out at Spence’s trial through other visitors to the park.

Kyles presents a wholly different picture on the cumulative effect of withheld testimony. As the court summed up the withheld evidence there:

But confidence that the verdict would have been unaffected cannot survive when suppressed evidence would have entitled a jury to find that the eyewitnesses were not consistent in describing the killer, that two out of the four eyewitnesses testifying were unreliable, that the most damning physical evidence was subject to suspicion, that the investigation that produced it was insufficiently probing, and that the principal police witness was insufficiently informed or candid.

— U.S. at -, 115 S.Ct. at 1575. The jury, the Court believed, might legitimately have been led to wonder whether Kyles or his erstwhile Mend “Beanie,” the voluntary police informant, was best situated to have committed murder. In this case, the withheld “evidence” is of a decidedly less significant east, and, even if disclosed, it would not have undermined Tony’s testimony and other evidence already recited. We do not believe that the cumulative effect on Spence’s defense of withheld information was reasonably likely to have affected the jury verdict.

4. Odontological Evidence

Spence argues that the district court erred in not holding a hearing on his challenge to the admission of testimony by the State’s forensic odontologist, Dr. Homer Campbell, and that it erred in excluding Spence’s expert repoi’ts challenging Dr. Campbell’s testimony. We reject these arguments.

To receive an evidentiary hearing, a habeas corpus petitioner must allege facts which, if proven, would entitle him to relief. Lavernia v. Lynaugh, 845 F.2d 493, 501 (5th Cir.1988). “The court need not blindly accept speculative and inconcrete claims as the basis to order a hearing.” Id. (internal quotations and citations omitted). “Nor is a hearing required when the record is complete or the petitioner raises only legal claims that can be resolved without the taking of additional evidence.” Id.

Spence contends that, because he submitted materials challenging Dr. Campbell’s methodology and conclusions, the district court should have held a hearing to determine whether the admission of Dr. Campbell’s testimony violated the Eight Amendment’s requirement of “heightened reliability” under Johnson v. Mississippi, 486 U.S. 578, 108 S.Ct. 1981, 100 L.Ed.2d 575 (1988). Johnson is, however, inapplicable to the instant case. In Johnson, the Supreme Court vacated the death sentence because the jury had been allowed to consider evidence that was false. Id. at 585, 108 S.Ct. at 1986 (emphasis added). 8 In the instant case, Spence does not raise a question over whether Dr. Campbell’s testimony is false, but rather over what weight the jury should have accorded his testimony. Spence’s argument that Dr. Campbell had misidentified the remains of another woman likewise does not expose his testimony against Spence as false.

Spence is simply trying to relitigate this aspect of his defense eleven years too late. At trial, Spence introduced his own forensic odontologist, Dr. Gerald Vale, a leading expert in the field. Dr. Vale spiritedly criticized Dr. Campbell’s methodology and conclusions, although, critically, Dr. Vale admitted he could not rule out Spence’s teeth as the source of the bite marks. Because this evidentiary issue was fully and competently aired in the state courts, no violation of fundamental fairness under the due process clause has been shown. Bailey v. Procunier, 744 F.2d 1166, 1168 (5th Cir.1984).

Alternatively, Spence argues that the federal district court erred in excluding reports from five other expert odontologists who concluded that Dr. Campbell’s testimony was unreliable. But because Spence filed these reports after the district court’s discovery deadline, without explanation for his untimely filing, the district court did not abuse his discretion in refusing to admit the reports.

5. Cumulative Error Doctrine

Finally, we reject Spence’s argument that he is entitled to relief for the “cumulative errors” in his trial. The cumulative error doctrine provides relief only when the constitutional errors committed in the state trial court so “fatally infected the trial” that they violated the trial’s “fundamental fairness.” Derden v. McNeel, 978 F.2d 1453, 1457 (5th Cir.1992) (en banc), cert. denied, 508 U.S. 960, 113 S.Ct. 2928, 124 L.Ed.2d 679 (1993). “Few defendants, however, will succeed in demonstrating on collateral review that their prosecutions merited such condemnation. ... We therefore have defined the category of infractions that violate ‘fundamental fairness’ very narrowly.” Id. (citations omitted). In determining whether the cumulative error doctrine provides relief, we must “review the record as a whole to determine whether the errors more likely than not caused a suspect verdict.” Id. at 1458.

In the instant case, Spence cannot demonstrate he was deprived of a fundamentally fair trial. First, the undisclosed evidence undermining Gilbert Melendez’s credibility is not material. The jury would not have reasonably believed Gilbert falsely incriminated himself in the murders, for which he accepted two life sentences in prison. Second, Spence failed to establish his claims that Gilbert’s testimony was fabricated or that the state knowingly presented false testimony of Puryear. Third, Spence failed to establish that police reports mentioning other suspects could reasonably undermine the jury’s verdict. Fourth, the admission of Dr. Campbell’s testimony did not violate the Eighth Amendment or due process. Fifth, the trial record reviewed as a whole does not suggest the jury’s verdict is suspect.

Therefore, we affirm the district court’s judgment denying Spence habeas corpus relief regarding his conviction and sentence for Kenneth Franks’s murder.

B. Spence’s Claims Regarding Jill Montgomery’s Murder

Because the evidence at this trial was different from that in the later capital murder trial, a brief summary is necessary to place Spence’s claims in perspective. No eyewitness or codefendant testimony was offered in the first prosecution; Gilbert and Antonio Melendez had not yet pleaded guilty. Instead, the state built its case around the theory of a failed murder-for-hire plot spawned by Muneer Deeb, co-owner of a Waco convenience store, to avenge his jealousy over 16-year-old Gayle Kelley. Evidence suggested that Spence, Deeb’s friend, was persuaded to attempt to kill Gayle so that Deeb could cash in an accidental death insurance policy Deeb had recently purchased on Gayle. Murder victim Kenneth Franks was in fact a close friend of Gayle and an antagonist of Deeb; had Gayle not unexpectedly been placed on restriction at the Methodist Home in Waco, she testified, she almost certainly would have been at Lake Waco with Kenneth on the fatal evening. Other testimony suggested that Gayle resembled victim Jill Montgomery and that Spence, though he knew of Gayle from Deeb, did not necessarily know her on sight.

As the defense rightly pointed out, no hair or fiber evidence from the victims was ever tied to Spence or his automobile. The only forensic evidence against him was the hotly disputed testimony of the state’s odontologist that Spence’s teeth produced the bite injuries on Jill and Raylene. Instead, the bulk of the state’s case rested on incriminating statements from Spence himself.

About 1:30 a.m. the night of the murders, Spence told two acquaintances that he had been at the lake or that he had been with Tony earlier that evening. After the murders, Spence appeared depressed, was drinking heavily and obliquely confessed to his friend Dorothy Miles that he cut someone and might have killed someone. Witnesses who contributed to the murder-for-hire theory included Karim Qasem, the co-owner of the convenience store with Deeb, and Spence’s co-worker Ray Payne and his wife. Spence and Deeb argued loudly in front of Payne and Qasem about the murder of a girl named Gayle later in the summer, 1992: Spence was accusing Deeb because he incorrectly thought the victim was Gayle Kelly. Spence told Regina Rosenbaum within a few weeks after the Lake Waco murders that he and his friends had raped two girls out at Lake Waco.

Significant reinforcement of the state’s case was provided by seven men who were inmates with Spence, five of them in the McLennan County jail from September 1982 to March 1983, and two others, Snelson and Ivy, at TDC later. At trial, the inmate witnesses all testified to highly incriminating statements of various sorts made by Spence, and they all denied receiving any promises or inducements to testify. Inmate David Pur-year described the bandana Spence asked him to decorate and said that Spence told him he had committed the Lake Waco murders and enjoyed doing it.

Spence’s defense consisted of an attack on the weaknesses of the murder-for-hire theory, e.g. the fact that the accidental life insurance policy would not pay Deeb in the event of Gayle’s murder, and the possibility that Spence knew Gayle on sight and could not have accidentally killed Jill instead. Spence’s attorneys tried to minimize the incriminating effect of statements to Dorothy Miles, Regina Rosenbaum and David Pur-year. They vigorously cross-examined the inmate witnesses about the likelihood of deals received from the prosecution. They showed, for instance, that Jennings just happened to keep notes of his conversations with David Spence — and David Spence only — on a legal pad in his cell. They intimated that Sypho had received a more lenient sentence than he ought to for being a four-time felony convict. They suggested that Jordan had been let out of prison soon after he gave a statement against Spence. And they used Puryear’s testimony about the bandana, among other things, to intimate that Truman Simons was planting rumors in the McLen-nan County jail that effectively acquainted many of the inmates with facts about the Lake Waco murders and enabled them to concoct stories about Spence. Finally, they pointed out the patent inconsistencies between the testimony of Snelson and Ivy concerning their motive for talking with Spence about the crimes. 9

Regarding his capital conviction for Jill Montgomery’s murder, Spence principally contends that (1) the State failed to disclose that inmates testifying against him received special privileges in exchange for this cooperation; (2) the State knowingly presented false testimony from the inmates; (3) the State suppressed evidence implicating other persons in the murders; and (4) the district court failed to evaluate the cumulative effect of the above violations. He also repeats the argument, dealt with above, that the federal district court erred in not granting a hearing on the admission of the State’s forensic odon-tologist’s testimony.

1. Inmate Witnesses

Spence contends that the State violated Brady by failing to disclose that the inmate witnesses were given special privileges for their testimony against him, and that the district court applied erroneous standards in evaluating the materiality of this suppressed evidence.

In its opinion in this case, more than in the companion ease, the district court seems to emphasize a result-oriented approach to materiality under Brady. As we have noted, Kyles re-emphasized that the proper question is whether there is a reasonable probability that the suppressed evidence could have affected the verdict. Kyles, — U.S. at -, 115 S.Ct. at 1566. To the extent the district court erred in applying the standard, however, the error is harmless.

Spence also contends the district court erred by conducting a “piecemeal” analysis of his Brady claims. While it is true that Brady violations must be considered for their cumulative effect on the jury verdict, Lindsey v. King, 769 F.2d 1034 (5th Cir.1985). Spence’s piecemeal argument is erroneously premised on his contention that all seven inmate witnesses received undisclosed privileges for their testimonies. The district court rejected Spence’s claims concerning Puryear and Sypho, and Spence omits to mention that he did not assert that James Jordan ever received undisclosed benefits or privileges. The magnitude of cumulative Brady error is much smaller than Spence would have us believe.

Spence alleges the State violated Brady by failing to disclose that the inmates who testified against him were given special privileges — such as conjugal visits, the ability to keep food, alcohol, and drugs in their jail cells, access to “free world” food and cigarettes, recommendations of leniency, and assistance in gaining admission into the Federal Witness Protection Plan — in exchange for their testimony. At trial, each of the seven inmate witnesses denied having received any privileges for his testimony. Six years later, two of these witnesses recanted their testimony, claiming in affidavits or depositions that they and other inmates received such benefits. The district court found that only David Snelson and Jesse Ivy received undisclosed privileges. The court specifically rejected claims that David Puryear and Charles Sypho received undisclosed inducements to testify. 10

At the outset, we reiterate that “recanting affidavits and witnesses are viewed with extreme suspicion by the courts.” May v. Collins, 955 F.2d at 314 (citations omitted). “In the [Fifth] [C]ireuit, a federal district judge, faced with a motion for a new trial predicated upon the contention that a witness has provided a recanting affidavit, must compare the trial record with the affidavit of recantation and determine for himself whether the affidavit is worthy of belief.” Id. In the instant case, the trial record shows that both the prosecution and Spence’s counsel questioned each inmate and former inmate extensively about any agreements with the State. 11 Each witness claimed he had received nothing for his testimony. That some of these witnesses — Snelson, Ivy, and some other inmates who did not testify, now claim that the testifying inmates received privileges for their testimony does not make it so. Further, the law enforcement officers’ testimony about special privileges is conflicting. Given the competing testimony, it would be difficult to hold the district court’s factual findings clearly erroneous.

The court’s treatment of Charles Sypho is illuminating: Spence alleged that Sypho received conjugal visits with his wife in exchange for his testimony against Spence. Spence bases this claim on affidavits from other inmates. Sypho, though, testified at Spence’s trial that he did not receive any inducements to testify. Thus, even if he enjoyed a few conjugal visits, the district court did not clearly err in finding they were not the inducement of his testimony. Given the competing sworn statements, we uphold the district court’s finding.

The district court also did not clearly err in rejecting Spence’s claim that David Pur-year received undisclosed assistance with his case for his testimony. This claim is identical to the one made regarding Spence’s trial for Kenneth Franks’s murder. It depends on the hearsay affidavit of Puryear’s former brother-in-law, which was contradicted by Puryear’s unsolicited letter to prosecutor Ned Butler volunteering to testify and disclaiming any “deal” for his testimony against Spence.

While we are bound by the district court’s findings that Snelson and Ivy received privileges that were undisclosed at trial, we also conclude this evidence is not material under Brady.

Both Snelson and Ivy were effectively cross-examined by the prosecution, although not on the subjects of conjugal visits or their potential eligibility for the federal witness protection program. But even if these inducements, made known to the jury, had caused them to discount the two inmates testimony, this would not have been reasonably likely to affect the verdict.

The impact of the undisclosed evidence in this case is distinct from the evidence that was held to require a new trial in Giglio v. U.S., supra. In Giglio, the evi dence undermined the credibility of the only witness tying the petitioner to the crime. To the contrary, the prosecution for Jill’s murder did not depend on the credibility of one or two witnesses, such as Snelson and Ivy. Besides his challenged admissions to the inmate witnesses, Spence made general admissions of guilt to other people. He told his neighbor, Dorothy Miles, that he “might have killed somebody” and that he “cut somebody.” Miles testified that Spence appeared “troubled” and “depressed” during July and August of 1982, the period right after the murders were committed. Regina Ann Rosenbaum testified that, in late July or early August of 1982, Spence recounted to her how he and “some friends” had gone to Lake Waco, come across some “chicks,” tied them up, and raped them. Spence has never attacked the testimony of former inmate James Jordan, to whom he confided not only his guilt but also striking details of the offense and the possible motive of jealousy. 12

Second, the State presented evidence consistent with the inmate witnesses’, Miles’s, and Rosenbaum’s testimony. The State’s pathologist testified that Jill’s and Raylene’s bodies had numerous stab wounds on the chest, shoulders, and breasts. She also testified that the girls’ bodies had internal bruises on the genital areas consistent with rape. Additionally, the State presented forensic odontological evidence demonstrating that the bite marks on the girls’ bodies were consistent with Spence’s dental impressions.

Third, the State presented evidence tying Spence to the time and location of the murders. Todd Childers testified that Spence told him he had been at Lake Waco the night of the murders. Clifford Oliver testified he remembered Spence saying he had “been with Tony” the night of the murders.

Given the voluminous amount of testimony and evidence in the trial record, we cannot conclude that the state’s failure to disclose certain privileges granted to some of the inmate witnesses was material.

2. Perjury Claims

Spence’s brief is long on vituperation and innuendo but precious short of facts supporting his contentions that the State violated Napue v. Illinois, supra, by knowingly presenting false testimony from all of the inmate witnesses that they had received no inducements and by assisting all of the inmates in fabricating their testimony. Spence also urges that the district court applied incorrect materiality standards in evaluating his Napue claims.

Notwithstanding Spence’s extremely one-sided characterization of the depositions and affidavits before the federal habeas court, the court did not accept his theory that Truman Simons orchestrated false inmate testimony by seven witnesses against Spence. At best, the court may have found that Snelson and Ivy, inmates who never encountered Spence until he was sent to TDC, might have delivered false testimony. The court’s opinion moves on to the materiality issue without expressly making findings that Snelson or Ivy perjured themselves at tidal or that the state knew of the perjury. As to Puryear, the court finds only “inconsistencies”, but no perjury or knowing presentation of perjury by the state. In its rehearing opinion, the court saw no need to modify these findings.

The court did not clearly err in rejecting the broad claim of mass-manufactured inmate testimony against Spence. There is no support for it in any of the numerous law enforcement officers’ depositions, 13 including that of Jan Price because her testimony, vehemently disputed by Si- mons in his deposition, does not deal with this ease. Spence also quotes selectively from depositions of other officers who appear to criticize Truman Simons; those quotations are not faithful to the overall context of their statements. It should finally be pointed out that a conspiracy to fabricate inmate testimony about incriminating statements made by Spence could surely have first been identified and challenged at trial rather than six years later in habeas proceedings.

But even if Snelson and Ivy testified falsely at trial, and even if the district court incorrectly employed an outcome-determinative approach to the materiality of their testimony, no constitutional error occurred. Under the proper materiality standard, it is not reasonably likely that Snelson’s and Ivy’s false testimony would have affected the jury’s judgment. Napue, 360 U.S. at 271, 79 S.Ct. at 1178. This standard, concededly less onerous than the Brady materiality standard, Kirkpatrick v. Whitley, 992 F.2d 491, 497 (5th Cir.1993), is not met here. The quantity and quality of the evidence against Spence, including his incriminating admissions to Rosenbaum, Miles and the other inmates, and the circumstantial and forensic odontolo-gy evidence, rendered Snelson’s and Ivy’s testimony cumulative.

3. Police Reports on Other Suspects

As in the Kenneth Franks case, Spence contends that the State violated Brady by failing to disclose Waco Police Department records implicating other persons in the murders, and that the district court applied an erroneous standard in evaluating this contention. We need not repeat the previous discussion concerning nondisclosure of reports about Tab Harper and Kenneth Franks’s alleged drug dealing. Those claims are also invalid here.

New to this case is Spence’s contention that the State violated Brady by suppressing police reports implicating James Russell Bishop and Ronnie Lee Breiten in the murders. Specifically, the State does admit it did not disclose a police report that one person stated he had seen a man resembling Bishop threatening Kenneth over drug debts and a report suggesting Bishop and Breiten were acquainted. However, we reject Spence’s Brady contention; we do not find the information in the reports exculpatory or material. 14

First, Spence’s materiality argument is premised on his theory that Kenneth was murdered over drugs. We have already rejected this assertion. Second, the State’s forensic odontological evidence eliminated Bishop as a suspect in the murders. The State made dental impressions of Bishop’s teeth, compared the impressions to the bite marks on the JiE’s and Raylene’s bodies, and concluded the two were not consistent. This inconsistent odontological evidence rules out Bishop as a suspect. As the Texas Court of Criminal Appeals explained,

[tjhere is, however, unanimous agreement in the field of scientific odontology that if even one point of dissimilarity is found between the suspect’s dentition and the bite mark, it may be said with certainty that the suspect did not make the bite mark. Thus, that suspect may be eliminated.

Spence v. State, 795 S.W.2d 743, 751 (Tex. Crim.App.1990) (en banc) (citations omitted), cert. denied, 499 U.S. 932, 111 S.Ct. 1339, 113 L.Ed.2d 271 (1991).

Third, the undisclosed police report indicating Breiten and Bishop may have been acquainted cannot be material. The report documents that, two days prior to the murder, Bishop cashed his paycheck at a store where Breiten’s wife worked. This fact is too tenuous to undermine confidence in Spence’s conviction. Further, Spence’s materiality argument regarding this report is premised on his assumption that Breiten conspired with Bishop to commit the murders. The odontological evidence eliminated Bishop as a suspect, and as such, undermines Breiten as a co-conspirator.

Because these reports were not exculpatory of Spence, no Brady violation occurred from their nondisclosure.

4.Cumulative Effect of the Undisclosed Information

So that no misunderstanding arises, we have considered cumulatively the significance of the undisclosed evidence — the special privileges given to Snelson and Ivy, and the police reports on Tab Harper, James Russell Bishop, Ronnie Lee Breiten, and Kenneth’s alleged status as a drug user— against the State’s total case against Spence. Unlike in Kyles, supra, this evidence was not the basis for the State’s case and, taken cumulatively, does not undermine our confidence in the jury’s verdict. The State’s case was based on incriminating statements Spence made to Miles, Rosenbaum, Childers, and Oliver. This testimony was reinforced with testimony from the inmate witnesses who, other than Snelson and Ivy, have not recanted their testimonies. The State also presented witnesses supporting its murder-for-hire theory. Spence has not attacked any of these witnesses. Additionally, the State presented the odontology evidence.

Thus, while the undisclosed evidence regarding Ivy and Snelson may have weakened their testimonies, Spence cannot undermine the testimonies of all the other witnesses. The undisclosed police reports similarly would not have weakened the State’s case. As discussed regarding Kenneth’s murder, the police did not find credible the reports on Harper, and there is no reason to believe Spence could have challenged them work based on the reports. There is similarly no credible support for Spence’s theory that Kenneth and Jill were murdered because Kenneth was a drug dealer. Further, the odontologieal evidence ruled out Bishop as a suspect, and thus the police report about him would not have affected the jury’s verdict. Finally, the police report suggesting Breiten may have been acquainted with Bishop hardly support’s Spence’s theory that Breiten and Bishop conspired to commit the murders.

5.Odontologieal Evidence

Spence contends that the district court erred in not holding a hearing on his challenge to the admission of testimony by the State’s forensic odontologist, Dr. Campbell, and that it erred in excluding Spence’s expert reports challenging Dr. Campbell’s testimony. This contention is identical to the one raised regarding Kenneth’s murder. We have already rejected it.

6.Cumulative Error Doctrine

Finally, we reject Spence’s contention that the district court erred in not evaluating the cumulative effect of his above four allegations. Taking his arguments as a whole, he was not deprived of a fundamentally fair trial. The undisclosed evidence undermining Snelson’s and Ivy’s credibility is not material. Spence cannot establish his Napue claims or his Brady claims regarding the police reports. We also concluded that the State’s admission of Dr. Campbell’s testimony did not violate the Eighth Amendment.

Therefore, we affirm the district court’s judgment denying Spence’s habeas corpus petition regarding Jill Montgomery’s murder.

IV. CONCLUSION

For the foregoing reasons, we AFFIRM the district court’s judgments.

1 . Gilbert testified, for instance, about Spence's use of the “lovestick", which inflicted injury consistent with the vaginal injuries observed on Ray-lene.

2 . At Spence’s direction, Puryear, a fellow inmate, decorated a bandana for Spence with pictures of a blonde and a brunette girl, like the victims, and a knife.

3 .We later address Spence's challenges to some of this evidence and his contention that Gilbert’s and Puiyear’s testimonies were fabricated.

4 . Repeatedly, Spence asserts that because the federal district court did not specifically address certain underlying facts related to his claims, e.g. the potential offer of immunity to Gilbert or the photos of Gilbert and his girlfriend in the D.A.'s office, the court did not adequately analyze the case. We strenuously disagree. The court took extraordinary measures in permitting Spence to embark on voluminous discoveiy pursuant to his motions for new trial after the court's first opinions in these habeas cases were written. Detailed additional briefing was undertaken. The court held a hearing on the post-trial motions and then concluded that his original opinions were still valid. The record reflects the court's efforts to uncover, not obfuscate, the facts. That his conclusions on rehearing do not mirror the vehemence of Spence's arguments is no criticism of their ultimate accuracy.

5 . Spence's brief does not charge that Puryear lied in quoting Spence's admission that he committed the Lake Waco murders.

6 . Contrary to Spence’s assumption, the district court did not make a factual finding that Moore’s account was the correct version of events or that Puryear lied at Spence’s trial.

7 . Spence also contends the state violated Brady by not disclosing police reports containing statements by persons in Koehne Park the night of the murder stating that they did not see Spence or the Melendez brothers or hear any screams. This evidence is not material because, at most, it would have been cumulative of defense testimony at trial from several people who had visited the park and neither noticed the defendants nor heard screams.

8 . Furthermore, in Johnson, the Supreme Court specifically noted that the false evidence was “the sole piece of documentary evidence of any relevance to [the State's] sentencing decision.” Id. at 585, 108 S.Ct. at 1986. In the instant case, much other evidence demonstrated Spence’s guilt.

9 . The defense attempted to offer testimony that two other men, James Bishop and Ronnie Brei-ton, could have committed the murders. This attempt fell apart because Breiten’s accuser, his stepmother, testified outside the presence of the jury one day and, recanting her earlier statements, refused to testify at all the next day. The state trial court was not persuaded that the defense had offered evidence sufficient to connect Bishop or Breiten to each other, much less to the Lake Waco murders.

10 . Spence's broad complaint that the district court could not make credibility findings about the inmate witnesses because it conducted a "paper hearing,” is without merit. Through his counsel, Spence had agreed the best way to develop the evidentiary record for the hearing was through depositions and affidavits. He cannot now complain about this method. We also note that “the concerns about the inadequacy of 'trial by affidavit' are even more diminished in the context of a factual dispute rooted in witnesses' claims that they perjured themselves at trial.” May v. Collins, 955 F.2d at 314.

11 . By the time of trial, Jennings, Jordan and Beckham had all been released from prison, while Puryear, Sypho, Ivy and Snelson were still incarcerated.

12 . Spence's statements to Ray Payne about the murder of "Gayle” and his loud arguments with Deeb about that murder implied that he knew of sinister plans or plans gone awry relating to Gayle Kelley. Qasem heard Deeb and Spence frequently discuss whether Gayle Kelley ought to be killed for insurance, but Qasem did not take these conversations seriously. Puryear’s testimony concerning the blonde and brunette on the bandana circumstantially incriminated Spence.

13 . Some officers were generally critical of Si-mons's investigative techniques and use of informants. Because of these troubling allegations, we have reviewed the record carefully. These officers had limited involvement with Spence's case in its critical stages. None have alleged that Simons "fed” information to the inmate witnesses or aided them in fabricating their testimonies.

14 . We agree that the district court erred in concluding the undisclosed police reports were not material because they would not have been admissible at Spence's trial. The Fifth Circuit has held that inadmissible evidence may be material under Brady. Sellers v. Estelle, 651 F.2d at 1077 n. 6. However, we also conclude this error was harmless.