OPINION OF THE COURT
Lazer, J. P.The question here is whether in an action involving comparative fault there must be a new trial because the appellant’s request for a special verdict was rejected by the trial court with the result that a general verdict simply stating a sum of money was rendered in favor of the plaintiff. Additionally, if the form of the verdict was fatally defective, we must determine whether the verdict could be resuscitated by juror affidavits explaining the thought processes that brought it about.
In this dental malpractice action, the plaintiff is a young man with a history of rheumatic fever with complications. He alleged that he sustained serious injuries due to an infection that followed the extraction of a tooth without appropriate special precautions by appellant, who was not aware of plaintiff’s medical problems. Liability was premised on appellant’s purported failure to take a proper medical history, and resolution of this issue turned upon several disputed questions involving the circumstances under which a medical history form was completed. The case presented a clear issue of comparative negligence in which the jury was required to consider both the alleged inadequacies of the procedure followed by appellant dentist and the plaintiff’s failure to inform the dentist of his past medical problems.
The court properly instructed the jury on the comparative negligence issue, but denied appellant’s request that the jury be instructed to return a special verdict specifying the comparative percentages of negligence, if any, attributable to either party and the total amount of damages before any reduction was made on the basis of comparative negligence. The court based its decision on two grounds: (1) a stipulation *572the parties had entered into which the court interpreted as authorizing a general verdict, and (2) the belief that a special verdict was necessary only in bifurcated trials, which this was not. The jury was instructed to return a general verdict, and, in accordance with this instruction, it rendered a verdict of $1,055,700 for plaintiff by a vote of 5 to 1. Plaintiff then moved that the jury be asked what their comparative negligence findings were, but the motion was denied. Appellant subsequently moved to set the verdict aside on a variety of grounds, of which the only one requiring a discussion was the denial of the request for a special verdict. Responding to the motion to set aside, plaintiff submitted a joint affidavit by five jurors, four of whom had voted for the verdict and one of whom had opposed it, declaring that the jury had found that the total damages suffered by plaintiff were $1,173,000, that plaintiff was responsible for 10% of his injuries, and that they had accordingly deducted 10% from that amount to arrive at their verdict. The motion to set aside was denied and the appeal is from the judgment subsequently entered on the verdict. We conclude that the refusal to submit the issues in the form of a special verdict constituted an abuse of discretion fatal to the verdict and that juror affidavits are not available to cure a defect of that nature.
General verdicts have been described as being "as inscrutable and essentially mysterious as the judgment which issued from the ancient oracle of Delphi” (Sunderland, Verdicts, General and Special, 29 Yale LJ 253, 258). Whatever the acceptability of such determinations in less complicated times, in this era of complex issues and comparative fault, the necessity that appellate bodies be provided with some illumination of the jury’s rationale has been rendered quite acute. In providing that illumination, the special verdict has the advantage of offering a more precise definition of the jury’s finding (see, Lipscomb, Special Verdicts Under the Federal Rules, 25 Wash U LQ 185, 213-214), and it is for that reason that the appellate judiciary and legal commentators have repeatedly suggested that special verdicts or general verdicts with interrogatories be utilized in comparative fault cases (see, Collins v Weinberg, 88 AD2d 1037; Radtke v Yokose, 87 AD2d 220; Noga v Monroe Medi-Trans, 78 AD2d 988, appeal dismissed 53 NY2d 916; 1 NY PJI 2d 2:36, at 76 [1984 Supp]; see, Siegel, NY Prac § 399, at 522; Prosser, Comparative Negligence, 51 Mich L Rev 465; Timmons and Silvis, Pure Comparative Negligence in Florida: A New Adventure in the Common *573Law, 28 U Miami L Rev 737; Note, Judicial Adoption of a Comparative Negligence Rule in Illinois, 1967 U 111 LF 351). Indeed, in some jurisdictions, the requirement is statutory (see, e.g., Kan Stats Ann § 60-258a [b]; NJ Stats Ann § 2A:15-5.2) or by court rule-making (see, NM Sup Ct Order No. 8000 Misc, Uniform Jury Instructions—Civil, Judicial Pamph 18 [Mar. 30, 1981]), but judicial gloss has achieved the same result in a number of other jurisdictions (see, Placek v City of Sterling Hgts., 405 Mich 638, 275 NW2d 511; Lawrence v Florida E. Coast Ry. Co., 346 So 2d 1012 [Fla]).
Our own attitude toward verdicts unsusceptible to intelligent review was quite recently reflected in Mertsaris v 73rd Corp. (105 AD2d 67, 88), where we "call[ed] to the attention of the trial bench the clear necessity for requiring special findings by the jury in cases such as this * * * where the basis of the jury’s determination cannot otherwise be determined and disclosure of the correct basis will be necessary for adequate appellate review”. Our current holding was also foreordained in Davis v Caldwell (54 NY2d 176), where the Court of Appeals dealt with a somewhat similarly mystifying general verdict. There, the jury rendered a general verdict in a case where the plaintiff relied on five separate theories of liability and the record supported the sufficiency of only three of them. The Davis court concluded that it was not possible to determine on appellate review whether the jury’s verdict was based on one of the insufficient theories and a new trial was granted. A special verdict asking the jury’s judgment on each of plaintiff’s theories would have avoided the necessity for the retrial in Davis. In the instant case, it is impossible for us to determine whether the apportionment and assessment of damages was within permissible limits; the verdict does not indicate either the total amount of damages or the percentages attributable to each party. Furthermore, the issue of the form of the verdict was preserved for our review by appropriate request to the trial court.
The rule, then, is that the denial of a request for a special verdict or a general verdict with interrogatories in a case which presents an issue of comparative fault generally will constitute an abuse of discretion and require a new trial if the result is a simple money award in favor of one of the parties. When a litigant in a comparative fault case timely contests the amount of the award or the division of liability before either the trial court or an appellate tribunal, the court has a duty to review the verdict in the light of the trial record to *574determine the validity of the objection. This duty is patently incapable of performance unless the verdict is rendered in a form which clearly denominates the percentages of fault attributable to each party by the jury. Meaningful postverdict and appellate review of these determinations is impossible unless the court knows precisely what those determinations were, and this information cannot be gleaned from a bare general verdict. Here, for example, the general verdict of $1,055,700 could have been reached by the jury on any of a number of bases, ranging from a finding that plaintiffs total injuries were worth $1,055,700 and appellant was responsible for 100% of the injuries, to a finding that plaintiffs total injuries were vastly higher in value but all but a small percentage was attributable to plaintiff himself. While the trial record might well support any number of these possible calculations, it certainly could not support them all.
Unlike the trial court, we see no reason to confine this rule to bifurcated trials. Although the necessity for a precise apportionment may be more obvious in a bifurcated trial, it is equally great in a nonbifurcated trial because of the review considerations discussed above. Nor can we agree that the parties to this action stipulated to a general verdict. Although such a stipulation could be entered into, thereby voluntarily limiting the scope of subsequent review, here it was not. Our examination of the record indicates that the stipulation referred to by the trial court pertained to the question of whether the jury would be instructed to itemize damages, as is required by CPLR 4111 (d) in medical malpractice cases, and had no relevance to the question whether a general verdict or a special verdict was to be rendered.
A more troublesome issue is presented, however, with regard to the plaintiffs posttrial attempt to clarify the verdict with an affidavit by five of the jurors in which the apportionment determination allegedly made by the jury is explained. Were we to accept this affidavit as a replacement for the special verdict, so plaintiff argues, we would then be able to engage in an effective review of the verdict despite the erroneous ruling. We decline plaintiffs invitation to indulge in a type of harmless error analysis based on juror affidavits.
It is hardly new law—although it was recently reiterated— that absent exceptional circumstances a juror’s testimony on affidavit may not be used to attack a jury verdict (see, Kaufman v Lilly & Co., 65 NY2d 449, 460; People v De Lucia, 20 NY2d 275). A long time ago, the Supreme Court concluded *575that posttrial challenges of such a nature "would * * * make what was intended to be a private deliberation, the constant subject of public investigation—to the destruction of all frankness and freedom of discussion and conference” (McDonald v Pless, 238 US 264, 267-268). Although the rule against jury impeachment can cause an unjust result at times, it is necessary to prevent "the posttrial harassing of jurors” and the "chaos” which would result from the instability of verdicts (see, People v De Lucia, supra, at p 278).
While the effort here is not a challenge to the verdict, of course, and it is true that juror testimony may be used in certain rare instance to correct what amounts to a ministerial error in reporting a jury verdict (see, Dalrymple v Williams, 63 NY 361; Rose v Thau, 45 AD2d 182, 184-185; Wirt v Reid, 138 App Div 760), plaintiff’s thrust is not directed at mere correction of a recording error (see, Pache v Boehm, 60 AD2d 867). The verdict as recorded is the verdict determined by the jury in accord with the charge. There is nothing to correct and it must stand or fall on its own. Nor is the case similar to those in which a juror was allowed to explain some purported misconduct (see, e.g., Dennis v Tishman, 141 Misc 847, revd on other grounds 235 App Div 722), for here there has been no allegation of misconduct outside the jury room. Rather plaintiff seeks to explore the deliberative process of the jury itself and this he may not do.
To allow the use of evidence such as this could subject jurors to continuing harassment at the hands of attorneys and investigators for both parties in an attempt to obtain the details of what is best shrouded in secrecy. Jurors could easily be subjected to persuasive and confusing interrogatories which might well result in conflicting or ambiguous responses from various jurors. Any discrepancy among the responses given by the several jurors would necessitate some sort of factual inquiry, and even a unanimous response in the face of unilateral questions could justifiably lead to a request for a hearing from the other party (see, McCabe v McMahon, 31 AD2d 734). Moreover, were we to accept such juror affidavits as an alternative to a special verdict in order to render harmless the refusal to require a special verdict, simple justice would then require us to use that same juror testimony to determine whether the verdict should be set aside on the merits, thereby effectively using juror testimony to attack the verdict. We deem it far better to avoid this morass by adhering to the basic rule against the admission of such evidence.
*576Accordingly, the judgment appealed from should be reversed, and a new trial granted, with costs to abide the event.
Gibbons, Brown and O’Connor, JJ., concur.
Judgment of the Supreme Court, Dutchess County, entered February 16, 1984, reversed, on the law, and new trial granted, with costs to abide the event.