(dissenting).
What influences juries, courts seldom know. Indeed, most courts (including the federal courts) not only do not diligently seek such knowledge but have a general policy of deliberate unwillingness to learn —and usually seal up the only possible sources from which they could learn — what occurred in the jury-room. As we recently said, per Judge Learned Hand, this policy stems from awareness that, were the full truth disclosed, it is doubtful whether more than 1% of verdicts could stand.1
*22This self-imposed judicial ignorance 1a— preventing the courts from knowing how a particular jury behaved and from attaining moderately well-educated guesses as to how juries in general probably conduct themselves — has had some strange corollaries: (1) In spite of — or perhaps because of — this ignorance, upper courts make this remarkable assumption: A very substantial mistake by the trial judge in the wording of a highly intricate substantive rule is assumed to have affected the jury and therefore (necessarily and logically) to have led to a wrong verdict — even when the difference between the correct and the mistaken wording of the rule is one that only a lawyer, after careful study, can detect, so that very probably the mistake could not actually have had any effect on the jury.2 In such a case, thanks to the unverified and usually unreal assumption, the upper court notes the error — although it was not called to the trial judge’s notice — and orders a new trial.3 (2) Yet, paradoxically, many an upper court will refuse to reverse for a mistake in the charge, unrelated to any intricate substantive rule largely unintelligible to the jury, but dealing with matters far more comprehensible by the jury and therefore far more likely to have influenced their verdict.4 Such errors, the very kind most to be suspected of working harm, are, I think, too often dismissed as harmless.5
My colleagues’ opinion in this case illustrates, I think, the second point. For the faulty “presumption of innocence” instruction — that the presumption was “designed for the protection of the innocent” and was “not intended as a bulwark behind which the guilty might hide” — may easily have misled the jury, to the grave prejudice of *23the accused. We do not know whether or not it did. But we do know that that “presumption” is much discussed popularly, out of court. Consequently, there prevails a belief, probably well-founded, that jurors listen attentively to, and take with peculiar seriousness, the particular part of the judge’s instructions defining that “presumption.” On that account, I think the Fifth Circuit, in Gomila v. United States, 146 F.2d 372, wisely held such an instruction, as that here given, reversible error, although there, as here, the defendant had not objected when the charge was delivered.6
My colleagues, however, differentiate the charge in the Gomila case because it contained a “cumulation of errors.”7 *But so, too, here.8 The trial judge, far from curing his “presumption of innocence” error by his comments on “reasonable doubt,” aggravated that error when he told the jury that such a doubt is one which a juror “can give a reason for entertaining.” This again was matter to which the jurors probably paid close attention, since the important “reasonable doubt” principle is also often popularly discussed out of court. The judge’s definition of that principle, therefore, most probably impressed the jury.
Now this very sort of instruction was severely criticized, and made the basis of reversals, by Judge Walter H. Sanborn of the Eighth Circuit.9 He, one of our greatest judges,10 was no soft “sentimentalist,” eager to absolve criminals, but he was sensitive to the possible tragic impact on men’s civil liberties of verdicts due to carelessly phrased statements by trial judges to juries. As he pointed out, the instruction he condemned means that the jurors must not consider any doubt about guilt unless they can articulate, at least to themselves, a sound “reason” for doing so.
Our own judicial experience teaches that sometimes even trial judges have difficulty in rationally explaining the grounds of their decisions. Some trial judges resent any obligation in jury-less cases, where the oral testimony is conflicting, to go beyond publishing laconic, unexplained, judgments. They even object to filing special findings of fact. Some of those judges complain that the formulation of fact-findings is too arduous and, more, bound to be artificial.11 I share the view that, on net-balance, that complaint is unjustified.12 There would be more justification for complaint, if a trial judge were required to reveal, or even to disclose that he himself knew, the “reasons” for his special findings. But there is no such requirement.
Surely, then, we ought not ask a juror, who has nothing like a trial judge’s experience and training, not merely to find facts13 but also to be able to recognize, and to make explicit to himself or the *24other jurors, the-“reason” for joining in a finding of fact or for any intermediate step in his mental processes. Mr. Justice Holmes, speaking for the Supreme Court, said the reasoning of an administrative body need not be thus articulate, since its decision expresses “an intuition of experience which outruns analysis and sums up many un- . named and tangled impressions * * * which may lie beneath consciousness without losing their worth.” 14- We should not exact more of a jury. Accordingly, I think the jurors were badly mi-sinstructed when told they must ignore any doubt which they could “give no reason for entertaining.”
My colleagues assume, 'in effect, that the jurors construed that instruction as would an intelligent law student or lawyer, well-versed in the traditional meaning of “reasonable doubt.” I cannot go along'with that assumption. Far ’more probably, the jurors thought the judge meant just what he said. One can, then, well imagine that one of the jurors, when in the jury-room, impressively remarked to his fellow-jurors: “Remember ■ that the judge especially warned us we must discard any doubt about guilt unless we can ‘give a reason’ for holding it.” So it seems to me not at all improbable that the jurors, acting under this erroneous instruction,' and incapable of formulating a rational explanation of a doubt they entertained, concluded that they had the duty to discard a doubt which, had they been properly instructed, would legitimately have yielded an acquittal. The defendants ought not bear the risk that the verdict derived from' such an error.
I should add 'that the decisions cited by my colleagues, of this court15 and of the Ninth Circuit,16 which (in whole or in part) disagreed with Judge Sanborn’s ’ ruling, were written when, previous to Kotteakos v. United States, 328 U.S. 750, 66 S.Ct. 1239, 90 L.Ed. 1557, those courts were misinterpreting the “harmless error” doctrine-The evidence in the present case does not suffice to render the errors “harmless”' within the ruling of the Kotteakos case-For it cannot be said, from merely reading the printed record, that the defendants are so unmistakably guilty that the jury would have been unreasonable if, guided by proper instructions,' it had found them innocent.17
. Jorgensen v. York Ice & Machinery Corp., 2 Cir.. 160 F.2d 432, 435: “The whole subject has been obscured, apparently beyond hope o£ clarification, by *22Lord Mansfield’s often quoted language in Vaise v. Delaval, [1 Term R. 11] that no evidence of misconduct was competent which came from the jurors themselves, although, as judges have repeatedly pointed out, it is impossible to see from what source better evidence could be obtained. On the other hand, it would be impracticable to impose the counsel of absolute perfection that no verdict shall stand, unless every juror has been entirely without bias, and has based his vote only upon evidence he has heard in court. It is doubtful whether more than one in a hundred verdicts would stand such a test *. *
See Sunderland, Verdicts, General and Special, 29 Vale L.J. (1920) 253: “The general verdict is as inscrutable and essentially mysterious as the judgment which issued from the ancient oracle at Delphi. * * * The court protects the jury from all investigation as fully as the temple authorities protected the priestess who spake to the suppliant votary at the shi-ine. * * * Since the case can ordinarily go to the jury only if a verdict either way is legally possible, whatever ,the jury does is presumed to be right, and this presumption excludes any inquiry from the jurors themselves. * * [In] the vast majority of cases the verdict is a complete mystery, throwing a mantle of impenetrable darkness over the operations of the jury. * * * The record must be absolutely flawless, but such a result is possible only by concealing, not by excluding, mistakes.”
See also Frank, Courts On Trial (1949) 115-116; cf. Johnson v. Hunter, 10 Cir., 144 F.2d 565, 567.
. I refer, of course, to the rule that, especially after a jury’s discharge, the testimony of no jurors will be received as to what went on in the jury room, except as to the grossest sort of misconduct. See, e. g., Jorgenson v. York Ice Machinery Corp., 2 Cir., 160 F.2d 432.
. As Judge Rossman observed, lawyers engaged in a jury trial frequently say they are not sure of the meaning of a substantive-rule instruction, until it has been written up, after the jury has been discharged. See Rossman, The Judge-Jury Relationship in the State Courts, 3: F.R.D. 98, 109. Cf. Chamberlayne, Evidence (1911) § 96.
. See, e. g., United States v. Ausmeier, 2 Cir., 152 F.2d 349, 357; United States v. Haug, 2 Cir., 150 F.2d 911, 914-915.
. See Skidmore v. Baltimore & Ohio R.Co., 2 Cir., 167 F.2d 54, note 25d.
. I do not mean that the upper courts are-wrong in reversing for the first kind of error. The theory of the general verdict requires them to assume that the jury (a) does understand the intricacies of the substantive rules set forth in the charge and (b) does apply those rules-For a criticism of general verdicts, in this and other respects, see Skidmore v. Baltimore & Ohio R. Co., 2 Cir., 167 F.2d 54.
. We cited the Gomila case in United States v. Modern Reed & Rattan Co., 2 Cir., 159 F.2d 656, 658, where wo held that a plain error should be noted on appeal although no objection had been made in the trial court.
. The “presumption of innocence” error, however, was the first error discussed in the Gomila opinion.
. Were it necessary, I would hold that the “presumption of innocence” error alone warranted reversal.
. See Pettine v. Territory of New Mexico, 8 Cir., 201 F. 489; Ayer v. Territory of New Mexico, 8 Cir., 201 F. 497.
. Of course, no invidious distinction of the present very able Judge John B. Sanborn is intended.
. “We all know, don’t we,” asked Judge McLellan, “that when we hear a criminal case tried, we get convinced of the guilt of the defendant or we don’t; and isn’t it enough if we say guilty or not guilty, without going through the form of making special findings of fact designed by the judge — unconsciously, of course — -to support the conclusions at which he has arrived?” Federal Rules of Criminal Procedure, With Notes and Proceedings of the Institute of Law of the New York University School of Law (1946) 173. See also Davis, Administrative Findings, Reasons and Stare Decisis, 38 Calif.L.Rev. (1950) 218, 221, 227.
. See Frank, Courts On Trial (1949) 183-185; Davis, 38 Calif.L.Rev. (1950) 218, 226.
. As to the desirability of special verdicts (i. e., fact-verdicts), see, e. g., Skidmore v. Baltimore & Ohio R. Co., 2 Cir., 167 F.2d 54.
. Chicago, B. & Q. Ry. Co. v. Babcock, 204 U.S. 585, 598, 27 S.Ct. 326, 329, 51 L.Ed. 636.
. United States v. Woods, 2 Cir., 66 F. 2d 262. There the acquiescence in the instruction was somewhat grudging and half-hearted.
. Louie Ding v. United States, 9 Cir., 246 E. 80.
. Of course, that is not the same as saying — and I certainly do not say — that there was such insufficient evidence of guilt that a jury, even if properly instructed, would have been unreasonable, had it found them guilty. For elaboration of this distinction, see my dissenting opinion in United States v. Antonelli Fireworks Co., 2 Cir., 155 F.2d 631, 642, 650-652. The decision there ante- ■ dated the Kotteakos decision.