In this case we consider whether G. L. c. 273, § 12A (1984 ed.), providing for admission of the results of a blood grouping test in a paternity proceeding to exclude the possibility of paternity of the alleged father, prohibits the introduction of inculpatory human leukocyte antigen (HLA) test results. Further, we consider whether HLA test results may be admissible to establish paternity, if not barred by G. L. c. 273, § 12A (1984 ed.).

The defendant, Michael Beausoleil, was charged by Sharon Burke with fathering her child, in a proceeding under G. L. c. 273, § 12 (1984 ed.). The defendant moved for a court order pursuant to G. L. c. 273, § 12A (1984 ed.), requiring that he, the mother, and the child submit to an HLA blood test to determine whether his nonpaternity might be established. The trial judge allowed the defendant’s motion, but proceeded to trial and took other evidence. Later, the University of Massachusetts Medical Center’s department of laboratory medicine submitted a report stating that, based on a statistical analysis of the results, the probability of paternity was 98.2%, which it termed “very likely.” Because the report did not exculpate the defendant, the judge entered a finding of guilty and ordered the defendant to pay child support of $75 a week. The defendant appealed for a trial de nova before a jury of six. G. L. c. 218, § 27A (1984 ed.).

Prior to the trial de nova, the Commonwealth filed a motion in limine requesting admission of the HLA test results. The judge denied the motion without a hearing.1 The Commonwealth then filed a petition for relief pursuant to G. L. c. 211, *208§ 3 (1984 ed.). Treating the denial of the Commonwealth’s motion as an allowance of a motion to suppress under Mass. R. Crim. P. 15 (b) (2), 378 Mass. (1979),2 a single justice of this court transferred the case to the Appeals Court. G. L. c. 211, § 4A (1984 ed.). Subsequently, we transferred the case to this court on our own motion.

The Commonwealth asserts on appeal that the judge improperly denied its motion in limine, arguing that G. L. c. 273, § 12A (1984 ed.), does not prohibit the introduction of HLA test results to prove paternity. The defendant argues to the contrary. The Commonwealth further contends that, in the absence of any statutory prohibition, inculpatory HLA test results are sufficiently reliable to satisfy the fundamental admissibility requirements for scientific evidence. The defendant argues that the record is barren of evidence of the scientific reliability of HLA test procedures. He further argues that he would be “highly” prejudiced if HLA test results were admitted as evidence of his paternity.

Before we address these issues, a brief discussion of the basic scientific principles and procedures involved in paternity testing is warranted. In view of the absence of evidence or judicial findings, we draw on the authorities cited herein for our generalized discussion.3

*209Paternity testing is based on the existence of genetic markers which are inherited from a child’s parents and are found in the various components of the blood. More than 260 genetic markers have been identified for red blood cells (red isoantigens) and over fifty genetic markers have been identified for white blood cells (white isoantigens). Paternity testing involves the identification of such markers followed by application of Mendelian rules of inheritance. Traditionally, exclusion of paternity has been the primary use to which these rules have been put. If a child lacks a genetic marker that a child of the accused must have, or if the child displays a marker that neither the mother nor the putative father has, paternity is conclusively excluded. Lee, Current Status of Paternity Testing, 9 Fam. L.Q. 615, 616-617, 621 (1975). In such cases the impossibility of the accused’s paternity is established to a medical certainty. See Commonwealth v. D’Avella, 339 Mass. 642, 645 (1959) (“The reliability of [blood] tests to prove nonpaternity is well established as a scientific fact”).

Six red blood cell tests most commonly have been employed in paternity testing. The first three were discovered by Dr. Karl Landsteiner and his colleagues and consist of the ABO, MNS’s, and Rh systems, known collectively as the Landsteiner series. Lemmon & Murphy, The Evidentiary Use of the HLA Blood Test in Virginia, 19 U. Rich. L. Rev. 235, 238-239 (1985). These three systems yield a cumulative probability of between 52% and 57%, depending on the race of the putative father, that at least one of them will exclude paternity of a falsely accused man. Joint AMA-ABA Guidlines: Present Status of Serologic Testing in Problems of Disputed Parentage, 10 Fam. L.Q. 247, 256-258 (1976) (hereinafter, Joint Guidelines). Even with the addition of three other red isoantigen tests, Kell, Duffy, and Kidd, now typically employed in conjunction with the Landsteiner. series (enhanced Landsteiner series), the probability of exclusion of a nonfather is still no greater than 63% to 72%. Id. Therefore, although exclusion of paternity by means of these tests is conclusive, nonexclusion *210is not, because there may remain a greater than 30% probability that a falsely accused man would not have been excluded.4

In recent years there have been claims of substantial advances in the field of paternity testing. HLA testing was developed in the 1960’s by Dr. Paul Terasaki to determine donor-recipient compatability of organ transplants, but more recently has been used for purposes of determining paternity. See Terasaki, Resolution by HLA Testing of 1000 Paternity Cases not Excluded by ABO Testing, 16 J. Earn. L. 543 (1978). The HLA test is based on the identification and typing of more than fifty antigens found in the white blood cells. Unlike the enhanced Landsteiner series which only can be performed on blood, the HLA test can be performed on certain body tissue in which HLA antigens appear. Consequently, it is often termed a tissue typing test, although for purposes of paternity testing, economic feasibility usually dictates that it be performed on blood.

HLA testing allegedly represents an improvement over red isoantigen blood tests in two respects. First, it has a greater capability than the enhanced Landsteiner series for excluding falsely accused men. When administered by itself, the HLA test excludes between 78% and 80% of all nonfathers; this figure increases to over 90% when it is administered in conjunction with the enhanced Landsteiner series. See Joint Guidelines, supra at 257, 258. Second, because of this increased exclusionary capability and the relative rarity of HLA markers, it is claimed that the HLA system can be utilized to calculate a reliable statistical estimation of a accused’s likelihood of paternity.5 This statistical calculation is “made on the basis of estab*211lished data regarding the frequency with which particular genes appear in certain segments of the population and the combinations in which these genes appear in the child and putative father.” Jones v. Robinson, 229 Va. 276, 280 (1985). This information is then used to derive the probability of paternity by use of Bayes’ Theorem, a mathematical formula which describes the way newly discovered statistical information alters a previously established probability.6 Peterson, supra at 681. In most cases in which the putative father is not excluded by HLA testing, the estimated probability of paternity will be high, typically exceeding 90%. Terasaki, supra.

We turn now to consideration of the issues raised on appeal. Statute 1954, c. 232, now G. L. c. 273, § 12A, is entitled, “An Act providing for blood grouping tests to aid in the determination of paternity.” The statute is a so-called “exclusionary” blood test statute, that is, one permitting the introduction of blood test results only to exclude the possibility of paternity. The statute provides: “In any proceeding to determine the question of paternity, the court, on motion of the alleged father, shall order the mother, her child and the alleged father to submit to one or more blood grouping tests, to be made by a duly qualified physician or other duly qualified person, designated by the court, to determine whether or not the alleged father can be excluded as being the father of the child. The results of such tests shall be admissible in evidence only in cases where definite exclusion of the alleged father as such father has been established. If one of the parties refuses to *212comply with the order of the court relative to such tests, such fact shall be admissible in evidence in such proceeding unless the court, for good cause, otherwise orders.” (Emphasis supplied.)7

The statute refers specifically to “blood grouping tests,” terminology that has come to be associated with the red blood cell isoantigen tests of which those comprising the enhanced Landsteiner series are typical. The HLA system, however, is a tissue typing procedure. See, e.g., Ellman & Kaye, Probabilities and Proof: Can HLA and Blood Group Testing Prove Paternity?, 54 N.Y.U.L. Rev. 1131, 1139 (1979) (“HLA testing can be thought of as tissue typing rather than blood group typing”); Terasaki, supra at 543 (discussing the “HLA system of tissue types”). Thus, the Supreme Court of Ohio, construing a statute similar to G. L. c. 273, § 12A, held that HLA tests “are basically genetic comparison examinations, rather than blood grouping tests” and are, therefore, not within the statutory proscription against admission of inculpatory blood grouping tests. Owens v. Bell, 6 Ohio St. 3d 46, 53 (1983).8

*213We are convinced by our review of the case law from other jurisdictions as well as the available scientific and medicolegal commentary that the HLA system is not a blood grouping test, but is a tissue typing system. Assuming, however, that the statutory language is ambiguous,9 we would be reluctant to ascribe to the Legislature by virtue of G. L. c. 273, § 12A, an intent to prohibit the introduction of evidence of paternity derived from a technique that was not in existence when the statute was enacted in 1954, long before the advent of HLA testing. At that time red blood cell isoantigen testing was prevalent, but was capable of disproving paternity only. Massachusetts, like many other States, enacted an exclusionary blood test statute to give effect to the scientific knowledge of the time. As the Supreme Court of Idaho stated, given a similar historical perspective, “it would be illogical to conclude that by enacting [the Idaho Paternity Act], the legislature expressed an intent to bar the use of evidence obtainable through scientific techniques not then known to it.” Crain v. Crain, 104 Idaho 666, 671 (1983). Accord Callison v. Callison, 687 P.2d 106, 110 (Okla. 1984); Phillips v. Jackson, 615 P.2d 1228, 1233 (Utah 1980). In our view, therefore, G. L. c. 273, § 12A, does not govern the admissibility of inculpatory HLA tests.10

*214We turn now to consider the admissibility of inculpatory HLA tests as a matter of the common law rules of evidence. We stated in Commonwealth v. Stappen, 336 Mass. 174, 176 (1957), that, even in the absence of statutory authority, blood grouping test results which exclude the possibility of paternity are admissible, “if otherwise competent.” The court held that exclusionary blood tests satisfy this standard, noting that there is substantial authority to support the reliability of blood grouping tests to prove the impossibility of paternity. Id. at 176-177. See Symonds v. Symonds, 385 Mass. 540, 542 (1982). It follows that, in the absence of a statutory directive, inculpatory HLA test results should likewise be admissible if “otherwise competent.”

This court has adopted the standard of admissibility for scientific evidence articulated in Frye v. United States, 293 F. 1013, 1014 (D.C. Cir. 1923), wherein the Court of Appeals stated: “Just when a scientific principle or discovery crosses the line between the experimental and demonstrable stages is difficult to define. Somewhere in this twilight zone the evidential force of the principle must be recognized, and while courts will go a long way in admitting expert testimony deduced from a well-recognized scientific principle or discovery, the thing from which the deduction is made must be sufficiently established to have gained general acceptance in the particular field in which it belongs.” Thus, we have said that “[j]udicial acceptance of a scientific theory or instrument can occur only when it follows a general acceptance by the community of scientists involved. When supported by substantial authority establishing scientific reliability, this court has not hesitated to accept the benefits of science.” Commonwealth v. Fatalo, 346 Mass. *215266, 269 (1963).11 Cf. Commonwealth v. Vitello, 376 Mass. 426 (1978); Commonwealth v. A Juvenile, 365 Mass. 421 (1974).

Because this case is before us, in effect, as an interlocutory appeal of a denial, without hearing, of a motion in limine, we do not have the benefit of expert testimony adduced at trial on which to make the requisite Frye determination. We may properly consider, however, the plethora of articles concerning HLA paternity testing, as well as the decisions by other courts addressing the issue now before us. See Commonwealth v. Kater, 388 Mass. 519, 527 (1983); Commonwealth v. Vitello, supra at 431-439; Commonwealth v. Lykus, 367 Mass. 191, 198-199 (1975). On review of these authorities, we conclude that HLA testing has been accepted as a reliable method of proving paternity both in the scientific community and in the courts.12

The American Medical Association, the American Association of Blood Banking, and the American Association of Histocompatibility have all approved HLA testing to determine paternity. See Haines v. Shanholtz, 57 Md. App. 92, 100 (1984). Moreover, while there is some disagreement among medicolegal commentators as to how inculpatory HLA test results should be presented to the fact finder in paternity cases, *216nearly all agree that such evidence is reliable and should be admitted in one form or another. See, e.g., Ellman & Kaye, Probabilities and Proof: Can HLA and Blood Group Testing Prove Paternity?, 54 N.Y.U.L. Rev. 1131, 1161 (1979); Page-Bright, Proving Paternity — Human Leukocyte Antigen Test, 27 J. Forensic Sci. 135, 143, 150 (1982); Peterson, A Few Things You Should Know About Paternity Tests (But Were Afraid to Ask), 22 Santa Clara L. Rev. 667, 676 (1982); Lemmon & Murphy, The Evidentiary Use of the HLA Blood Test in Virginia, 19 U. Rich. L. Rev. 235, 238-239 (1985). Finally, those courts in other jurisdictions that have determined the admissibility of inculpatory HLA test results by reference to Frye have concluded unanimously that such evidence is generally accepted as reliable in the scientific community.13 See Crain v. Crain, supra at 672; Tice v. Richardson, 7 Kan. App. 2d 509, 513 (1982); Perry v. Commonwealth ex. rel. Kessinger, 652 S.W.2d 655, 661 (Ky. 1983); Haines v. Shanholtz, supra at 98; Imms v. Clarke, 654 S.W.2d 281, 284 (Mo. App. 1983); Callison v. Callison, supra at 111.14

Although we conclude that HLA testing is generally accepted in the scientific community as a reliable means of determining paternity, and that inculpatory evidence derived from HLA *217testing is therefore admissible, we think it prudent to adopt the following limitations on the affirmative use of such evidence.

First, inculpatory HLA test results may not be presented to the jury expressed simply in the form of a probability of exclusion calculation, that is, as a stated percentage of the population of nonfathers who would have been excluded by the specific array of tests conducted. The probability of exclusion may be relevant to the issue of paternity as it tells the jury that the defendant falls within the relatively small per cent of the male population who could have fathered the child; however, “it does nothing to distinguish the true father from the perhaps millions of men who fall into this group.” Peterson, supra at 680. Furthermore, the jury are apt to confuse the probability of exclusion with the likelihood of paternity even though there is no direct relationship between the two. See notes 4 & 5, supra. Thus, even if the probability of exclusion is considered relevant, “the potential for confusing and misleading the fact-finder is so great that the court should exclude it under its general power to exclude evidence which creates a substantial danger of prejudice, of confusing the issues, or misleading the fact-finder” (footnote omitted). Peterson, supra. See Proposed Mass. R. Evid. 403. See also State ex rel. Hausner v. Blackman, 233 Kan. 223, 229 (1983) (probability of exclusion is not probative on the issue of paternity and its introduction is prejudicial); Imms v. Clarke, supra at 285 (probability of exclusion, unless in “extreme range,” is of little aid to jury in determining paternity issue); Davis v. State, 476 N.E.2d 127, 137 (Ind. App. 1985) (probability of exclusion, without more, has been viewed as incomplete and misleading).

Second, we believe it preferable that the HLA test results be presented to the jury in terms of the probability of paternity,15 *218as that has been termed “the most accurate way of conveying the significance of the test data.” Ellman & Kaye, supra at 1146. It appears to be generally accepted in the scientific community that a statistically meaningful probability of paternity calculation can only be derived from a combination of HLA and red blood cell tests yielding a mean probability of exclusion of 90% or more. See Joint Guidelines, supra at 256-258; Imms v. Clarke, supra at 286. Accordingly, as a further limitation on the affirmative use of HLA evidence, we conclude that the jury may be presented with a statistical estimate of the putative father’s likelihood of paternity only where the combined tests administered would exclude at least 90% of the nonfathers tested.16

Third, we think it necessary to establish a minimum threshold for admissibility of probability of paternity estimates. The Joint Guidelines recommend that, if the blood test results show a “strong likelihood” of paternity, the evidence should be presented to the jury. This is to be determined by reference to a table in the Joint Guidelines matching various probabilities of *219paternity with corresponding “verbal predicates.” Probabilities of paternity between 90% and 95% are said to make paternity “likely,” while those between 95% and 99% make it “very likely.” Joint Guidelines, supra at 262, 282. Thus, we read the Joint Guidelines as recommending that probabilities of paternity below 95% be inadmissible. This accords with the view of many commentators. As one stated: “The significance of a ‘high’ probability of paternity is difficult to evaluate. It is clear that nonexcluded nonfathers can score comparatively high probabilities of paternity (since most nonfathers are excluded, the rest must of necessity ‘fit’ reasonably well). In one study, as many nonexcluded nonfathers scored over 88% as scored under 88%. On the average, fathers have higher probabilities of paternity than nonexcluded nonfathers (in the same study, the mean for fathers was 98.04%), so a paternity index must be rather high before it really differentiates the nonexcluded nonfathers from the (obviously nonexcluded) fathers. There is no general agreement in the scientific community that a probability below 95% has significance. Scientists generally agree that a probability greater than 95% ... is significant, so significance should be attached only to those probabilities exceeding 95%.” (Footnote omitted.) Peterson, A Few Things You Should Know About Paternity Tests (But Were Afraid to Ask), 22 Santa Clara L. Rev. 667, 707-708 (1982).17 We agree, and require that inculpatory blood test evidence shall be admissible only if the probability of the putative father’s paternity is 95% or greater.

Fourth, though admissible, inculpatory blood test evidence establishing a likelihood of paternity equal to or greater than 95% should not' be conclusive on the issue of paternity. It is *220merely some evidence of paternity which the jury must consider, along with all of the other evidence presented in mating their determination.18 See, e.g., Crain v. Crain, supra at 672-673; Turek v. Hardy, 312 Pa. Super. 158, 163 (1983); South Carolina Dep’t of Social Servs. v. Bacot, 280 S.C. 485, 489 (1984). The jury are free, for example, to disbelieve that the putative father ever had intercourse with the mother, or that he had intercourse with her during the period of probable conception, despite the blood test evidence indicating a high likelihood of paternity. In such a case, the jury would be warranted in concluding that paternity had not been established. See, e.g., Commissioner of Social Servs. of the County of Erie ex rel. Mannion v. Murray, 112 A.D.2d 724 (N.Y. 1985) (finding of nonpaternity affirmed despite a greater than 90% probability of paternity because Family Court judge was warranted in disbelieving mother’s testimony). We emphasize that HLA evidence is but one factor for the jury’s consideration.

Finally,19 inculpatory HLA blood test evidence shall be admissible only if a proper foundation is laid establishing that *221proper testing procedures were employed in the particular case and that the expert witnesses through whom the evidence is sought to be introduced are qualified properly. See Commonwealth v. Vitello, 376 Mass. 426, 456 (1978); Tice v. Richardson, 7 Kan. App. 2d 509, 513 (1982); Phillips v. Jackson, 615 P.2d 1228, 1235 (Utah 1980). This determination is, of course, for the trial judge to make in the first instance as a threshold inquiry whenever inculpatory blood test results are sought to be introduced.

The estimated probability of the defendant’s paternity in the instant case is 98.2%, above the minimum threshold for admissibility established today. Although the record indicates that this figure was derived from HLA and red blood cell typings, it is not clear exactly what red blood cell tests were performed and whether the cumulative probability of exclusion for all the tests administered was at least 90%. Normally, we would remand the case to the trial judge to make this factual determination. However, we conclude that a remand is unnecessary in this instance.

*222Considerations of fairness dictate that the results of the blood tests sought to be introduced to prove the defendant’s paternity be excluded. The defendant requested that the HLA blood test be performed pursuant to G. L. c. 273, § 12 A, citing the statute in his motion to the judge. It was clearly his belief that, under the terms of this statute, the results of the test would be admissible only if exculpatory. No judicial decision had given him cause to assume otherwise. See Commonwealth v. Blazo, 10 Mass. App. Ct. 324 (1980).20 We think the defendant’s reliance on G. L. c. 273, § 12A, reasonable, and we decline to apply today’s decision retroactively to this defendant.

Last, because of its implications in future paternity cases, we address briefly the defendant’s claim that court-ordered HLA blood testing, against a defendant’s wishes, raises constitutional issues under the Fourth and Fifth Amendments to the United States Constitution.21 In Schmerber v. California, 384 U.S. 757, 766-772 (1966), the United States Supreme Court held that the extraction of a blood sample from a nonconsenting defendant charged with driving an automobile while *223under the influence of intoxicating liquor was reasonable in the circumstances and did not violate the defendant’s right against unreasonable searches and seizures under the Fourth Amendment, or his privilege against self-incrimination, guaranteed by the Fifth Amendment. As to the Fourth Amendment, the Court noted that there was probable cause to arrest the defendant for the crime charged, and stated that the extraction of blood is a commonplace procedure involving virtually no risk, trauma, or pain. Id. at 768, 771. With regard to the Fifth Amendment claim, the Court held that the privilege against self-incrimination only protects against testimonial compulsion and does not extend to compulsory blood testing, because blood-testing evidence is physical in nature rather than testimonial. Id. at 764.

HLA and red blood cell testing procedures are no more intrusive or testimonial than those at issue in Schmerber. In addition, probable cause to believe that the test results will be inculpatory exists when a complaint is issued on probable cause, as in the instant case, charging a putative father with paternity. Consequently, the Fourth and Fifth Amendments provide no grounds for prohibiting court-ordered HLA and red blood cell testing.22 See State v. Meacham, 93 Wash. 735 (1980) (court rejected right to privacy and Fourth Amendment challenges to court-ordered HLA testing); Albany County Dep’t of Social Servs. ex rel. Sousis v. Seeburger, 112 A.D.2d 674, 675-676 (N.Y. 1985) (court rejected right to privacy, Fourth Amendment, and Fifth Amendment challenges to court-ordered HLA testing).

*224In summary, we hold that HLA tests are not blood grouping tests within the meaning of G. L. c. 273, § 12A. Exclusionary results relevant at common law (see supra at 222 & note 20) may, however, be admitted. Further, we conclude that HLA testing is accepted generally in the scientific community as a reliable procedure for proving or excluding paternity. On offer of the prosecution, statistical evidence of the accused’s probability of paternity shall be admissible provided the HLA test on which the calculation was derived, either alone or in conjunction with whatever standard red isoantigen tests were performed, yields a cumulative probability of exculsion of 90% or greater, and provided further that no evidence of probability of paternity under 95% shall be admissible. Such inculpatory evidence is not conclusive on the issue of paternity, and its admission is subject to normal foundational requirements for the admission of scientific evidence.

The inculpatory HLA test results sought to be introduced in the instant case shall be excluded, and the defendant may not be compelled to undergo additional blood tests. After the date of this opinion, an alleged father (as well as the mother and child) may be required to submit to such test procedures (see supra 222, 223). Trial on the paternity charge is to proceed in accordance with this opinion.

So ordered.

(concurring in part and dissenting in part). I concur in the result reached by the court because I agree that it would be unfair to apply retroactively to the defendant the evidentiary rule announced today.

The bulk of the court’s opinion addresses the future admissibility of probability of paternity estimates based on HLA test results to establish paternity, and it is to that subject that I direct my attention. I agree that the HLA procedure is a tissue test, not a blood test, with the consequence that the admissibility of HLA test results and probability of paternity estimates based thereon is not controlled by G. L. c. 273, § 12A (1984 ed.). *225I agree too, that, probable cause having been found, there is no constitutional impediment to a court order requiring a defendant to submit to HLA testing. Furthermore, I accept the court’s conclusion that HLA testing has gained general acceptance in the appropriate scientific community as a reliable indicator of the probability of paternity if it be established by other evidence that the defendant had intercourse with the mother at a time and under conditions that reasonably might have lead to the conception in question. I conclude, however, differently from the court, that, because the probability of paternity yielded by a statistical analysis employing HLA test results is premised on an assumption that that intercourse occurred, the test results should not be admitted in evidence unless that assumed fact is first established by a finding made by the trier of fact. That, of course, would require bifurcation of the trial, a procedure which is neither alien to our practice nor unduly burdensome, and, in my view, is dictated in paternity proceedings by considerations of fairness.

The court notes, ante at 211, that “the probability of paternity [is derived] by use of Bayes’ Theorem, a mathematical formula which describes the way newly discovered statistical information alters a previously established probability.” The court explains, ante at 211 n.6, that “[apparently, the most widely accepted way of calculating the probability of paternity involves an assumption that without regard to the HLA or other blood test results there is a 50% chance that the defendant is the father. This is termed the prior odds of paternity and is a constant. Bayes’ Theorem is then used to determine how the HLA test results alter this prior probability” (emphasis added). The court further explains, ante at 220-221 n.19: “There has been criticism of the probability of paternity statistic because it is based on a preliminary assumption of a 50% likelihood of the accused’s paternity. In other words [emphasis added], the calculation starts with the premise that the accused and one random male both had intercourse with the mother ‘at a time, and under circumstances, in terms of timing, fertility, frequency of coition, and use of birth control, making them both equally likely to have fathered the child’ (emphasis in original). Peterson, *226[A Few Things You Should Know About Paternity Tests (But Were Afraid to Ask), 22 Santa Clara L. Rev. 667,] 685 [1982].” A relevant portion of Peterson’s article supporting the court’s understanding and giving enlightenment about the test’s assumption is set forth in the margin.1

*227No test tends to prove a fact it assumes. The HLA test, therefore, has no relevancy to the question, typically the one critical question in a paternity case, whether the defendant had intercourse with the mother at a time and in circumstances that would explain the conception of the child. Apparently, the court agrees, for it states that “HLA test results are not evidence of intercourse,” see ante at 220 n.18. The court also states that “the judge should instruct the jury, if he is so requested, that they may not consider HLA test results as evidence of intercourse, and that they may not consider such evidence of paternity unless they have found, beyond a reasonable doubt, that sexual intercourse at or about the time of conception had taken place between the mother and the alleged father.” Id.2

Although HLA test results are not legitimately helpful to a resolution of the question whether a defendant had intercourse with the mother in relevant circumstances, it would be naive *228in the extreme to expect fact finders, particularly inexperienced jurors, to ignore evidence of probability of paternity estimates based on HLA test results, presented through an expert, in determining the usually contested issue whether such intercourse had taken place. In the typical case, the mother asserts, and the defendant denies, the defendant’s involvement. The Commonwealth has the burden of proving the defendant’s involvement beyond a reasonable doubt. See Commonwealth v. Lobo, 385 Mass. 436, 445 (1982); Commonwealth v. MacKenzie, 368 Mass. 613, 619 n.5 (1975). It cannot be said reasonably and with confidence that an expert’s testimony that the probability of the defendant’s paternity is 95% will not be misused by the fact finder to resolve the credibility issue in the Commonwealth’s favor. Even if an assumption be gratuitously made that the expert’s testimony will be fairly and clearly given and skillfully tested on cross-examination, the danger that the fact finder will be moved to decide between the conflicting claims on the basis of the expert’s testimony that the defendant “fits the mold” is inescapable. The portion of Peterson’s article quoted in n. 1 of my opinion supports that proposition.

The court recognizes the risk of which I speak, but, in my view, does not adequately appreciate or deal with it. The court counters the risk by requiring the judge, if requested in a jury case, to instruct the jury not to consider HLA test results as evidence of intercourse, and not to consider that evidence “unless they have found . . . that sexual intercourse at or about the time of conception had taken place between the mother and the alleged father.” Ante at 220 n. 18. Because the interests at risk are so substantial, the incidence of risk is so high, and a way to avoid the-risk is so simple, I am not content to meet the risk by simply requiring jury instructions.

“The naive assumption that prejudicial effects can be overcome by instructions to the jury ... all practicing lawyers know to be unmitigated fiction.” Krulewitch v. United States, 336 U.S. 440, 453 (1949) (Jackson, J., concurring) (citation omitted). See Bruton v. United States, 391 U.S. 123, 135 (1968) (“[T]here are some contexts in which the risk that the jury will not, or cannot, follow instructions is so great, and *229the consequences of failure so vital to the defendant that the practical and human limitations of the jury system cannot be ignored”); Commonwealth v. DiMarzo, 364 Mass. 669, 681-682 (1974) (Hennessey, J., concurring) (“It is reasonable for us to be confident that in most cases limiting instructions accomplish their intended purpose. Nevertheless, in cases . . . where the evidence subject to limitations has an extremely high potential for unfair prejudice, we have a duty to be skeptical as to the effectiveness of limiting instructions .... The reasoning of the Bruton case is highly relevant. . . since that case recognized the futility, even the absurdity, of expecting a jury in some circumstances to conform to limiting instructions”).

Because of the futility in expecting triers of fact, especially juries, to comply with an instruction to ignore the probability of paternity factor based on HLA testing in determining whether the defendant had intercourse with the mother at or about the time of conception, I would hold that HLA based probability of paternity estimates are inadmissible except in the second segment of a bifurcated trial following a finding on that critical issue by the judge or the jury. After an affirmative finding, and within the limitations expressed by the court, ante at 217-221, I would hold the evidence to be admissible to show the probability that the defendant is the father of the child. Obviously, then, if the court were to accept my reasoning, the utility of HLA-based probability paternity estimates would be quite limited. Such evidence not only would be inadmissible in the absence of the above finding, but it would be superfluous in any case in which the fact finder (perhaps the jury in response to a special question) finds beyond a reasonable doubt that the defendant was the only one to have had intercourse with the mother at a relevant time. The test results would have value only in those cases in which it is open to the defendant to contend that, even though his act could have resulted in the conception, another man or other men, perhaps unidentified, also had sexual relations with the mother at a relevant time. It is at that time, and only then, that a probability of paternity estimate meeting the requirements articulated by the court is highly probative without being unfairly prejudicial to the defendant.