18 DR. CROW: This report is pretty extensive. We
19 responded -- I did -- to quite a number of requests to make
20 this user friendly. So it starts out at the third grade
21 level. It doesn't end up there. It becomes complicated by
22 the time it's over.
23 I would welcome advice from people in the room as
24 to how to make this more acceptable. I am also interested
25 in finding things that are in error, and I am interested in
1 things that are omitted and that should be included.
2 Our assignment was to look ahead five or ten years
3 and anticipate what technological developments would occur
4 during that time and what the impact of that might be. I am
5 as aware as you are of the difficulties in looking ahead in
6 a rapidly moving field, and I will remind you of one
7 terrible example of a person who tried to look ahead.
8 This was Alfred Russel Wallace, the co-discoverer
9 with Darwin of the theory of evolution. At the turn of the
10 last century he made some predictions about what would
11 happen in the 20th century, and they were atrocious. He
12 said that phrenology will finally come into its own. A
13 second thing that he predicted was that vaccination would be
14 shown to be a fraud. That sort of makes me pause somewhat
15 in making too many predictions.
16 Nonetheless, there are things that can be
17 predicted. One of them is, almost certainly, that 13 core
18 loci are here to stay for a while. There is enough
19 investment in time and effort and equipment on the part of
20 the people that use this that they are not likely to change
21 lightly to another system even if the other system is quite
22 a bit better. So one of our more confident predictions is
23 that we can have the 13 core loci for some time.
24 That is really pretty powerful. With a group of
25 13 core loci the match probability is about one in 200
1 trillion. That is considerably larger than the size of the
2 world population. I will remind you that the world
3 population passed the 6 billion mark last October. I find
4 it interesting as a geneticist that finally we have more
5 people in the world than we have basis in a single DNA cell,
6 which happens to be 6 billion also.
7 Actually, of course, that is a poor estimate,
8 because there is structure in the population and there are
9 likely to be close relatives. So I can't really say that
10 it's almost impossible for two people to have the same set
11 of 13 core loci, but unless they are close relatives, it is
12 exceedingly improbable.
13 Another thing that is a very certain kind of
14 prediction are the kind of technological advances that will
15 make these 13 core loci work better, more efficiently, more
16 rapidly, more miniaturized, and more automated. These are
17 all happening.
18 The speed of the operation is going up, or going
19 down, whichever way you like to say it, and the
20 miniaturizing is taking place rapidly. I've seen a number
21 of articles based on future predictions and in fact present
22 workings of systems that have hand-held computer systems
23 that use a very tiny amount of material and do the work very
24 rapidly.
25 For example, one by Ehrlich at MIT. It can handle
1 8 loci. It happens to be 8 of the 13. It can go through
2 the process in two minutes and classify all of these 8.
3 That can be done today in a laboratory. I don't want to
4 guess how soon this becomes robust enough that you can carry
5 it in your pocket, but it's pocket sized at least, and maybe
6 in two years, maybe in five years, certainly in ten years it
7 would be possible to take something like this to the crime
8 scene.
9 It is an exciting marriage of Silicon Valley and
10 microbiology in using the electronic tricks of one and the
11 microbiology of the other.
12 One of the 13 loci, THO, has a subset. So there
13 is a unit that isn't repeated exactly the same number of
14 units in the unit. To resolve that took ten minutes rather
15 than two minutes. The point is that this system looks
16 accurate enough to be put to use almost any time provided it
17 can be made robust enough to be of immediate use.
18 The other things that we can certainly predict
19 will happen.
20 There will be much more emphasis on mitochondrial
21 DNA in the future and it will require databases. It
22 requires a large database for mitochondrial DNA. Because
23 you can't multiply the components together, the probability
24 of finding it is limited by exactly the size of the
25 database.
1 So I think we can predict some large databases for
2 mitochondrial DNA, which is particularly useful, as
3 everybody in the room knows, because there are many
4 mitochondrial particles in a cell, and therefore a much
5 smaller amount of material can often be resolved.
6 It is also useful because it can trace the female
7 line of ancestry, because mitochondria is always transferred
8 from the mother to the children.
9 The Y chromosome has finally ceased being inert.
10 I have memorized for years and taught for years that the Y
11 chromosome doesn't have any genes except those that
12 determine sex. Actually, it has a very large number of
13 discernible DNA sites, SNPs and STRs. So the Y chromosome
14 is just about as good as any other chromosome now from this
15 standpoint, and it has some unique properties that make it
16 useful; it has some properties that make it confusing.
17 What makes it useful is that it traces the line of
18 the male ancestry from father to son. And it has been
19 particularly useful in ways that are not particularly
20 relevant for forensics.
21 I emphasize that, because probably we could be
22 misled if we wanted to use it in this way. The Y chromosome
23 is a good marker of human population structure. The Y
24 chromosome varies much more from population group,
25 geographical group to other geographical group, much more

1 than the X chromosome does, much more than the autosomes do.
2 Why is that? I think it's because of this human habit that
3 the females do the migrating. The males stay home and carry
4 females to the next colony, and the net effect of this is
5 that the female traits, mitochondrial DNA, are much more
6 uniform throughout the world than the Y chromosome.
7 I say this partly because it's interesting, but
8 partly because we might be pretty badly misled if we tried
9 to determine an individual's ancestry from what his Y
10 chromosome is. One example of this was there is a group of
11 people in Africa who claimed to have Jewish ancestry. I
12 think anthropologists to a person simply refused to believe
13 this.
14 But it has actually turned out that there is a Y
15 chromosome in this population that is the same as the Y
16 chromosome in the Egyptian area, in northern Africa. So the
17 tradition of father to son transmission has carried on this
18 particular gene along with its religion, but all the rest of
19 the genome is fairly scrambled. If you took this particular
20 Y chromosome as evidence of a person's skin color, you would
21 be pretty badly misled. You would say this person is
22 Jewish, whereas in fact he is black in that particular
23 population.
24 If we are going to continue to use the 13 loci for
25 database purposes, there is going to continue to be a need
1 for further refinements for case studies. One way to do it
2 would be just to expand the 13 to more and more loci. One
3 suggestion would be that particularly useful from this
4 standpoint are penta-nucleotide, 5-unit repeats. They have
5 less stutter than the smaller ones do, less ambiguity in
6 interpretation, but particularly nice from the standpoint of
7 analysis is that they tend to have a larger number of
8 alleles and therefore much more heterozygotic than the
9 smaller number of repeat units.
10 Certainly there is going to be much more use of
11 SNPs, single nucleotide probes, in the future, but I think
12 our best guess is that these will be used as an adjunct to
13 the STRs rather than as replacements. Maybe ten years down
14 the road these will be so much better that they will replace
15 the others, but I suspect that replacement will be gradual
16 and reluctant on the part of a great many people.
17 There are millions of sequences in the human
18 genome. We call them Alu. They originated probably with
19 some kind of a virus infection millions of years ago. They
20 are no longer changing now, I might add. But these leave
21 their track from a previous infection that happened sometime
22 in the human population. As I said, there are millions of
23 these. They could be exploited for forensic uses. I see no
24 immediate need for it right now, but there is a possibility
25 there for the future.
1 Finally, just as physical methods tend in the long
2 run to replace chemical methods, I think the physical
3 methods for analyzing DNA molecules will replace the
4 chemical methods. Mass spectroscopy methods are being
5 worked on right now. I don't understand them in the least,
6 so I'm not going to discuss any more about it, but I'm told
7 that they offer considerable promise for the future.
8 A bit about the population structure and
9 relatives. One principle that the committee likes, at least
10 I like, and I think the committee is persuaded, and that is
11 that we would really like to be able to have a sample from 2
12 people and compare it to 2 samples and would like to be able
13 to say that they are different even if there are close
14 relatives involved. One of the difficulties with the
15 present situation is we assume a random population, which
16 can't strictly be true, and we assume no structure in the
17 population, or we have taken it into account with a rather
18 crude correction.
19 It would be nice to have a system that is robust
20 enough that you don't have to worry about it. That is, a
21 system that would distinguish any two persons however
22 closely related they are. There is a property of sibs,
23 brothers and sisters, that makes it very convenient from
24 this standpoint, and we are finally beginning to have enough
25 loci to make this useful. For any particular locus, the
1 probability that two brothers would match exactly at that
2 locus is exactly one fourth and it is one fourth for each
3 locus.
4 I emphasize that this one fourth comes not from
5 any considerations of population genetics or population
6 structure; it comes from simple Mendelism. There are four
7 different ways of putting these genes together and a fourth
8 of the time they will be the same. Furthermore, the
9 different loci are independent of each other whether there
10 is linkage equilibrium in the population or not.
11 What I am saying is that if we use this factor of
12 one fourth, we are dealing with something that is very
13 robust with respect to departures from traditional
14 assumptions. One fourth is not the whole story because
15 there are gene frequencies that add to this also, but they
16 are small, and the one fourth dominates the equation.
17 So I think in the future we are going to find it
18 possible to tell any two individuals apart even if they are
19 close relatives, even if they are brothers, which is the
20 closest relative from this particular standpoint.
21 What does that do to the match probability? With
22 13 loci, the match probability becomes about 1 in a million.
23 If you raise that to 21 loci, it's about 1 in a billion.
24 What I am suggesting here is that maybe the courts would
25 prefer robust evidence that nobody could question: 1 in a
1 million as having more persuasive power than 1 in a billion
2 where you are not quite so certain about the assumptions.
3 Anyhow, I think we can predict that this will be a
4 way of the future. Maybe not the only way to do the
5 calculation, but one way in which it could be done.
6 This also means that racial differences are less
7 important. If I ask for this calculation, thinking of
8 brothers, whether I've taken the database from the right
9 racial group makes much less difference than it does
10 ordinarily just because of the dominance of this factor of
11 one fourth that intrudes.
12 The much maligned 1992 NRC report introduced what
13 it called the ceiling principle, which is roundly denounced.
14 The purpose of that was a laudable one, and that was to make
15 the process such that you wouldn't have to take into account
16 geographical or racial or ethnic differences, and we can
17 approach that by having a system that is robust enough to
18 detect any two people however closely related.
19 This is for the future, but I don't think it's
20 very far in the future at the rate at which new loci are
21 being discovered.
22 I want to talk a little bit about identifying
23 group membership or individual traits from a DNA sample. I
24 will talk first just about the 13 core loci.
25 Charles Brenner, a few years ago, back in the
1 classical antiquity of VNTRs, had a sample of 5 VNTRs from a
2 serial rapist. For some strange reason, they didn't know
3 what racial group this person belonged to. He found that
4 the likelihood ratio of white as compared to black in the
5 population was 45. In other words, this particular set of
6 VNTR loci was 45 times as likely that it came from a white
7 person as it came from a black person. That turned out to
8 be correct, and the person did turn out to be white.
9 I want to ask what happens not with VNTRs, but
10 what happens with the 13 loci. Charlie Brenner,
11 fortunately, has done such a calculation. The average
12 likelihood ratio is something like 70. In other words, if I
13 have a sample that comes from a black person tested with a
14 white database, what is the expected average likelihood
15 ratio? It's about 70. So right now with the 13 loci that
16 are being done routinely one could make a pretty strong
17 guess as to the racial makeup of the person that left that.
18 It's much less effective in distinguishing
19 Hispanics, to nobody's surprise. That is not a very good
20 biological classification, because it's based on language,
21 and there has been no research as far as I know about
22 orientals in this regard.
23 That is what could be done just using the existing
24 13 loci. If we are going to ask further questions about
25 group identification, then we will have to go outside this
1 particular system.
2 But there are other loci. There is the Duffy
3 blood group that is associated with malaria resistance and
4 therefore is very common in parts of Africa and virtually
5 unknown elsewhere. So it's a pretty strong indicator of
6 African ancestry.
7 Again, I would caution that finding one gene
8 pointing to African ancestry doesn't tell you very much
9 about the person's phenotype. A person could have a very
10 light skin color and still have one gene that had its
11 origination in Africa.
12 So our committee looks forward to the time in
13 which we can pay less attention to racial and group
14 differences and more attention to individual trait
15 differences.
16 How are things coming in this field? I don't
17 know, because most of the work that is going on here is
18 isolated individuals studying individual traits and I'm not
19 aware of much in the literature, but certainly it should be
20 possible, as I look forward 5 or 10 years, to identify from
21 a DNA sample whether a person had red hair or not, or was
22 bald if it's a male, had straight hair or dark skin, or was
23 color blind.
24 But once I go through a list like this, I've taken
25 the obvious ones, and there are not an awful lot more of
1 conspicuous traits. So if we are going to try to identify
2 racial features or height, weight, things like this, then we
3 are getting into the realm of polygenic multifactorial
4 traits, with all of the statistical problems that that
5 entails. I don't doubt that within 10 years one could say
6 quite a bit about a person from the DNA profile, but I think
7 it's pretty hard to predict right now how you go about doing
8 it.
9 What about identical twins? Every time we talk
10 about this subject we say all this works except for
11 identical twins. I think it's quite likely that in the next
12 10 years, and maybe sooner, it will be possible to identify
13 identical twins. We could use two kinds of principles.
14 There may be more, but two obvious principles.
15 One is to find regions of the genome that are very
16 highly mutable so that therefore one twin would have mutated
17 more than the other. The immunoglobulin loci have that
18 property, and it should be possible right now in principle
19 to distinguish a pair of twins by the fact that they would
20 have different antibody repertoires.
21 Another possible way is that each of us is unique
22 in the parasites that we carry. Our intestinal bacterial
23 and the various kinds of other organisms that we carry
24 around with us are presumably unique and they are different
25 between identical twins. So maybe we can identify a person
1 not by his genes, but by his parasites. I think it could be
2 done. It's just a matter of how much effort that people
3 want to put into that.
4 This is high tech virology now, but there are
5 viruses now that enter into the organism, join in with the
6 DNA and become incorporated in the DNA. The retroviruses do
7 this. If each of these two twins has a different set of
8 infections, or not necessarily infections but just viruses
9 that cause no problems, these viruses might intrude
10 themselves into the DNA, and they won't do it in the same
11 spots in the two individuals. So by identifying the site of
12 introduction of the virus into the tissue it might be
13 possible to separate twins.
14 Whether any of these three has any practical value
15 or will in 10 years, I don't know, but it's not beyond the
16 realm of possibility that identical twins, like the rest of
17 us, turn out to be resolvable.
18 Another point I want to mention briefly is partial
19 matches. It is alluded to in David's report, and that is a
20 case in which the DNAs were not identical but very, very
21 similar and therefore likely to have come from close
22 relatives.
23 One can say quite a bit qualitatively and
24 certainly by use of likelihood ratios quantify this. It
25 would be a property of brothers and sisters that they will

1 agree exactly, that is, both allele locus, a substantial
2 fraction. It's one fourth on the average, as I said a while
3 ago. So a way to look for sibs would be to look at
4 collections of loci that match exactly. That has happened
5 in the past and it will certainly happen more often in the
6 future, and the larger number of loci that are concerned,
7 the greater the probability that this could happen.
8 Another relationship that strikes you in the eye
9 the minute you see it is parent and child. As you realize
10 from just knowledge of parents and children, a parent and a
11 child always share exactly one gene at any particular locus
12 but never more than one. So you can look for a series of
13 single gene matches with very few double gene matches and
14 identify parents and children. So we can tell sibs; we can
15 tell parent and offspring pretty clearly with, say, 13 or
16 more loci.
17 Other degrees of relationship it is not so
18 obvious. There is no way that I'm aware of in which I can
19 distinguish between half sibs and an uncle and nephew,
20 because they share the same fraction of their genes.
21 Cousins are 50 percent as large a fraction. It's probably
22 impossible to make the distinction.
23 Probably this means that we would have sort of
24 like the so-called primitive number system, one, two and
25 many. We could say sibs, parent and child, and all others,
1 and probably make a pretty good guess as to what other
2 possibilities there are but not specifically what they are.
3 I have the impression from Paul that Virginia
4 prohibits the use of this kind of information, and it is
5 very likely to differ from state to state. But at least the
6 technical possibility is there now and could be used on a
7 wide scale if it was so desired.
8 What about database searches?
9 In the 1992 report the report worried about
10 finding suspects by searching a database and suggested that
11 you don't use the database in the case study but pick up
12 other loci which would be independent of how you found the
13 data. The 1996 report, which David and I are well aware of,
14 suggested a numerical correction for this, taking into
15 account the database size.
16 One thing that doesn't come out in discussions of
17 this, at least a problem I've had in talking about this with
18 people, is that both of these studies were thinking of
19 databases that were equivalent to a random sample of the
20 population.
21 We were not thinking of databases from convicted
22 felons. Taking the recidivism rates that are given in your
23 report, it looks as if the prior probability of finding a
24 match from a felon database isn't much different than the
25 prior probability from a case that is picked up from
1 eyewitness identification or some other mechanisms. So I
2 think probably the mechanism that tends to be done -- it's
3 certainly done in Britain -- simply ignoring the fact that
4 this came from a database, probably isn't too bad.
5 But if we have a database of the entire population
6 and then find a suspect by searching through what amounts to
7 a random sample of the population, then you will certainly
8 have to have some kind of a correction for the fact that
9 this was done.
10 This is a topic of great dispute among the members
11 of our committee over a fine point that most of you don't
12 really care about, whether the likelihood ratio is better
13 than the match probability as a way of dealing with these
14 and how you make the correction. What I would say is that
15 the likelihood ratio is in the data, but the prior
16 probabilities will differ in accordance with the manner in
17 which you got the data.
18 I think it may be that if we ever find people from
19 the entire population databases that maybe the much maligned
20 recommendation in the 1992 report that you don't test on the
21 same individuals that you discovered on the principle that
22 you don't test a hypothesis on the data that generated it --
23 that is pretty far in the future and highly uncertain
24 because we don't know what is likely to happen with respect
25 to general population databases.
1 One final remark. Our report says nothing about
2 laboratory standards, error correction and error protection
3 on the assumption that this is the business of the DAB, but
4 I would reiterate the main conclusion of our 1996 report,
5 and that is that the best protection that an innocent person
6 has is not higher and higher standards, which of course are
7 necessary, but is really the opportunity for a retest.
8 That's it.
9 CHIEF JUSTICE ABRAHAMSON: Thank you, Jim.
10 Somebody had a comment.
11 MR. REINSTEIN: Just from the idiot judge
12 standpoint. You asked some questions in the report about
13 what we thought about redundancy. For example, the history
14 and the glossary, and what not. I think that is good. We
15 have a glossary and we have a biological issues portion in
16 the post-conviction report, but somebody who wants to look
17 at technological advances in the future may not even be
18 concerned about post-conviction. I think that's good.
19 Also, I thought the historical approach was really
20 valuable for people who just are not in the field, judges
21 and prosecutors who don't deal with this on a regular basis.
22 I think it's a really good primer.
23 So aside from me reading the report and my brain
24 shuts down when I see a formula, I thought everything else
25 was great. I think the population issues are really
1 important, but most lay people are going to shut down on it.
2 MR. CROW: There was a lot of blood, sweat and
3 tears that went into those introductory elementary sections,
4 but I hope it has helped. We certainly do invite readers to
5 skip that part.
6 MR. GAHN: When I read this a few days ago, I
7 thought this was terrific. However, I got a little
8 concerned when I got to page 44 and page 45. I would like
9 to talk just a little bit about the uniqueness issue and the
10 database search.
11 Correct me if I am wrong. As far as addressing
12 the uniqueness issue, are we backpedaling a little bit from
13 NRC II? I got the sense reading NRC II that they put out
14 the probability that the time will come when uniqueness will
15 be able to be stated. I think they even threw out that once
16 these 13 core loci are in place -- the numbers of years that
17 I've been at the PROMEGA conferences and just talking to
18 those in the field, I always had a sense that these 13 core
19 loci was sort of going to be the answer to the uniqueness
20 issue, that that's the person. I know in the state of
21 Wisconsin when our crime lab does these cases the numbers
22 are in the sextillions or quintillions now.
23 It seems like we are backpedaling a little from
24 NRC II. Can't we make a stronger statement about uniqueness
25 than what is said here?
1 MR. CROW: I'm sensitive to this. The real issue,
2 as we all know, is that uniqueness is a scientific question
3 and we really want a political solution to it.
4 I didn't talk about it now, but it is in here.
5 What the FBI advocated is if you ask if the probability is
6 less than the reciprocal of the United States population and
7 then put in some fudge factors taking care of the
8 uncertainty -- and these are large fudge factors; they are
9 factors of 10 in 2 different cases -- that if the
10 probability is so small that it's less than the reciprocal
11 of the population of United States and you are rather sure
12 of that, then there is every reason to say that person is
13 unique.
14 To get a group of statisticians to make this
15 statement is something quite different. I think what has to
16 happen is that some political or legal group has to just say
17 that if the probability is less than such and such we will
18 call it unique and make that as a political statement.
19 MR. GAHN: Could we make a stronger statement as a
20 Commission about uniqueness?
21 MR. CROW: Maybe the Commission could. Just don't
22 ask my working group to do that.
23 MR. GAHN: If we were to survey all the crime
24 labs, shall we say, in the country, I suspect most of the
25 ones that get a match with the 13 core loci are calling it
1 source or source attribution.
2 MR. CROW: The 13 loci has a probability of about
3 2 trillion. That is 3 or 4 or 5 orders of magnitude smaller
4 than the world population. It is what almost any of us
5 would call unique but we don't want to say it.
6 MR. GAHN: Maybe this is an area we could look at
7 again and that the Commission should decide. We still seem
8 to be on the fencepost here with this uniqueness issue.
9 MR. CROW: I would like to pass it on to the
10 Commission as we have done with everything else that is
11 hard.
12 [Laughter.]
13 MR. SCHECK: With all respect, Norman, I think it
14 is a strategic error looking at it from the point of view of
15 prosecutors and judges. The numbers, no matter how you
16 express it, likelihood ratios, frequencies, everything else,
17 are overwhelming, particularly in a case where there is any
18 kind of corroborative evidence. That is where the focus
19 ought to be.
20 What I get from this is there is a sufficiently
21 vigorous debate within the statistical community that if you
22 try to engage in fudge factors and everything else to use
23 uniqueness, which Dr. Crow is characterizing as a political
24 solution so you can say it, you are opening yourself up to
25 more battles of experts, more unnecessary money spent on
1 debating arcane statistical issues, and it is a mistake.
2 I would actually say, with respect again to this
3 same issue of database searches, that the 1992
4 recommendation and the way that even in the UK that you get
5 a hit on 13 core loci, we've got plenty of STRs. You are
6 going to bring a guy to trial for a serious crime that
7 actually is contesting it, which is going to be a marginal
8 number of cases. You are going to have plenty of samples.
9 Just do it again and don't engage in a statistical debate
10 where there are some very significant statisticians and
11 geneticists who will come in and argue with you. It's nuts.
12 It's a huge waste of time. I would sure like to see the
13 court money spent on other things.
14 In terms of kicking it to the Commission, I tried
15 to wrap my brain around this and look at a lot of the issues
16 for a while. I don't feel competent to evaluate it. I've
17 been litigating this statistic since 1989. I honestly don't
18 feel comfortable. I still don't understand half these
19 formulas.
20 MR. CROW: You understand half of them?
21 [Laughter.]
22 MR. SCHECK: I don't really think in some ways we
23 are competent even to make that decision. It only skews the
24 debate in the sense that we all know that what is going to
25 happen is that the ultimate issue in court is going to be
1 "who made this decision?" when they start talking about it.
2 With the statisticians and population geneticists, it's a
3 bunch of us, and how much significance that has.
4 Frankly, for all practical purposes we are stating
5 what the positions are and ducking.
6 MR. CROW: Statisticians are trained to make fine
7 distinctions, and they do. Let me say this, Barry. Even if
8 one understands every one of these formulas in full detail,
9 the uncertainty is still just as great in this regard.
10 CHIEF JUSTICE ABRAHAMSON: Norm.
11 MR. GAHN: Also in the database search, as I read
12 this, there is a hit in the data bank with the convicted
13 offender at the 13 core loci. We go back. We do a search
14 warrant and get the blood from the convicted offender and go
15 back and look at the evidence. What went to the CODIS was
16 semen from a vaginal swab. We take the cervical swab now,
17 do the semen and test for those 13 core loci and come up
18 with a match.
19 Are you stating that we should be telling our
20 crime labs, now I want you to give me 5 penta-nucleotides
21 and 7 SNPs? Isn't that good, those 13 core loci?
22 MR. CROW: Let me say something. This is on my
23 own, but I would like to present it to the working group,
24 which is essentially what I said a while ago.
25 We weren't -- I wasn't at least -- aware of the

1 recidivism rates that are reported in David Kaye's statement
2 from laboratories of convicted felons. That means that if
3 you find a match, the prior probability is somewhere between
4 5 percent and 95 percent or 1 percent and 99 percent. In
5 any case, the likelihood ratio of 10 to the 15th will swamp
6 that.
7 I think that one should, if the data come from the
8 convicted felon database, simply take the likelihood ratio
9 at face value.
10 To be clear, if we ever go into having databases
11 of the general population, that changes the situation.
12 MR. CLARKE: I have the concern that Norm did. I
13 don't understand as a lay person what is wrong with
14 repeating the same markers. In other words, the database
15 match which has been used for investigation purposes. Then
16 when a redraw is made, what is wrong with using the same
17 markers?
18 MR. CROW: You are testing against the accuracy of
19 the previous test. You are not testing against the
20 representativeness of that. You really have to take 13
21 different ones.
22 MR. CLARKE: But isn't what law enforcement is
23 doing independent of any investigative lead and they are now
24 attempting to determine the likelihood, now that we have
25 completed the independent test, that this person or a random
1 person could have left that sample? That's what I don't
2 understand. In other words, what's wrong with that
3 independent figure, which may be 1 out of a million or may
4 be 1 out of a trillion?
5 MR. CROW: I think technically it's not so good,
6 but I think practically it's the right thing to say.
7 MR. SCHECK: First of all, Norm, just to take your
8 example, I wouldn't retest the cervical swab. I'd go back
9 to the original vaginal swab, because on the cervical swab
10 you have a mixture or some other kind of problem.
11 I'm not saying we should necessarily take
12 positions on these things, but we could throw out
13 possibilities. Why wouldn't you as a prosecutor want to say
14 we took part of the vaginal swab, we got a hit on 13 loci,
15 we have now identified an individual and arrested him on
16 that basis? Okay, defense, we are proposing (a) to send out
17 the second part of the swab for additional testing with
18 different loci and/or have you do it.
19 That gets rid of the replication problem. That
20 sinks the defense. That is more persuasive to the jury, and
21 it avoids statistical issues.
22 MR. CLARKE: I just don't think the statistical
23 issue is there as an end user. For instance, how is Jan
24 going to start operating a laboratory doing not only the 13
25 loci? Now she is going to have do so many more in addition
1 to that.
2 MS. BASHINSKI: You have just, with that
3 statement, doubled the workload and the developmental
4 overhead of all the laboratories. That would turn around
5 everything we are trying to do, which is to increase the
6 capacity of the laboratories to do more cases without
7 suspects and to deal with the data bank.
8 I think what we need to really pursue is this a
9 practical and/or a legal issue or political issue. There
10 are distinctions and very fine things that are stated based
11 on population genetics. As a matter of practicality, when
12 you apply them to everyday life, it's a distinction without
13 a difference. I think we really need to figure out if that
14 is where we are here, and I think that is where we are here.
15 Because if that's the case, you are asking to derail the
16 progress we have made so far.
17 MR. CROW: It's almost like taking a fingerprint
18 search. Let's say by computer search there are 10 points of
19 identification that match, using that term loosely, and then
20 telling the examiner in the reexamination, you have to
21 ignore those now.
22 MS. BASHINSKI: Right. Find another 10 or another
23 finger.
24 MR. CROW: I was just saying that if these came
25 from a felon database search, you probably wouldn't need to
1 do that.
2 How about this, David? You've thought about this.
3 I'm using partly your data here. Not your data, but your
4 compilation of it.
5 MR. KAYE: I guess I would reinforce one thing
6 that was said earlier. A month ago I was at the Fourth
7 International Conference on Forensic Statistics listening to
8 various statisticians disagreeing with the 1996 and the 1992
9 report arguments that there needs to be any adjustment for
10 database search.
11 Taking the position that Norm and Woody have taken
12 here, I still don't think there is closure within the
13 statistical community, but I don't think it's crucial. I
14 don't think it's that important what this Commission does or
15 doesn't do, because frankly I think experts can come into
16 court on the basis of their understanding of statistics,
17 present a result, and say that when you have got enough
18 matches even in a very large database, as we start talking
19 about the 13 and more loci, that this is secure evidence it
20 is this person by virtue of uniqueness, and that kind of
21 thing.
22 I think the courts are going to work it out
23 regardless of this group. To really understand this
24 statistical issue would require an awful lot of effort.
25 It's a conceptual point that I gave up trying to understand
1 when I was on the committee.
2 MR. CROW: In a way we have been through this war
3 before in the previous committee.
4 CHIEF JUSTICE ABRAHAMSON: David, would you state
5 your full name and affiliation for the record.
6 MR. THOMA: Sure. The last comments can be
7 attributed to David Kaye. I'm the reporter for the Legal
8 Issues Working Group.
9 CHIEF JUSTICE ABRAHAMSON: Thank you.
10 Jeff, you have been very patient.
11 MR. THOMA: I'm just trying to add my two cents
12 from a practical perspective. I really have to agree here
13 that the issue is probability. It's not uniqueness.
14 Uniqueness is not something that has been proved, will
15 necessarily be proved. Until and unless you absolutely had
16 every single person and an inordinate amount of loci, it
17 clearly cannot happen for Norm to ask -- I appreciate the
18 vigor with which he is asking -- this Commission to overturn
19 what research and development and those people have a better
20 handle on than the members of this Commission.
21 Several of us do this practically. I have been on
22 the other side of Woody doing it, and I know Barry has as
23 well. But it is probability. That is what we are left
24 with, and you can't just override that and say X is Y
25 because I want it to be, because it would resonate louder in
1 jurors' minds. You are really going to have a backlash of
2 the greatest order and you are going to have more of a
3 fight, as Barry states.
4 So leave it at probability. Go to whatever amount
5 of probability that you are comfortable with or you believe
6 jurors are comfortable with ultimately and leave it at that.
7 I would vehemently object to trying to call it uniqueness.
8 MR. CROW: That seems wise to me.
9 Do you think that as a result of all this
10 discussion of statistics that that will filter over or
11 spread over or diffuse over into other kinds of
12 identifications which do not use statistics which ought to,
13 in my opinion? That is not anything that we need to be
14 saying, but it is a concern for me.
15 MR. SCHECK: Let me address one thing, Jan. I
16 wasn't suggesting that you necessarily do it in every case
17 such that your lab would have to have this huge capacity of
18 some other markers. Where this issue is going to arise as a
19 practical matter is when the only evidence is a database
20 hit.
21 In other words, it's your classic cold hit case
22 where they are bringing some guy in who says, I was in
23 Minnesota and you said this happened in Chicago. And nobody
24 has any other evidence that links this person to the crime.
25 Then all of a sudden it becomes a serious issue, and in that
1 kind of rare case, why not just get rid of it by doing
2 additional markers that can be found?
3 MR. CROW: Let me agree with Barry and add one
4 point. There are going to be instances arise in the future
5 in which 13 loci aren't enough. Mixed samples, for example,
6 can be confusing. There are going to be special cases, but
7 I think these are going to be rare.
8 MR. SCHECK: Mixed samples are not rare. I get
9 these calls all the time on multiple offenders with the
10 mixtures. There are plenty of state and federal cases
11 dealing with these mixtures that are driving people crazy.
12 MR. CROW: One of the things that we predicted
13 with great confidence is that computer programs are going to
14 be smarter than people in distinguishing mixtures probably
15 before very much longer. These are complicated problems.
16 I'm sure there will be good work on that.
17 CHIEF JUSTICE ABRAHAMSON: Paul.
18 MR. FERRARA: Let the record show that I've
19 managed for almost an hour and 15 minutes not to say
20 anything; I keep my mouth shut.
21 A couple points I want to make or emphasize. A
22 couple of years ago I wouldn't have agreed with Barry, but I
23 do agree with him now with regards to the issue of the
24 individuality and the uniqueness.
25 Now that the courts are getting over the phase of
1 having statisticians in the courtroom arguing ad nauseam
2 about whether the random match probability or likelihood
3 ratio is 1 in 100 million and not 1 in 10 billion, the
4 courts have come to accept an 8 locus match as almost a de
5 facto identification for all intents and purposes, as well
6 they may. Of course we do have access to additional
7 megaplexes to round out that to the full 13.
8 I had one case where an FBI expert tried to
9 testify in a Virginia case to uniqueness and all of his
10 testimony was struck because there was no basis for it.
11 The thing I'm all about and I know Jan is about
12 and the lab directors and the people in the laboratories and
13 the Commission really is work flow, getting work in and
14 getting work out. I for one don't want to go through the
15 terrible drain of staff on admissibility issues or debates
16 with independent experts, discovery motions and pretrial
17 hearings just to be able to get to say "that's the guy" as
18 opposed to "he cannot be excluded as a possible
19 contributor."
20 I have to agree.
21 With respect to the data bank searches, I've got
22 some experience on this that I will talk about. We have to
23 keep in mind that when we report a database hit to a law
24 enforcement agency there are no statistics. In fact, we
25 tell the investigative agency this is an investigative lead.
1 Go check this guy out, because we did a search in our data
2 bank and we made a hit.
3 If you, based on that information, can develop
4 probable cause to get an arrest warrant and get a sample
5 from him, then we will do a direct comparison, and when that
6 comparison is done, admittedly at the same loci, what we are
7 then reporting is either the random match probability or
8 likelihood ratio of finding that individual's genetic
9 profile in a population at random.
10 There is really a two-test here. Quite frankly,
11 from a practical standpoint, in Virginia, in most of the
12 forensic science laboratories it's almost a two-phased
13 approach. We do searches on using 8 loci. We have not made
14 a decision to run every piece of crime scene sample on all
15 13 loci. We will screen with one multiplex of 8, and then
16 when we want to determine if in fact there is a match, then
17 we have the power, the ability, the option and inevitably
18 the necessity when we make a match to go ahead and run the
19 full 13. So in effect there is already a two-phased
20 approach to the searching.
21 Of all of the data bank hits that we have had in
22 Virginia, which has gone through VNTRs as well as STRs, the
23 search hasn't necessarily had to be based on the total full
24 complement of loci. So I don't think it's a problem.
25 The other thing is, with the issue of

1 individuality, I'm not sure that the FBI DNA Advisory Board
2 isn't wading in on that one themselves.
3 MR. CROW: One thing I wish is that the FBI, if
4 they are going to advocate their particular statement, would
5 write it out and let us know what it is. I have access only
6 to one press release, which is clear enough, but that was
7 three years ago.
8 CHIEF JUSTICE ABRAHAMSON: We will try to get
9 that.
10 Chris.
11 MR. ASPLEN: The issues that Norm brought up and
12 Barry talked about and I think David Kaye crystallized have
13 also been expressed but kind of from a different end that I
14 think as executive director it's my job to communicate to
15 you.
16 In speaking at the CODIS Users Group meeting a
17 couple of months ago, in the public comment section after my
18 discussion about the Commission's activity the issue was
19 raised but on a different issue, kind of from the other
20 side. It had more to do with the statistical issues that
21 are talked about, and the sib principle.
22 The question was whether or not by nature of this
23 report addressing issues like that and how we address that
24 becomes something other than what the report was originally
25 intended to be, i.e., the difference between something that
1 professors just speak about, the technological future of DNA
2 evidence, as opposed to something that may well be used as a
3 tool in court to either support the proposition that we
4 should be taking a different statistical approach in a case
5 than what was actually taken on one end of the spectrum, or
6 on the other end of the spectrum, a tool that should be used
7 to make the point that we should be able to define
8 uniqueness or call uniqueness in this particular case.
9 So the question that really arises to the
10 Commission is, what is the Commission's intent on the
11 utilization of this particular document and how can we
12 affect that, if you will?
13 CHIEF JUSTICE ABRAHAMSON: Barry.
14 MR. SCHECK: I have a question that is sort of an
15 informational question that I think relates to this issue.
16 Dr. Crow, you mentioned this business with Brenner
17 and the idea that right now using the 13 core loci you could
18 develop an average likelihood ratio of something like 70 to
19 distinguish whether the person who left the unknown sample
20 is black or white. Do I understand that correct?
21 MR. CROW: Yes.
22 MR. SCHECK: Where is the underlying work or data
23 for that? Is that available anywhere?
24 MR. CROW: I got it by personal communication.
25 I'd have to get it from Brenner. I know how he did it.
1 MR. SCHECK: And what it is for Hispanics.
2 I don't think the DAB has to worry about something
3 like this. The point of Dr. Crow's recommendation, as I
4 understand it, in terms of distinguishing sibs, is pointing
5 towards a direction where you can get out of some very
6 difficult political and privacy problems and policy issues
7 which I think are uniquely the province of this Commission
8 as opposed to a purely technical issue about how you crunch
9 numbers or how you do a protocol in a lab, which I think is
10 probably more properly their jurisdiction.
11 Things like this, I think, should be of greater
12 concern to us, frankly. I'm a bit worried about those
13 numbers. I think before we even venture forth with that
14 other than stating it in the most general terms we should
15 know exactly what the status of that research is, because
16 this is one of those areas where obviously it could be a
17 valuable investigative tool. On the other hand, it could be
18 mishandled.
19 Frankly, for my money, I think the one failing we
20 have collectively is I don't think we have really addressed
21 these privacy and political issues carefully enough. Let me
22 give you one example.
23 There was a case recently decided by the 2nd
24 Circuit dealing with a town. The case went like this.
25 There was a crime in a white community, and the only
1 description they had was that the crime was committed by a
2 black person. It went out over the radio. They then
3 started stopping every black person in this town. The 2nd
4 Circuit upheld that on the grounds that if it had been a
5 description of a white person in a black community that had
6 committed the crime, they would justify the stops as well
7 because in this area blacks were comparatively less likely
8 to be and vice versa. That sort of supposedly evenhanded
9 this.
10 I can easily see where this kind of statistic or
11 this use of an investigative lead could justify all kinds of
12 things that may be more problematical socially and
13 politically than it's worth. These, I think, are the issues
14 where we may solve them, but I think we have to put out some
15 clear indicators that these things are going to become
16 issues sooner rather than later given the power that we have
17 here.
18 MR. CROW: I think we have a tough job pointing
19 out what could be done, which is not the same as saying that
20 it ought to be done.
21 MR. SCHECK: For example, when you were talking
22 about the retroviruses, and I know we can do it with HIV,
23 the first thing that occurred to me is, well, we can end a
24 lot of these things, because every time a convicted violent
25 felon leaves jail, why don't we just inject them with a tag?
1 CHIEF JUSTICE ABRAHAMSON: We'll mark you down as
2 saying that.
3 [Laughter.]
4 MR. SCHECK: I think that is going to be the next
5 suggestion.
6 MR. CROW: I am worried about talking too much
7 about retroviruses.
8 Maybe what we should do is really do this right.
9 Instead of taking just one particular example, what Brenner
10 did was take a random sample from the black population, test
11 that against the black population, and then repeated this
12 experiment over and over again. That's where this
13 likelihood ratio came from. One ought to do this not only
14 for that particular set but for all other large groups. It
15 wouldn't be that hard to do. It would be hard for me to do,
16 but it wouldn't be that hard for a computer expert to do.
17 CHIEF JUSTICE ABRAHAMSON: I would like to wind
18 this up if I can. I know we are running late on the break.
19 Woody had one addition. If it continues on, we will break
20 and come back to it.
21 MR. CLARKE: Actually, as far as the issue of
22 uniqueness, I think it is to some extent a practical
23 two-edged sword as well, because there have been instances
24 in my own county where an opinion is offered about
25 uniqueness. It doesn't carry as much weight as providing a
1 statistical estimate did. In fact, in one instance, a
2 lawyer said, that came across so poorly, I went into the
3 statistics, and I had no intention of doing so. So there
4 are at least two schools of thought about that.
5 The sib principle is the only other thing I wanted
6 to return to. I am concerned about the description. I
7 guess it is a subheading: A more reliable match criterion;
8 the Sib Principle. Is it correct to say more reliable or
9 perhaps more conservative and universal? Something like
10 that.
11 MR. CROW: Maybe robust is the word I want.
12 MR. CROW: I am concerned that the way that is
13 couched, Jim, is that carries a certain criticism that may
14 not be appropriate.
15 MR. CROW: Yes.
16 CHIEF JUSTICE ABRAHAMSON: Any other discussion on
17 this?
18 I think this is an excellent report, Jim. Not
19 because you are from Madison. It has been very helpful.
20 The issue may be whether to have a short foreword to this
21 that might raise some of the problems from a legal aspect or
22 policy aspect that have been raised here without necessarily
23 answering them. But we will talk further about that.
24 MR. SCHECK: I have one practical suggestion about
25 this. As with our post-conviction recommendations, it might
1 be useful to try to send out this draft, and it will
2 probably be true of some of the other drafts we will
3 discuss, to various groups of scientists, ethicists, lawyer
4 types, to get some kind of peer review and feedback on it.
5 MR. CROW: There is nothing very secret about
6 this.
7 MR. SCHECK: We might be able to get some good
8 responses.
9 CHIEF JUSTICE ABRAHAMSON: Why don't you give
10 Chris names of entities.
11 MS. FORMAN: People or organizations.
12 CHIEF JUSTICE ABRAHAMSON: People or
13 organizations.
14 We are in recess. We will come back at 3:25.
15 [Recess.]
16 CHIEF JUSTICE ABRAHAMSON: Paul, we are ready for
17 you, the Laboratory Funding Working Group report.

Previous	Contents	Next
Back to National Commission Main Page