FRANK, Circuit Judge.
In opinions holding that the findings of trial judges were not "clearly erroneous" within the meaning of Rule 52(a),1 Federal Rules of Civil Procedure, 28 U.S.C.A., we have often stressed the importance of a trial judge's advantage over us when he saw and heard the witnesses as they testified. We have pointed out our inability to appraise the cogency of demeanor evidence, lost to us because it cannot be captured in the witness' words as recorded on paper.2 In so holding, we may perhaps, at times, have overlooked distinctions described in United States v. United States Gypsum Co., 333 U.S. 364, 394-396, 68 S.Ct. 525, 542, 92 L.Ed. 746.
There the Court made it clear that Rule 52(a) merely adopted the equity practice then prevailing in the federal courts.3 The Court said a finding of a trial court, if it be by a judge, "is `clearly erroneous' when although there is evidence to support it, the reviewing court on the entire evidence is left with the definite and firm conviction that a mistake has been committed." The Court pointed out that a trial judge's finding does not have the dignity which jury verdicts derive from the Constitution nor the dignity which some statutes confer on findings of some administrative agencies.
We think that United States v. Yellow Cab Co., 338 U.S. 338, 70 S.Ct. 177, did not modify what was said in the Gypsum case. In Gypsum, an anti-trust action, the defendants testified orally that they had not acted in concert, and the trial judge so found. The Supreme Court rejected that finding, remarking that "the witnesses denied that they * * * had agreed to do the things which in fact were done. Where such testimony is in conflict with contemporaneous documents we can give it little weight, * * *." In Yellow Cab, where the Court affirmed the trial judge's findings against the government in an antitrust action, the oral testimony was not incompatible with inferences which could reasonably be drawn from the documentary evidence.4
In the light of the Gypsum case, we may make approximate gradations as follows: We must sustain a general or a special jury verdict when there is some evidence which the jury might have believed, and when a reasonable inference from that evidence will support the verdict, regardless of whether that evidence is oral or by deposition. In the case of findings by an administrative agency, the usual rule is substantially the same as that in the case of a jury, the findings being treated like a special verdict.5 Where a trial judge sits without a jury, the rule varies with the character of the evidence: (a) If he decides a fact issue on written evidence alone, we are as able as he to determine credibility, and so we may disregard his finding.6 (b) Where the evidence is partly oral and the balance is written or deals with undisputed facts, then we may ignore the trial judge's finding and substitute our own, (1) if the written evidence or some undisputed fact renders the credibility of the oral testimony extremely doubtful,6a or (2) if the trial judge's finding must rest exclusively on the written evidence or the undisputed facts, so that his evaluation of credibility has no significance.6b (c) But where the evidence supporting his finding as to any fact issue is entirely oral testimony, we may disturb that finding only in the most unusual circumstances.6c
It follows that evidence sufficient to support a jury verdict or an administrative finding may not suffice to support a trial judge's finding.7 So in the instant case, perhaps, on the record evidence, we might have affirmed a jury's verdict or an administrative agency's finding in plaintiff's favor. That, however, we need not decide. For here the finding is that of a trial judge, and the evidence consists in large part of facts neither side disputes, in circumstances such that the trial judge's evaluation of credibility becomes unimportant. In short, for reasons we shall state, the undisputed facts are such that we have a "definite and firm conviction" that the trial judge was mistaken in finding that Orvis and Mrs. Orvis "each pursued an independent course" in creating the 1934 trusts, and that no reciprocity was intended. We therefore hold that finding "clearly erroneous," and hold, rather, that each of those trusts was made in consideration of the other.
In so holding we assume that, because of the "evanescent factor which cannot come before us"7a — i. e., the demeanor of the witnesses — the trial judge fully believed everything they said. Even so, there is nothing in the testimony which in any manner offsets what we believe to be the virtually irresistible inference drawn from the undisputed facts. To offset that inference, the trial judge relied on no positive testimony that Mr. and Mrs. Orvis acted independently but relied merely on negative testimony as to the absence of an expressed intention to act reciprocally, i. e., that "no witness even intimated that the decedent acted with Mrs. Orvis' intention in mind or she with his." But those same witnesses, unequivocally and without contradiction by any other witnesses, testified to the following facts which we think can lead to but one reasonable conclusion: Mr. Orvis, in the Fall of 1934, spoke to his son, Warner, about creating a trust. Subsequently, in November 1934, Mrs. Orvis also spoke to Warner about creating a trust. He told his mother that Mr. Orvis had been looking into the subject of trusts and had already talked to his lawyer, Merritt, about his own proposed trust. Mr. Orvis thereafter talked to his wife about her proposed trust. On the advice of Warner, Mrs. Orvis also consulted Merritt, concerning her trust, in November, at the Orvis apartment; Mr. Orvis was then in the apartment, and may "have stepped into the room * * * and stayed for a minute." There was not, Merrill testified, "any secretiveness by her" with respect to her intention to create a trust. The only testimony on which the trial judge relied, or could have relied, to counterbalance the foregoing is that of Mr. Merritt as to the reasons respectively given to him by husband and wife for the creation of their respective trusts. Not only were those respective reasons strikingly similar but none of them sufficed or purported to explain why each of the trusts set up a life estate; nor did the expression of those reasons at all negative the existence of an intent to make the trusts reciprocal. The finding of an absence of such an intention must, then, depend not on an inference drawn from anything positive in the testimony concerning statements of intention made by Mr. and Mrs. Orvis, but on an inference from their conduct. And that inference, in turn, must rest on a belief in the purely chance concurrence of several events, although the coincidental occurrence of those events would ordinarily be highly improbable.8 Such a belief ought not to be the foundation of a trial judge's finding on a fact issue, in favor of that side having (like plaintiffs here) the burden of proof as to that issue, unless the purely chance character of those events is positively confirmed by clear evidence.8a There is no such confirmatory evidence here.
In rejecting the trial judge's finding, we do not, therefore, reject his estimate of the credibility of any orally testifying witnesses. We differ from him solely about the inference which may reasonably be drawn from the evidence, assuming without question that the witnesses he credited spoke the truth.9 In a case like this, on an issue relating to intention, what we said in E. F. Drew & Co. v. Reinhard, 2 Cir., 170 F.2d 679, 684 is applicable: "When an appellate court is faced with that question, it is in substantially as good a position to answer it as the trial judge, provided it accepts as true all the oral testimony as we do here, so far as it was relevant. We yield to findings of fact, so far as those parts of the evidence which cannot come before us may have controlled their decision; we must assume that these evanescent factors may have been persuasive, unless what does come before us rationally forbids the conclusions, no matter what the unknown factors were. After giving them every possible probative force they might have, our problem therefore becomes the same as that before the trial court."
Since we find that reciprocity was intended, the situation is just as if Mr. Orvis had been the settlor of the 1934 trust created by his wife.9a That trust must, then, be regarded as one he made in which he retained a life income. His estate is therefore subject to tax on the value of the corpus of that trust, i. e., $250,466.46.10
We have not discussed the "contemplation of death" defense because, were we to sustain it, the amount of the judgment resulting from our sustaining the "reciprocal trust" defense would not be affected. For only the value of the corpus of the wife's trust was subjected to tax by the Commissioner, and the time when he could assess an additional tax on plaintiffs (were we to sustain the "contemplation of death" defense) has long expired.
Reversed and remanded.
CHASE, Circuit Judge (dissenting).
I cannot agree with my brothers that these findings are "clearly erroneous." The trial judge saw and heard witnesses who testified concerning matters which had a direct tendency to explain plausibly how these two trusts might have been created when, and as, they were, without any so-called "reciprocal" taint. That judge, as his reasoned opinion shows, believed them and gave effect to their testimony in weighing all the relevant facts proved. His findings support the judgment and those findings, as this record unmistakably shows, are the result of a careful study of evidence which he might reasonably believe, and believing, might reasonably use as the basis for drawing the inferences he did.
This is a typical instance for the application of Civil Rule 52(a). Though trial judges may at times be mistaken as to facts, appellate judges are not always omniscient.
I would affirm.