OPINION OF THE COURT
Wachtler, J.During a trial a juror conducted a "test” which bore on the credibility of a key prosecution witness. The issue on this appeal is whether that conduct and the communication of the results of the "test” to other jurors during their deliberations, constitutes reversible error.
The defendant Brown was convicted after trial of one count of robbery, first degree, and two counts of reckless endangerment, first degree, in connection with the robbery of a Queens candy store. The defendant was accused of being the driver of the escape car; his two codefendants were alleged to have conducted the actual robbery and to have fired shots at pursuing police.
Brown’s first trial ended in a hung jury. The evidence at the second trial, with which we are concerned here, was as follows:
About 6:45 p.m. in late October of 1974, three plainclothes policemen were cruising in an unmarked General Motors van, when they observed a car slowly moving in the opposite direction with its three occupants looking into store windows. The police made a U-turn and followed. When the car stopped at a red light, the police pulled the van alongside the driver’s car on the driver’s side.
While the two policemen in the front seat of the van looked straight ahead so as not to arouse suspicion, the third, Ford, who was seated in the second seat of the van, looked at the driver of the car for the duration of the red light, a maximum of 37 seconds. The police then followed the car for some 20 blocks until it parked and two men, neither of whom was the defendant, got out and entered a candy store. Ford followed on foot, and when the two suspects exited from the store and *392began to run, Ford signaled his partners in the van to pursue them.
The suspects re-entered the escape car and, followed by the van, engaged in a running gun battle that ended when the escape car rammed another vehicle. Ford’s partners testified that they actually saw only two persons run from the escape car, and neither of them was the defendant. All three suspects escaped on foot. The defendant was arrested later that evening after police traced the escape car to him. The defendant, however, had reported the car stolen about 7:30 p.m. that evening, approximately a half hour after the robbery.
At trial, Ford, who testified that the defendant was the driver he observed at the stoplight, was the only witness who placed Brown at the scene of the crime, and the defense concentrated heavily on the accuracy of that identification. At issue were Ford’s opportunity to observe, including the length of the observation, the lighting conditions, and whether the angle of sight was such that a passenger in the second seat of a van could observe the face of the driver of an adjacent car which is approximately three feet lower, as well as Ford’s personal powers of observation and general credibility.
After the jury returned its guilty verdict counsel for the defendant moved to set it aside having learned that one juror had conducted a "test” of the visibility afforded by her own Volkswagen van, which is of a different design from the General Motors van used by the police, and had informed the jury panel of her "test” and her conclusion that it was possible to see the face of the driver of an adjacent car. In other words, her "test” had supported the accuracy of Ford’s testimony.
At a hearing on the motion to set aside the verdict, both sides were afforded an opportunity to present evidence and examine witnesses. The juror in question acknowledged conducting the "test”. She testified, in part, that the exercise was not preplanned and did not occur at the site described in the trial testimony, but that she had been curious about whether it was possible to see the driver of an adjacent car. To satisfy this curiosity, when a similar occasion presented itself, she sat in the second seat of her van to look out the window at an adjacent driver.
The trial court concluded that the results of the "test” had been conveyed to other members of the jury before the verdict was reached, but that the test was not preplanned or ar*393ranged, had not occurred at the scene of the incident in question, and that there was no evidence that the jury had been affected by an inherently prejudicial outside influence.
The Appellate Division questioned whether the evidence given at the hearing proved that the "test” results had been told to the jury before the verdict was reached, but concluded that there was no reversible error in any case as the juror who conducted the "test” merely "spontaneously drew on a common everyday situation with which she was familiar, in order to test the detective’s veracity.”
At the outset we must consider whether the juror’s conduct in question is to be considered improper. Although the juror’s actions have been characterized as a "test”, merely to label them does not necessarily resolve the issue. With some logic it is contended here, as both the trial court and the Appellate Division found, that the juror’s "test” was no more than the application of everyday perceptions and common sense to the issues presented in the trial.
Jurors, of course, do not live in capsules. It is not expected that their selection as jurors should cripple their cognitive functions. On the contrary, the leavening accomplished by the application of a lay jury’s collective intelligence and experience to the tasks of sifting evidence and reaching a verdict is justly regarded as a hallmark of our judicial system.
Generally, a jury verdict may not be impeached by proof of the tenor of its deliberations, but it may be upon a showing of improper influence (Parker v Gladden, 385 US 363). Improper influence, of course, embraces not merely corrupt attempts to affect the jury process, but even well-intentioned jury conduct which tends to put the jury in possession of evidence not introduced at trial (United States v Beach, 296 F2d 153). Thus, in People v Crimmins (26 NY2d 319) and People v De Lucia (20 NY2d 275), this court set aside verdicts where jurors had made unauthorized visits to locations described in trial testimony, finding the jury conduct under those circumstances to be inherently prejudicial to the defendant’s historic rights of confrontation and cross-examination of witnesses against him.
The inherently prejudicial character of the jury conduct in De Lucia (supra) arose from the fact that the jurors actually re-enacted the crime; in Crimmins it inhered in the jury conduct because the credibility of a key witness turned on her ability to see and hear the events to which she testified, and *394the unauthorized viewing might well have permitted the jury to resolve questions about lighting in the area. Of course, not every misstep by a juror rises to the inherently prejudicial level at which reversal is required automatically. Because juror misconduct can take many forms, no ironclad rule of decision is possible. In each case the facts must be examined to determine the nature of the material placed before the jury and the likelihood that prejudice would be engendered.
In the case now before us, the "test” conducted may not properly be classed as an application of everyday experience. This is demonstrated by the fact that the juror felt compelled to conduct the exercise at all, despite the fact that she owned and regularly operated a van, i.e., what she could see from the second seat was not within her experience. At the hearing she testified, in part, that "I had to make the test to verify it, and I was still unsure at the conclusion, and the summation convinced me on evidence it was possible.”
We need not consider now what the result might be if the juror had been placed by pure coincidence in a position to make her observations, because in the case at bar the conduct was conscious, contrived experimentation.
Second, the "test” was directly material to a point at issue in the trial. At stake was whether Detective Ford, from where he sat, could have observed the driver of an adjacent car. Clearly, testimony that a passenger in the second seat of a van could see the face of the adjacent driver made Ford’s account more believable. The juror testified that "I had questioned angles and the light in my own mind, and I could see that it was plausible” and that her test "was a piece in a puzzle”.
Third, the juror’s "test” created a substantial risk of prejudice to the rights of the defendant by coloring the views of the other jurors as well as her own. Ford’s credibility was not only at issue, but was crucial to the prosecution’s case against the defendant. As Ford was the sole witness to directly link the defendant with the crime, the case necessarily failed if his identification could not be believed. As Ford’s credibility was bolstered by the juror who was generally experienced with vans, it seems reasonable to assume that the jury might give her view extra weight.
The respondent urges that because the "test” did not relate specifically to this defendant and was not conducted at the location described, it should be considered an application of *395general experience. However, the impact of the "test” could only have been aggravated by these circumstances. It put before the jury evidence that could never have been admissible at trial because it was not comparable in location, lighting, or type of van.
Finally, having found that the juror’s "test” and its relation to the rest of the panel constituted error prejudicial to the rights of the defendant, we find that the error cannot be excused. This court in People v Crimmins (36 NY2d 230, 241), has carefully set out the various tests to be applied to determine whether a violation of a Federal constitutional right, a State right, or a general fair trial right, may be considered harmless error. However, we need not parse the error to fix its precise classification. Because the degree to which the jury process was compromised in this case can neither be tolerated nor be deemed harmless.
Inasmuch as we find that the juror’s conduct of a "test” requires reversal here, we do not pass on the prosecutor’s cross-examination of the defendant in such a way as to imply a consciousness of guilt from his exercise of the right to counsel, a matter conceded to be error and one which might well constitute independent grounds for reversal.
The order of the Appellate Division should be reversed and a new trial ordered.