(dissenting). In affirming the judgment of the Alcoholic Beverages Control Commission (ABCC) the court permits the ABCC to rely upon the uncorroborated hearsay statement of a witness in a criminal trial for a finding essential to its judgment (that the licensees served alcohol to a minor) where the fact of the sale was not a contested issue in the criminal trial; where the licensees were not participants in the criminal trial or in privity of interest with any of the parties in the criminal trial; where it was in the witness’s interest to suggest that she was under the influence; and, where the ABCC has made no ruling on why it relied on this evidence. In so concluding, the court has blurred the distinction between credibility and reliability and overruled a long-standing decision of this court. I believe the court reaches this erroneous conclusion by failing to establish correctly what quantum of evidence is necessary for it to be given probative effect within the meaning of G. L. c. 30A, § 11 (2) (1986 ed.).
It is indisputable that administrative agencies need not observe the rules of evidence that pertain in the courts. It does not follow, however, that the power of an agency to accept or reject evidence is totally uncircumscribed and insulated from effective appellate review. The question upon review until now has been whether this is “the kind of evidence on which reasonable persons are accustomed to rely in the conduct of serious affairs.” G. L. c. 30A, § 11 (2). See Dwyer v. Commissioner of Ins., 375 Mass. 227,236 (1978); AAA Movers v. Department of Pub. Utils., 354 Mass. 390, 393 (1968). The court today adds the consideration whether the evidence is so inherently unreliable as to fail to meet that test. It appears to me that the addition of the modifying words “so inherently unreliable” adds a presumption of reliability to hearsay not previously thought to be necessary or appropriate.
The question of the extent to which hearsay evidence might constitute substantial evidence is by no means a new one. Briefly summarized, various jurisdictions have adopted one of three basic approaches designed to ensure the trustworthiness or reliability of the evidentiary basis of agency decisions. The first approach has been to require corroboration regardless of *533the character of the hearsay. See, e.g., Goodman v. Commonwealth, 98 Pa. Commw. Ct. 371, 375-376 (1986) (arrest and police reports submitted in evidence without direct testimony or independent corroboration did not provide adequate support for denial of application for renewal of firearms permit). The second, and until recently, more common approach has been adherence to the “residuum rule.” As stated in the landmark case which originated the rule, “[the New York statute] may be taken to mean that while the commission’s inquiry is not limited by the common law or statutory rules of evidence or by technical or formal rules of procedure, and it may in its discretion accept any evidence that is offered, still in the end there must be a residuum of legal evidence to support the claim before an award can be made.” Carroll v. Knickerbocker Ice Co., 218 N.Y. 435, 440 (1916).
Both approaches have been criticized by Professor Davis, who delineates the third approach to ensuring the reliability of evidence. “Rejection of the residuum rule does not mean that a reviewing court must refuse to set aside a finding based upon incompetent evidence; it means only that the court may set aside the finding or refuse to do so as it sees fit, in accordance with its own determination of the question whether the evidence supporting the finding should be deemed reliable and substantial in the circumstances” (emphasis added). 4 K.C. Davis, Administrative Law § 16.6 (1980 ed.). Accordingly, courts are reluctant to let stand agency decisions that depend upon evidence that, although arguably admissible under agency practice, lacks a firm basis of reliability. See Credit v. Whitfield, 488 So.2d 1064 (La. Ct. App. 1986) (although hearsay is generally admissible in unemployment compensation proceedings, where employer bears burden of proving employee misconduct, employer may not base entire case on hearsay when claimant offers direct, contradictory evidence); Higley v. Edwards, 67 Or. App. 488 (1984), disapproved on other grounds, Wiggett v. Oregon State Penitentiary, 85 Or. App. 635 (1987) (in administrative hearing on allegations of deputy sheriff’s misconduct by sexual contact with prisoner, unsworn and uncorroborated written statement of prisoner was inadmissible *534hearsay, where it was only evidence on whether there had been sexual contact and prisoner did not appear at hearing).
The Massachusetts cases decided after adoption of G. L. c. 30A, § 11 (2), pretermitted reconsideration of the rule enunciated in Sinclair v. Director of the Div. of Employment Sec., 331 Mass. 101 (1954), only by adopting an ad hoc approach in tune with the general principles outlined above. See Goodridge v. Director of the Div. of Employment Sec., 375 Mass. 434, 436-437 (1978); United Food Corp. v. Alcoholic Beverages Control Comm’n, 375 Mass. 238, 244 (1978); Dwyer v. Commissioner of Ins., supra at 236; Olde Towne Liquor Store, Inc. v. Alcoholic Beverages Control Comm’n, 372 Mass. 152, 154 (1977); Western Mass. Bus Lines v. Department of Pub. Utils., 363 Mass. 61, 63 (1973). In each case the court was either able to find independent corroboration or, in fact, made an independent evaluation of the reliability of the hearsay evidence, in essence the substitute for independent corroboration. Here the court apparently concludes that independent corroboration exists and that, on the basis of its own independent evaluation, the evidence is reliable.
I disagree with respect to both conclusions. First, the critical evidence of where Kozec became intoxicated is not corroborated except by additional hearsay which I would submit is less reliable than the hearsay of the transcript which was, at least, under oath and subject to some form of cross-examination, albeit not by a party with an interest in the case comparable to the plaintiffs’. Second, in its reliability analysis the court’s reliance on Richardson v. Perales, 402 U.S. 389 (1971), is misplaced. The Federal court decisions make it clear that, under the Federal Administrative Procedure Act, courts will evaluate testimony or evidence for indicia of reliability on a case-by-case basis. Factors to be considered include independence or possible bias of the declarant, the type of hearsay materials submitted, whether statements are sworn to, whether statements are contradicted by direct testimony, availability of the declarant, and credibility of the declarant. Calhoun v. Bailar, 626 F.2d 145, 149 (9th Cir. 1980), cert, denied, 452 U.S. 906 (1981). In holding that the hearsay evidence (written med*535ical reports) relied upon in Perales could constitute substantial evidence, the Supreme Court stressed that “the specter of questionable credibility and veracity is not present; there is professional disagreement with the medical conclusions . . . but there is no attack here upon the doctor’s credibility or veracity.” Perales, supra at 407. Moreover, unlike the situation here, the Court noted that written medical evaluations were the sort of evidence traditionally relied upon in the agency’s proceedings. Id. at 405-406. See also Hoska v. United States Dept. of the Army, 677 F.2d 131 (D.C. Cir. 1982) (in determining validity of revocation of public employee’s security clearance and resulting dismissal from his civilian position, mere hearsay lacking sufficient assurance of its truthfulness was not substantial evidence to overcome sworn testimony of employee); Johnson v. United States, 628 F.2d 187 (D.C. Cir. 1980) (hearsay statements could be relied on where the three out-of-court declarants were disinterested witnesses, their accounts were consistent, and appellant’s counsel had access to the statements before agency’s hearing); Duvall v. United States, 647 F.2d 131 (Ct. Cl. 1981) (absence of credibility determination affects weight to be accorded hearsay evidence where there is dispute over its truthfulness).
Duvall underscores one of the key problems in this case. In assessing the reliability of Kozec’s testimony one must consider her credibility. Yet neither the ABCC nor the licensing board of Salisbury was in any position to assess her credibility and, therefore, could, at best, make only an abstract determination of the inherent credibility of her testimony in light of the circumstances surrounding that testimony. Although a reviewing court is never in a position to assess credibility, where a reliability determination is affected by the credibility of a witness who has not testified before the agency, the reviewing court’s independent assessment of reliability should at least discount the agency’s credibility role. Therefore, to the extent that the court has relied on the traditional credibility assessment of a fact finder, here that reliance is misplaced.
The court further confuses the issue by essentially making dispositive the licensees’ failure to subpoena Kozec for cross-*536examination and ignores the key fact that Kozec was not subject to cross-examination at her criminal trial by the licensees. Compare Eichberg v. Maryland Bd. of Pharmacy, 50 Md. App. 189, 194 (1982) (testimony of witness at previous criminal trial was admissible in license revocation hearing where board’s charges were essentially the same as those tried at criminal case and witness was subject to cross-examination by defendant/appellant). It is true that the Perales court pointed to the appellant’s failure to make use of its subpoena power. There, however, the Court also noted that the burden was on the appellant to establish a right to a benefit. Here we are dealing with a license suspension which, as the licensees correctly argue, requires the ABCC to prove is warranted. Where the burden was on the agency to produce substantial evidence, the licensees’ failure to subpoena Kozec does not somehow render unreliable hearsay now reliable. This court’s holding in United Food Corp. v. Alcoholic Beverages Control Comm’n, supra at 243, that “[tjhere was nothing wrong in receiving the BLB transcript supplemented as it was by other evidence, all open to challenge and contradiction by the appellant armed with a right of subpoena,” is not contradictory. There, the issue was the procedural fairness of the ABCC’s proceedings rather than the evidentiary support for its decision.1
I, therefore, conclude that uncorroborated hearsay testimony at a criminal trial on an issue not contested at the criminal trial where the witness has a motive to lie and was not subject to cross-examination by a person with the same or similar interests as the person against whom the evidence is being used is not sufficiently reliable to be given probative effect under G. L. c. 30A, § 11 (2), and, therefore, cannot constitute substantial evidence within the meaning of G. L. c. 30A, § 1 (6).
Similarly, I would agree that the licensees’ constitutional due process arguments should fail here as a result of their failure to subpoena Kozec. This issue is completely separate from the question whether there was substantial evidence.