(dissenting). For the first time, we are allowing a theory of third-party culpability to go forward based on an offer of proof consisting entirely of hearsay. Defendant’s offer of *142proof fell far short of that presented in People v Primo (96 NY2d 361 [2001]), the only case before today in which this Court has held that a trial court abused its discretion as a matter of law in precluding such evidence. This decision imposes an undue restraint upon the discretion of trial courts to weigh the probative versus prejudicial value of third-party culpability evidence and make a reasoned decision regarding its admissibility. I respectfully dissent.
Howard Gombert’s alleged statement to Joseph Santoro was required to surmount a threshold evidentiary hurdle by satisfying the four-pronged test of the hearsay exception for declarations against penal interest. The first three elements of that test were met. I disagree with the majority, however, that Gombert’s alleged statements to Santoro satisfied the final prong — that defendant presented “supporting circumstances independent of the statement itself ... to attest to its trustworthiness and reliability” (People v Settles, 46 NY2d 154, 167 [1978]). Even under the more lenient standard applied when a defendant seeks to introduce a declaration against interest as exculpatory evidence (see People v Soto, 26 NY3d 455, 462 [2015]), the fourth factor of the test, which is “the most important aspect of the exception” (.People v Shabazz, 22 NY3d 896, 898 [2013] [internal quotation marks omitted]), was not met.
Defendant’s proffered evidence of third-party culpability fell into four categories. First, defendant sought to introduce Gombert’s alleged statements to Santoro. Second, defendant sought to introduce other evidence allegedly connecting Gombert to the crime, which consisted of the statements of Anita Albano and Gombert’s former girlfriend. Gombert’s former girlfriend told police that shortly after the victim’s disappearance, Gombert told her that he had given the victim a ride to some unspecified place on some unspecified day around the time that the victim went missing. Gombert did not tell her that he had committed a crime or that he was involved in the victim’s disappearance. Further, although Albano testified at the pretrial hearing, her testimony did not inculpate Gombert, inasmuch as she testified that the red car she saw the victim enter on the day of her disappearance was not the car of Gombert’s former girlfriend, and that Gombert was not the young man whom Albano saw driving the red car. Third, defendant offered statements from certain witnesses who told police that Gombert knew the victim and that the victim had *143offered to babysit for Gombert. Finally, defendant sought to introduce “reverse Molineux” evidence regarding Gombert’s history of sexual assaults.
Every piece of evidence that defendant presented in an attempt to demonstrate the reliability of Gombert’s statements to Santoro was hearsay. Gombert’s statement to his former girlfriend that he had given the victim a ride was hearsay, as was his former girlfriend’s statement to police itself. Statements from witnesses asserting that Gombert knew the victim and that she had offered to babysit were hearsay. The proffered “reverse Molineux” evidence was also hearsay. Defendant offered that proof through a defense investigator, an obviously interested party. Defendant did not offer any statements, sworn or unsworn, from Gombert’s victims themselves.
To be clear, I am not suggesting that a defendant seeking to introduce evidence of third-party culpability must present all evidence in support of that application in admissible form at the pretrial stage. Here, nothing was in admissible form, except for the testimony of Albano, which did not support defendant’s application.
The quality of the proof the defendant offers must be considered in evaluating whether the trial court abused its discretion as a matter of law in precluding the evidence. Defendant did not submit an affidavit from Gombert’s former girlfriend, or any of the witnesses who defendant asserted could establish Gombert’s relationship with the victim. Defendant did not present an affidavit from any of Gombert’s victims to support his reverse Molineux application, and he further failed to present an affidavit from the defense investigator who allegedly interviewed these victims. In short, defendant asked the trial court to admit Gombert’s statement to Santoro not based upon “sufficient competent evidence independent of the [statement itself] to assure its trustworthiness and reliability” (People v Brensic, 70 NY2d 9, 15 [1987]), but rather based upon questionable hearsay.
Defendant has acknowledged that he will be required to offer the live testimony of Gombert’s victims and any other appropriate witnesses upon retrial. In his pretrial application, however, defendant did not indicate that any of these witnesses were willing to testify at trial. Faced with the unreliable and speculative nature of the evidence defendant offered in an attempt to support the trustworthiness of Gombert’s statement *144to Santoro, the trial court did not abuse its discretion as a matter of law in concluding that defendant failed to satisfy the fourth prong of the Settles test (see Settles, 46 NY2d at 167).
Even assuming Gombert’s alleged statements to Santoro qualified as a declaration against penal interest, I nevertheless cannot conclude that the trial court abused its discretion as a matter of law in precluding evidence of third-party culpability.
In Primo, we rejected the “clear link” standard for admissibility of third-party culpability evidence that had been employed by the Appellate Division, which generally required the defendant to show a “clear link” between the third party and the crime in question, in favor of a test “described in terms of conventional evidentiary principles” (see Primo, 96 NY2d at 354-355). Under that test, the trial court should conduct a discretionary balancing of the probative value of the evidence against its potential for “trial delay, undue prejudice to the opposing party, confusing the issues or misleading the jury” (id. at 355). We noted that, “[t]o the extent that the ‘clear link’ standard implies no more than an abbreviation for the conventional balancing test, it presents no problem” (id. at 356).
Even under the proper test, however, “remote evidence of a third party’s culpability — though relevant — will not be sufficiently probative to outweigh the risk of trial delay, undue prejudice or jury confusion” (id. at 356). We cautioned in Primo that, in the context of third-party culpability evidence, the “risks of delay, prejudice and confusion are particularly acute,” and that if these risks are not “weighed against the probative value of the evidence, the fact-finding process would break down under a mass of speculation and conjecture” (id. at 356-357). For these reasons, “[t]he admission of evidence of third-party culpability may not rest on mere suspicion or surmise” (id. at 357).
The evidence of third-party culpability defendant offered here was far weaker than the evidence offered by the defendant in Primo. There, the shooting victim acknowledged that the third party was at the crime scene at the time of the shooting. Significantly, a ballistics report linked the bullets recovered from the crime scene to a gun used by the third party two months later in an unrelated crime (see Primo, 96 NY2d at 353-354). We concluded that the probative value of the ballistics report, “coupled with proof that [the third party] was at *145the scene of the shooting, . . . plainly outweigh [ed] the dangers of delay, prejudice and confusion” (id. at 357). We therefore held that the trial court improperly precluded the ballistics report as evidence of third-party culpability (see id.).
Here, by contrast, there is no physical evidence connecting Gombert to the crime, and no witness placing Gombert with the victim on the day of her disappearance, let alone at the crime scene. The reverse Molineux evidence presented by defendant, although it points to Gombert as a heinous individual, does not persuade me that the trial court abused its discretion as a matter of law. Defendant’s investigator allegedly spoke to several of Gombert’s victims, and they told the investigator that Gombert only sometimes restrained their hands behind their backs or shoved articles of clothing in their mouths during his sexual assaults (cf. People v Beam, 57 NY2d 241, 252 [1982] [“Although there are minor differences in each attack, the pattern of the initial encounter and the specifics of the sexual attacks, all of which followed the same pattern, indicate a unique modus operandi”]).
This Court has generally left the resolution of issues surrounding third-party culpability to the discretion of the trial court, unless the evidence of third-party culpability was so unmistakably strong that it was an abuse of discretion for the trial court not to allow the jury to hear it (see Primo, 96 NY2d at 357). Where the defendant’s proffer has fallen short of the strong proof of third-party culpability presented in Primo, this Court has not interfered with the discretionary determination of the trial court (see People v King, 27 NY3d 147 [2016] [decided herewith]; People v Gamble, 18 NY3d 386, 398-399 [2012], rearg denied 19 NY3d 833 [2012]; People v Schulz, 4 NY3d 521, 529 [2005]). This Court should not interfere in those discretionary determinations unless the evidence, like that presented in Primo, is so compelling that the trial court clearly abused its discretion as a matter of law in precluding it. I conclude that the evidence offered by defendant here does not rise to that level.
Although defendant will finally be required to present proof of third-party culpability in admissible form at a new trial, the trial court must otherwise exercise its discretion to ensure that “the fact-finding process [does not] break down under a mass of speculation and conjecture” (Primo, 96 NY2d at 357). The determination of admissibility can only be answered at trial. The trial court should be allowed to exercise its discretion upon *146retrial with respect to these issues. Faced with the prospect of another appellate reversal, however, the trial court here, and other trial courts attempting to comply with the Court’s ruling in the future, will be reluctant to do so.
Defendant’s remaining contentions are without merit. I would affirm the order of the Appellate Division.
Judges Pigott, Rivera and Abdus-Salaam concur; Judge Fa-hey dissents in an opinion; Chief Judge DiFiore and Judge Garcia taking no part.Order reversed and a new trial ordered.