This is an appeal from a denial of a petition for a writ of habeas corpus brought by a state prisoner to challenge his conviction essentially on the ground that his constitutional rights were denied by the prosecutor’s inclusion of improper racial remarks in the summation.
Petitioner was found guilty by a jury of criminal sale of a controlled substance (heroin) in the second degree, N.Y. Penal Law § 220.41, and sentenced on June 2, 1976 in the New York Supreme Court (Monroe County) to a term of eight years to life. The Appellate Division affirmed without opinion, People v. McFarland, 59 A.D.2d 1067, 399 N.Y.S.2d 828 (4th Dept. 1977), and the New York Court of Appeals denied permission to appeal. People v. McFarland, 43 N.Y.2d 836, 402 N.Y.S.2d 1042 (1977). A petition for a writ of habeas corpus was denied on May 30, 1978 by the United States District Court for the Western District of New York (Hon. Harold P. Burke, Judge).
At trial, the State’s case depended almost entirely on the testimony of Patricia Dorman, a Rochester undercover police officer. She testified that she purchased $450 worth of heroin from petitioner in the bedroom of a second-floor apartment. She recognized petitioner as a person she had known in high school and had since seen occasionally. The defense case depended entirely on the testimony of petitioner’s friend, Isaac Singletary. He testified that he and petitioner had come to the apartment house to see two prostitutes with whom they had earlier made a date. According to Singletary, he and petitioner went upstairs to the second-floor apartment together with a Puerto Rican man who had entered the building just after they did. Singletary further testified that he waited in a front room, petitioner used the bathroom, and the Puerto Rican man entered the bedroom along with a Black woman (Dorman) and another Puerto Rican man. Singletary heard a brief discussion in the bedroom, after which the Black woman left the building. Singletary said petitioner emerged from the bathroom, they both asked the Puerto Ricans where the girls were, and when they were told there were no girls, both left. The inference from Singletary’s testimony was that Dorman had purchased narcotics from the first Puerto Rican male, and not from petitioner.
Not surprisingly the summation of defense counsel contended vigorously that Officer Dorman’s version was false and Singletary’s version was true.
In the course of the prosecutor’s summation the following occurred:
*416Mr. Pappalardo [the prosecutor]: . The officer herself being, by the book/1! a young woman, black woman, by the way this Defendant is black also.
Mr. King [defense counsel]: Objection to the racial connatation [sic] of individuals.
The Court: Of course I’ll instruct the jury now they shall not take into consideration to any extent and use that against any individual race, color, creed makes no difference whatsoever. You may continue.
Mr. Pappalardo: I’ll also instruct the jury—
Mr. King: Objection.
The Court: Yes, that’s improper. You cannot instruct the jury.
Mr. Pappalardo: Excuse me, I seem to be interrupted before I finish my statement because the interruption is what the People believe the People’s position, as in every single case, it makes no difference what color the Defendant is. I’ll finish my point. Don’t you convict anyone on color or race. It makes no difference. It makes no difference to me. I hope it makes no difference to Mr. King and anybody else, but the fact is that Officer Dorman is black and the Defendant is black. That’s a fact. That’s a fact like you consider any other fact. If she’s lying she’s lying against a member, a person that [sic] is black.
Mr. King: Objection.
The Court: Overruled.
Mr. Pappalardo: That is a proper consideration for you to examine, to think about and now she’s lying against another black person. You think about it because that’s what Mr. King is telling you that she’s lying. Someone she knows and that’s [sic] a member of her own race. You use your common sense to think about that.
(Tr. 369-71).
The prosecutor thus urged the jury to credit Officer Dorman’s testimony on the theory that the probability of truthfulness was increased by the circumstance that a Black person was testifying against another Black person. The trial judge’s overruling of defense counsel’s objection assured the jury that the Court accepted the propriety of this argument.
In United States ex rel. Haynes v. McKendrick, 481 F.2d 152 (2d Cir. 1973), this Court ruled that racial remarks in a prosecutor’s summation can constitute a violation of a defendant’s right under the Due Process Clause to a fair trial. Judge Oakes’ opinion drew upon the line of fair trial cases beginning with Moore v. Dempsey, 261 U.S. 86, 43 S.Ct. 265, 67 L.Ed. 543 (1923), and the line of equal protection cases beginning with Strauder v. West Virginia, 100 U.S. 303, 25 L.Ed. 664 (1879), and. noted that when racial prejudice is injected into a criminal trial, “the due process and equal protection clauses overlap or at least meet . .” 481 F.2d at 159 (footnote omitted).
The Office of the Monroe County District Attorney, which has intervened to uphold petitioner’s conviction, contends that the racial remarks of the prosecutor, while “imprudent” (Intervenor’s Br. 12), were not racial slurs. The remarks in Haynes involved racial slurs, and the District Attorney argues that only remarks of that category are appeals to racial prejudice that can render a conviction invalid under the Fourteenth Amendment.
Neither Haynes nor the lines of authority on which it drew set the constitutional limits for improper prosecution argument at racial slurs. Race is an impermissible basis for any adverse governmental action in the absence of compelling justification. When a prosecutor’s summation includes racial remarks in an effort to persuade a jury to return a guilty verdict, the resulting conviction is constitutionally unfair unless the re*417marks are abundantly justified. To raise the issue of race is to draw the jury’s attention to a characteristic that the Constitution generally commands us to ignore. Even a reference that is not derogatory may carry impermissible connotations, or may trigger prejudiced responses in the listeners that the speaker might neither have predicted nor intended.
This is not to say that every race-conscious argument is impermissible. Indeed, in Haynes, defense counsel, with apparent court approval, had attacked identification testimony on the ground that the eyewitness, being White, was unlikely to be able to discern distinguishing characteristics of the face of the criminal, who was Black. 481 F.2d at 160. These remarks were race-conscious, but race-neutral, since presumably an argument could be made with equal force that a Black eyewitness would have difficulty discerning the features of a White criminal. And there is some basis for accepting the validity of both contentions. Chance, Goldstein & McBride, Differential Experience and Recognition Memory for Faces, 97 J. Soc. Psych. 243 (1975); Malpass, Racial Bias in Eyewitness Identification, 1 Personality & Soc. Psych. Bull. 42 (1974); Malpass & Kravitz, Recognition for Faces of Own and Other Race, 13 J. Personality & Soc. Psych. 330 (1969); Shepherd, Deregowski & Ellis, A Cross-Cultural Study of Recognition Memory for Faces, 9 Int’l J. Psych. 205 (1975). But given the general requirement that the race of a criminal defendant must not be the basis of any adverse inference, any reference to it by a prosecutor must be justified by a compelling state interest. The issue in this ease is whether the racial remarks, even if not overt racial slurs, were sufficiently justified to be countenanced.
In People v. Hearns, 18 A.D.2d 922, 923, 238 N.Y.S.2d 173, 174-75 (2d Dept. 1963), the Appellate Division reversed a conviction because, as in this case, the prosecutor had urged the jury to credit the testimony of Black police officers partly on the basis of their membership in the same racial group as the defendant. That argument, the Court concluded, is predicated on a false and illogical premise and constitutes an appeal to racial prejudice. Some analysis is warranted to explore that conclusion. Since the prosecutor in this case did not spell out his reasoning, one is left to consider what possible lines of reasoning might support a valid argument that the testimony of Officer Dorman is entitled to some degree of enhanced probability of truthfulness because her race is the same as the defendant’s.
The analysis may begin by recognizing the obvious fact that from any group, racial or otherwise, some persons called as witnesses will testify helpfully to a defendant and some will testify accusingly.2 It may well be that testimony is more frequently helpful than accusing when the testimony is given within group lines (witness and defendant members of the same group) than when testimony is given across group lines (witness and defendant not members of the same group). Two circumstances would seem to support this thesis. First, alibi and character witnesses normally come from those with whom the defendant spends time, and there is a reasonable likelihood that members of his group are a disproportionately large segment of his friends and associates. Victims and by-stander witnesses who testify accusingly are less likely to be drawn disproportionately from the defendant’s group (though for some crimes, victims may be). Second, when testimony is given within rather than across group lines, the incidence of helpful testimony may be further increased because of lying. Of course, of all witnesses who testify helpfully, some percentage are lying, reflecting at least whatever extent mendacity is prevalent in the total population. But in the category of helpful testimony within group lines, an extra increment of lying might occur because of the tendency of some small percentage of the members of any group to *418lie in an effort to be helpful to a fellow member of their group.3
The prosecutor in this case might have believed that both of these circumstances operate to make the incidence of helpful testimony higher within group lines than across them, and conversely that the incidence of accusing testimony is lower within group lines than across them. In other words, if 100 instances are randomly selected where a witness and a defendant are members of the same group, and another 100 instances are randomly selected where a witness and a defendant are not members of the same group, the percentage of witnesses giving accusing testimony may well be lower in the first 100 than in the second 100.
If this is what the prosecutor believed (and was urging the jury to believe), the premise might be sound, but the conclusion — that Officer Dorman’s accusing testimony is more likely to be credible because given within group lines rather than across them — is completely illogical. All the premise indicates is that testimony within group lines, compared to testimony across group lines, is less likely to be accusing. But this premise provides no basis whatever for reaching any conclusion as to the likelihood that accusing testimony within group lines is credible. Specifically, it provides no logical basis for concluding that accusing testimony within group lines is more likely to be truthful than accusing testimony across group lines. Reduced frequency of occurrence is no indicator of credibility. The pertinent analysis is not a comparison of the incidence of accusing testimony within and across group lines, but a comparison of the truthfulness of accusing testimony within and across group lines.
As with witnesses giving helpful testimony, some percentage of all witnesses giving accusing testimony are lying. But when accusing testimony within and across group lines is compared, another circumstance may well be at work that might affect the likelihood of credibility. This is prejudice— the hostility of some few members of any group against members of a different group to such a degree that they are willing to accuse falsely: It may well be that prejudice increases the probability of lying when accusing testimony is given across group lines to a greater degree than when accusing testimony is given within group lines. To whatever extent this is so, the converse effect would be to increase the probability of truthfulness when accusing testimony is given within group lines to a greater degree than when accusing testimony is given across group lines.
If the prosecutor was basing his argument on this reasoning, his argument might have some slight logical validity,4 but is *419nonetheless constitutionally impermissible for two reasons. First, the degree of validity is highly uncertain and may well be extremely slight. It is one thing to permit race-conscious arguments to be made when comparing the reliability of facial identifications within and across racial lines, but quite another to permit such arguments with respect to comparative rates of false accusations. While there is some reason to believe that identifications are more reliably made within racial lines than across them, there is no comparable basis for confidence in comparisons about false accusations. A race-conscious argument is not constitutionally permissible unless the basis for it has a sufficiently high degree of reliability to warrant the risks inevitably taken when racial matters are injected into any important decision-making. A major risk here is that the jury will totally fail to follow the narrow reasoning process that lends any possible validity to the prosecutor’s argument and instead simply be influenced adversely to the defendant because of the prosecutor’s reference to his race. A further risk is that the jury will wrongly conclude that the argument draws its validity from the previously discussed premise concerning the reduced incidence of accusing testimony within group lines. If the jury accepts that reasoning, it will be accepting an argument that is, as previously pointed out, completely illogical.
There is a second reason for disallowing the argument, to whatever extent it might have logical validity. The increased credibility of accusations within group lines compared to accusations across group lines results, if at all, from the degree to which some members of one group are so prejudiced against another group that they are willing to make false accusations. When a prosecutor argues for enhanced likelihood of credibility because the accusation is within group lines, he is asking the jury to give his witness some extra credit simply because the witness is lacking the prejudice that might prompt a witness of another group to accuse falsely. The credibility of the state’s witnesses should depend on an assessment of many pertinent factors, but the state should not be entitled to have its witness’s credibility enhanced simply because they are not members of a group that might be prejudiced against the defendant.
This point can best be appreciated by contemplating the minor premise the prosecutor would have to explain to the jury in order to develop his reasoning fully. If a Black officer is logically entitled to any enhanced credibility when testifying against a Black defendant, it can only be because White police officers are more likely than Black police officers to give false accusing testimony against a Black defendant. If the difference is true at all, it presumably is true for all police departments, including Officer Dorman’s. If to the prosecutor’s knowledge some White officers of her department would falsely accuse a Black defendant, such an outrageous circumstance surely cannot be a constitutionally valid basis for enhancing the credibility of this witness for the prosecution.
Furthermore, to whatever extent prejudice increases the incidence of false accusing testimony across compared to within group lines, this circumstance supports an argument that is not race-neutral. As the intervenor acknowledged at oral argument, it is inconceivable that a prosecutor would argue to a jury that one reason to believe the accusing testimony of a police officer is that both the officer and the defendant are White.
These considerations lead to the conclusion that the prosecutor’s argument is constitutionally impermissible. It invokes race for a purpose that is either illogical or of very slight and uncertain logical validity, and does so at a distinct risk of stirring racially prejudiced attitudes. The evils of racial prejudice lurk too frequently throughout the administration of criminal justice. They must be condemned whenever they appear. The Constitution forbids the racial remarks in the summation that preceded petitioner’s conviction.
The District Attorney contends that if constitutional error occurred, it was harmless beyond a reasonable doubt. See Chapman v. California, 386 U.S. 18, 87 S.Ct. *420824, 17 L.Ed.2d 705 (1967). As both the prosecutor and defense counsel recognized, the case against the defendant depended on the credibility of Officer Dorman as opposed to the defendant’s witness, Singletary. The corroborating details of the prosecution’s case confirmed a narcotics sale and petitioner’s presence at the apartment house, facts he did not contest, but did not significantly prove that petitioner was the seller. When the District Attorney contends that the case against the petitioner was strong, he is assuming the truth of Officer Dorman’s testimony. A constitutional error that taints the proper assessment of her credibility cannot be considered harmless.
The judgment is reversed and the cause remanded with instructions to issue a conditional writ discharging the petitioner from custody unless the District Attorney within sixty days files with the District Court a statement of intention to retry petitioner and thereafter proceeds to retrial with reasonable promptness.
. Later in his summation the prosecutor explained that in characterizing Officer Dorman as someone who goes “by the book,” he meant that she follows correct procedure for an undercover officer in insisting that she has the purchased narcotics in her hands before she pays out any money. (Tr. at 371-72).
. The percentage of each group that testifies accusingly may not be the same, but the differences do not matter to this analysis.
. There is no need in this case to decide whether the incidence of such helpful lying within group lines is of sufficient certainty and extent and sufficiently race-neutral to be the basis of a permissible argument by a prosecutor that helpful defense testimony should be disbelieved because a defendant and his witnesses are members of the same group. Whatever the infirmities of such a summation, see Allison v. State, 157 Tex.Cr. 200, 248 S.W.2d 147 (1952); Arnold v. State, 96 Tex.Cr. 214, 256 S.W. 919 (1923), it poses issues different from those that arise in this case where the prosecutor argues that accusing testimony should be believed because the defendant and a prosecution witness are members of the same group. To the extent that the jury may have taken the prosecutor’s remark as a veiled suggestion to disbelieve Singletary’s helpful testimony, the vice of injecting improper racial considerations into an assessment of Officer Dorman’s accusing testimony was compounded, as Judge Kearse’s concurring opinion points out.
. In some contexts, another phenomenon may be at work that would lessen or even totally undercut any logical validity the prosecutor’s argument might otherwise have. This is the tendency among members of some groups to be so personally embarrassed by wrongdoing by members of their group that they are overly quick to condemn. To whatever extent that phenomenon might increase the rate of mistaken accusations within group lines compared to accusations across group lines, it would be expected to manifest itself only in circumstances where the witness was perceiving ambiguous events and too quickly and erroneously concluding that wrongdoing was occurring. It is difficult to imagine that this phenomenon would have any bearing on the relative incidence of deliberately false accusations within group lines compared to deliberately false accusations across group lines.