CHAIRPERSON ABRAHAMSON: The Commission is back in session. And we have a very important announcement about dinner.

DIRECTOR ASPLEN: Very important dinner announcement. For those interested in going to dinner, meet in the lobby at 7:15. We have reservations at 7:30 at La Cassa Senna, a half a block from the plaza. It's a very old Santa Fe style restaurant, ten minute leisurely walk, at most.

The other thing I would add is that at 5:00, we'll be breaking here. However, we are then going to stretch for a few minutes and then go to the next room over, where we're going to have a demonstration by an NIJ grantee on some of the technology that he is developing. It won't be more than 20 minutes, on a video camera that he is using to detect biological components, and I think you'll find it very interesting. It's truly futuristic.

CHAIRPERSON ABRAHAMSON: Back to dinner.

DIRECTOR ASPLEN: Back to dinner. Casual dress for dinner.

CHAIRPERSON ABRAHAMSON: Okay. We are now ready for Dr. Crow, who is going to talk about the research and development working group report, with discussion. Jim?

COMMISSIONER CROW: I want to note that I'm glad that Lisa is here, or at least I thought she was here -- yes, she is. There are likely to be technical questions that she's in a better position to answer than I am. The committee was chosen for technical expertise, with my exception. I was selected for whatever reason they did, but not for technical expertise.

There are likely to be some legal ramifications that come up and, again, she has given thought to that.

What I want to do, though, and perhaps for CODIS, too, I'm not going to say anything about CODIS, but that still leaves quite a bit to talk about. I would like to pretty much go through the report and if you'll permit me, I guess since I have the floor here, you have to permit me, I'll approach this historically and say some things that most people know, but think it's nice to have the history of this as a way of judging what the future is likely to be. We can use it as a guidance to future extrapolation.

So to go back to the very early days, the earliest methods were blood groups. I find it interesting. Blood groups were discovered in 1900. They were discovered by three different investigators, each of which named them one, two, three and four, which was fine, except two and three were interchangeable between two of these groups. So hospitals that used different systems of blood transfusions had what was then called transfusion confusion. Now that's been settled in an unambiguous notation.

Anther curious thing about this to me, the methods must not have been very good in the earlier days of the century because the theory of how these were inherited was wrong, turned out to be wrong, and yet it lasted 24 years. It finally took -- and it took a population geneticist to clear this up. He analyzed data from soldiers in World War I that enabled the proper analysis of this.

And it's interesting to go to the old literature. There are certain kinds of matings that can't happen under the correct hypothesis, but which can happen under the wrong one, and the number of those substantially diminished in 1924, the year the correct hypothesis was worked out. All of which tells you what you already know, that the methods were not really very reliable back in those times.

Then by 1950, new blood groups kept being discovered. By 1950, there were a dozen or more of these. Still not enough, though, to be used for inclusive purposes or not at the moment at that time at least, and blood groups were used primarily just for exclusion.

Then in the `70s mainly, then people started thinking of using DNA to give inclusive probabilities, such as probabilities of a match, and this was all pretty well worked out, very well worked out, for paternity cases, again, using blood groups, but it was not -- we're dealing here in probabilities of the order of one in a thousand, which is all right for paternity testing, at least it was accepted, but in forensic work, we demand a stronger criterion than that.

The DNA methods suddenly appeared in 1985 with Jeffreys, who just found these what we now called VNTRs, and the discovery that you could use DNA itself was really a remarkable thing.

Most of us that are old enough to grow up before we knew what DNA was, we knew what DNA was, but we didn't know what it did, it's totally remarkable to find out that something as simple as DNA could, in fact, be the gene. But it's not simple in one sense; it's astonishingly long.

If I put all the DNA in a -- well, there are three billion bases of DNA. The amount of DNA in a single cell may be a yard or so in length, and if you like to play these kind of games, if you ask how long the DNA in your body would be, if stretched end to end, it would go to the moon and back a great many times. I haven't calculated exactly how many.

It reminds me of what Francis Krig said. He said imagine how short, in fact, you must look to a DNA molecule. Anyhow, it's these three billion bases that really makes this so powerful, once you could analyze all three billion of them or even one percent of them, which would still be definitive for any two individuals.

But what we have to be content with, less and less each year though, is a sample that's not very big, but our ability to sample the DNA gets bigger and bigger each year and it's growing very rapidly right now.

The first to be used were the RFLPs or VNTRs, as they're usually called in these circles. They have many advantages. You're acquainted with them, I think. There are many alleles at each locus. There are not very many loci, it would be nice if there were more, if we were going to continue to use this. They are -- the large databases are available, thanks to many people, and summarized in these big volumes by the FBI, but there are disadvantages, too, and one of them was the necessity for binning, which always led to statistical difficulties and disagreements, and the limited number of loci, as I said before, and then single bands can be ambiguous.

A single band may be a homozygote or it may be one of two in which you didn't see the other one, and they're time-consuming, especially when radioactive methods were used. They've been largely replaced now, so that the time is no longer in months or weeks, but now in days, at most a week.

Then came what's getting to be and soon will be, in a couple of years, I suppose, almost universal, those that depend on the polymerase chain reaction.

I wish somebody that I had more respect for had discovered the polymerase chain reaction, but I guess you have to take them where you find them. Anyhow, he made this maybe the nicest discovery in molecular biology in years and years, just a way of taking a very minor amount of DNA, in principal, a single molecule, and expanding it to large enough amounts that they can be analyzed by ordinary laboratory chemistry.

What this does is do pretty much what happens in the cell anyhow, and what Cary Mullis did was simply confine this to a part region and you amplify the region that you're interested in.

So with the coming of this, then the tide has changed very rapidly in the direction of getting more and more systems that use the PCR.

The first one to use it effectively was the DQ alpha system or DQA. That has some advantages of being fast and is still being used and may be able to continue to be used for a while for that reason. There is about a five percent chance, or maybe, to say it the other way around, about a 95 percent chance of exonerating an innocent person who is wrongly accused by just this one test, so it provides a very quick, somewhat dirty, but very quick way of getting an answer of an exclusion source.

Then there are five more loci added to this that make the polymarker system. That has a total discrimination capacity of about one in 4,000. We really want more than that, more precision than that for forensic purposes.

So the STRs have finally come into this. They are essentially like the VNTRs, except the individual units are smaller and the number of units in a locus are smaller. That matters because PRC, at the moment at least, can only be used for relatively small molecules, and that means that the older VNTRs were not amenable to that. But the STRs are.

So there is no necessity for binning, because the number of types is discreet enough. There are always problems and I guess I should -- I'm enough of a biologist to know that no statement really holds up universally. But for the most part, you can avoid any problems of classification of individual alleles.

The number of loci is essentially unlimited. So we could go on and on and on getting more and more, as long as there is a need for it.

With the canonical 13 loci that the FBI has decreed, the average match probability is one in 180 trillion. So with 13 loci, you have all the power that you need. Of course, there are going to be individual frequencies that are going to amount to that, but this is the average probability.

I'm going to come back to this and ask how long this is likely to persist as better and better methods are discovered later, but I'll defer that for a moment.

I want to say a little bit about some population issues -- oh, no -- first, about mitochondrial DNA. That has the capacity for being extremely sensitive. There are -- that is a very small amount of material can generate information. Whereas there's only one copy or maybe two copies of each gene in a cell, there are thousands of mitochondria. So a few cells will produce enough information that can be used.

So the great use of mitochondria is and will continue to be, I think, the capacity to generate information from very small and very unpromising degraded samples.

The Y chromosome, for a while, we didn't get any information from. It was pretty much inert, except for one function; that is, determining males. But if we ignore that capacity, the Y chromosome had very little on it. That has changed enormously in the last year. There are now, the last I heard, more than 150 loci on the Y chromosome.

Most of these are snips, a few of them are STRs, and I'll come back to another property of the Y chromosome in a minute that makes it particularly interesting for a particular purpose, perhaps troublesome for some other reasons.

Now, some of the population issues that I'd like to discuss and the committee is discussing and hasn't fully agreed on. We need to make some -- in our NRC report, we made suggested recommendations for how to handle substructure. As far as I know, that's still pretty well accepted as a way of doing it by using this quantity data, which is a measure of the amount of structure subdivision within the population.

For the four major racial groups in this country, the studies have shown the value of that to be considerably less than one percent, so probably our earlier recommendation of one percent as a safe value to use still stands.

If you're particularly cautious, and perhaps the population is likely to be particularly structured, maybe the .03 is better.

Now, I want to raise something that hasn't been raised generally before. We have now reached the power in these systems that we don't have to worry so much about the possibility of undetected relatives in the population or of the population structure, population subdivisions that you didn't know about.

What I have in mind here is it might be possible, with the amount of power we have now, to just make the calculations on the basis that the individuals are sibs. Now, why are sibs particularly useful or interesting in this regard?

It's because the probability of identity of two sibs is one-fourth, just by the nature of sibs. I'll draw you a picture, you like. But there are small increments beyond this that are based on the gene frequencies, and those increase in gene frequency size, but the one-fourth so dominates this quantity that there is very little difference in -- depending on ancestry, racial group, other kinds of relatives and such.

So I think maybe we'll approach -- I don't think we're there yet, but I think in a few years, we'll approach a situation in which maybe we can say that you measure the two individuals that you're interested in, two samples, and ask what's the probability of a match if they were brothers, and if they weren't brothers, the probability of a match, anything anywhere else is a fortiori less than for brothers, then you have an ultra-safe criterion. Let me emphasize, that makes -- at least that automatically you can stop worrying about undetected half-sibs in the population and by other kinds of complications by population structure.

What kind of numbers am I talking about? With the 13 CODIS loci, the average match probability for sibs is about one in 300,000, and that's not very big. But by going to 21 loci, which is quite feasible now, that becomes one in 600 million and with additional loci it goes up in very rapid proportions.

So I want to suggest that one possibility for the future, I don't know whether this would be acceptable or not, but it strikes me as a rather intelligent approach to the problem, I have friends, and you do, too, who would like very much to get rid of any racial designations in our discussions of this issue and if you don't have to ask what particular race a person belongs to, you make the same kind of calculation irrespective of their race, that has, in some people's minds at least, an ethical and a social benefit.

The original NRC committee, NRC-1, presented what it called the ceiling principle, which was roundly denounced, and has pretty largely been forgotten, but it had a high motive and the principal motive of that was to stop having to take races and other ethnic divisions into account.

Another thing I want to say a bit about is the possibility of inferring racial makeup or ethnicity from DNA samples. That's increasingly possible and increasingly troublesome. Even if we take the 13 STRs that exist right now, it's very easy to find a combination of a half a dozen of those, with the likelihood ratio coming from the white population versus coming from the black population, is 1,000 or 100,000, and so there are cases right now in which you could take a blood sample and just with ordinary STRs state a strong likelihood that that it had to come from a certain racial pool.

I think you'd want to, if you're going to use this for investigatory purposes, to narrow the range of suspects. I think you ought to take advantage of the prior probability of the racial groups; that is, natural frequencies of these, and rate your probabilities accordingly, as a good Bayesian would urge.

There is at least one blood group that's found essentially 100 percent in some parts of Africa and essentially zero elsewhere. It's called Duffy null. So if you find a person with the Duffy allele, that almost certainly means that that person has African background.

I caution you, though, that it doesn't tell you how much, nor does it tell you what the person will look like, because there are people who are very nearly white in skin color and people who are very dark in skin color, who would have -- who would show the same, who would be detected by this, the same way.

But by adding more and more characteristics, it's soon -- by soon, I mean it's just a few years, I think, we'll say a lot more about the geographical origin of a person or the person's ancestors, maybe I should say, from the DNA sample.

I don't have to tell you that this is going to raise ethical issues, that I pass on to the Commission for resolution.

There is one X chromosome locus that's sharply divided between the African population and the rest of the world, but the most interesting in this regard is the Y chromosome. Ordinarily, if we analyzed the variability of a population from ordinary genes, that is, for ordinary forensic genes, we find that maybe 95 percent or more of the population variability is between individuals within a group and very small part of the variability is determined by the mean differences between groups.

That's one reason that's made these things so powerful for forensic purposes, because they're telling you individual differences, which is what you want. The Y chromosome is different. The Y chromosome, more than 50 percent of the variability is between geographical regions, by which I mean Africa, Asia and North America and Europe.

Why the Y chromosome should be different? You might think that it's, of course, transmitted from father to son. Mitochondria are transmitted from mother to all of the children. Why don't mitochondria show this same kind of difference? I don't think anybody knows, but there is a good hypothesis that seems to fit the anthropological literature, and that is that most of the amalgamation that happens is by migration and it turns out that it's the women that migrate and the men stay home, and why I don't know, and maybe this is wrong, but that's about the only explanation one has for this otherwise rather remarkable finding.

It could be used, though, if it were desired, it could be used as a way of racial classification of people.

I emphasize, perhaps I don't need to, but I emphasize that this is going away from the usual directions of forensics, because I'm now talking about loci that are used for particularly for individual phenotypic purposes, whereas for forensic purposes, we try to stay away from discernable traits.

So if these things come into practice, it's because of a change in the emphasis of ordinary forensic usage. Right now, the gene for red hair I think has been analyzed. The gene for color blind has. So it's possible to detect a sample of blood and say this came from a person with color blindness and that could be used as a way of aiding the search.

Another issue that has -- and none of these are old, but they're coming to the fore business they've become the kind of things one can do a great deal with right now, and these are partial matches. With 13 loci, if they examine all of these loci, it will be the same person almost always, but if they match it, a fraction of these, but not of the others, that probably means they're relatives.

So it's worth asking what the analysis would yield for sibs. Again, I'm particularly interested in sibs because of this particular relationship that they have.

If I take a pair of brothers -- well, what I did, which is get a collection of data from sets of brothers and sisters and just at them from the STR standpoint, the first one I analyzed turned out to be a little better example than average, but anyhow, it had a match at, I believe, five of the 13 loci, and a partial match -- that is, one unit fit and the others not, at another six or seven, and then there was one complete non-match.

But when I did my test of this, it turns out that that particular profile has more than a million times the probability, if it comes from a sib than if it comes from unrelated people, so that would be likely to be very convincing in a court, but especially convincing as a way of identification.

This particular -- you might say, well, father and son share a lot of their genes, too. In this case, you ca rule it out, in this particular example, and oftentimes you can, because if the parents -- if these two individuals by both being homozygous for different genes, they can't be the parent of offspring. So in this case, I can say it's a very high probability that they're sibs, zero probability that they're parent and offspring, and a rather low probability that they are half-sibs or less closely related individuals.

That does mean, and it's going to be increasingly true with the CODIS databases, that I'm going to find examples in searching the database of a person who is the brother of the person you want. Then I ask the legal and ethical question. What rights have you to search the brothers of this particular person in order to find the guilty party?

There is the uniqueness question. It's been proposed, but, by no means, settled, or individualization question. The FBI has announced its procedure a year or so ago of how to define, when two samples are unique or an individual is unique. They define it as not finding it in a population as large as the United States, which they took at that time as 260 million.

So if the probability of a match is less than one in 260 million, it's regarded as this individual is unique. Except they hedged, as they ought to have, of course, they made two conservative modifications. They asked not just for the average probability of this happening, but for the 99 percent confidence limits of this happening. That cuts the numbers down considerably.

Then they said pick, of the four databases, whichever one is larger and then they multiplied the whole probability by ten. Where did this ten come from? It's a figure that I think we can justify, but it doesn't arise out of any theory. It's simply the empirical level of agreement between two individuals from different parts of the world tested with the same sets of VNTRs. And by looking at the data, this is our '96 committee, we decided that a factor of ten pretty well encompassed the range of error that one is likely to find; I mean, a factor of ten in either direction, which is the 100 counting in both directions.

So this means that if you estimate the probability as one in a million, it might be as small as one in 100,000 or as large as one in 100,000 or as small as one in ten million, and that's an empirical measure.

I prefer this to better statistical methods, because this is empirical and it's based on what the actual data look like, and the statistics ordinarily don't take all kinds of errors into account.

The FBI procedure -- that would be to taking 3.9 as ten-to-the-minus-11th and as the cutoff point. So if you calculate your probability as less than that, then you're able to say that to a reasonable degree of scientific certainty, these two samples came from the same person.

Within our committee, there is considerable vociferous criticism of this statement. It's not very good statistics, it's not very good genetics, but it's not intended as being good statistics and not intended as being good genetics. It's a pragmatic agreement that maybe the courts or the forensic community would decide, but it's clearly not a scientific question and, therefore, it's one that I could be a narrow scientist and toss off to other people to discuss.

The FBI procedure does not consider relatives, but they do -- it does -- if you make this same kind of calculation for relatives, you will hardly ever reach the size of the United States population.

So if you think or are worried about the possibility of undetected relatives in the system, then this doesn't stand up very well.

So I really don't know what is going to be the outcome of that, except I think it has actually been practiced in some courts, some of you can help me with that particular point.

Then there is the authority question, at least amongst statisticians, and that is that when the suspect is identified by a database search, the 1992 committee regarded this as an important issue and they said if you find too many suspects by database search, you're increasing the risk of prosecuting an innocent person.

So they suggested what you should do is find your suspect using whatever markers you use for that and then test it for court purposes with other loci.

In our '96 report, we approved of that, but it didn't seem practical because there are only half a dozen loci to use and if you use half of them up in finding the suspect, there are not enough others to give any statistical power.

That's no longer true. With the STRs now, if you used half a dozen to identify the subject, that leaves seven of the 13 and then most laboratories can go beyond that. So I think in the near future, it would be possible to have this kind of a criterion, that as databases get to be large enough, there is a legitimate fear of finding an innocent person too often. Then the numbers will soon be large enough to apply this by about the time the databases get to be large.

However, there is an influential group of statisticians who feel quite differently about this and they say that the way in which you got the subject doesn't matter, that what you do is analyze this particular person, ask how likely that probability is, and calculate accordingly. Then everything is simple. You treat the individual from the database the same as if it were gathered otherwise.

And I don't mind saying privately, this isn't very private, I don't mind saying publicly, but not to my committee, that I think there really is a difference between gathering one individual from the sample of a thousand and finding one individual from a sample of one, and that your probability calculation should take that into account.

But we are wrestling with this and I have on this committee two statisticians, both of which are rather strong-willed, so likely to prevail. One of them -- in one respect, the report will certainly be better because one of our statisticians is a stickler for using the language correctly and about six times he has uprooted me for saying probability when I meant likelihood or likelihood when I meant probability.

Which reminds me of the statement from Pygmalion that it doesn't care what you say actually, as long as you pronounce it correctly, and I'm sure our report is going to be accurate from the standpoint of language because of this one member.

Now, part of our business, a big part of our business is technology projections, and I'll say a little bit about what we think is likely to come up in the next two, five and ten years.

For the two-year period, by two years from now, which we would say is 2002, that the STRs will surely have pretty much dominated the situation by that time. We saw in these surveys that the -- I read it just last night, that the FBI did, as to how rapidly the laboratories are changing over into STRs. That will certainly dominate case applications at that time.

The database will certainly be much enlarged and not only be enlarged, but made more specific, so there would be better data on finer subdivisions of the population. And if we want to understand people from a particular side of the country, we'll have enough data to say something about that.

I think the data are going to be published, but if we have any recommendation to make, it's going to be for the FBI to get busy and publish these data that exist, but are not in public form.

One of the things that will come with larger databases is making the mitochondrial DNA and the Y chromosome more useful. The mitochondria, we can't use Mandalian principles to expand the database by multiplying numbers together.

All the information you have is just the number in the database. So if your database is 1,000, the smallest probability you can have is 1,000 or zero, and it's expedient, therefore, to get larger databases for mitochondrial DNA especially; perhaps also for the Y chromosome, except I don't think it's likely to have quite as much forensic use, except for rather special kinds of uses in which you want to separate two different males. There are times in which -- two different males from different sources.

Of course, Y chromosomes won't distinguish between the male descendents of a single male, but they will distinguish between the male descendents of two different males.

And certainly within two years we're going to have improvements in the collection of evidence and in the isolation of DNA and in the automation and in the miniaturization. There is a group at Harvard that announced that they were -- or at Cambridge, I should say, that announced that they were working on a chip, I guess we could call it a chip, it's the same principal, a tiny unit, that just does STR analyses, the same way you would do them macroscopically, except this does it in a minimum scale. I think Lisa interviewed those people, but my information comes from newspapers.

It's at least possible, in principal, to just miniaturize the use of the existing techniques and by miniaturization, they'll not only become cheaper, because the materials cost less, but faster, and if it works, much more efficient.

This group in Cambridge, at the Whitehead Institute, said they'll have this ready in two years. But I caution you that there is quite a difference in having something that works approximately well in a laboratory in two years and having something that will withstand the rigors of the forensic community in two years time. But certainly in your lifetime, if not mine, there is going to be a miniaturization of the 13 CODIS loci, and that ought to come fairly soon, while the loci are still being used.

Then what can we expect in five years? Well, the genome sequence is going to be complete by that time. In fact, they promised it in considerably less than that. And we'll have much more computerized analyses. I think they will be retained.

And one of the things that's troublesome now is mixed samples with STRs that could be complicated, but they shouldn't be too hard to program. So that machines can do these fairly complicated calculations.

We will certainly have genetic and racial markers for investigation and society has to decide or the forensic community has to decide how to us these identifiers that can identify from a blood sample the racial or ethnic or even phenotypic data that the person has.

There's certainly going to be chips and miniaturization and the chips that are now being used for various kinds of genetic research are immensely impressive. I've enjoyed seeing demonstrations of them. They are also immensely expensive and they have to be specialized for rather particular tasks and I'm not sure that every state laboratory is going to be able to afford the kind of instrumentation that this calls for, unless there are improvements in the technology itself that makes the instrumentation cheaper.

I think that's a real tradeoff for the future as between these very rare powerful systems that are very, very expensive and something that's not quite as powerful, but not so expensive.

STRs will surely continue to dominate for the next five years and probably longer and the 13 standard loci will presumably be adopted by virtually all laboratories by that time, probably sooner, and most laboratories by then will have a substantial number of additional loci to use for troublesome cases or where you particularly want to do a better than usual analysis.

One thing to emphasize in looking at the new kinds of technologies that are emerging is that some of them simply improve the efficiency of doing what we're now doing; that is, using the same 13 CODIS loci for doing them more accurately, more rapidly, more efficiently, more cheaply, and in a miniature scale.

Others, like snips, involve a new technique and before they become widely used, they would replace or be used in conjunction with the STRs. And if we look a further on, for down ten years, that's a little harder to predict. The investment involved in getting the 13 core loci up and running in many different laboratories is a big investment and it works very well.

My guess is that there will be a very great reluctance to change even for something that's demonstrably better just because of the big investment that's been made in a system that works awfully well and especially works still better with the modifications that are being developed.

We can certainly look forward in ten years to a miniaturization and speeding up in such a way that probably that the investigation can be done right at the crime laboratory and that would, of course, have tremendous advantages in quickly clearing an innocent, for example, but, also, it will shorten the chain of custody and make less likely, therefore, the kinds of errors that can arise in any system where multiple humans are involved in the operation.

I suspect ten years from now that old and new systems will coexist. We're using snips already, which most of the Y chromosome is done by snips, and there are -- and they probably will introduce elsewhere, too. So to just hazard a guess, probably if you look down ten years, we'll still be using these 13 loci, but various others will be added to it, sometimes in the same laboratories, sometimes in other laboratories that specialize in particular kinds of cases.

The last part of our report is a series of appendices, or an appendix with a series of parts, and these are all written by the people who specialize in that particular subject, and I'm neither going to talk about them nor answer questions about them, first, for choosing not to, and the second because of being unable.

But by ten years, we're going to have allele sequences, mass spectroscopy, which brings us into the physical rather than the chemical level, and, of course, direct sequencing.

Most of the people who are making predictions, in the conversations I have at least, think that the most promising among all these are snips. They're her already and probably the technique will improve and I hope the cost of using them. There are snips and snips and they range in cost from very expensive to just expensive for use.

I suspect or our committee suspects really that much of what we'll be using in ten years from now will be determined by what medical research yields in the intervening ten years, because it's bound to turn up interesting things and perhaps useful and perhaps important.

We will sure here, within ten years, of more instances in which animals and plants, or even wild animals, are involved in DNA cases. That raises difficulties and it has simplicities, too. One difficulty is that most of these organisms are likely to be highly inbred and that complicates the genetic analysis. It means that it takes larger numbers to get the same information.

But there is another concern that doesn't exist, though. You don't have to look for markers that are not associated with phenotypes and diseases I don't think any cat is going to sue because its privacy was invaded, but maybe I'm wrong. But in any case, one can get a lot more information by not having to worry about getting traces that are not used as phenotypic markers.

I also want to, if you look at other organisms, you can look at bacteria and protozoa and worms and other kinds of parasites. We all carry our own constellation of parasites, bacteria, viruses, and it's just possible that there would be forensic uses of not of our bone genes, but of the genes of the organisms that we carry. I don't predict this, but I don't predict it not happening either, and it might very well.

As far as our behavior is concerned, we don't have in mind any very specific recommendations regarding research and I think the reason for this is that the research is driving itself. The universities are competing with each other and laboratories and the new techniques are going to be developed considerably faster than they can be put to use.

It's hard to predict what's going to happen with costs. You would like to think there would be economies with more efficiency, but on the other hand, you're discovering new and better techniques, which usually turns out to be more expensive.

But I have heard reports about the cost of doing individual tests coming down, if they can be automated or mass produced or there are economies of scale that have come into this, and I hope those apply to the backlog that I've been hearing so much about in the last day.

Some social issues that came up in our committee, but don't really belong there. We haven't talked about laboratory standards, largely because that's pretty well dealt with in other places in the community. So unless we're instructed to the contrary, I think we won't talk on that particular point.

There are questions about long-term storage, both the technologies and the proprieties, and a question about whether databases can or should be used for research, whether one can invent or make up the required privacy considerations that would permit useful kinds of research.

So far, it's a little hard to think of good questions, important questions that could be answered from a database that wouldn't infringe on the kinds of liberties that all of us take very seriously. But it's a question worth raising.

And one other point that came up in our committee, but it really belongs in the parent body, this is my last, and that is whether scientific questions in general are better handled by greater use of court-appointed experts than is now the case.

Lisa is sitting back in the back of the room, not necessarily eager to talk, but I've offered the opportunity.

DR. FORMAN: But here if you need me.

COMMISSIONER CROW: She's there to answer questions at least. So let me invite them.

CHAIRPERSON ABRAHAMSON: Questions, comments.

COMMISSIONER GAHN: What about the population frequency issue and racial differences? There are obviously cases where other evidence in a case makes one racial category -- that is, a frequency in a particular racial category -- more important than others, and perhaps to the exclusion of all others from the rest of the evidence, and the case indicates the person, the attacker, let's say, was of a specific racial and/or ethnic group.

In that case, or, I should say, that type of case, isn't it important to provide those specific frequencies? I mean, specific to a particular race.

COMMISSIONER CROW: To a particular race, yes. I think that would go on. We do that right now. But everybody knows that racial classifications are fuzzy at the edges and there are always going to be people that either classify themselves differently than their appearance or just by the appearance. In the Hispanics, it's really not a biological classification at all. It's a language classification. And there are always some people, including me a bit, that were using these seemingly rather artificial measures, and yet they're real and they have predictive value and they've been very useful.

One of my reasons, though, it's obvious that I'm enthusiastic about this, and it would be nice just not to have to ask that kind of question, that every individual could be treated as an individual.

If we use more race-specific markers as opposed to just statistical aggregates, if we use more race-specific markers, I think we're going to find more and more cases in which we identify a possible bit of racial ancestry, without it being very useful in telling you what the person looks like or acts like. But maybe more loci would change that.

COMMISSIONER REINSTEIN: Dr. Crow, where do you see court-appointed experts being used most frequently in trial courts?

COMMISSIONER CROW: I don't have very serious ideas about this, but it seems to me that there are likely to be situations in which two experts brought in without a bias being forced on them, let me say it that way, maybe can -- one or two can quickly come to agreement on quite a number of points. Those can be stipulated, and then you go on from there and debate the others.

I doubt that -- well, I think you're in a better position to say than I am. I doubt very much now that the expert would be definitive in the final disposal of it, maybe is the word I want to say.

COMMISSIONER REINSTEIN: I think you can be -- if the judges make a determination, if it's an admissibility hearing, I think that if it's a trial in front of a jury, they're still going to be calling in their hired guns.

COMMISSIONER CROW: You reminded me of what I meant to say and forgot, too. I think if the court-appointed experts have -- initially, it's likely to be an admissibility.

COMMISSIONER REINSTEIN: I would be happy to hear Lisa, who is testifying in my court next week, as my court-appointed expert.

COMMISSIONER CROW: Good. It's clearly the result of our recommendation.

I want to say but just one more thing, then. This report has been in an earlier stage, it's in a late stage now, earlier with regard to the date of the completion, and I do welcome suggestions from any and all of you.

The one question that constantly arises is what level this should be, how technical. We're trying to write the report itself in a way that can be understood by the so-called intelligent layman, but big words do creep in and we'll try to put in a glossary for that.

Then in the technical part of it, that can be as technical as it wants. Now, with our -- as David Kaye will help me remember, in our 1996 NRC report, we had a public version and then a more or less technical version, much longer. The public version was deemed too difficult or too long, so we had a special short version for people with particularly poor attention spans that could be read in a minimum time, and I've mentioned it to David, because he wrote most of that part of it.

COMMISSIONER REINSTEIN: Well, the technical version, when you get into the appendix and you get into these formulae, somebody will need to translate that or say ignore.

COMMISSIONER CROW: Those particular ones are written, as I say --

COMMISSIONER REINSTEIN: Is that Dr. Weir?

COMMISSIONER CROW: Actually, it's me, but some of the formula come from Weir. And if Weir were doing it, let me assure you that they would be more complicated.

Thanks for your attention.

CHAIRPERSON ABRAHAMSON: Any other comments? So what's the working group going to go to now, defining this?

COMMISSIONER CROW: Yes, our next step, and if the Commission wants to give us other assignments -- I took as our main assignment these predictions. If we do that, are we off?

CHAIRPERSON ABRAHAMSON: Any other comments? Thank you very much.

[Applause.]

DIRECTOR ASPLEN: I don't know if there -- Lisa, could you see if Cohen is ready for us?

DR. FORMAN: Let me check.

DIRECTOR ASPLEN: Dave, did you have some comments about the earlier discussion we had that you wanted to --

MR. COFFMAN: Dave Coffman again. I did want to make a statement that it concerns me a little bit, just as a spectator, to hear that we feel like we don't need to do the research to come up with true numbers to get our message across.

I'm speaking not only on the arrestee issue, I'm speaking on the cases that are out there not being worked, because I can just speak from personal experience in Florida that our legislature was totally unaware, they have this idealized view of what's going on in the agencies they're funding and they truly felt like we were working every crime in the State of Florida, first of all, that was being submitted, and that we were working every crime for DNA.

It's great that we all know that that's not true, but we have to do the research to prove to them that that's not true, so they'll fund it.

We were audited in 1995 and the work, the verbiage in the pre-audit report was the auditors were shocked and appalled that not every case, sexual assault case was submitted to the laboratory and that it was worked for DNA.

And out of that audit report came 69 new positions for the Florida Department of Law Enforcement and also some funding for some lab expansion.

So I think if you can show the numbers and do some sort of research, that even though it may not be accurate to the nth degree, I think it's at least an estimate of what the problem is out there and it can help get you funding in your local state, because that's what it did in Florida, and I just felt like I needed to say that.

CHAIRPERSON ABRAHAMSON: Thank you. We are ready to go next door. So we're in recess, until next door, and then tomorrow morning at nine, except for dinner at 7:15.

[Whereupon, at 4:55 p.m., the meeting was recessed, to reconvene at 9:00 a.m., Saturday, May 8, 1999.]

Previous	Contents	Next
Back to National Commission Main Page