GOLDBERG, Circuit Judge:
This appeal is from an order denying to Don Garriga Chapman post-conviction relief pursuant to 28 U.S.C. § 2255. Chapman is currently serving a ten year sentence for unlawfully attempting to enter a federal banking institution with intent to take money therefrom, in violation of 18 U.S.C. § 2113(a). This court affirmed Chapman’s conviction on direct appeal.
United States v. Chapman,
435 F.2d 1245 (5th Cir. 1970),
cert. denied,
402 U.S. 912, 91 S.Ct. 1392, 28 L.Ed.2d 654 (1971). For purposes of this collateral attack, Chapman argues that two constitutional errors fatally tainted his trial.
The first line of attack, which we rejected on direct appeal, is that the prosecution improperly elicited evidence of Chapman’s post-arrest, post-Miranda-warning silence as part of its case in chief. The second line of attack is premised on the prosecution’s cross-examining Chapman regarding two prior uncounselled felony convictions. Although we deflect without difficulty the second attack by invoking the harmless error rule,
the first requires us to assess the impact of recent Supreme Court and Fifth Circuit cases handed down after Chapman’s 1970 conviction had become final. Ultimately we find that this challenge as well fails to meet the threshold requirement of the harmless error rule, and consequently we affirm.
I. Facts
A police officer was directed to the Parkway National Bank. Upon his arrival, he
heard a scraping noise- in the stairwell behind the bank. He then saw Chapman removing a crowbar from the door of the bank. Chapman crouched in the corner of the stairwell, concededly attempting to hide, but the officer arrested him.
The officer escorted the appellant from the stairwell. When they reached the top of the stairs, the officer read Chapman his
Miranda
rights. He searched Chapman, then led him to the officer’s patrol car.
In the course of presenting its case at trial, the prosecution asked the arresting officer the following question:
“Did Mr. Chapman say anything else to you immediately after you arrested him as he was coming up out of the stairwell or during the time you were taking him to the car and after you put him in the car, did he say anything else to you?”
The defense objected. The prosecutor told the witness to answer yes or no. After the trial judge had excused the jury, government counsel explained: “ . . . Now, the reason I asked him the question is I want to point out that he did not say anything, he did not give any explanation.” The judge said he would allow the question. When the jury returned, the prosecutor repeated his question. The arresting officer answered, “no.”
The defense subsequently called its only witness, the appellant. Chapman denied that he had been attempting to enter the bank. He claimed that two hitchhikers had borrowed his car, parked it near the bank, and, without his prior knowledge, used his crowbar in an attempt to enter the bank. When he learned that his companions had left the crowbar wedged in the bank’s rear door, Chapman averred, he immediately sought to retrieve the implement, knowing that it bore his fingerprints. Chapman claimed he was apprehended while trying to recover the crowbar.
The jury chose to disbelieve Chapman’s account and found him guilty as charged. Chapman does not challenge the sufficiency of this evidence to support his conviction.
II. The Use of Post-Arrest,
Fost-Miranda
Warning Silence
On direct appeal, this court held,
inter alia,
that the testimony regarding Chapman’s silence was not prejudicial and did not infringe Chapman’s fifth amendment rights. The court reasoned that the prosecutor asked a single question requiring merely a yes or no answer and that he received the answer without comment.
United States v. Chapman, supra,
435 F.2d at 1247.
A defendant is entitled to relitigate on a motion to vacate judgment an issue decided on direct appeal where there has been an intervening change in the law.
See Davis v. United States,
417 U.S. 333, 342, 94 S.Ct. 2298, 2303, 41 L.Ed.2d 109, 116 (1974);
Kaufman v. United States,
394 U.S. 217, 230, 89 S.Ct. 1068, 22 L.Ed.2d 227 (1969).
We shall consider whether an intervening change in the law that makes clear that the tactic employed against Chapman was a violation of due process enables his collateral attack to succeed where his direct appeal failed.
A. Intervening Law
Had we Chapman’s direct appeal before us today, there can be no doubt that we
should find the reference to the appellant’s post-arrest silence to constitute a violation of due process. Quite apart from whether any error was harmless the conclusion that error was committed follows a
fortiori
from our decision in
United States v. Impson,
5th Cir., 531 F.2d 274,
rehearing denied,
535 F.2d 286 (5th Cir. 1976), and the Supreme Court’s decision in
Doyle v. Ohio,
426 U.S. 610, 96 S.Ct. 2240, 49 L.Ed.2d 91 (1976).
The
Impson
court relied on the balancing approach taken in
United States
v.
Hale,
422 U.S. 171, 95 S.Ct. 2133, 45 L.Ed.2d 99 (1975),
to conclude that admitting evidence of the defendant’s post-arrest silence as part of the prosecution’s case in chief was error because its great potential for prejudice outweighed the slight probative value of defendant’s silence.
The court did not reach the question whether admission of such evidence as part of the prosecution’s case in chief was a constitutional violation under
Miranda.
In
Doyle,
the Supreme Court faced the question whether, although evidence of a criminal defendant’s post-arrest,
post-Miranda-warning
silence could not be admitted for substantive purposes, it could be used for impeachment purposes. The Court held that the due process clause forbids the prosecution to use such evidence for purposes of impeaching an exculpatory story offered at trial.
See also United States v. Luna,
539 F.2d 417 (5th Cir. 1976);
United States v. Harp,
536 F.2d 601 (5th Cir. 1976).
Hale
and
Doyle
condemn what this court had allowed in a series of decisions subsequent to Chapman’s trial but prior to
Hale. See e. g., United States v. Quintana-Gomez,
488 F.2d 1246 (5th Cir. 1974);
United States v. Ramirez,
441 F.2d 950 (5th Cir.),
cert. denied,
404 U.S. 869, 92 S.Ct. 91, 30 L.Ed.2d 113 (1971).
B. Applicable Law
Doyle
did not spring full-blown from the Court’s collective brow, but rather traces its impressive lineage to fundamental principles of our jurisprudence. In
Miranda
v.
Arizona,
384 U.S. 436, 468 n. 37, 86 S.Ct. 1602, 1625, 16 L.Ed.2d 694 (1966), the Supreme Court announced in dictum that “it is impermissible to penalize an individual for exercising his Fifth Amendment privilege when he is under police custodial interrogation. The prosecution may not, therefore, use at trial the fact that he stood mute or claimed his privilege in the face of accusation.” Although that admonition refers primarily to the prosecution’s affirmative use of the defendant’s prior silence rather than its use for impeachment purposes, even before
Doyle
some courts, though not the Fifth Circuit, had construed
Miranda
to prohibit its use for the latter purpose as well.
See United States v. Anderson,
162 U.S.App.D.C. 305, 498 F.2d 1038 (1974);
Johnson v. Patterson,
475 F.2d 1066 (10th Cir.),
cert. denied,
414 U.S. 878, 94 S.Ct. 64, 38 L.Ed.2d 124 (1973).
United States v. Brinson,
411 F.2d 1057 (6th Cir. 1969).
But see Sharp v. United States,
410 F.2d 969, 972 (5th Cir. 1969) (Brown, C. J., dissenting)-
Regardless of whether such impeachment evidence was prohibited under the law applicable at the time of Chapman’s trial, however, the trial court arguably should not have allowed this substantive use of the defendant’s silence. Throughout the briefs and oral argument of this case, the parties have assumed that only
Hale
and
Doyle
could show a violation. But unlike those cases, in the case at bar the prosecution arguably did not use Chapman’s silence to impeach but to support an inference of guilt. When the prosecutor elicited testimony regarding Chapman’s post-arrest silence, Chapman had not yet offered his exculpatory story. Impeachment evidence is ordinarily elicited on cross-examination or rebuttal. The timing of the prosecutor’s questions make clear that this was substantive, not impeachment testimony. Indeed, when the prosecutor defended the propriety of a question he knew would result in evidence of Chapman’s silence, he said nothing about impeachment; he said only that he wanted to show that Chapman did not give an explanation. This suggests that the prosecutor intended to bring before the jury the fact of Chapman’s silence as a tacit admission — a tactic this court properly condemned long before
Hale. See, e. g., Walker
v.
United States,
404 F.2d 900 (5th Cir. 1968). The failure to give a limiting instruction should prevent the subsequently acquired impeachment value of defendant’s silence from saving the impropriety.
The prohibition against the use of a defendant’s silence as a means of implying his guilt obtained at the time of Chapman’s trial. We have recognized that
“Miranda
establishes that the prosecution may not use as part of its case in chief a criminal defendant’s silence following his arrest and
warning.”
United States v. Fairchild, supra,
505 F.2d at 1383.
Chapman squarely raised and this court squarely rejected this claim on direct appeal, however, and we are bound by the law of the case. The present difficulty is understanding precisely the nature of our earlier ruling and determining whether the intervening law would change the result.
On direct appeal, we considered whether the reference to Chapman’s silence was “prejudicial.” We apparently concluded that because the reference was not “prejudicial,” it did not infringe Chapman’s “fifth amendment rights.”
See
note 2,
supra.
It is not clear whether we were analyzing the problem as a
Miranda
violation and concluding that it was harmless beyond a reasonable doubt, or weighing the prejudicial impact of the reference as an evidentiary matter. If our intention was the former, then assuming the harmless error doctrine is applicable to
Doyle
the retroactive application of that case could not change the result. We could but echo our earlier ruling of harmless constitutional error.
On the other hand, assuming that we found no constitutional violation at all, this could only have been because we neglected to apply
Miranda
or thought the prosecutor’s use of defendant’s silence as part of the ease in chief was permissible under
Miranda,
perhaps because we supposed the prosecutor was using it solely for impeachment purposes.
In either case, it follows that we have never decided whether the prosecution’s use of Chapman’s silence was harmless beyond a reasonable doubt. Applying the intervening law of
Doyle
would force this question upon us.
To be sure, it remains true, assuming we thought the prosecutor’s conduct permissible under
Miranda,
that we have once rejected Chapman’s claim that the prosecutor’s use of Chapman’s silence on direct testimony was unconstitutional. Therefore, it might be said, because it has always been understood that such conduct was prohibited under
Miranda
and because
Doyle
addresses only the use of a defendant’s silence for impeachment purposes,
Doyle
adds nothing to the law that obtained at the time of Chapman’s trial as it applied to Chapman. We reject this argument.
First, it is not clear whether the court thought the reference to Chapman’s silence was justified by its impeachment value. If it did, then clearly
Doyle
applies. But second, even if the court thought the reference was used solely for substantive purposes,
Doyle
would still have had an important message for the court. It must be recognized that the Supreme Court emphatically prohibited the use of a defendant’s post-arrest silence for impeachment purposes despite the value placed on cross-examination as a means, often the only means, of rebutting an exculpatory story.
See Doyle v. Ohio, supra,
426 U.S. at 634-636, 96 S.Ct. at 2252-53 (Stevens, J., dissenting). This makes it abundantly clear that when the value of cross-examination is not a factor, as when the prosecutor uses defendant’s silence in its case in chief, the Court would not hesitate to find this use of a defendant’s “insolubly ambiguous” silence a violation of constitutional dimension.
See United States v. Stevens,
538 F.2d 1203, 1205-6 (5th Cir. 1976). Therefore we think that in either case, had
Doyle’s
clarification and reaffirmation of this principle been available to the panel that heard Chapman’s direct appeal, the court would have found a constitutional violation.
At least as regards the impact of
Doyle
on a prior holding founded on an impeachment justification for admitting evidence of Chapman’s silence, we must confront the
question of
Doyle’s
retroactivity. Even if the court reached its decision without an impeachment justification but applied the law regarding substantive uses of a defendant’s silence, however, it would seem difficult to circumvent the question of retroactivity. For it would be troublesome indeed to allow Chapman to assert that
Doyle
effected an intervening change in the law for purposes of deciding whether our decision on direct appeal may be collaterally attacked, but to deny that
Doyle
constituted a “new rule” for purposes of posing the question of its retroactive application. We can consider Chapman’s collateral attack only if we assume there was an intervening change in the law, and this requires that we consider the matter of retroactivity.
We shall in any event treat our ruling on Chapman’s direct appeal as though it rested on the assumption that the reference to Chapman’s silence was saved from constitutional error by its value as impeachment evidence — the assumption
Doyle
directly contravenes.
C. Retroactivity of
Hale
and
Doyle
We first faced the question of the retro-activity of
Hale
in our reconsideration of
United States v. Impson, supra,
535 F.2d 286. We said that because the Supreme Court had vacated a previous panel decision in light of
Hale,
we specifically declined to deny retroactive application of
Hale
to
Imp-son’
s case.
See United States v. Impson,
506 F.2d 1055 (5th Cir.) (per curiam),
vacated and remanded,
422 U.S. 1031, 95 S.Ct. 2647, 45 L.Ed.2d 688 (1975),
on remand,
531 F.2d 274 (5th Cir. 1976),
rehearing denied,
535 F.2d 286 (5th Cir. 1976). Under similar circumstances this court subsequently applied
Doyle
itself retroactively to a case on direct appeal.
See United States
v.
Harp,
513 F.2d 786 (5th Cir. 1975),
vacated and remanded,
423 U.S. 919, 96 S.Ct. 258, 46 L.Ed.2d 246,
on remand,
536 F.2d 601 (5th Cir. 1976) (per curiam).
Even so,
Impson
and
Harp
establish only that
Hale
and
Doyle
apply retrospectively to cases on direct review, and it might be argued that the latter cases do not apply retroactively to a conviction under collateral attack. Generally speaking, however, the Court has declined to adopt for purposes of determining the retroactivity of new rules of criminal procedure the distinction between cases on direct review and those arising on collateral attack.
See Williams v. United States,
401 U.S. 646, 651-52, 91 S.Ct. 1148,1151-1152, 28 L.Ed.2d 388 (1971) (plurality opinion).
But see Bradley
v.
School Board of Richmond,
416 U.S. 696, 94 S.Ct. 2006, 40 L.Ed.2d 476 (1974);
United States v. Fitzgerald,
545 F.2d 578 (7th Cir. 1976),
reh. denied
(1976) (when a lower court relies on a principle or decision that is changed prior to direct review by the Supreme Court, an appellant court must apply the current law rather than the law applied by the lower court).
Moreover, in
Meeks v. Havener,
516 F.2d 902 (6th Cir. 1975) (per curiam),
vacated and remanded,
428 U.S. 908, 96 S.Ct. 3215, 49 L.Ed.2d 1213 (1976),
on remand,
545 F.2d 9 (6th Cir. 1976) (per curiam) [20 Cr.L.Rep. 2237, December 15, 1976], the Supreme Court vacated a denial of habeas relief in light of
Doyle,
and the Sixth Circuit found a violation of due process, albeit a harmless error. To be sure, it would be improper to conclude merely from the Supreme Court’s remanding in light of
Doyle
that it intended
Doyle
to be fully retroactive.
Nevertheless, the Sixth Circuit decision in
Meeks
represents a fully retroactive application of
Doyle.
Neither
Meeks
nor any of this circuit’s decisions, however, have explicitly considered the merits of according retroactive application to
Doyle
or
Hale
in light of the criteria established by the Supreme Court.
We need not decide whether decisions applying
Doyle
and
Hale
retroactively to cases on direct appeal imply that
Doyle
and
Hale
should be retroactively applied to a conviction under collateral attack, or if not, whether the Supreme Court’s test for determining retroactivity independently compels a conclusion of retroactivity here. Assuming
Doyle
and
Hale
are fully applicable to the case at bar, we think Chapman’s collateral attack fails because any violation of due process was harmless beyond a reasonable doubt.
D. Harmless Error
In
Doyle
itself, the Supreme Court held it would reverse because the prosecution’s use of defendant’s silence for impeachment purposes constituted a due process violation and because the state had failed to argue that the violation was harmless error. 426 U.S. at 618-619, 96 S.Ct. at 2245. This allusion to harmless error is not authority for applying that doctrine, but rather expressly reserves the issue. Nonetheless, we
decide today to apply the doctrine to Chapman’s case.
In
Meeks v. Havener, supra,
545 F.2d at 10, the Sixth Circuit relied on the doctrine of harmless error to deny relief to a habeas petitioner at whose trial the prosecution had committed a
Doyle
violation. In two decisions of this court, we have found it appropriate to reject claims of harmless error not because the doctrine was thought inapplicable to a
Hale
or
Doyle
case, but because, as the panels took care to show, the particular errors were not in fact harmless.
See United States v. Impson, supra,
531 F.2d at 278
(Hale), United States v. Harp, supra,
536 F.2d at 603
(Doyle
).
We agree with the Sixth Circuit that the harmless error doctrine is applicable to the kind of constitutional violation at issue in
Doyle. See Chapman v. California,
386 U.S. 18, 24, 87 S.Ct. 824, 828, 17 L.Ed.2d 705 (1967).
The next question is whether the constitutional violation in the case at bar was harmless beyond a reasonable doubt. We find transparently frivolous Chapman’s wholly uncorroborated story that he had loaned his car to two hitchhikers whom he had just met, that upon hearing that they had left his car at the bank and his crowbar in the bank’s rear door he went not to the police but to retrieve the crowbar, and that when discovered by a police officer holding on to a crowbar wedged into the rear door of the bank he was merely in the process of extricating the crowbar from the door. Chapman adduced absolutely no evidence that might have lent some plausibility to the story.
On the other hand, we have held that where the prosecution used the defendant’s post-arrest silence to defeat even a transparently frivolous exculpatory story, reversible error resulted. In
United States v. Harp, supra,
the defendants had been discovered in “an obvious escape attempt,” concealed in a boxcar about to leave the penitentiary and armed with various tools, clothing, and two “authentic looking handguns made of wood.” At trial for attempted escape, two of .the accused presented a “transparently frivolous duress defense,” claiming that they had been kidnapped by a codefendant. 536 F.2d at 602.
The court said that in light of
Doyle,
the prosecution’s emphasis during closing argument on the defendants’ post-arrest silence violated due process. The court added that “[bjecause the prosecutor’s comments struck at the jugular of [the defendants’ exculpatory] story, these remarks cannot be classified as harmless.”
Id.
at 603.
The facts of
Harp
made it evident that whatever remote plausibility the duress defense might have had was destroyed by the prosecutor’s closing argument. He repeatedly and explicitly contrasted the unlikely story with defendant’s failure to say anything about that story upon their arrest.
That is what the court meant in
writing that the prosecutor's comments “struck at the jugular” of the defendants’ story.
By contrast, in the case at bar the prosecutor asked one question eliciting the fact of Chapman’s silence, and that fact was never alluded to again. Neither the prosecutor nor any prosecution witness tied together the fact of Chapman’s silence with his improbable story. The jury was never told that silence could be used for impeachment purposes. Neither in cross-examination nor in argument did the prosecutor suggest that silence impeached Chapman’s trial testimony.
Cf. Mercado v. Massey,
536 F.2d 107 (5th Cir. 1976) (prosecutor’s remark adverting to defendant’s silence not construed as suggestion that silence could be used for impeachment purposes).
On the other hand, 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, supra,
532 F.2d at 278-79. Although
Impson
did not reach the constitutional ground of decision, after
Doyle
the conduct condemned in
Impson
clearly constitutes a due process violation. Under the facts of
Impson,
moreover, we would not require more than a single reference on direct examination to defendant’s silence, such as transpired in the case at bar, in order to reverse on due process grounds.
See
also
United States
v.
Hale, supra,
422 U.S. at 174, 95 S.Ct. at 2133.
The exculpatory story offered in
Impson
was, however, far more plausible and the indicia of guilt far weaker than in the case at bar. Impson was a passenger of a car that was stopped and searched. The police found counterfeit money in the car but no counterfeit bills on Impson’s person. Imp-son was charged with possessing counterfeit bills; accordingly, the prosecution was obliged to show that Impson knew and had constructive possession of the currency. Impson claimed in defense that he did not know that the driver of the car had such currency in the automobile and that he was in the car only because his own automobile had been wrecked earlier that day. The jury could not agree regarding guilt or innocence at Impson’s first trial, possibly because of the plausibility of the story, and Impson was convicted only after a retrial. In short, the story was not totally implausible and the evidence of guilt less than overwhelming. We properly rejected the government’s invitation to apply the harmless error doctrine.
In the case at bar, however, Chapman was discovered tugging at a crowbar lodged in the rear door of a bank. Evidence of his guilt was so strong and his exculpatory story so implausible and totally lacking in evidentiary support that we cannot say there is a reasonable possibility that the single reference to his silence contributed to his conviction. In short, we hold that it was harmless beyond a reasonable doubt.
See Chapman v. California, supra,
386 U.S. at 24, 87 S.Ct. at 828.
The results in these cases are easily harmonized. When the prosecution uses defendant’s post-arrest silence to impeach an exculpatory story offered by defendant at trial and the prosecution directly links the implausibility of the exculpatory story to the defendant’s ostensibly inconsistent act of remaining silent, reversible error results even if the story is transparently frivolous.
See United States v. Luna,
539 F.2d 417 (5th Cir. 1976);
United States
v.
Harp, supra.
When the prosecutor does not directly tie the fact of defendant’s silence to his exculpatory story,
i. e.,
when the prosecutor elicits that fact on direct examination and refrains from commenting on it or adverting to it again, and the jury is never told that such silence can be used for impeachment purposes, reversible error results if the exculpatory story is not totally implausible or the indicia of guilt not overwhelming.
See United States v. Impson, supra.
When there is but a single reference at trial to the fact of defendant’s silence, the reference is neither repeated nor linked with defendant’s exculpatory story, and the exculpatory story is transparently frivolous and evidence of guilt is otherwise overwhelming, the reference to defendant’s silence constitutes harmless error. Chapman’s fate is to instantiate this third rule.
The infusion of “harmlessness” into error must be the exception, and the doctrine must be sparingly employed. A miniscule error must coalesce with gargantuan guilt, even where the accused displays an imagination of Pantagruelian dimensions.
The judgment of the district court is AFFIRMED.