concurring in part and dissenting in part:
I respectfully dissent with respect to Part II of the majority opinion. There is apparently no dispute between the members of the majority and myself that it was error for GBI agent Keadle to testify that during his questioning of Alderman the latter decided to exercise his right to an attorney and to remain silent. The majority holds that the error is harmless beyond a reasonable doubt and that the district court misconstrued our holding in Chapman v. United States, 547 F.2d 1240 (5th Cir.), cert. denied, 431 U.S. 908, 97 S.Ct. 1705, 52 L.Ed.2d 393 (1977). I find the discussion of the harmless error point by District Judge B. Avant Edenfield very helpful:
Even though Doyle and Hale were violated, this can be a basis for reversal only when the error is not “harmless.” For an error to be harmless, it must appear to the Court beyond reasonable doubt that the evidence complained of did not contribute to the Petitioner’s conviction. Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967). This difficult standard was found met by the Fifth Circuit in Chapman v. United States, 547 F.2d 1240 (5th Cir. 1977). In that case, the court relied upon several factors. There was only a single reference at trial to defendant’s silence. That reference was neither made nor elicited by the prosecution, but instead resulted from a witness’s spontaneous remark. The reference was neither repeated nor linked to defendant’s exculpatory story. Furthermore, the court found that this story was “transparently frivolous” and the evidence of guilt was otherwise overwhelming. 547 F.2d, at 1250. See also United States v. Sklaroff, 552 F.2d 1156, 1162 (5th Cir. 1977). However, Chapman also notes that “even a single reference on direct examination to defendant’s silence carried an intolerably prejudicial impact, where the defendant’s exculpatory story was not totally implausible and the government’s inculpatory evidence was not overwhelming.” 547 F.2d, at 1249. Moreover, the Court indicates that reversible error can result where prosecutors purposefully employ a defendant’s post-arrest silence to defeat even a transparently frivolous exculpatory story. Id., at 1248.
Applying these guides to the facts at hand, the Court must conclude that the error complained of was not harmless beyond a reasonable doubt. Mr. Alderman testified that he had quarreled7 with his wife early in the evening of her death. Later on, he drove his motorcycle to her grandmother’s home believing that Mrs. Alderman had probably gone there. Petitioner claimed that he saw the family car with lights on in the creek near the grandmother’s home. His pants became stained when he went down from the road to investigate and, subsequently, put her head in his lap. However, the shock of the discovery and fear of his in-laws’ reaction caused him to *565flee the scene almost immediately and to block all recollection of the event out of his mind for some time thereafter.
Obviously, Petitioner’s explanation is not easily accepted.7 8 However, it does not appear that the inculpatory evidence in the present case was nearly so overwhelming as was true in Chapman. There police discovered the defendant holding a crowbar wedged in the front door of a bank. He explained only by claiming that he was recovering his crowbar after it had been placed there by two unidentified hitchhikers to whom he had loaned his car. In the present case the major direct link of Petitioner to the crime was the testimony of John Arthur Brown, an acquaintance and alleged accomplice who, quite unlike Petitioner, had a long history of drug abuse and mental instability. T. 383. This witness admitted to experiencing periods when he could not distinguish reality from fantasy and other occasions when he drank so heavily that he was unable to recall what he had done. T. 393. He also indicated that he had used numerous drugs the night before the crime, and consumed at least sixteen drinks the evening of the alleged murder. T. 454.
Even though Petitioner’s story is hardly subject to easy understanding, many aspects of Brown’s account' were similarly open to doubt. For example, he testified that he stalked Mrs. Alderman around a small apartment for over half an hour holding the twelve-inch crescent wrench allegedly used to strike her without arousing suspicion or even curiosity in the decedent. T. 431. Brown also described a violent struggle during which Mrs. Alderman was first struck on the head with this wrench.9 She was then tackled by Mr. Alderman and knocked to the floor after Mr. Brown’s blow proved inadequate. Brown stated that he and Petitioner then strangled her into unconsciousness and placed her in the apartment bathtub where she drowned. However, the state medical examiner found no indication whatever of strangulation. He also indicated that it was all but impossible for Mrs. Alderman to have been attacked as Brown described without leaving some such evidence. T. 297. Moreover, the examiner indicated that, while there had been a blow to the decedent’s head, it was not of a serious nature, and not sufficient to make him doubt that she was anything other than a typical drowning victim. T. 297. Furthermore, because no tests were ever made on the water in Mrs. Alderman’s lungs, there was never firm demonstration that she was drowned in her bathtub and not Dasher’s Creek. T. 318.
It is apparent then that significant aspects of the prosecution’s case were far from overwhelming. It is also apparent that the state’s role in the error complained of was far different from the one described in Chapman. Here the witness’s statement was in fact produced by the direct inquiry of the prosecution. Far from being spontaneous, it was the result of a carefully-phrased question which obviously attempted to keep the GBI agent’s reference to Petitioner’s silence outside the protections discussed in Doyle and Hale. The prosecutor asked, “Was Mr. Alderman in custody ... or were you simply discussing with him. . .. ” The witness replied referring to seeking only “A little bit of background.” T. 111. As I have already indicated, the circumstances surrounding this questioning do not at all support an inference of “preliminary investigation.” Moreover, the fact that the prosecutor ventured into this area and risked the very result reached here suggests that he did not regard his own case as otherwise “overwhelming.”
Finally, the Court must note that the jury which reviewed this evidence was already tainted by the Witherspoon error noted above. It was not an impartial cross-section of the community. Instead, three persons who exhibited particular sensitivity to the gravity of the juror’s responsibility were improperly excluded. The Court must therefore evaluate not merely the possible impact of the improper reference on the actual panel but on one that was never in fact developed. The Court must consider the role of the prosecution in the error and the nature of the evidence presented, all in the context of the Witherspoon problem. In the face of all these facts, I cannot conclude that no “reasonable doubt” exists as to the harmlessness of the disputed testimony. In these circumstances, I must also reverse and remand for retrial on the issue of guilt.
*566The majority states that the district court erred in applying the Chapman test1 conjunctively rather than disjunctively. The majority opines that the trial judge believed that he had to find the defendant’s exculpatory story not totally implausible and the indicia of guilt not overwhelming, to deny habeas under Chapman. The majority says that a finding of either is sufficient to deny the writ of habeas corpus. The majority then opines that the indicia of guilt were overwhelming and thus, since one part of the disjunctive test was met, the habeas corpus writ should be denied.
There was no error in applying both prongs of the Chapman test. Chapman requires reversal if either test is met. The majority says that a finding of guilt must be affirmed if either test is met and that here the indicia of guilt are overwhelming. The majority turns the test on its head. Reversal is required if either test is met. Here the district court found that both tests were met; thus he had two reasons for reversing.
Lastly, the majority refers to United States v. Shavers, 615 F.2d 266 (5th Cir. 1980). There we said the following:
Our standard for determining whether prosecutorial comment on defendant’s silence for substantive or impeachment value is harmless has been somewhat uncertain. In Chapman v. United States, 547 F.2d 1240, 1249-50 (5th Cir.), cert. denied, 431 U.S. 908, 97 S.Ct. 1705, 52 L.Ed.2d 393 (1977), we attempted to harmonize our decisions concerning Doyle violations and the harmless error test by placing cases into three distinct categories.
Unfortunately, as was noted recently in United States v. Dixon, 593 F.2d 626 (5th Cir. 1979), many cases lie somewhere in between the categories discussed in Chapman. In such situations we must seek refuge in the case by ease rule of United States v. Davis, 546 F.2d 583, 594-95 and n.31 (5th Cir.), cert. denied, 431 U.S. 906, *56797 S.Ct. 1701, 52 L.Ed.2d 391 (1977). “The decision requires an examination of the facts, the trial context of the error, and the prejudice created thereby as juxtaposed against the strength of the evidence of defendant’s guilt.” United States v. Meneses-Davila, 580 F.2d 888, 890 (5th Cir. 1978).
We have held that even a single reference on direct examination to defendant’s silence carried an intolerably prejudicial impact, where the defendant’s exculpatory story was not totally implausible and the government’s inculpatory evidence was not overwhelming. United States v. Impson, 531 F.2d 274 (5th Cir. 1976), cert. denied, 434 U.S. 1050, 98 S.Ct. 900, 54 L.Ed.2d 803 (1978).
615 F.2d at 269 (footnote omitted).
The district court found that the defendant’s exculpatory story was not totally implausible and the government’s inculpatory evidence was not overwhelming. I agree. I dissent.
There was no serious argument made at trial that this misunderstanding motivated the slaying. Neither was there any showing of particular marital discord. Instead, it was asserted that Petitioner killed his wife for the *566proceeds of a $10,000 insurance policy she held as an employee of the City of Savannah. However, there was no hint that Mr. Alderman was in any special need of money. He was regularly employed in a managerial job at a Garden City supermarket. There was no indication that he had a criminal record or was otherwise in any trouble.
It should be pointed out that Mr. Alderman’s story was supported by the testimony of an expert witness. This doctor concluded that Petitioner had suffered a “dissociative reaction” caused by the shock of the discovery. This reaction was, the Psychiatrist concluded after examination of Mr. Alderman, the cause of his selective amnesia. Petitioner’s account also comports with police descriptions of the bloodstains on his clothing as light and “watery.” T. 65, 174. Moreover, this state of “shock” seems to account for the fact that Petitioner did not simply remove his wet, stained, and obviously highly incriminating clothing. Certainly he had ample opportunity. Mrs. Alderman was discovered around 11:00 P.M. T. 27. Mr. Alderman was found at his apartment around 3:00 A.M. T. 236. Thus, he had several hours to change clothes, and he was not unaware of the problem. According to his chief accuser Brown, he had already changed once after his clothing was bloodied in the struggle to subdue Mrs. Alderman. T. 346. (There was apparently no attempt made to locate the clothing allegedly worn during the struggle to subdue Mrs. Alderman. Nor did police perform any tests on the clothing Mr. Brown put on after the murder to determine whether it too was stained. Brown indicated that it was still in his jail cell when he took the witness stand. T. 440.) Furthermore, the prosecution accorded the account sufficient weight to justify questioning the expert about his use of hypnosis in direct disregard of the order of the trial court that it not be brought up in any way. The prosecution asked Dr. Smith, the defense expert witness, whether hypnosis had been used and whether it could be “faked.” Both answers were in the affirmative. When the defense sought to question the witness on techniques that had been used to detect “faking,” the trial judge enforced its earlier pronouncement and refused to allow specific questioning by the defense. T. 736. Error is claimed on the basis of this ruling. However, because I find Doyle and Hale dis-positive, the issue is not reached below.
A wrench such as Brown described was found among Petitioner’s tools after the alleged crime. However, no tests were ever performed on it to determine whether it might have been used to strike Mrs. Alderman. There was also no explanation for why she was not rendered unconscious or at least much more severely injured when Brown struck her on the head with this presumably quite heavy implement.
. The Chapman test is “reversible error results if (he exculpatory story is not totally implausible or the indicia of guilt not overwhelming.” Chapman v. United States, 547 F.2d 1240 (5th Cir.) cert. denied, 431 U.S. 908, 97 S.Ct. 1705, 52 L.Ed.2d 393 (1977). The entire paragraph appears on page six of the majority opinion. Actually, there are three tests in Chapman and this is the second one.