On the night of January 17, 1958, two men committed a robbery in a liquor store operated by Patrick and Francis O ’Brien. The appellant Jackson and another defendant were indicted for the crime and at the trial, held some eight months after its commission, both O’Briens identified Jackson as one of the robbers. Significantly, however, Patrick O’Brien had not picked Jackson out of a line-up which he had viewed seven months earlier and only two weeks after the robbery.
Whether the appellant was one of the robbers was, so far as he was concerned, the only issue in the case, and the evidence *144that Patrick O’Brien, who identified him at the trial, had failed to select him from the line-up was unquestionably a factor of prime importance to the defense. However, the effect of this earlier failure to identify him was dissipated, completely destroyed, by a statement of the prosecuting attorney during his summation. Without any evidence whatsoever to support him, he blandly assured the jury that Patrick had “ recognized ” the appellant in the line-up and that, although “he didn’t ” pick him out, there was good reason for this, adding,
‘ ‘ I will be more than happy to tell you * * * why after this trial is over.”
This was a serious impropriety, an error both egregious and inexcusable, whose potential for prejudice was incalculable. (See, e.g., People v. Lovello, 1 N Y 2d 436, 438-439; People v. Tassiello, 300 N. Y. 425, 430; see, also, People v. Ochs, 3 N Y 2d 54, 57-58.) On the only issue in the casé, identification, the district attorney proceeded to place his own veracity and official position in the scales against the accused. Only one of the three trial witnesses had been requested to view the line-up and he had failed to pick the appellant from it. This might have given the jury pause as to the accuracy and validity of the trial identification and created considerable doubt as to whether the appellant was one of the robbers. But what possible chance could he have had once the prosecutor had assured the jurors that O’Brien had in fact recognized the appellant and that he would tell them the reason why the witness did not pick him out, after the trial was over?
It is urged that this impropriety was ‘1 harmless ’ ’ because it had not been repeated, because it had been “ confined to a single isolated instance ” in a case where there was £ £ overwhelming evidence of guilt” (opinion of Dye, J., p. 146).
We cannot agree with this evaluation of the proof. It is quite true that the two O’Briens testified at the trial that they were sure that the appellant was one of the men who held them up; but, as we have already observed, Patrick, the only witness who viewed a line-up, and that shortly after the commission of the crime, admitted that he had failed to pick out the appellant Jackson. It is also true that another witness, a patrolman, testified that Jackson was the man whom he had pursued on the night of the robbery; but he acknowledged that his only view *145of the suspect was “ a glimpse * * ' * for not more than two or three seconds ’ ’ and, in addition, he admitted that he had expressed a doubt, immediately after the chase, as to his ability to identify that person. This is hardly ‘‘ overwhelming ’ ’ proof on the issue of identification.
However, even if the proof could be so considered, we do not see how we may treat the impropriety complained of as harmless error. Although section 542 of the Code of Criminal Procedure directs an appellate court to disregard “technical errors” not affecting “ substantial rights ”, it could never have been the legislative design that a court should regard errors as technical, however grave they may be, upon the hypothesis that, in any event, the jury correctly decided the case. Observing that “ Such a course would tend to abolish ‘ all forms of law taught by experience to be necessary to the protection of the innocent ’ ’ ’, this court recently wrote, 1 ‘ convincing though the evidence of guilt may seem to be, we could affirm only if we were to announce a doctrine that the fundamentals of a fair trial need not be respected if there is proof in the record to persuade us of defendant’s guilt. We are not prepared to announce such a doctrine.” (People v. Mleczko, 298 N. Y. 153, 162-163.)
Nor is it possible to hold that the instructions given by the court erased the impression made on the jurors’ minds or eliminated the harm thereby occasioned. (See People v. Carborano, 301 N. Y. 39, 42; People v. Robinson, 273 N. Y. 438, 445-446.) If instructions could have such an effect, then, we have declared, “the prosecution would be in a position to violate the rules of fair conduct with impunity, secure in the thought that the verdict, if one of guilt, would not be upset as long as the judge simply directed the jury to disregard what had occurred.” (People v. Carborano, 301 N. Y. 39, 43, supra.)
Although we may not be able to say that the prosecutor’s remark on summation actually brought about the appellant’s conviction, there can be little doubt that it must have had a tremendous impact upon the jurors in their deliberations. Some indication of this is found in the fact that the jury did not return its verdict of guilt against the appellant until more than eight hours after it had retired and in the interim, between retiring and verdict, it had actually reported to the court that, *146while a verdict had been reached as to his codefendant, ‘1 agreement ” on a verdict with respect to the appellant himself “ looks doubtful
In the light of all of the circumstances, we may not, therefore, treat the error as technical or harmless and disregard it.
The judgment should be reversed and a new trial ordered.