OPINION OF THE COURT
Chief Judge Lippman.At the conclusion of a long and complex trial during which plaintiff sought to prove that defendant physicians had committed malpractice by failing to detect, diagnose and properly treat a tumor situated in her pelvis, the jury returned a verdict evidently exonerating defendants while purporting to award damages “for the plaintiff” in the amount of $1.5 million. Shortly after the verdict, plaintiff requested that the jury be polled to ascertain whether “each juror consents this is the verdict as read by the foreperson.” The request was denied as “unnecessary” and the jury discharged. The trial court would, however, subsequently acknowledge that it had erred in denying the poll and, on that ground, grant plaintiffs posttrial motion to set aside the verdict and declare a mistrial (2007 NY Slip Op 34280[U]). In the ensuing litigation, defendants have conceded *173that it was error not to poll the jury, yet urge that harmless error analysis may save the verdict.
In the decision and order now before us, brought up for review by plaintiff’s appeal as of right, pursuant to CPLR 5601 (d), from an Appellate Division order affirming the dismissal of her remaining claims (50 AD3d 319 [2008]), a divided Appellate Division panel agreed with defendants, and denied the previously granted motion to set aside the verdict. While the entire panel acknowledged that plaintiff had been entitled to have the jury polled and even that the entitlement was “absolute,” the majority viewed the error as one of form only, since, in its estimation, “the objective facts set forth amply demonstrated] that polling the jury would not have resulted in a different verdict” (Duffy v Vogel, 49 AD3d 22, 25 [1st Dept 2007]). This view was, at bottom, premised upon the manner in which the verdict had been rendered: each member of the jury had signed the verdict sheet in response to each of the answered interrogatories, the jury’s responses to the interrogatories had been uniformly unanimous, and during the foreperson’s recitation of the verdict in open court, no juror cried out in protest. The dissenters countered that “[i]t is simply impossible, short of clairvoyance, to decide that no juror in this case would change his or her mind upon being polled facing the litigants in open court” (id. at 32), and, accordingly, would have held that the trial court’s failure to honor plaintiff’s absolute right to have the jury polled required that the verdict be set aside, as the trial court had done.
We now reverse.
At common law, it was recognized that although jurors had agreed upon a verdict within the confines of the jury room and announced their verdict in open court, they might yet be examined by poll to determine whether they remained wedded to their verdict, and if it appeared that the verdict had been due to mistake or partiality, the jurors would, prior to the recording of the verdict, have an opportunity to “go together and consider the better of it, and alter what they have delivered” (2 Hale, History of the Pleas of the Crown, at 299-300 [1800 ed]). That jurors “should be enabled to avail themselves of the locus penitentiae, and correct a verdict which they have mistaken, or about which, upon further reflection, they have doubt” (Blackley v Sheldon, 7 Johns 32, 33-34 [1810]) has been recognized in the vast majority of jurisdictions, and the polling of jurors in open court has been viewed as essential to the provision of such *174a locus, since, without the device, the reservations of individual jurors about a verdict to which they may have assented in the enforced privacy of the jury room under misapprehension, pressure from fellow jurors or out of sheer exhaustion, would likely never gain timely expression.
In New York, we have long recognized that affording jurors a last opportunity individually to express agreement or disagreement with the reported verdict, is, when requested by a litigant, indispensable to a properly published, and thereby perfected, verdict:
“It is a general rule, that no verdict is of any force but a public verdict given in open court; until that is received and recorded there is no verdict. When the jury come to the bar to deliver their verdict, all or any of them have a right to dissent from a verdict to which they had previously agreed. (Root v. Sherwood, 6 John. 68.) A verdict is not recognized as valid and final until it is pronounced and recorded in open court: the jury may change their mind and disagree as to their verdict after they have pronounced it in open court before it is received and entered on the minutes. After a verdict is rendered or announced and before it is entered, the jury may be examined by the poll, if the court please, and either of them may disagree to the verdict. ([.Blackley] v. Sheldon, 7 John. 32.) The expression in the last case, ‘if the court please,’ would seem to imply that the polling of the jury was in the discretion of the court; but in the case of Fox v. Smith, (3 Cowen, 23,) and Jackson ex dem. Fink and others v. Hawks, (2 Wend. 619,) it is decided to be the absolute right of a party to have the jury polled on their bringing in their verdict, whether it be sealed or oral, unless he has expressly agreed to waive that right” (Labar v Koplin, 4 NY 547, 550-551 [1851]).
That a verdict may not be deemed “finished or perfected” until it is recorded, and that it may not be validly recorded without a jury poll where one has been sought* (see Warner v New York Cent. R.R. Co., 52 NY 437, 442 [1873]), have been uncontroversial propositions. Although we have not had the *175opportunity recently to reconsider them, they have been applied over the years with axiomatic force by New York’s intermediate appellate courts (see Brigham v Olmstead, 10 AD2d 769 [3d Dept 1960] [“Not until inquiry is made of the whole jury, not merely the foreman, as to their verdict, and it is duly entered by the Clerk is the verdict complete. Until it is thus announced there is no verdict”]; Muth v J & T Metal Prods. Co., 74 AD2d 898 [2d Dept 1980], lv dismissed 51 NY2d 745 [1980] [reinstatement of verdict following appellate denial of previously granted motion to set aside the verdict not possible because, in light of the trial court’s denial of defendant’s request to poll the jury, there was no valid verdict to reinstate]; see also Luppino v Busher, 119 AD2d 554, 556 [2d Dept 1986]; Ricchueto v County of Monroe, 267 AD2d 1012 [4th Dept 1999]; Matter of National Equip. Corp. v Ruiz, 19 AD3d 5, 12-14 [1st Dept 2005]). No cogent, principled argument is made for their revision here.
Inasmuch as, under New York law, the honor of a request for a jury poll is a necessary condition of a “finished or perfected” verdict, it follows that in this State’s courts the failure to poll a jury may never be deemed harmless. Harmless error analysis is a judicial device employed to sustain an already perfected verdict, not to perfect a verdict in the first instance.
A verdict in a jury trial is emphatically not a judicial construct. Indeed, it is hard to know upon what empirical basis a verdict untested in open court by direct inquiry of the individual jurors might be judicially deemed to be the final expression of the jury’s true intention. Here, it is important to understand that the relevant question is not whether the trial evidence was sufficiently persuasive to impel a hypothetical reasonable juror to vote in favor of the announced verdict, but rather whether each juror chosen by the parties to hear the case would, upon reflection, publicly affirm that the verdict agreed to in the jury room was the one he or she actually intended. The exercise of individual conscience involved is one whose outcome defies prediction.
Certainly, the prediction called for may not be based, as it was by the Appellate Division, on the circumstance that jurors have signed a verdict sheet. The proper publication of a verdict in open court, so long deemed essential to assure the integrity of the verdict, is not to be cast aside as a mere formality on the theory that jurors are prospectively bound to act in accordance with their verdict sheet signatures. Indeed, in Root v Sherwood (6 Johns 68, 68 [1810]), where the argument, here exactly *176reiterated, was made that “[t]he signing of the verdict is an express assent by each juror to the verdict, and is equivalent to a polling of the jury,” the court responded,
“There is no verdict of any force but a public verdict, given openly in court; until it was received and recorded it was no verdict, and the jury had a right to alter it as they may a private verdict. The previous agreement, that the jury might seal up their verdict, did not take away from the parties the right to a public verdict, duly delivered. There being then no legal verdict in this case, a new trial must be awarded.” (Id. at 69; see also Dore v Wyer, 1 AD2d 973, 974 [2d Dept 1956] [denial of plaintiff estate administrator’s request for a jury poll as “unnecessary” upon the ground that 10 jurors had signed the verdict sheet required reversal and a new trial since “(t)he administrat(or) had an absolute right to have the jury polled, and the denial of that right was serious error”].)
Nor can there be any conceptually viable contention that the outcome of the public polling sought was foregone since the jurors in the present case privately signed the verdict sheet in numerous places, particularly since, had they followed instructions and concluded their deliberations after finding that defendants had not been negligent, their signatures would have been significantly less numerous.
Finally, we do not think it sensible to expect that a juror would in open court spontaneously pipe up his or her disagreement with an announced verdict. Jurors who have been pushed to a verdict about which they have serious reservations are not likely moments later in the solemn and intimidating atmosphere of the courtroom attending the announcement of the verdict to feel free to express their reservations unless it is made clear that it is permissible to do so and an opportunity is provided.
It is doubtless true, as defendants point out, that the absolute right of a party to have a jury polled is rooted in the common law, and not in statutory or constitutional enactment. It is also true that in some jurisdictions, albeit a minority, there exists no such right, and in others the right is not viewed as absolute, but as one which may in limited circumstances be denied without impairing the verdict. None of this argues persuasively for treating the right of a litigant to have a jury polled as less than absolute in this jurisdiction. Jury polling ordinarily entails little *177additional burden in the conduct of a trial, yet is demonstrably efficacious in assuring that the trial’s outcome is in fact the true verdict of the particular jury chosen by the parties to hear the case. A jury verdict, once properly delivered may, of course, be subsequently sustained against claims of error upon evidentiary analyses that courts are equipped to perform. No court, however, may claim to know each juror’s conscience so as to retrospectively offer assurance that the verdict was in its initial iteration what its authors had actually intended. Only timely inquiry of jurors will disclose whether their announced verdict truly expresses their will, and it is for this reason, and not out of unreasoned devotion to antique forms, that the common-law insistence upon jury polling has persisted. Harmless error analysis in this context would amount to no more than a speculative exercise, impermissibly substituting the judgments of judges for those that would have been made and disclosed by jurors had their verdict been properly pronounced in open court.
Even if we were writing upon a clean slate, which we are not, we would not concur in the dissent’s view that it would be a good idea to treat the denial of a litigant’s right to have the jury polled as a kind of potentially harmless error. Contrary to the premise of the dissent, jury polling has never been justified on the ground that there is a high probability that it will uncover disparity between the announced verdict and what the jurors intended. Its justification rests instead upon the right of a litigant to a public verdict demonstrably that of the particular jurors chosen in the case. Long, and we think indispensable experience, has shown that that basic entitlement, so closely enmeshed with and protective of the right to trial by jury, may not be deemed secured in any individual case simply upon the foreperson’s announcement of a verdict—even one multiply subscribed—and that the claim to be able reliably to distinguish in hindsight the case in which the failure to honor the entitlement was or was not harmless is highly suspect and should not be adopted as a basis for law. While it is, from defendants’ perspective, doubtless nightmarish to face a new trial of this matter, sparing them a new trial upon the approach advocated by the dissent is an even less attractive option, prospectively involving courts too confident of their ability to discern what is in a juror’s mind in the unwitting validation of false verdicts and the concomitant deprivation of true verdicts.
Accordingly, the order of the Appellate Division should be reversed, with costs, and a new trial ordered.
*178Smith, J. (dissenting). The Court decides today that a failure to poll a civil jury on request can never be harmless error. This result is compelled by no statute and supported by no binding precedent. It results in a gross injustice in this case, and will no doubt do so in some future cases. And it is highly unlikely to do any practical good. Because I see no justification for this decision, I dissent.
Doctors James Vogel and Allan Jacobs were sued for medical malpractice. The case was tried for two weeks, and a jury decided in the doctors’ favor, filling out an 11 page verdict sheet containing 21 interrogatories. Five inapplicable questions were not answered, but the answers to the remaining 16 were all consistent, and every one of the 16 answers was signed by every juror. All the answers were unanimous, though a 5-1 verdict would have been valid (CPLR 4113 [a]). The foreperson announced all the answers in open court, and declared as to each one that the jury had answered unanimously.
After this, plaintiffs counsel asked for a poll of the jury, and the trial court mistakenly denied the request. For this reason, the majority holds, the whole trial was a waste of time and Doctors Vogel and Jacobs must undergo another one, running the risk that the second jury will disagree with the first.
The majority’s rationale for this nightmarish result seems to have three parts. One is verbal: the right to have a jury polled on request has been described as “absolute” (Labar v Koplin, 4 NY 547, 550 [1851], quoted in majority op at 174). The second is formal: “the honor of a request for a jury poll is a necessary condition of a ‘finished or perfected’ verdict” (majority op at 175). And the third claims to be practical: it is impossible to know, the majority says, what the result of a poll will be unless the poll is conducted (majority op at 175-177). I find all these reasons unconvincing.
We have indeed said that the right to the poll of a jury on request is “absolute.” But it is clear that we meant only that a denial of the request is always error—we have not answered the question whether the error may be harmless. The context in which we used “absolute” in Labar confirms this. After quoting from a case that had said “the jury may be examined by the poll, if the court please,” we went on to say:
“The expression in the last case, ‘if the court please,’ would seem to imply that the polling of the jury was in the discretion of the court; but in the case of Fox *179v. Smith, (3 Cowen, 23,) and Jackson ex dem. Fink and others v. Hawks, (2 Wend. 619,) it is decided to be the absolute right of a party to have the jury polled . (4 NY at 550-551.)
Thus, the right to a jury poll is “absolute” only in the sense that it is not a matter of discretion with the court. The “absolute” nature of the right does not imply that the denial of the right can never be harmless error. Indeed, most cases of harmless error involve the denial of rights that are “absolute” in that sense. If a court has discretion to grant or refuse a party’s request, the denial of it is ordinarily not error at all, and there is no occasion to consider whether the error is harmless.
The assertion that, where a request for a jury poll is not honored, the verdict is not “finished or perfected” is an abstraction, incapable of proof or disproof. It is true because the majority says it is true, and for no other reason. The formal logic on which it seems to rest is not even consistently applied. If the majority were to take that logic to its limit, it should hold that the poll is essential, whether a party requests it or not. If the verdict is not “finished or perfected” once it is read in open court, how can a party’s failure to request a poll finish or perfect it? Or, if it can, why cannot the harmlessness of an error in refusing a request finish or perfect it equally well? Whether to apply harmless error analysis should not turn on the answer to such scholastic riddles, but—since this is a situation in which our conclusion is compelled by no statute and no precedent—on whether permitting harmless error analysis would be a good idea.
The majority suggests it would not be a good idea because it is always possible that the error was not really harmless. The majority says, in substance: You never know until you ask whether there is a juror (though in this case it would be inconsequential unless there were two) who would not say “yes” to the question: “Is this your verdict?” According to the majority, “The exercise of individual conscience involved is one whose outcome defies prediction” (majority op at 175).
I will make a prediction: In almost every case, the poll will confirm the verdict as announced by the foreperson. There are experienced trial lawyers who have never seen a verdict upset by a jury poll—and others who have seen it once, and tell the story to the end of their days. The possibility that that rare event would occur in this case—a case in which the jury’s verdict was detailed, clear and specifically endorsed by every juror’s *180repeated signature, and in which defendants even had a vote to spare—is vanishingly small.
Of course, I cannot say I am absolutely certain that the trial court’s error here was harmless—but that can never be said, of any error. The sort of fanciful possibility that the majority relies on exists in every case, and if it were given the kind of weight the majority gives it here, the harmless error doctrine would not exist. We have many times found errors to be harmless where the chance that the error determined the result was significantly greater than it is in this case.
There is a fourth possible reason that might support the result the majority reaches—one the majority does not mention, but one that gives me a bit more pause. It is the possibility of abuse. As I have pointed out, a failure to poll a jury will usually— indeed almost always—in fact be a harmless error. Is there not then a risk that trial judges will be tempted to reject requests for jury polls, knowing that the harmlessness of the error will protect them from reversal?
I do not think so. I might feel differently if a jury poll were a burdensome procedure, or if there were some other reason trial judges might seek to avoid it—but it is short and simple, and I do not think that a jury poll is ever likely to be denied for any reason except the reason we have here: an honest mistake. Still, as a precaution, I might consider limiting the application of the harmless error doctrine to cases where there is a particularly strong reason to think the jury poll would not have changed the result. This is certainly such a case.
So far as I know, we have never before today held that a particular kind of error in a civil case cannot be harmless. That does not mean that no such errors exist. No doubt, for example, the wrongful denial of a trial by jury in a civil case would not be subject to harmless error analysis. Still, I find it puzzling indeed that, by virtue of today’s decision, we have afforded a sacrosanct status never before conferred on any right of a civil litigant to the quasi-medieval ritual of the jury poll.
Judges Ciparick, Graffeo, Read, Pigott and Jones concur with Chief Judge Lippman; Judge Smith dissents in a separate opinion.
Order reversed, etc.
Although the early cases, such as Labar v Koplin (supra), in fact, required an express waiver if the jury was not to be polled, the entitlement is now deemed waived in the absence of a request.