Plemel v. Walter

735 P.2d 1209, 303 Or. 262

• Court: Oregon Supreme Court
• Decided: April 21, 1987
• Docket: 549; CA A34015; SC S33117
• Status: Published

Opinion

LENT, J.

In this filiation proceeding the jury, by a 9-3 vote, found that Brent Walter was the father of Dena Plemel’s child. Plemel’s expert witness testified that blood tests of Plemel, Walter and the child did not exclude the possibility that Walter was the father. Walter did not challenge the validity of the tests or the failure of the tests to exclude him but did object, on grounds of irrelevance and prejudice, to testimony by the expert regarding the probability that he was the father. The trial court permitted the testimony, and the Court of Appeals affirmed, Plemel v. Walter, 80 Or App 250, 721 P2d 474 (1986). Because we hold that the testimony was inadmissible in the form that it was presented, we reverse and remand for a new trial.

I.

Plemel’s child was born on June 9,1983. Pursuant to ORS 109.125, she initiated filiation proceedings against Walter, alleging that he was the father of the child. Subsequently, the state intervened as a petitioner. Walter denied paternity.

Plemel testified that she and Walter had intercourse on one occasion, September 11,1982. Walter admitted having intercourse with Plemel but testified that it had occurred on the night of August 13-14, 1982. Both Plemel and Walter introduced the testimony of witnesses to corroborate their respective versions of when they had been together. A nurse practitioner who had examined Plemel testified that conception of the child would have been impossible in mid-August; she estimated that conception occurred between September 9 and September 16.

Plemel also called as a witness Dr. E. W. Lovrien, the director of the Oregon Health Sciences University Phenotype Laboratory. The laboratory had conducted blood tests of Plemel, Walter and the child. Lovrien testified that: (1) the tests did not exclude the possibility that Walter was the father; (2) the probability that the tests would have excluded a “falsely accused father” was 97.5 percent; (3) Walter’s “paternity index” was 178; (4) Walter’s “chance of paternity” was 99.4 percent; (5) Walter’s “chance of nonpaternity” was 0.6 percent; and (6) it was “extremely likely” that Walter was the father. The last three statements were essentially equivalent to, and derived from, the paternity index. The “chance of paternity” was the “paternity index” stated as a percentage chance (i.e., odds of 178 to 1 are equal to a 99.4 percent chance); the “chance of nonpaternity” was the “chance of paternity” stated negatively; the expression “extremely likely” was taken from a table developed by a joint committee of the American Medical Association and the American Bar Association 1 to express the significance of any given “chance of paternity.”

Walter, by way of a motion in limine and an objection, sought to exclude that portion of Lovrien’s testimony related to the paternity index and its equivalents. He argued that the testimony was irrelevant because the “probability of excluding a falsely accused father” provided the jury with all of the information that could be obtained from the blood tests. The argument was premised on his contention that in order to derive the paternity index, Lovrien had made an arbitrary assumption about the strength of the other evidence presented in the case. Walter also argued that the testimony was prejudicial in that the jury would be too confused by it to understand its proper significance. The trial court, without stating its reasons, denied Walter’s motion in limine and overruled his objection. The Court of Appeals affirmed, holding that the testimony regarding the paternity index helped the jury to interpret the blood test results and that Lovrien’s testimony minimized any potential for jury confusion by clearly explaining the limited significance of the statistics he presented. 80 Or App at 253-54.

Because courts have so frequently misinterpreted the meaning and significance of paternity test results, see, e.g., McCormick, Evidence § 211 (3rd ed 1984) (and cases cited therein); Ellman & Kaye, Probabilities and Proof: Can HLA and Blood Group Testing Prove Paternity?, 54 NYU L Rev 1131 (1979) (and cases cited therein), we believe that some background is appropriate before we analyze Walter’s arguments. 2

Lovrien referred to the group of blood tests performed in this case as an “extended red cell enzyme” test. This group of tests reveals the presence of various antigens, red cell enzymes and plasma proteins in the blood. By knowing which of these substances are in the blood of the mother, child and putative father, a geneticist can state whether it is possible for the putative father to be the true father.

Each person has a large number of inherited traits, such as eye color and facial features. Different versions of a particular trait are known as phenotypes. For the inherited trait of eye color, for example, blue eyes are one phenotype, brown eyes are another. The antigens, enzymes and proteins in a person’s blood are also inherited traits. One set, or “system,” of antigens are the ABO antigens, which have been widely used in the classification of blood. Because no individual possesses every antigen in the ABO system, the ABO antigens that an individual does possess determine that individual’s “blood type,” or phenotype, for the ABO system. The ABO system can be divided into six phenotypes: O, A1; A2, B, AjB or A2B. See Joint AMA-ABA .Guidelines: Present Status of Serological Testing in Problems of Disputed Parentage, 10 Fam LQ 247, 263 (1976). If a person has the phenotype Aj, this means that a blood test revealed the presence of the Ax antigen. Similarly, if a person has the phenotype AXB, the blood test revealed the presence of both the A1 antigen and the B antigen. If the person’s phenotype is O, the blood test was unable to detect the presence of any of the antigens for the ABO system.

The specific set of phenotypes that a person possesses is determined by that person’s genes. Genes occur in pairs that contain one gene from each parent. A specific gene pair or a specific group of gene pairs control a particular body trait, and an individual’s phenotype for that trait will reflect the nature of the genes that make up the controlling pair or pairs. If a person has the phenotype AXB, that person’s ABO gene pair consists of an A1 gene and a B gene. If the phenotype is A2, the pair consists of an A1 gene and either another Ax gene or an O gene. This is because the 0 gene does not produce a detectable antigen. See Reisner & Bolk, A Layman’s Guide to the Use of Blood Group Analysis in Paternity Testing, 20 J Fam L 657, 662 (1982). If the phenotype is O, both genes are O.

Information about a person’s phenotypes, then, can be used to derive information concerning that person’s corresponding gene pairs. Moreover, because each parent contributes one gene to a pair, knowledge of a person’s phenotypes can also be used to derive information about that person’s parents’ or children’s phenotypes.

For example, Plemel’s phenotype for the ABO system was found to be Ax. Thus, she had an Ax gene and either another Ax gene or an O gene. The child’s phenotype was found to be AXB, requiring that the child have an Ax gene and a B gene. Because the child received one gene from Plemel and one from its father, and because Plemel did not have a B gene, the child had to have received the A1 gene from Plemel and the B gene from its father. The child’s father’s phenotype would therefore have to be one of the “B phenotypes”: B, AXB or A2B. See Joint AMA-ABA Guidelines, supra, at 264. If Walter’s phenotype for this system were O, A1 or A2, he could not have been the father. Because Walter’s phenotype was found to be A2B, a phenotype consistent with paternity, the test for this system could not exclude him as a potential father.

In addition to the ABO system, Lovrien’s laboratory determined Walter’s, Plemel’s and the child’s phenotypes for 18 other blood systems. Lovrien testified that for each of these systems, Walter’s phenotypes were consistent with the accusation of paternity. The “extended red cell enzyme” test did not exclude the possibility that Walter was the father of Plemel’s child.

If the frequency with which phenotypes occur in the population of potential fathers is known, more can be said than that the paternity test has failed to exclude the putative father. It is also possible to derive statistics about the ability of the paternity test to exonerate men falsely accused of paternity and the relative likelihood that the putative father is the true father.

Probability of Excluding a Falsely Accused Father. This statistic, sometimes termed the “prior probability of exclusion” or, simply, the “probability of exclusion,” measures the ability of a paternity test to exclude men falsely accused of paternity. McCormick, supra, at 657 n 4. The statistic is calculated as follows: For each mother-child combination of phenotypes, certain phenotypes will be inconsistent with paternity. For example, if the mother is Ax and the child is B, the father cannot be Aa. The frequency with which such combinations occur in the population is the probability of excluding a falsely accused father. See Reisner & Bolk, supra, at 670-71. If there are no such combinations in the population, e.g., if everyone is A1; then the probability of excluding a falsely accused father is zero. A test with this probability would be useless.

As more blood systems are tested, the probability of excluding a falsely accused father can become quite high. The test in this case used 19 systems and had a probability of excluding a falsely accused father of 97.5 percent. Expressed in another way, in 39 instances out of 40 in which a man is falsely accused of being the father of a child, the tests conducted in this case would prove that the man did not father the child.

This statistic frequently is confused with the percentage of men in the population whose phenotypes are inconsistent with paternity, a percentage that is also often called the “probability of exclusion.” The parties in this case, as well as the Court of Appeals, appear to have fallen into this confusion. Although the “extended red cell enzyme” test had a probability of excluding a falsely accused father of 97.5 percent, this does not imply that the proportion of the male population capable of fathering Plemel’s child is 2.5 percent. 3 The probability of excluding a falsely accused father is a measure of the ability of the paternity test to exclude falsely accused fathers without reference to the phenotypes of any particular mother-child combination; the probability will be 97.5 percent for everyone tested. The results from the ABO system in this case provide a good illustration of the distinction. Lovrien testified that the ABO system alone will exclude falsely accused fathers in about 17 percent of cases. The father of Plemel’s child, however, must have a B gene, which, according to Lovrien, occurs in only about 5 percent of the population. Thus, although the ABO system has a probability of excluding a falsely accused father of 17 percent, 95 percent of the male population has been excluded by the ABO system in this case, 4

No statistic on the percentage of the relevant population capable of fathering Plemel’s child was presented by Lovrien.

Paternity Idex. This statistic, also known as the “likelihood ratio,” the “chance of paternity” and the “likelihood of paternity,” measures the putative father’s likelihood of producing the child’s phenotypes against the likelihood of a randomly selected man doing so. Despite the alternative labels, it is not the probability that the accused is the father. McCormick, supra, at 659 n 17; Reisner & Bolk, supra, at 671-74.

Within the group of men genetically capable of fathering a particular child, some will be genetically more likely to have done so than others. For example, we noted above that Walter’s phenotype for the ABO system was A2B and that the child’s father had to have one of the “B phenotypes.” Walter could transmit to his child either his A2 or his B gene. The chance that he would transmit the B gene would be 50 percent. A man who had a B phenotype and two B genes (as opposed to a man with a B phenotype and one B gene and one O gene), however, would have a 100 percent chance of transmitting a B gene. Other things being equal, this man would be more likely to be the father of Plemel’s child than would Walter.

The paternity index, which is a ratio that compares the putative father’s likelihood of producing the child’s phenotypes with the likelihood of a randomly selected man doing so, is not the likelihood of producing the child in question, but the relative likelihood of producing a child with the same phenotypes. The numerator of the ratio is the probability that a man with the phenotypes of the putative father and a woman with the mother’s phenotypes would produce an offspring with the child’s phenotypes. The denominator is the probability that a randomly selected man and a woman with the mother’s phenotypes would produce an offspring with the child’s phenotypes.

For example, as we noted above, the father of Plemel’s child had to transmit a B gene to the child. We also noted that the chance that someone with Walter’s phenotypes would do so was 0.5. This is the numerator of Walter’s paternity index for the ABO system. The chance that a randomly selected man would produce an offspring with the child’s phenotypes is simply the frequency with which the B gene occurs in the relevant male population, as compared to other genes. In the white male population this frequency is approximately 0.0658. (The percentage of white males with B genes will be somewhat less because some men will possess two B genes rather than only one.) This number is the denominator of the paternity index. Dividing the numerator by the denominator yields approximately 7.6, which is Walter’s paternity index for the ABO system. Multiplying together his paternity indexes for all 19 serologic systems tested yields his overall paternity index, 178.

From the example above, it can be seen that the denominator of the paternity index will be the same for every putative father. This is because the denominator is the gene frequency in the population. The numerator, however, will vary from putative father to putative father because their phenotypes will vary. In the ABO system, some men will have two B genes, some, such as Walter, will have one B gene, and some will have no B gene. For each combination of genes, there will be a different probability of producing the child’s phenotypes. Because the numerator varies from putative father to putative father, the paternity index will also vary. Thus, even though all men not excluded by a paternity test are capable of fathering the child, they will have different paternity indexes and thus different relative likelihoods of .having fathered the child.

Because Walter’s paternity index was 178, Walter was 178 times more likely than a randomly selected man to have fathered a child with the phenotypes of Plemel’s child. Converting this into a percentage resulted in a “chance of paternity” of 99.4 percent and a “chance of nonpaternity” of O. 6 percent. The AMA-ABA “verbal predicate” used to describe a “chance of paternity” of 99.4 percent is “extremely likely.” See Joint AMA-ABA Guidelines, supra, at 262. Again, this does not mean that it is “extremely likely” that Walter is the father of Plemel’s child, only that, compared to a randomly selected man, it is “extremely likely” that Walter is the father of PlemePs child. 5

Probability of Paternity. In order to convert the paternity index and its equivalents into a probability of paternity, 1. e., the actual likelihood that this putative father is the father of the child at issue, some estimate of the strength of the other evidence in the case must be made. If Walter were sterile, he would still have a paternity index of 178 and a “chance of paternity” of 99.4 percent, but obviously his probability of paternity would be zero. Similarly, barring divine intervention, if Walter were the only person to have had intercourse with the mother, his paternity index would be 178 and his “chance of paternity” would be 99.4 percent, but his probability of paternity would be 100 percent. Of course, most cases will fall between these extremes.

The usual method for calculating the probability of paternity is Bayes’ formula. See McCormick, supra, at 659; Ellman & Kay, supra, at 1147-52. This formula demonstrates the effect of a new item of evidence on a previously established probability. In this instance, the new item of evidence is the blood test result (i.e., the paternity index), and the previously established, or prior, probability is the probability of paternity based on the other evidence in the case. The calculation is simplified if the probabilities are expressed as odds. The odds of paternity using Bayes’ formula are simply the product of the prior odds and the paternity index. See McCormick, supra, at 660. For example, if the other evidence in this case had established that Walter was sterile, his prior odds of paternity would have been zero. His odds of paternity, then, would have been zero times his paternity index of 178, which equals zero. On the other hand, if the other evidence in the case had led one to believe that the odds that Walter was the father were one to one, his odds of paternity would have been one times 178, which equals 178 or a probability of paternity of 99.4 percent. It can be seen, then, that the paternity index will equal the probability of paternity only when the other evidence in the case establishes prior odds of paternity of exactly one. This is because the paternity index’s comparison of the putative father with a randomly selected man is mathematically equivalent to the assumption that the prior odds of paternity are one to one, or that the prior probability is 50 percent. Reisner & Bolk, supra, at 674.

Lovrien did not derive a statistic for probability of paternity, although, as we note below, his testimony frequently referred to the paternity index and its equivalents in language suggestive of a probability of paternity. If the paternity index or its equivalents are presented as the probability of paternity, this amounts to an unstated assumption of a prior probability of 50 percent. Reisner & Bolk, supra, at 674.

II.

Whether the paternity index and its equivalents were admissible is controlled in the first instance by ORS 109.258, which in relevant part provides:

“If * * * the blood tests show the possibility of the alleged father’s paternity, admission of this evidence is within the discretion of the court, depending upon the infrequency of the genetic marker.” 6

The words “this evidence” refer to “the possibility of the alleged father’s paternity,” but it is not clear whether this phrase is limited to the bare possibility of paternity or includes a statement of the degree of possibility.

ORS 109.258 is virtually identical to section 4 of the Uniform Act on Blood Tests to Determine Paternity (UABT), which was enacted in Oregon in 1953. 7 Or Laws 1953, ch 628, § 4. When the UABT was proposed in 1952, courts admitted only evidence of exclusion; blood test results consistent with paternity were inadmissible. See 9 Uniform Laws Annotated 103-04 (1957) (commissioners’ prefatory note 2). One objective. of the UABT was to make results consistent with paternity admissible. The commentary to the UABT states:

“[I]f the [blood] test showed that the alleged father could be the father of the child and if the blood type discovered disclosed that the type of blood and the combination in the child was of a rare type and that it would be infrequent to find such a combination of blood, * * * such evidence ought to be admissible as evidence of proof of paternity.”

9 Uniform Laws Annotated 103 (1957) (commissioners’ prefatory note 2). We infer that “such evidence” refers to the infrequency of the blood type. This interpretation is consistent with the statutory language, which directs the court to exercise its discretion to admit blood test evidence in accordance with the “infrequency of the genetic marker.” No mention is made of calculations of relative or absolute probabilities of paternity. For these reasons, we conclude that whether a putative father’s paternity index, probability of paternity and similar statistical calculations are admissible is a matter to be determined by generally applicable laws of evidence.

The admissibility of expert testimony is governed by three general constraints. First, expert testimony must be relevant. See OEC 402. Relevant testimony is testimony

“having any tendency to make the existence of any fact that is of consequence to the determination of the action more probable or less probable than it would be without the evidence.”

OEC 401. Second, expert testimony is subject to OEC 702:

“If scientific, technical or other specialized knowledge will assist the trier of fact to understand the evidence or to determine a fact in issue, a witness qualified as an expert may testify thereto in the form of an opinion or otherwise.”

Finally, expert testimony must not be unduly prejudicial, confusing or time-consuming:

“Although relevant, evidence may be excluded if its probative value is substantially outweighed by the danger of unfair prejudice, confusion of the issues, or misleading the jury, or by considerations of undue delay or needless presentation of cumulative evidence.”

OEC 403. Put more succinctly, in evaluating the admissibility of expert testimony, “this court must identify and evaluate the probative value of the evidence, consider how it might impair rather than help the factfinder, and decide whether truthfind-ing is better served by exclusion or admission.” State v. Brown, 297 Or 404, 409, 687 P2d 751 (1984).

Walter contends that Lovrien’s testimony regarding the paternity index and its equivalents was not probative because, it provided the trier of fact with no information beyond that provided by the “probability of excluding a falsely accused father.” This contention is not correct and is based on a misunderstanding of these statistics.

As we noted above, the “probability of excluding a falsely accused father” measures only the power of the test to exclude any falsely accused father; it does not describe the percentage of the relevant male population that could be the father in a particular case. It also does not reveal the relative likelihoods of paternity among the men not excluded by the blood test. This relative likelihood is conveyed by the paternity index, and for this reason the paternity index has probative value beyond that provided by an exclusion statistic. For example, even if the blood test reveals that the putative father is a member of a small percentage of the population capable of fathering the child, his paternity index, and hence his relative likelihood of paternity, may also be very small.

In addition, the presentation of paternity test results in terms of probabilities or likelihoods of paternity rather than population percentages may enable the trier of fact to understand better the significance of the test results, see Ellman & Kaye, supra, at 1146; Joint AMA-ABA Guidelines, supra, at 260-63; Finkelstein & Fairley, A Bayesian Approach to Identification Evidence, 83 Harv L Rev 489, 502 (1970), and thereby satisfy OEC 702. First, a population percentage may be too easily confused with the probability of paternity. From the statistic that 97.5 percent of the population is incapable of being the father, the trier of fact may wrongly jump to the conclusion that there is a 97.5 percent probability that the putative father is the true father. 8 Second, even if it is explained to the trier of fact that the population percentage is not the probability of paternity, the trier of fact may be left wondering what the probability of paternity is and what significance the population percentage has for it. See Ellman & Kaye, supra, at 1146.

We conclude that the paternity index and its equivalents are probative. This conclusion, however, does not end our inquiry. We must still assess whether, as asserted by Walter, the probative value is substantially outweighed by any prejudice and confusion engendered by its presentation.

The fundamental problem in the presentation of blood test results is conveying to the trier of fact the need to integrate the blood test results with the other evidence presented in the case. Although there have been many recent advances in paternity testing, leading to very high probabilities of excluding falsely accused fathers, paternity testing alone cannot yet prove paternity. Other evidence must narrow the number of potential fathers so that the test results become meaningful. The presentation to the trier of fact of the putative father’s “paternity index,” “chance of paternity” or “probability of paternity” creates several difficulties for conveying to the trier of fact the need to integrate these statistics with the other evidence in the case.

First, the expert is unqualified to state that any single figure is the accused’s “probability of paternity.” As noted above, such a statement requires an estimation of the strength of the other evidence presented in the case (i.e., an estimation of the “prior probability of paternity”), an estimation that the expert is in no better position to make than the trier of fact. 9 If the expert were to make such an estimation, it could not satisfy OEC 702’s requirement of assistance to the trier of fact. 10 Cf. Tiedemann v. Radiation Therapy Consultants, 299 Or 238, 243-44, 701 P2d 440 (1985) (expert opinion that medical treatment was “not negligent” did not comply with OEC 702).

Second, the paternity index and its equivalents frequently are confused with the probability of paternity. See McCormick, supra, at 661; 1A Wigmore, Evidence § 165b, 1821 (Tillers rev 1983). This case is an example. Lovrien did not calculate a “probability of paternity,” but his testimony regarding Walter’s paternity index and its equivalents could easily have been confused with a probability of paternity. Lovrien correctly described the paternity index as the chance that the child received its “genes from this man [Walter] compared with just an average man.” Lovrien also testified, however, that the index meant that the odds were “178 times to 1 that he [Walter] is the right father,” that a paternity index of 178 meant that “the chance that Brent Walter is the father is extremely likely,” that Walter’s “chance of paternity” was 99.4 percent, and that the “chance he is not the father based upon these [test results] is 0.6 percent.” We doubt that any of the jurors would have made a distinction between the likelihood that Walter, rather than a randomly selected man, was the father (which is what the paternity index and its equivalents measure) and the probability that Walter was in fact the father.

Finally, the paternity index’s comparison of the putative father with a randomly selected man is only indirectly relevant to the issue the trier of fact must decide. “The [trier of fact’s] function is not to compare a defendant with a person selected randomly but to weigh the probability of defendant’s [responsibility] against the probability that anyone else is responsible.” Finkelstein & Fairley, supra, at 502 (emphasis in original). The comparison with a random man is obviously probative, but there is a danger that the trier of fact will accord it too much weight when it comes to decide whether the putative father is the true father. For example, if an individual purchases 178 of a total of 100,000 raffle tickets, and if the average person purchases one ticket, that individual will be 178 times more likely to win the raffle than a randomly selected person. The individual’s chances of winning the raffle prize, however, are still minuscule because the proper comparison is not with the number of tickets purchased by a randomly selected person, but with the total tickets purchased by everyone. Similarly, Walter is 178 times more likely to be the father of Plemel’s child than a randomly selected man, but this figure standing alone is not particularly meaningful. It is important for the trier of fact to understand that the paternity index must not be taken as practically conclusive evidence of paternity but should be considered in conjunction with other evidence presented in the case. 11

Where the determination whether the probative value of evidence is substantially outweighed by the dangers set forth in OEC 403 must be made on a case-by-case basis, we ordinarily defer to the determination of the trial court. See, e.g., State v. Johns, 301 Or 535, 555, 725 P2d 312 (1986) (admissibility of prior acts to be determined on a case-by-case basis). We conclude that this is not such a case. The probative value of the statistics derived from blood test results and the dangers in their presentation to the trier of fact will be substantially the same in every case. 12 This court, as an appellate court, should determine the admissibility of this evidence. Cf. State v. Brown, supra, 297 Or at 442 (probative value of polygraph evidence outweighed by reasons for its exclusion).

Evidence of the putative father’s paternity index and its equivalents is highly probative but also presents a substantial danger of misleading the trier of fact. For that reason, we conclude that this evidence should be admissible, but only subject to certain conditions. The purpose of these conditions is to convey to the trier of fact the significance of this evidence and the need to weigh it with the other evidence presented in the case.

First, the paternity index is admissible so long as the expert explains that the index is not the probability that the defendant is the father, but measures only the chance that the defendant is the father compared to the chance that a randomly selected man is the father. The expert should also not be allowed to use misleading formulations of the paternity index such as “the chance of paternity” and “the chance of nonpaternity” without making this qualification. If the expert’s testimony does not satisfy this condition, the objecting party is entitled to have this testimony stricken and the jury instructed to disregard it.

Second, the expert, whether testifying in person or by affidavit, see ORS 109.254(2), should never be allowed to present over objection a single figure as “the” probability of paternity. If the expert does so, the objecting party is entitled to have that testimony stricken and the jury instructed to disregard it. The reason for this limitation is that the probability of paternity cannot be stated mathematically without making certain assumptions concerning the strength of the other evidence presented in the case. Similarly, the expert should not be allowed to make statements such as “it is extremely likely” or “it is practically proven” that the defendant is the father.

Finally, as a corollary to the above conditions, if the expert testifies to the defendant’s paternity index or a substantially equivalent statistic, the expert must, if requested, calculate the probability that the defendant is the father by using more than a single assumption about the strength of the other evidence in the case. The expert may also so testify without request or without testifying as to the paternity index. This condition is not at odds with the second condition because this condition requires the expert to use various assumptions about the strength of the other evidence in the case rather than making a single assumption and presenting the probability calculated from that assumption as “the” probability of paternity. In this way the strength of the blood test results can be demonstrated without overstating the information that can be derived from them. If the expert uses various assumptions and makes these assumptions known, the factfinder’s attention will be directed to the other evidence in the case, and it will not be misled into adopting the expert’s assumption as the correct weight to be assigned to the other evidence. The expert should present calculations based on assumed prior probabilities of 0,10,20,30,40,50,60,70,80,90 and 100 percent. 13 If the expert is requested to do so and fails to make these calculations, the trier of fact should be, instructed to ignore the paternity index and its equivalents. Other statistics, such as the probability of excluding a falsely accused father and the proportion of the relevant population excluded by the blood tests, would still be admissible.

III.

With respect to the disposition of this case, we conclude that it is necessary to remand for a new trial. Lovrien testified that Walter’s paternity index of 178 meant that he was 178 times more likely to be the father of Plemel’s child than “an average man.” While that testimony was accurate, other statements made by him, as we noted above, could have led the jury to infer that the paternity index was the probability that Walter was the father. Lovrien also testified that the paternity index was based on the assumption that “the most logical person” had been identified and accused of being the father and that if there was a zero prior probability that Walter was the father, the paternity index would be meaningless. From these statements the jury could have inferred that the significance of the paternity index would depend to some extent on the other evidence presented in the case, but this relationship was not made clear by Lovrien. Perhaps particularly confusing to the jury was the following statement made by Lovrien near the close of his testimony:

“[Assuming that the right man has been accused or investigated, then the laboratory is used as a means of investigating, if he denies fatherhood, then you want a good laboratory test to say well if he is not the father, let’s show that he is not. And that is what we did. We used a good laboratory test here. And if he is not the father, he should have been excluded, but he wasn’t.”

On the record before us, we cannot say that the result would have been the same if Lovrien’s testimony had been presented under the conditions set forth above.

The decision of the Court of Appeals is reversed. The judgment of the trial court is reversed, and the case is remanded to the trial court for a new trial.

1 See Joint AMA-ABA Guidelines: Present Status of Serologic Testing in Problems of Disputed Parentage, 10 Fam LQ 247, 282 (1976).

2 Our discussion of paternity testing is drawn from McCormick, Evidence § 205B (3rd ed 1984); Ellman & Kaye, Probabilities and Proof: Can HLA and Blood Group Testing Prove Paternity?, 54 NYU L Rev 1131, 1132-52 (1979); Joint AMA-ABA Guidelines, supra; Reisner & Bolk, A Layman’s Guide to the Use of Blood Group Analysis in Paternity Testing, 20 J Fam L 657 (1982); see also American Association of Blood Banks, Inclusion Probabilities in Parentage Testing (1983); American Association of Blood Banks, Paternity Testing (1978).

3 One could infer from Lovrien’s response to a question on cross-examination that 2.5 percent of the male population was capable of fathering the child, but careful reading of his testimony shows that this was not the case.

4 If the testing excluded a total of 99 percent of the population as the potential father, it might be tempting to conclude that the probability that the putative father was the true father was also 99 percent. It is easy to see, though; that this is not the case. If the relevant population were 1,000,000, and if 99 percent were excluded by the testing, then 990,000 would have been excluded, leaving 10,000 as potential fathers. On the basis of the blood tests alone, then, the probability of the putative father’s paternity would not be 99 percent, but 1/10,000, or 0.01 percent. See 1A Wigmore, Evidence § 165b, 1819-20 (Tillers rev 1983); Ellman & Kaye, supra, at 1141; see also Tribe, Trial by Mathematics, 84 Harv L Rev 1329,1355 (1971).

5 The joint guidelines, however, appear to contemplate that these predicates will be used to describe the probability that the putative father is the real father, not just the probability that he is the father compared to a randomly selected man. For example, for a “likelihood of paternity” greater than 99.8 percent, the verbal predicate is “practically proved.” Because a randomly selected man is not the subject of paternity litigation, this predicate can only mean that it is “practically proved” that the accused is the true father.

6 At the time of trial, ORS 109.258 referred to “blood type” instead of “genetic marker.” The law was amended by Oregon Laws 1985, chapter 671, section 44b, to conform with advances in paternity testing, which is no longer limited to tests for blood types.

7 At the time of trial, ORS 109.258 was identical to section 4. In 1985 the legislature substituted in the portion quoted above the words “genetic marker” for “blood type.” See note 6, supra.

The UABT was proposed in 1952 and was adopted by only a few states. Even fewer adopted the quoted language from section 4. Section 4 became section 10 of the Uniform Act on Paternity in 1960. This act was also not widely adopted and was withdrawn in 1973 in favor of the Uniform Parentage Act. Section 12 of that act provides: “Evidence relating to paternity may include: * * * (3) blood test results weighted in accordance with evidence, if available, of the statistical probability of the alleged father’s paternity.” This section goes much further than section 4 of the UABT in that it explicitly permits admission of probability of paternity calculations. The Uniform Parentage Act has been adopted in approximately 15 states.

8 This confusion probably could be dispelled easily by the putative father’s attorney. See Ellman & Kaye, supra, at 1146 & n 78. The attorney need only note that, if the relevant population were 100,000, exclusion of 97.5 percent would still leave 2500 potential fathers. Absent other evidence, the probability of paternity would not be 97.5 percent, but 1 in 2500, or 0.04 percent.

9 The standard assumption in calculating the probability of paternity is that the prior probability of paternity is 50 percent. See McCormick, supra, § 211 at 659; Ellman & Kaye, supra, at 1149-51; Reisner & Bolk, supra, at 674. With this assumption the probability of paternity is equal to the paternity index. Reisner & Bolk, supra, at 674. Studies in Poland and New York City have suggested that this assumption favors the putative father, because in an estimated 60 to 70 percent of paternity cases the mother’s accusation of paternity is correct. See Ellman & Kaye, supra, at 1150-51. Of course, the purpose of paternity litigation is to determine whether the mother’s accusation is correct, and for that reason it would be both unfair and improper to apply the assumption in any particular case.

We also note that the Court of Appeals’ justification of the assumption of a 50 percent prior probability of paternity is erroneous for another reason. The court stated that the “paternity index shows * * * that [the] assumption was, if anything, unduly favorable to father.” 80 Or App at 253. The paternity index has absolutely no bearing on the prior probability of paternity; the prior probability is independent of the blood test results. See Tribe, supra, at 1366-68. t

10 We emphasize that the estimation is objectionable under OEC 702. Such an estimation would not necessarily violate OEC 703 or OEC 705. Under OEC 703, other evidence presented in the case could be made known to the expert, and the expert could then form an opinion as to the proper prior probability of paternity to use in calculating the probability of paternity. Under OEC 705, the expert could testify to the probability of paternity without initially disclosing the basis for the calculation. Such disclosure, however, must be made at the direction of the court or on cross-examination.

11 For a comprehensive critique of the use of statistical methods in the conduct and design of trials, see Tribe, supra. Because of the unique circumstances of paternity litigation, as opposed to other efforts at legal identification, many of Tribe’s criticisms are inapplicable to this case. See Ellman & Kaye, supra, at 1155-56.

12 The trial court, of course, retains its discretion under ORS 109.258 to refuse to admit evidence of the possibility of paternity where the blood test is unable to exclude a sufficient proportion of the relevant population. Given the strength of modern paternity tests, the need to exercise this discretion should rarely arise.

13 This final condition is derived from proposals made in Ellman & Kaye, supra. These proposals are discussed favorably in McCormick, supra, at 662.

At least two paternity testing laboratories in California have regularly reported calculations of probabilities of paternity based on assumptions of prior probabilities of paternity ranging from 10 percent to 90 percent in 10 percent increments. Peterson, A New Things You Should Know About Paternity Tests (But Were Afraid to Ask), 22 Santa Clara L Rev 667, 691 n 74 (1982). This requirement would impose a relatively minor burden on testing laboratories. Additional calculations based on other assumed prior probabilities may be presented with the permission of the trial court.