(dissenting). I dissent and vote to affirm because, in my view, there was no legal error in the trial court’s evidentiary rulings requiring a new trial.
I agree with the majority that application of the 1984 *67amendment to Penal Law § 130.16 to this case does not violate the constitutional prohibition of ex post facto laws (US Const, art I, § 10). An ex post facto law is one "which punishes as a crime an act previously committed, which was innocent when done; which makes more burdensome the punishment for a crime, after its commission, or which deprives one charged with crime of any defense available according to law at the time when the act was committed” (Beazell v Ohio, 269 US 167, 169). Simply put, repeal of the corroboration requirement formerly contained in Penal Law § 130.16 accomplished none of these forbidden objectives (see, People v Spearman, 128 Misc 2d 112). Thus, the evidence was sufficient to sustain the conviction, even though corroboration was lacking with respect to many of the charges.
The majority nonetheless reverses the conviction, holding that the Trial Judge erred in two respects: (1) allowing the People to elicit testimony that defendant had committed similar uncharged acts of sexual abuse with Domenick M. and (2) precluding defendant from examining the two investigating officers about the manner in which the child-witnesses were first questioned. In my view, both of these rulings involved exercises of the trial court’s discretion with which this court cannot and should not interfere.
Turning first to the proposed testimony by the investigating officers, I agree that, because the evidence concerned the child-witnesses’ reason to fabricate, it was not collateral. That does not mean, however, that its exclusion was error.
A trial court may not exclude all such evidence, but it may, in its discretion, limit its quantity (see, People v McDowell, 9 NY2d 12, 15). Here, the witnesses that defendant sought to impeach with the proffered evidence admitted on cross-examination that they had at first told the investigating officers that defendant had not touched them but changed their stories after further questioning. This was the very evidence that defendant sought to elicit from the investigating officers.
Where the facts underlying a witness’s reason to fabricate are admitted by the witness, extrinsic proof of those facts may properly be excluded, in the court’s discretion, on the ground that it would be cumulative (People v Brooks, 131 NY 321, 326-327; McCormick, Evidence § 40, at 88-89 [3d ed]). Thus, the trial court acted within the bounds of its discretion in excluding the officers’ testimony and that ruling provides no basis for upsetting the conviction.
*68Neither was the admission of Domenick M.’s testimony reversible error. Of course, evidence of uncharged crimes must be examined carefully to ensure that it bears sufficient relevance to a material issue in the case and is not offered simply to demonstrate the defendant’s bad character. The rule of People v Molineux (168 NY 264) calls for such care. It does not call for a phobic reaction to all evidence of uncharged crimes. It is such a reaction, it appears to me, that blinds the majority to the obvious relevance of Domenick’s testimony in rebuttal to the defendant’s rumor theory.
We usually begin our analysis of such matters, appropriately, with the general rule that all evidence is admissible if it is relevant — that is, if it has any tendency in reason to prove any material fact (People v Lewis, 69 NY2d 321, 325). It is also elementary that, though relevant, evidence may nonetheless be excluded if its probative value is outweighed by other considerations, such as undue prejudice, delay, or needless presentation of cumulative evidence (see generally, McCormick, Evidence § 185, at 544-545 [3d ed]).
The Molineux rule is concerned with evidence that the defendant committed crimes for which he is not on trial. It is generally recognized that such evidence is not without probative value. In human reason, such evidence has a tendency to prove that the defendant is guilty of the crime charged, because one who has a record of bad acts is, according to human experience, more likely to commit another than is a person who has led a blameless life (1A Wigmore, Evidence § 58.1, at 1211, n 2 [Tillers rev ed]).
When offered for the sole purpose of demonstrating the defendant’s bad character or his propensity towards crime, however, evidence of uncharged crimes is per se inadmissible. That is the thrust of the Molineux rule. It is a rule of policy, not logic (People v Allweiss, 48 NY2d 40, 46). As Wigmore notes (op. cit., § 58.2, at 1212): "It may almost be said that it is because of the indubitable relevancy of specific bad acts showing the character of the accused that such evidence is excluded. It is objectionable not because it has no appreciable probative value but because it has too much. The natural and inevitable tendency of the tribunal — whether judge or jury — is to give excessive weight to the vicious record of crime thus exhibited and either to allow it to bear too strongly on the present charge or to take the proof of it as justifying a condemnation, irrespective of the accused’s guilt of the pres*69ent charge.” Thus, the Molineux rule is a policy-based exception to the general rule that relevant evidence is admissible. It is, in effect, a ready-made balancing of prejudice versus probative value. If the only relevance of the evidence is to show defendant’s propensity, no further inquiry is necessary — the unfair prejudice is too great.
None of this is new, of course. We have stated these principles repeatedly in recent years (see, e.g., People v Alvino, 71 NY2d 233, 241-242; People v Lewis, supra; People v Beam, 57 NY2d 241, 250-251; People v Ventimiglia, 52 NY2d 350, 359; People v Allweiss, 48 NY2d 40, 46-47, supra). But it is important to remember that the exclusionary rule of Molineux is itself an exception to the general rule. Thus, the so-called "exceptions” that Molineux recognized (intent, motive, identity, etc.) are not really exceptions at all. They simply illustrate instances in which proof of other crimes may be probative of something other than propensity and where, accordingly, the ready-made balance does not apply and the general rules of admissibility should prevail.
It may seem that this distinction is a matter of mere semantics, but it is deeper than that. It affects how the question is approached — whether the Molineux exclusionary rule is regarded as the starting point, presuming inadmissibility until the evidence can be pigeon-holed into one of the recognized "exceptions” or whether, instead, the analysis begins by asking whether the exclusionary rule applies at all. I believe that the latter approach is preferable, because it asks at the outset whether the evidence is relevant only to propensity. This, I think, more accurately reflects the policy considerations which gave rise to the Molineux rule. The policy judgment underlying the rule is not that all evidence of uncharged crimes is bad (if so, there would be no "exceptions”), but rather that a conviction based on the defendant’s propensity to commit crime must be avoided.
To be sure, extrinsic offense evidence carries "propensity” baggage with it, even when it is relevant to some material issue in the case, and it therefore harbors the potential for undue prejudice. But that problem is accounted for in the second step of the Molineux analysis — the discretionary balancing which weighs "the probative value and the need for the evidence against the potential for delay, surprise, and prejudice” (People v Alvino, supra, at 242). Actually, this second step is no different than the test by which all relevant *70evidence is measured (see, McCormick, op. cit., at 545; Fed Rules Evid, rule 403). That underscores the point that Molineux supplies an exception to the general rule. If the exception does not apply, the general rule governs, requiring the Trial Judge to balance probative value against prejudice, etc. The only difference in the Molineux situation is that the Judge will approach the normal balancing process with a heightened awareness of the unique kind of prejudice that extrinsic offense evidence can produce.
Thus, Molineux analysis should proceed in two steps. The first step is to determine whether the extrinsic offense evidence is relevant to something in the case other than the defendant’s propensity to commit the crime charged. If not, then the evidence is per se inadmissible. If the evidence is relevant to something else, however, then the inquiry proceeds to the second step or, more accurately, reverts to the general rule. The initial inquiry presents a pure question of law, fully reviewable by this court. The second step, however, is classic discretionary balancing.
At this stage, the Trial Judge must assess the proffered evidence in the context of the case as it has developed up to that point and as it is expected to develop, given the information gleaned from pretrial motions, conferences, opening statements and similar sources. With all this information in mind, the Trial Judge must determine, among other things, how probative of the point urged the evidence is, how necessary it is, to what extent the point is really disputed, whether the proof will cause delay, confusion or will unfairly surprise the defendant, and finally, to what degree the proof will result in undue prejudice of any kind, including the danger of a "propensity” conviction.
Having made these assessments and many others, the Judge must then weigh them all and reach a conclusion as to whether the evidence should be allowed in. This decision, like all matters of discretion, is entitled to our deference: "[A]l-though there may be room for a difference of opinion * * * the wide latitude and the broad discretion that must be vouchsafed to the trial judge, if he is to administer a trial effectively, precludes this court, in the absence of 'plain abuse and injustice’ [citation omitted], from substituting its judgment for his and from making that difference of opinion, in the difficult and ineffable realm of discretion, a basis for reversal” (People v Sorge, 301 NY 198, 202).
*71This two-step approach, one legal and one discretionary, is consistent with our recent decision in Alvino. We stated there that the question whether prior crimes evidence may be admitted, i.e., whether it is automatically barred under the Molineux rule, is a question of law. But, "[i]f the evidence of prior crimes is probative of a legally relevant and material issue before the court, and for that reason not automatically barred * * * admissibility turns on the discretionary balancing of the probative value and the need for the evidence against the potential for delay, surprise and prejudice” (People v Alvino, supra, at 242). Significantly, we upheld the conviction in Alvino because we found that the prior crimes evidence was probative of the defendant’s guilty knowledge. We refused to second-guess the Trial Judge’s ruling, once that legal determination was made, stating, "[although the courts below could have found the evidence * * * prejudicial and of questionable probative value * * * we find no abuse of discretion in the court’s reception of the evidence” (id., at 247).
Accordingly, the legal question before us is simply whether the evidence had some relevance to the case other than to show that the defendant was inclined to sexually abuse small children. I believe that it did. As the majority notes, defendant’s theory was that the accusations had started in October 1985 when Matthew L. was asked by his mother to defend a story he had told his little brother. Matt perpetuated the story rather than admit that he had lied to his brother; his mother spread the story by calling her neighbor; the neighbor’s son carried the story to the rest of the school; and ultimately a number of children got caught in the momentum.
Faced with this theory, the prosecutor sought to call one additional witness after the eight children named in the indictment had testified. This witness, Domenick M., had been in defendant’s class as a fourth-grader during the 1983-1984 school year. He testified that defendant had fondled him several times during that school year. He further testified that he had told his grandmother, with whom he lived, about the touching. His mother found out and wrote a letter to defendant and the school psychologist asking that the touching stop. Domenick testified that the defendant did not touch him any more after the letter was delivered.
This testimony, of course, is the evidence that gives rise to the supposed Molineux problem.
Assuming for a moment that Domenick’s testimony was *72probative only of defendant’s propensity to commit this kind of crime, I would have a difficult time believing that, in the context of this case, defendant’s conviction was the product of prejudice resulting from that evidence. All eight victims had already testified to dozens of incidents; if propensity figured in the jury’s deliberations it was because of the completely proper evidence of charged crimes, not because of some marginal prejudice contributed by Domenick’s testimony. Thus, even assuming that it was error to admit the testimony, I would have to conclude that the error was harmless.
I am not trying to downplay the effect of this evidence. I concede that it damaged defendant’s case and was, in that sense, prejudicial. The point is that the damage resulted from something other than its tendency to prove propensity. That suggests that the evidence must have been relevant to something else, as indeed it was.
Domenick’s testimony was offered to rebut the defendant’s theory that all of these claims were the product of rumors circulating through the school in October 1985. All of the other witnesses attended the school during that time and were, therefore, tainted, according to defendant’s theory. But Domenick had moved away in November 1984. He had no contact with the school or any of the students during the time when the rumors were circulating. He was, therefore, an untainted witness who had been subjected to the defendant’s sexual abuse and had reported it to his family long before the "snowball” began rolling. If this evidence was damaging, it was because it had the intended effect — it undermined the defendant’s theory.
The majority admits as much under the guise of harmless error analysis. The court states that "the Domenick M. evidence was highly prejudicial” because it was "qualitatively different from the testimony of the other boys, in that it was not subject to impeachment under the defense’s 'rumor’ theory.” (Majority opn, at 56.) This "qualitative difference”, however, is precisely what made the evidence relevant to something other than propensity and, therefore, admissible.
Molineux prohibits the following kind of reasoning: Defendant abused Domenick before; if he did it once, he will do it again; therefore he probably abused the other children (or deserves punishment anyway, whether he did or not). But the logic at work here is different and, I think, permissible under these circumstances: Several children claim that defendant *73abused them; one of them, Domenick, makes these allegations free of the influence that may taint the others; therefore, it is more likely that the others’ claims are not the product of such influence. In short, the evidence tended to show that the children’s charges were not simply the product of a rumor.
It was precisely on this basis that the Trial Judge allowed the evidence in. It should be noted that the Judge was not insensitive to the Molineux problem. He rejected other rationales offered by the People during the course of an extensive colloquy. And once he recognized that the evidence was relevant to rebut the defendant’s rumor theory, he properly moved to the second step of the analysis by weighing the probative value against the potential for unfair prejudice.
In sum, because the evidence was relevant to something other than propensity, the Molineux rule did not automatically require its exclusion. Therefore, the question of admissibility was a discretionary ruling. The Trial Judge exercised that discretion and did so very carefully, noting that very little prejudice of an impermissible sort would result, inasmuch as the jury had already properly heard of dozens of similar incidents. Thus, there was no abuse of discretion in admitting the disputed testimony.
Finally, the record indicates to me that the trial court did a commendable job the first time around in protecting the defendant’s right to a fair trial while at the same time allowing the People leeway, within proper bounds, to overcome the difficulties of proof inherent in prosecutions of this nature (see, People v Yonko, 34 NY2d 825, 827; People v Henson, 33 NY2d 63, 72-74).
Accordingly, because I perceive no legal error in the trial court’s evidentiary rulings, I vote to affirm.
Judges Kaye, Alexander and Hancock, Jr., concur with Judge Titone; Judge Alexander concurs in a separate opinion; Chief Judge Wachtler dissents and votes to affirm in another opinion in which Judges Simons and Bellacosa concur.
Order reversed, etc.