Professor Speed, yesterday a series of quotes from the NRC report that I've just asked you about were flashed up on the screen and you were asked whether or not you agreed with them. Do you recall that?
And approximately half of those comments were from chapter 3 of the report "DNA technology in forensic science," were they not?
Okay. And chapter 3 was the source or the area that you made your comment in the letter to Bruce Weir that is dated September 29, 1992, was it not?
Okay. And the areas from which quotes were elicited yesterday had to do with laboratory error rate; isn't that true?
Now, when you wrote this letter to Professor Weir in September of 1992, when you said, quote, "It amazed me how its assumptions escaped the notice of all the eminent people on the NRC panel," parentheses "Sarcasm intended," end parenthesis, you were criticizing the people, whoever they were, that produced part of chapter 3, were you not?
Objection. I would just ask that he have a chance to look at the letter before he answer the question.
May I point out that the assumptions I pointed out referred to the so-called ceiling principle.
And the eminent people that you pointed out with sarcasm intended were the eminent people in chapter 3, weren't they?
We just discussed the names of the people I know. Do you want me to go through the list of names again?
I'm sorry. Professor Speed, the first case that you became involved in was a capital case in San Francisco, People versus Briggs; isn't that true?
Okay. You have some extreme views about capital punishment or death penalty in the United States, don't you?
It was a case with Contra Costa County and I believe the Defendant's name was Robert Taylor.
None deal directly with the application of statistics in forensic settings, do they?
I would like you to refer to the item that is numbered 111 on your CV, if you would.
Okay. Would you tell the ladies and gentlemen of the jury what the role of statistics plays in that article?
I was assisting in the analysis of data that had been collected for a study on aids, so statistics was involved in the analysis of the data.
Well, it was a group of male homosexuals, some of whom were HIV positive at one time and not at a previous time, some of whom were HIV negative throughout the entire period, and there was--this group of male homosexuals were cross-classified according to various sexual practices that they had admitted indulging in or that had taken place or that they had been involved in over the period that would have related to when they might have gone from being HIV negative to HIV positive, if that was that group, or when they remained HIV negative throughout the entire period, if that was the other group.
And in what way does your--or does your contribution to that article demonstrate any connection between HIV and aids?
It is true, is it not, that of the 120 articles and chapters, you have never expressed any concerns that you seem to have expressed yesterday about the role of forensic DNA typing in the legal system? Isn't that true?
In papers. As I said earlier, I expressed these views in a letter to the second NRC report on this same topic.
Would you briefly describe the role of an editor and the responsibilities that are attendant to that role.
Well, I've mainly been associate editor, not the top editor, so the associate editor's role is usually to look at papers as they come in, as they are submitted for publication, and choose referees, people who will review the papers, get the reports from the referees, synthesize them into an overall report on the paper, which is then transmitted to the editor with a recommendation for the treatment of that paper.
And what would happen if something fell through the cracks, a published article fell through the cracks without your careful review?
Sure. If you hadn't carefully reviewed something and then an article emerged in the peer review literature with your name on it as an editor.
Well, usually articles don't appear in the literature that I am associated with with my name on it as editor, but I'm afraid I still don't understand what you mean.
Okay. If you were not careful in your review of an article and the article were published, what are the consequences of the implications of an article with faulty scientific data in it appearing in the peer review literature?
Well, usually if an article appears in the peer review literature, it is regarded as the responsibility of the author. The editors may be unhappy that they let it slip through, but ultimately authors are responsible for their publications. And if there are errors, that responsibility goes home to the person that made those errors, though it may cast some doubt on the quality of the reviewing process.
Umm, I think about April this year, either I was briefly in Los Angeles or--not Mr. Scheck on this occasion--Mr. Neufeld came to Berkeley. Those two communications occurred rather close together. One of them was when I was here and another was when Mr. Scheck--excuse me--Mr. Neufeld and Mr. Thompson came to Berkeley.
This case has given me an opportunity to get much more deeply into the issues than I might have otherwise, there is no doubt about that, so I couldn't say that I would have done it had the case not been around, no.
You made no notes of any of the sessions that you had with Defense counsel in this case; is that true?
I have written things. They are not notes of sessions and they are not reports, but I have written brief memos way, way back last--before Christmas.
Well, they were being prepared at the time I thought I was being involved in a Kelly-Frye hearing, and I prepared rather small number of notes on what I considered important issues, the usual sort of thing, you know, the product rule, that--just two or three of the contentious areas which I think--well, you would be familiar with. If you would like me to enumerate them, I will try to remember.
That is okay. Now, you knew, before you testified yesterday, that I would be cross-examining you; is that true?
And the first letter I asked for an opportunity to meet with you about your views; isn't that true?
And I had advised you that I had no idea what you were going to testify about; isn't that true?
And I would like--I wanted an opportunity to sit down and talk with you ahead of time; isn't that true?
Isn't that true? And you faxed me a cordial response declining the opportunity to talk to me; isn't that true?
Well, he is obviously a professional colleague who works in the area of statistics and genetics and that is not a very large field, so we cross paths not terribly often, but from time to time. I regard him as a friend as well. I have visited him and stayed in his home in North Carolina.
Okay. Now, in the second letter I propose that you discuss whatever you would testify about with Professor Weir, didn't I?
And you never responded to that letter or acknowledged that letter; isn't that true?
It was faxed to the number you get when you ring the address that you had on your letter and just I brought a copy of the acknowledgment that the fax was sent to that place and went to the--
It went to the Los Angeles County District Attorney's office bureau of central operations special trials and that is the fax number I got when I phoned because the previous fax number didn't work for some reason. (Reading.) Dear Mr. Harmon: Thank you for your fax dated July 27, 1995, requesting that I discuss my anticipated testimony with Dr. Bruce Weir. With respect I decline. Yours sincerely, Terry Speed."
Now, knowing that you are a personal friend of Dr. Weir and knowing you have a professional respect for him, why did you decline to talk to him so that we might know what you would testify about in advance?
Well, the reason was that I would have felt very uncomfortable because I know that my testimony would have been disagreeing publicly with positions he has held publicly, for example, reading out a quote that you heard me attest to that I disagreed with of Dr. Weir, that sort of thing. Honestly don't feel very comfortable chatting on the phone with a professional colleague just before I'm about to go into this sort of situation and disagree with him. I would have thought that is not a very unusual human reaction.
Now, this is the man who invited your comments on his review article, Professor Weir?
Okay. Now, you criticized yesterday Professor Weir's calculations for assuming that there was no error in the tests? Do you recall that?
You criticized Professor Weir's population frequency calculations simply because they assumed that there was no error in the testing methods. Do you recall that yesterday?
I said that I thought the assumption is inappropriate in this context. I made no reference that I was aware of to the actual figures. I was criticizing the assumption, because it seems to me an improper assumption to make in this context when there clearly is a possibility of lab error and other kind of error, that anything predicated on the assumption there was not--seems to me to be suspect.
Now, your views on the role of assumptions in statistics are far from mainstream and you have found that many times with census adjustment reactor safety, haven't you?
Professor Speed, is it true that your views on the role of assumptions in statistics are far from mainstream?
That is a correct statement that I put in my letter to Professor Weir, yes. I certainly feel that not enough attention is paid to the role of assumptions in doing statistical calculations and in particular evaluating probabilities, such as associated with nuclear reactors, evaluating probabilities such as those associated with incorrectly concluding that a given individual is the source of a stain. Assumptions are critical in my view, and that is why I said that.
In this particular context I've become more convinced that I have broad support. I was prevented from mentioning names yesterday of many eminent statisticians who agree with me.
I will ask a question, professor. Let's go back to what Professor Weir did when he presented his population frequency calculations to this jury. Okay. How else could the jury appreciate the significance of a genetic match if they decided that no error was in fact performed?
Well, in that case Professor Weir's calculations would probably be the most relevant thing.
And the quote from page 88 of the national research council report prohibits combining the population frequencies estimate and the laboratory error rate; isn't that true?
Now, assumptions do play a role in all sorts of statistical endeavors, don't they?
Let me give you an example. You know the kind of map, I think it is called a Mercator projection map that takes the world as a glove and flattens it out--I think that is the right term for it--
Okay. So if I said to you, Professor Speed, assume the earth was flat, as one of these maps represent, and I was in Chicago and I could walk westbound directly, you could tell me how long it would take me to fall off the edge, if I walked westbound at five miles an hour, couldn't you?
Well, I think if it was made a little more precisely, I probably could. I'm not quite sure I'm following exactly.
But if I forced you to make that assumption, you could do that sort of calculation, couldn't you?
Now, you have not spent any time reviewing the Los Angeles Police Department files on this case, have you?
I have spent a little time, but as you would know, that is outside my area of expertise, so I haven't made any attempt to master it completely.
Okay. And you heard him discuss the possibilities that cross-contamination might have occurred in various places?
And were you here for the demonstration which showed on People's exhibit 564, which demonstrated the history of samples 48, 50 and 52, the replicate testing demonstration?
Now, you don't know whether any error actually occurred with the handling of those samples, do you?
If you knew, if you were able to trace the history of those samples, you would be able to try to troubleshoot the likelihood that those samples might have gotten cross-contaminated, wouldn't you?
Common mode failure, somebody traced the electrical system to find out that all three of these systems were wired through the same box?
They did that after. Could you just define that common mode failure again, please.
Well, it is used in a context where two or more systems can be--in the reliability context, can fail as a result of some common cause, even though it is usually used in a context where beyond that common cause they are actually operating independently, so that the only real common connection between them is this initial one.
And do you know whether--have you evaluated the flow of samples in this case to determine whether there are bottlenecks or common comes in the handling of any of the samples?
Well, I have looked at it closely enough to see the possibility, but as I say, evaluating them in a professional sense is outside my area of expertise.
Well, when you know enough about a process to see that things which might initially be together, then get split and have an independent existence or an independent processing thereafter, when I have seen this situation often enough to know that that--the probabilities get reduced if there is consistency in the later analyses, but that the possibilities of some--in this case we are talking contamination occurring at the earliest stages before that was split--is an instance of this common mode that I've been talking about, so I have seen enough, without being an expert, to see the possibility.
Well, information about some of the, if you like, life histories of some of these samples.
I'm afraid I don't have a detailed recollection of any of these. I have just seen enough to see the possibility.
Well, what have you seen that is enough to contradict the possibility of those samples suffered from common mode failure?
Well, simply listen to Dr. Gerdes' testimony is enough to see the possibility of something happening at the beginning before things part ways. That is enough for me. I don't have to, at least from my point of view, to recognize this possibility, I don't have to read LAPD files or understand the testimony, as far as I'm concerned.
So let me get a clear idea what you heard of Dr. Gerdes' testimony. You heard his direct examination; is that true?
Did he directly address whether common mode failure or anything analogous like that might have occurred with 48, 50 and 52 specifically?
I don't recall that, but he may have. I don't claim to have followed every detail of the testimony.
But whatever he said was enough for you to recognize the possibility, is that what you are--
I'm not making statements specific to items 49, 50 and 52. I was talking about processes in general. And I think I know enough about processes in general to see that there are initial phases where things might be together and then later phases where they are being treated independently, and that the behavior of errors and the overall probability of an inconsistency has to have that sort of temporal and spacial flow. That is about the level at which I'm testifying, not in detail about particular drops or particular samples.
So that temporal and spacial flow that you have perceived through Dr. Gerdes' testimony, that in fact does not relate to any specific item of evidence in this case, does it?
No, it is not--I'm not standing up here testifying to probabilities relating to items of evidence specifically, no.
Sure, sure. I want you to assume that all three of those samples which you know are from the Bundy walkway were typed by the three different laboratories in this case, the Los Angeles Police Department, California Department of Justice DNA lab and Cellmark, and that they all produced the same DQ-Alpha results. Okay?
And in your jargon, the common mode failure where these samples seem to have diverged would be on June 14th in the evidence processing room; isn't that true?
And if those samples were all--all produced the same type, that suggests that if there was common mode failure it had to have happened on June 13th or June 14th?
I'm sorry, objection to the way he said 13th and then 13th or 14th and then ask him to rephrase the question.
That if there was common mode failure it would have to have happened on June 14th when they were all together for the last time?
I'm not making statements with that degree of specificity. In particular, it would seem to me I would need to be a microbiologist or somebody very knowledgeable in this to say when errors occurred. All I was talking--speaking to was the possibility of errors in that initial phase before they were split being analogous to common cause errors, and if you wanted to get the overall probability of the final consistent results being erroneous results, then you would have to consider that step in the process.
I'm certainly not saying that that is the only way in which an error could have affected all of them, but it is a way, as I think should be clear from the flow chart.
KEY QUOTEWell, if no error was made on June 14th in the sampling in the evidence processing room or on June 13th in the evidence processing room, then there is no reason to believe that the results produced by those three labs are incorrect; isn't that true?
Again, I'm afraid I have to be general. You said there was no reason to believe. Nuclear reactors have done very bad things because people had no reason to believe that whatever mechanical function actually happened might have happened. In other words, the fact that you have no reason to believe it doesn't actually mean it is true. I'm not here testifying that it is impossible or that it is necessary that errors may or may not happen in some places. We are talking about probabilities, about chances of errors occurring, and they exist regardless of whether in any particular instance somebody may have no reason to believe they occurred. The chance of errors is something which is ever present in this sort of human activity.
You could walk out in the street and get hit by a car crossing the street, couldn't you?
What significance do you attribute to the fact that given the flow of those samples each of the three labs produced the same result?
Well, that suggests--but again this is an area of professional expertise beyond mine, because I speak to statistics and probability, whereas you are talking now about interpretation of results, but it would suggest that what went out was the same. I mean, an error calculation which only considered errors in the individual labs and there was a consistent result would suggest that they received the same thing. It would certainly tell you nothing about errors that might or might not have occurred before they were sent out.
Professor Speed, in fact there are many other biological samples in this case that did not go through the June 13th and June 14th, let's call it, common mode, aren't there?
Professor Speed, you don't know whether or not there are many other biological samples that did not have the same history of sample 48, 50 and 52 on June 13th and 14th, do you?
Professor Speed, you do not know whether or not there are many other biological samples that did not go through the evidence processing room on June 13th and June 14th, do you?
I know what they are. I don't have an immediate recollection of their history, processing history.
Do you have a photo? It is hard to read from the monitor. Do you have an additional photocopy for Professor Speed?
It would be quicker if we just took that and photocopied. Mrs. Robertson, could we have a photocopy of that real quick, please.
Showing you a copy of exhibit 565, did you see that the other day with Dr. Gerdes?
And in fact those samples, if the dates are correct, have a totally different history than samples 48, 50 and 52, do they not?
And if they have a different history, then they share no common mode with 48, 50 and 52, do they?
And would it be helpful in assessing whether a common mode failure occurred in trying to track the history of specific sample in this case?
I wasn't doing any analyses, as you know, of specific situations in this case.
KEY QUOTEAnd is this one of the reasons you didn't want to talk to him on the phone, because you didn't want to discuss this with him?
Well, I really didn't want to enter into discussions about things on which I was about to testify where I would be--I don't really have much to add to what I said before except that I'm repeating myself. Publicly disagreeing with his point of view makes me feel slightly uncomfortable. I would rather have done it in an informal interchange well in advance of this situation and suddenly a few days before I'm about to start up and testify it is not a situation I relished. And I took your statement that it wasn't necessary that I do this at its face value and declined.
Well, you know, we have busy lives. I've been in Australia a bit. I've had a very busy semester and I don't know where Bruce has been other than here, but we haven't been having regular informal exchanges on these matters since the trial began. And, well, it doesn't seem to me too surprising, but perhaps we should have.
Okay. Now, in fact, if one were to--did you cull that out of his testimony yourself or did somebody present that snippet of his testimony to you to comment on?
I have read the testimony and obviously I was very interested in Bruce's views on something that I was testifying about. It wasn't my suggestion that we actually state it, though. It was something that I preferred to do, rather than have it done some other way.
Do you mean to imply that Professor Weir does not recognize the possibility of error?
Do you mean to imply that he rejects the notion that one might consider laboratory error rate?
It is my impression that Dr. Weir does not accord laboratory error rate the importance that I do and that seemed a fairly clear statement of it, and I haven't actually seen him write anything that is fundamentally contradictory to that statement, so that seemed to be his view.
So you disagree with him on the significance that should be accorded to a laboratory error rate? Is that what you have just said?
He has not said specifically that laboratory errors are impossible, yet he has, to my knowledge, never said they should play a role in the statistics of assessing the chance of incorrectly concluding somebody is the source of a DNA sample.
Okay. I would like you to actually look at the broader--and I would like to show you 37--I'm sorry, pages 33733 and 33734 of Professor Weir's testimony.
Does that help refresh your recollection about the full passage that Professor Weir made those statements in?
Yes, but I certainly would appreciate a copy if you want to question me more about it.
Let's take a recess. All right. Ladies and gentlemen, we are going to take our mid-morning recess. Please remember all my admonitions to you. Dr. Speed, you may step down. You are ordered to return in fifteen minutes. All right. We will take 15.
Dear Mr. Harmon: Thank you for your fax dated July 27, 1995, requesting that I discuss my anticipated testimony with Dr. Bruce Weir. With respect I decline. Yours sincerely, Terry Speed.
Nuclear reactors have done very bad things because people had no reason to believe that whatever mechanical function actually happened might have happened.
I'm certainly not saying that that is the only way in which an error could have affected all of them, but it is a way, as I think should be clear from the flow chart.
I wasn't doing any analyses, as you know, of specific situations in this case. I was speaking in general.