Appellant Schiller Toto was convicted of distributing marihuana in violation of the Virgin Islands Controlled Substances Law, 19 V.I.C. § 604(a)(1). He testified at his trial and, in response to the prosecutor's questioning on cross-examination, admitted that he had previously pleaded guilty to the misdemeanor of petit larceny. Defense counsel timely objected. The government seeks to justify the questioning on the ground that the prior conviction was brought out to impeach appellant. In the alternative, it argues that any error was harmless because the trial judge charged the jury to disregard the testimony concerning the prior conviction. We hold that appellant's conviction for petit larceny does not come within the ambit of the rule, long established in this circuit, that a witness may be impeached by evidence of a prior conviction only if the conviction is for a felony or for a misdemeanor in the nature of crimen falsi. United States v. Evans, 398 F.2d 159, 164 (3d Cir. 1968); United States v. Montgomery, 126 F.2d 151, 155 (3d Cir.), cert. denied, 316 U.S. 681, 62 S. Ct. 1268, 86 L. Ed. 1754 (1942). We also hold that the error in allowing evidence of the prior conviction was not cured by the trial judge's charge, and requires that appellant's conviction be reversed.1

The facts are neither complicated nor controverted. During appellant's testimony at the trial the following colloquy occurred:

Q. (The prosecutor:) Sir, you stated that at one time you lost your job at Sparky's when you were arrested.

A. (Appellant:) Yes.

Q. Now, are you referring to this arrest?

A. No.

Q. A previous arrest?

A. Yes.

Q. With respect to that arrest, sir, were you convicted of a crime?

MR. ALEXIS: Objection.

THE COURT: I overrule the objection. The witness may answer.

A. I plead guilty.

Q. And do you recall what you pleaded guilty to?

A. Petit larceny.

There was no further questioning on the point.

Following the close of testimony, the district judge discussed his proposed jury charge with counsel. It was not disputed that petit larceny, under Virgin Islands law, is a misdemeanor. 14 V.I.C. §§ 2, 1084. The government contended, however, that the petit larceny conviction was properly admitted on the theory that any conviction, felony or misdemeanor, involving moral turpitude was admissible to impeach a witness. The district judge disagreed and, in the body of his jury charge, specifically instructed the jury as follows:

(Appellant) testified that he entered a plea of guilty to the offense of petit larceny. Here I would think that the court erred in permitting that testimony because it serves absolutely no useful purpose in this trial. Counsel, I have to say was quite correct when he said it is as though those words had never been uttered. He is not on trial for petit larceny. The fact that he was convicted on his plea of guilty for petit larceny has no bearing on whether he is to be believed by you in this case or not, because the law says you may discredit the testimony of a witness by a showing that he has been convicted of a felony or a lesser crime, including falsehood or false statements. Petit larceny is just not that.

Petit larceny is the theft of personal property of a value less than a hundred dollars, so his conviction for that is not to be taken into consideration by you. Even had he been convicted of a felony, it would not render him incompetent as a witness. So you will kindly disregard totally testimony that he pleaded guilty to petit larceny. Under no circumstances should you consider that fact.

On appeal the government appears to concede that a conviction of petit larceny in the Virgin Islands does not come within the rubric of a misdemeanor in the nature of crimen falsi, and presents alternative contentions. First, it asks this court to depart from its traditional rule and to adopt the rule of those circuits2 which are more liberal in the reception of impeachment evidence. Second, because the district court attempted to fashion a curative instruction, the government urges that the reception of the conviction evidence can be construed as harmless, and not reversible, error.

The term crimen falsi has roots in the common law doctrine that persons convicted of certain kinds of crimes were disqualified from testifying. See Ex parte Wilson, 114 U.S. 417, 5 S. Ct. 935, 29 L. Ed. 89 (1885); 2 Wigmore, Evidence § 520 (3d ed. 1940). While the doctrine of testimonial disqualification has withered from our law, the term crimen falsi has retained vitality in the context of impeachment. The established law in this circuit is that a witness may be impeached by evidence of a prior conviction only if it is for (a) a felony or (b) a misdemeanor in the nature of crimen falsi. United States v. Gray, 468 F.2d 257, 262 (3d Cir. 1972); United States v. Remco, 388 F.2d 783, 785 (3d Cir. 1968). The specific contours of crimen falsi are uncertain. Crimen falsi describes crimes involving, or at least relating to, communicative, often verbal, dishonesty; we have said that they are 'crimes which touch the question of the honesty of the witness,' United States v. Evans, supra, at 164. For our purposes here, we have no difficulty in accepting the government's formulation of the concept: 'Although the term 'Crimen Falsi' has been subject to many definitions, the generally accepted scope of the term would be crimes that are in the nature of perjury or subornation of perjury, false statement, criminal fraud, embezzlement, false pretense or any other offense the commission of which involves some element of deceitfulness, untruthfulness, or falsification bearing on the accused's propensity to testify truthfully.' Appellee's Brief at 4. Absent special circumstances,3 and as the district court so aptly put it: 'Petit larceny is just not that.'

The question remains whether we ought to embrace the government's invitation to change the law of the circuit in favor of a more liberal impeachment standard. We decline to do so. In 1968 we were invited to broaden our rule and we rejected that invitation. United States v. Evans, supra. Today we reaffirm that rejection, for the reasons Judge Freedman articulated succinctly:

This is not the universal rule and many jurisdictions permit impeachment by showing conviction of a crime which is not a felony if it involves moral turpitude. . . . (W)e believe it undesirable to broaden the rule to allow merely generally discrediting evidence to be introduced as any indication of testimonial unreliability. Whatever residuary relevance a witness's conviction of some other violation of law may have on the question of his credibility is far out-weighed by the prejudicial effect of discrediting him for conduct which is extraneous to the proceeding. In the practical reality of a trial the use of such convictions to aid the jury in determining the honesty or trustworthiness of a witness is a device for the far greater advantage of dishonoring him.

Ibid. at 164--65.

We are further motivated to retain our traditional rule by Congress' adoption of a closely parallel formulation in Rule 609(a) of the recently enacted Federal Rules of Evidence:

(a) General rule. For the purpose of attacking the credibility of a witness, evidence that he has been convicted of a crime shall be admitted if elicited from him or established by public record during cross-examination but only if the crime (1) was punishable by death or imprisonment in excess of one year under the law under which he was convicted, and the court determines that the probative value of admitting this evidence outweighs its prejudicial effect to the defendant, or (2) involved dishonesty or false statement, regardless of the punishment.

The Report of the House and Senate Conferees discloses precisely what is meant by 'dishonesty or false statement':

By the phrase 'dishonesty and false statement' the Conference means crimes such as perjury or subornation of perjury, false statement, criminal fraud, embezzlement, or false pretense, or any other offense in the nature of crimen falsi, the commission of which involves some element of deceit, untruthfulness, or falsification bearing on the accused's propensity to testify truthfully.

(quoted in K. Redden & S. Saltzburg, Federal Rules of Evidence Manual 192 (1975)). Thus, it would appear the congressionally enacted Federal Rule tracks the substance of our long-standing practice.

The new Federal Rules of Evidence took effect on July 1, 1975; appellant's trial took place in March, 1975. The new rules, therefore, can have no literal application to this appeal. Nevertheless, the enactment of Rule 609(a) buttresses our conclusion that the law of this circuit with respect to impeachment by prior convictions is not unduly restrictive. Furthermore, because trials after July 1, 1975, will be governed by the new statutory rules, a judge-made change in the law at this time would be not only improvident, but also utterly futile.

This does not end our inquiry. We must now decide whether the district court's later instruction to the jury cured the earlier, improper reception of evidence of the prior petit larceny conviction.

In United States v. Clarke, 343 F.2d 90 (3d Cir. 1965), the prosecutor elicited testimony that the defendant had participated in a crime unrelated to the offense for which he was being tried. Upon objection, the trial court immediately ordered the jury to 'disregard anything they have just heard in the last minute.' Ibid. at 91. We held that even such prompt remedial action was insufficient to wipe clean the stain of the improperly received evidence.

'The general rule is that if evidence which may have been taken in the course of a trial be withdrawn from the consideration of the jury by the direction of the presiding judge, that such direction cures any error which may have been committed by its introduction. . . . But yet there may be instances where such a strong impression has been made upon the minds of the jury by illegal and improper testimony, that its subsequent withdrawal will not remove the effect caused by its admission, and in that case the general objection may avail on appeal or writ of error.' Throckmorton v. Holt, 180 U.S. 552, 567, 21 S. Ct. 474, 480, 45 L. Ed. 663 (1901).

In Clarke, we concluded that even though the trial judge took immediate remedial action, the subsequent withdrawal of the evidence could not remove the deleterious effect caused by its introduction. We drew upon the teachings of Kotteakos v. United States, 328 U.S. 750, 764--65, 66 S. Ct. 1239, 90 L. Ed. 1557 (1946), where inadmissible evidence had been submitted to the jury:

To weigh the error's effect against the entire setting of the record without relation to the verdict or judgment would be almost to work in a vacuum. . . . In criminal causes (the) outcome is conviction. This is different, or may be, from guilt in fact. It is guilt in law, established by the judgment of laymen. And the question is, not were they right in their judgment, regardless of the error or its effect upon the verdict. It is rather what effect the error had or reasonably may be taken to have had upon the jury's decision. The crucial thing is the impact of the thing done wrong on the minds of other men, not on one's own, in the total setting. . . .

. . . (I)f one cannot say, with fair assurance, after pondering all that happened without stripping the erroneous action from the whole, that the judgment was not substantially swayed by the error, it is impossible to conclude that substantial rights were not affected. The inquiry cannot be merely whether there was enough to support the result, apart from the phase affected by the error. It is rather, even so, whether the error itself had substantial influence. If so, or if one is left in grave doubt, the conviction cannot stand.

Ibid. at 764--65, 66 S.Ct. at 1247 (emphasis added).

While it may be true that each case requires an ad hoc weighing of 'the error's effect', Clarke reaffirms this court's tradition of protecting the presumption of innocence that accompanies a defendant throughout the trial. The accused is not only presumed to be innocent of the crime with which he is charged, but our legal tradition protects him from the possibility of guilt by reputation. Evidence received by a jury in a criminal prosecution must pass stringent tests of competency and relevancy. A defendant's previous brushes with the law, in and of themselves, are simply irrelevant to his guilt or innocence of the crime with which he is charged. Evidence of previous convictions may be admitted only under very special circumstances, and subject to stringent instruction at the time of its reception. Such evidence may be received to show intent, plan, scheme, design or modus operandi, see cases collected in United States v. Klein, 515 F.2d 751, 755--56 (3d Cir. 1975), and United States v. Chrzanowski, 502 F.2d 573, 575 (3d Cir. 1974), and such evidence is admissible to impeach a witness' credibility. Even where such evidence is relevant for these purposes, the trial judge in the exercise of his discretion may exclude it if its probative value is outweighed by its prejudicial effects. But we unerringly refuse to admit evidence of prior criminal acts which has no purpose except to infer a propensity or disposition to commit crime. See Michelson v. United States, 335 U.S. 469, 475, 69 S. Ct. 213, 93 L. Ed. 168 (1948). When such evidence inadvertently reaches the attention of the jury, it is most difficult, if not impossible, to assume continued integrity of the presumption of innocence. A drop of ink cannot be removed from a glass of milk.

We are thus required to apply in this case the test of Kotteakos. We must decide 'whether the error itself had substantial influence (on the minds of the jury.) If so, or if one is left in grave doubt, the conviction cannot stand.' 328 U.S. at 765, 66 S.Ct. at 1248. But stating the test is easier than applying it. To decide whether the error did in fact substantially influence the jury, the court must, perforce, resort to probabilities.

Roger J. Traynor, the distinguished former Chief Justice of California, offers the wisdom of profound experience in approaching the basic problem from the viewpoint of harmless error. He suggests that an appellate court realistically has three options in choosing a standard for reviewing trial error. The reviewing court might affirm if it believes: (a) that it is more probable than not that the error did not affect the judgment, (b) that it is highly probable that the error did not contribute to the judgment, or (c) that it is almost certain that the error did not taint the judgment. Justice Traynor chooses the middle ground: 'Unless the appellate court believes it highly probable that the error did not affect the judgment, it should reverse.' R. Traynor, The Riddle of Harmless Error 35 (1970). See also Brown v. United States, 411 U.S. 223, 231--32, 93 S. Ct. 1565, 36 L. Ed. 2d 208 (1973); Schneble v. Florida, 405 U.S. 427, 430--32, 92 S. Ct. 1056, 31 L. Ed. 2d 340 (1972).

We believe that this formulation was the unstated premise of our disposition in Clarke, supra, and find its application most appropriate in the resolution of this appeal. We are impressed by the reasons Traynor sets forth for choosing the middle standard:

Any test less stringent entails too great a risk of affirming a judgment that was influenced by an error. Moreover, a less stringent test may fail to deter an appellate judge from focusing his inquiry on the correctness of the result and then holding an error harmless whenever he equated the result with his own predilections.

There are objections also to the two tests that are more stringent than that of high probability. If the test were the mere presence of error, appellate courts could reverse, as many did in the nineteenth century, for any error, no matter how trivial. The end result was public disaffection with the judicial process. Almost as stringent is a test that would require reversal unless the court was almost certain that the error did not affect the judgment. . . .

. . . Advocates of the test contend that it is an impossible task to determine whether error influenced the trier of fact. The argument runs that a judge, purporting to evaluate the influence of error as he reviews the record, is realistically driven to evaluate instead the correctness of the judgment, perforce acting as trial judge or jury. Hence, the only safe course, whenever the record would support a judgment either way, is to hold the error prejudicial.

Ibid. at 35--36 (footnotes omitted).

We cannot say that it is highly probable that the evidence of the prior conviction--even though, as in Clarke, it was subsequently ordered removed from the jury's consideration--did not contribute to the jury's judgment of conviction.

The judgment of the district court will be reversed and the proceedings remanded for a new trial.