(dissenting):
The majority opinion is properly critical of the government’s performance in this case, both at trial and on appeal. *1175The prosecutor’s action in pointing toward Warf and the references to Warf’s prior incarceration were obviously improper. But these improprieties are reversible error only if there is a significant possibility that they had a substantial impact on the jury’s verdict. E. g., United States v. Constant, 501 F.2d 1284, 1289 (5th Cir. 1974), cert. denied, 420 U.S. 910, 95 S.Ct. 830, 42 L.Ed.2d 840 (1975); United States v. Arenas-Granada, 487 F.2d 858 (5th Cir. 1973). I respectfully dissent because I believe the majority opinion attaches insufficient weight to the totality of the evidence and consequently errs in performing the “harmless error” balance.
Leaving aside Hartman’s testimony, the majority opinion describes the evidence against Warf as “thin” and “circumstantial.” In the first place, the identification of Warf by Renee Lanier is direct evidence of Warf’s guilt. The uncertainty which the majority purports to find in that identification emerged under relentless cross-examination, and any weakening of her testimony is clearly attributable to the skill of defendant’s attorney rather than to any inherent weakness in her identification. See United States v. Bonds, 526 F.2d 331, 339 (5th Cir. 1976). The majority’s summary of the “circumstantial” evidence omits the significant fact — referred to earlier in the opinion — that Warf’s automobile, identified as similar to the one used by the robbers, was found on the day of the robbery, abandoned near the scene of the crime. If this be added to the summary, we see a defendant with a ear and a gun like those used in the crime, identified at the scene of the crime, apparently on the run with his car abandoned near the crime, and caught with some of the loot. With deference, this is not what I call thin.
Most significantly, the majority’s assessment of the possible impact of the prosecutor’s indiscretions virtually ignores the strongly incriminating statements attributed to Warf by Hartman. The weight of these statements is admittedly no greater than the ability of Hartman to identify the defendant as the declarant, but the majority ignores the strong indicia of reliable identification provided by the circumstances of this ease. Hartman identified the defendant by name and testified to an acquaintance with the defendant spanning about a year. Unlike Renee Lanier, Hartman was not identifying Warf as a previously unknown bank robber, seen on one occasion only and that the occasion of the crime for which Warf was being tried. As to her, the question was whether she could pick him out at all from among others present. By contrast, there is no real doubt or dispute that Hartman had known the defendant Warf previously and known him well. He "Was not identifying him in the sense that Ms. Lanier was, the sense of trying on the basis of a single earlier and imperfect viewing of the robber to determine whether her recollections were sufficient to fix upon Warf or, indeed, on anyone. Hartman was merely stating whether one whom he could without doubt identify was present in the courtroom. Thus, the cases are not parallel; and the prosecution is in large degree correct that no identification of Warf by Hartman, in any strict sense, was either necessary or intended. Given its context, Hartman’s testimony may be considered as evidence of Warf’s guilt for the purpose of determining the possibility that the prosecutor’s indiscretions had a substantial impact on the verdict. Considering all the evidence and the fact that the trial judge gave prompt instructions to disregard both improprieties, I believe that the improprieties created only “harmless error.” 1
With respect to the references to Warf’s prior incarceration, the persistence of the prosecutor in asking a follow-up question exacerbated the impropriety, but it did not automatically cre*1176ate reversible error. See United States v. Barcenas, 498 F.2d 1110, 1112-14 (5th Cir.), cert. denied, 419 U.S. 1036, 95 S.Ct. 521, 42 L.Ed.2d 312 (1974). The forceful instruction to the jury to disregard these references- mitigated their impact, and any error remaining was harmless.
. The majority quotes and refers to the trial judge’s “razor edge” and “thread” comments, but it should not be lost that after making them he denied a mistrial.