Introduction

            On May 21, 2000, Dr. Lawrence Farwell prepared a Forensic Science Report on a Brain Fingerprinting Test on Terry Harrington regarding the murder of John Schweer (Re: State of Iowa vs. Terry Harrington In the Iowa District Court for Pottawattamie County at Council Bluffs).  This Supplement provides additional scientific and technical detail and perspective, and is an integral part of said report. 

            In preparing scientific reports, there are always tradeoffs.  One tradeoff, particularly in forensic applications, is between using the most advanced and up-to-date science and technology, which will yield the highest accuracy and effectiveness, versus using somewhat older techniques and concepts that have had time to accumulate a longer track record, including more testing, peer-reviewed publications, and acceptance in the scientific community. 

            This supplement seeks to address this tradeoff in the following way.  The previous report emphasized cutting-edge science and technology.  This report will fill in the perspective and data provided by more traditional and more well established science and technology.  The conclusion that this will lead to is that all of the same scientific determinations and conclusions reached in the previous report can be substantiated on the basis of the most traditional and well-accepted methods and scientific theories in the field.  The statistical confidence is higher with the newer discoveries and techniques developed by the author, but the scientific determinations and conclusions are identical to those that are reached by using more traditional techniques that have been extensively tested, have achieved extensive peer review and voluminous publication, and enjoy unquestioned validity and acceptance in the scientific  community.  The accuracy rates achieved using traditional methods, though not quite as high as those achieved using the innovations described by the author in recent publications and in the initial report, are nevertheless near perfect. 

            A second tradeoff is between expressing the science and technology in terms understandable to a layman versus including all of the scientific and technical details that will allow one's scientific peers to understand and appreciate the scientific merit and technical soundness of the science and technology described.  This Supplement addresses both sides of this tradeoff.  We have included herein both more extensive explanations of the science and technology of Brain Fingerprinting in lay terms and more detailed technical and scientific descriptions of the scientific method, apparatus, and procedures employed.

            In order to bring out the relevant information in a format that is accessible, clear, appropriate, and applicable to legal proceedings, much of this supplement has been constructed in a question-and-answer format.

Questions and Answers

Research, Publications, and Funding

QUESTION: Have you published in the field of Brain Fingerprinting or brain-wave-based information detection?

DR. FARWELL: I have published in the leading peer-reviewed journals in the relevant scientific field, including Psychophysiology, Electroencephalography and Clinical Neurophysiology  or EEG Journal, and with Sharon Smith of the FBI in the Journal of Forensic Sciences.  I have also published in other prestigious publications such as Advances in Brain Research, Progress in Brain Research, Annual Review of Gerontology and Geriatrics, and others.  My CV is a part of the original report.

 QUESTION: Where have you conducted research?

DR. FARWELL: In addition to the University of Illinois and Harvard, I have conducted research on FBI agents at the FBI Academy in Quantico, research under contract with the CIA, and research at the US Navy.

 QUESTION: Did you receive funding from the CIA or the FBI, and how much?

ALLEGED WITNESS: My laboratory was funded by the CIA for three years, for a total of about a million dollars. My research at the FBI was approved by the FBI but not funded by the FBI. (I didn't need or apply for FBI funding at the time, because I was funded by the CIA.) The Navy study was done in collaboration with the CIA, and was funded by the CIA.

 QUESTION: Does you CV contain all of the jobs and professional activities you have participated in throughout your life?

ALLEGED WITNESS: No. It contains those professional activities that I deem relevant to my expertise in Brain Fingerprinting.  I have not included professional activities and jobs that are not relevant. For example, I have a black belt in kung fu, and have taught kung fu professionally for many years.  I have not included this because it is not relevant. 

What is Brain Fingerprinting Testing?

QUESTION: What is Brain Fingerprinting, and what does it measure?

DR. FARWELL: Brain Fingerprinting is based on the principle that the brain is central to all human acts.  In a criminal act, there may or may not be many kinds of peripheral evidence, but the brain is always there, planning, executing, and recording the crime.  The fundamental difference between a perpetrator and a falsely accused, innocent person is that the perpetrator, having committed the crime, has the details of the crime stored in his brain, and the innocent suspect  does not.  This is what Brain Fingerprinting detects scientifically.

            Brain Fingerprinting matches evidence from a crime scene with evidence stored in the brain of the perpetrator, similarly to the way conventional fingerprinting matches fingerprints at the crime scene with the fingers of the perpetrator, and DNA fingerprinting matches biological samples from the crime scene with the DNA in the body of the perpetrator.

            Brain Fingerprinting accomplishes this by measuring electrical brain responses to words, phrases, or pictures presented on a computer screen.   Details of the crime that would be known only to the perpetrator and investigators are presented in a sequence, along with other words or pictures that are not relevant to the crime.  We can determine by the brain response if the suspect recognizes the crime-relevant stimuli or not.  This reveals whether or not he has details of the crime that would be known only to the perpetrator stored in his brain.

            The brain waves are measured non-invasively from the scalp with a headband equipped with sensors.  The signals are amplified, digitized, and analyzed by a computer.

Accuracy of Brain Fingerprinting Testing

QUESTION: How accurate is Brain Fingerprinting?

DR. FARWELL: Brain Fingerprinting makes a determination as to whether or not the crime-relevant information is stored in the brain.  The determination can be "information present," "information absent," or, if insufficient data are available to make a clear determination, the result is "indeterminate" -- i.e., no clear determination can be made.  The system also provides a statistical confidence for each determination.  100% of the determinations (i.e. information present or information absent) have been correct, and we have made such a definite determination in 97% of cases.  About 3% of cases have resulted in an indeterminate result.  "Indeterminate" is not an error; it simply means there is insufficient data for a clear determination.  All of the indeterminate results were in the early research, without some of the more modern refinements.

 QUESTION: If 100% of the determinations made were correct, does this mean that Brain Fingerprinting is "100% accurate," that it will always solve every crime and can never make a mistake?

DR. FARWELL: In science, nothing is absolutely 100%.  There is always a level of uncertainty.  Fortunately, when we encounter a situation where no definite determination can be made with a high statistical confidence, then the system produces an "indeterminate" result.  This is done mathematically.  This  avoids making an error when there is insufficient data, and also avoids the scientifically untenable position of claiming that the system is perfect and could never fail to give us a definite and accurate solution to any crime.  Like any scientific evidence, the information provided by Brain Fingerprinting is not be taken as an absolute.  It must be evaluated and weighed along with all of the other available evidence.

Scientific basis of Brain Fingerprinting Testing

QUESTION: Describe the scientific basis of Brain Fingerprinting.

DR. FARWELL: Brain Fingerprinting is an application of the science of cognitive psychophysiology.  The neurons in the brain fire electrically.  When the brain processes information, neurons fire in specific patterns.  These patterns of electrical impulses can be measured non-invasively from the scalp, and from that we can tell what information processing tasks the brain is undertaking.  We put a headband with sensors on the head, amplify the signals, feed them into a computer, and analyze the brain waves.

            The brain responses that measure specific cognitive actions by the brain are called event-related brain potentials.  That is, they are related to an event.  For example, let's say an elephant came in the door of this room.  We would notice it.  We wouldn't have a choice about that.  Once we noticed it, we would have lots of choices what to do -- we could feed it, run away, etc.  But we would have no choice about noticing it.  When we take note of something there is a specific brain-wave pattern that occurs that can be measured from the scalp and analyzed by computer. 

 QUESTION: What do you mean by taking note of something?

DR. FARWELL:  Let's use an example.  Right now we're in a courtroom.  We know what's in the room and what kind of events generally happen in this room. If we close our eyes, we can't see the courtroom, but we still know we're here, because we have this information stored in our brain. In other words, we know the context we are now in; this context is stored in our brain.  That internal representation of the current operating environment is called a schema, sort of a map of where we are, not just physically, but it includes the rules we know about how things happen in this kind of environment. 

            Say the door opens over there and an elephant comes into the room.  Now the context has changed.  We now have a room with an elephant.  We need to be able to update the schema we have in our brain to include this change.  This is known as context updating.  It's a fundamental information-processing function that we all do.  We have to do it to function and survive in a changing environment.

            When we engage in this information-processing function, the neurons in our brain fire in certain patterns, and this creates electrical activity that can be measured at the scalp.  The particular pattern of electrical brain activity that takes place during context updating -- when we recognize the elephant -- was discovered in 1965 by Sutton and colleagues, and he called it a P300. Later we'll get into some finer distinctions in brain responses and terminology, but the fundamental scientific phenomenon that has been well established since the 60's is context updating and the P300 that it produces on the scalp.  P300 is also known as P3, P3b, late positive complex (LPC), or the positive aspect of the MERMER.

Electroencephalography (EEG) and event-related brain potentials (ERP)

QUESTION: How do scientists measure this brain response?

DR. FARWELL: We attach sensors to the head that pick up electrical signals.  These signals are amplified by an amplifier, converted to a series numbers by a digitizer, recorded, and analyzed by computer.

            Now the brain is not only doing the thing we are interested in, it's also doing lots of other things at the same time.  So scientists, starting about half a century ago, had to develop a way of isolating the signal from the noise -- that is, the response we're looking for from everything else the brain is doing. 

QUESTION: How did they do this?

DR. FARWELL: Let's use an example.  When you hear a click in the ear, the signal goes to seven different parts of the brain in the first 1/100^th of a second.  There is a specific recognizable blip in the brain-wave pattern for each of these, but you can't see them from looking at the raw signal because they're tiny, and there is so much else going on.  So what scientists did is to present the click many times, and average the response. This is called signal averaging. Then everything that is not time-locked to the signal averages out to zero, and you can clearly see the response.

            These responses are called event-related potentials, because they are related to an event, in this case hearing a click in the ear.  Detecting event-related potentials through signal averaging is the most well established method for detecting specific, short-term brain activity, and has been for half a century.

            Initially, the event-related potentials studied were measures just of sensory processing of a stimulus, like the clicks.  These are called exogenous event-related potentials, or evoked potentials.  They just measure sensory activity.  At first, scientists only measured a very brief time after the stimulus, and these sensory potentials were all they detected.

            But the brain does much more interesting things than just sensory activity.  When we perceive a meaningful stimulus, we process it cognitively.  After the sensory input comes in, we recognize it, we take note of its significance in the current context.  This is accomplished by the brain.  As science progressed, scientists discovered the event-related potentials that manifest at the scalp when this kind of process is undertaken by the brain.  This is a second type of event-related potentials.  They are called endogenous event-related potentials, or cognitive event-related potentials, because they are a manifestation of cognitive information-processing activity in the brain, not just sensory processing.  Instead of just the first few hundredths of a second, these potentials take place on the time scale of up to a second or two.  They are the manifestation on the scalp of the process that the brain undertakes when the brain processes specific information.

            We talked before about context updating, when we take note of something that is significant in the current context.  This is manifested by a P300. The P300 that was discovered in the 1960's is probably the most thoroughly researched event-related brain potential.

The well established scientific phenomenon applied in Brain Fingerprinting Testing

QUESTION: How well established is this as a scientific phenomenon?

DR. FARWELL: It is at least as well established as any phenomenon in this realm of science.  This was first discovered in the 1960's.  In 1965 Dr. Sam Sutton and his colleagues first reported a specific brain response that takes place when you take note of something.  They called it a P300.

            Since Sutton et al. discovered the P300, there have been hundreds, perhaps over a thousand articles published in peer-reviewed journals on the P300.  It is probably the most thoroughly researched and published phenomenon in the entire field of cognitive psychophysiology.  It has been found to be an extremely reliable indicator of when someone takes note of something, or context updating.

QUESTION: Has this been independently verified in other laboratories, and have their results been published?

DR. FARWELL: Yes, extensively. As I said, the science of this is perhaps the most well established phenomenon in all of cognitive psychophysiology, with hundreds or perhaps even thousands of studies from many different laboratories published frequently in every relevant journal over the last several decades. 

            This specific application has been published in prestigious peer-reviewed journals not only repeatedly by myself and my colleagues, including Sharon Smith of the FBI, but also repeatedly by several others, including Iacono, Allen, and colleagues, Rosenfeld and colleagues, and Rapp and Bashore.  I have published this specific application not only in Psychophysiology, a leading journal in this kind of brain research, but also to be published in January in the Journal of Forensic Sciences, a leading journal in forensic science.  Iacono and Rosenfeld have published in Psychophysiology and other journals.  Bashore and Rapp published in Psychological Bulletin, another prestigious peer-reviewed journal.

Fundamentally different from lie detection

QUESTION: What does this have to do with lie detection?

DR. FARWELL: Absolutely nothing.  Brain Fingerprinting detects information stored in the brain.  The results are exactly the same whether the person lies or tells the truth about this information or any other subject.  If a person's fingerprints or DNA match (or don't match) the fingerprints or DNA at a crime scene, this fact does not change in any way if the person lies or tells the truth about it.  The same is true of Brain Fingerprinting.  All it detects is the presence or absence of information, and this is completely independent of the process of lying or telling the truth.  Lying or telling the truth in no way affects the outcome of a Brain Fingerprinting test.  Brain Fingerprinting detects information, not lying. The fundamental differences between Brain Fingerprinting and psychophysiological detection of deception are covered in more detail in Appendix 1.

How the system works

QUESTION: How does Brain Fingerprinting use this science to detect criminals and exonerate innocent people?

DR. FARWELL: What we want to determine is whether or not someone has committed a crime.  The fundamental difference between a perpetrator and an innocent suspect, of course, is that the perpetrator actually committed the crime, so he has a record of that crime stored in his brain, while the innocent suspect does not.  Brain Fingerprinting detects this record stored in the brain by measuring brain responses to stimuli -- words, phrases, or pictures flashed on a computer screen.

QUESTION: How is this done?

DR. FARWELL:  Recall that we have a brain response, the P300, that the brain emits when it recognizes and processes a stimulus that is significant in the current context.  This has been well established over the last 35 years.  All we do is to present stimuli that will be recognized by the perpetrator as significant, but will not be recognized by an innocent suspect.  So we flash a series of words, phrases, or pictures on a computer screen.  Some of these stimuli are relevant to the crime and will be recognized by the criminal, but are not known to an innocent suspect.  We mix these crime-relevant words or pictures in with other stimuli that would be equally plausible for an innocent suspect. In Brain Fingerprinting, a computer analyzes the brain response to detect the P300, and thus determines scientifically whether or not the specific crime-relevant information is stored in the brain of the suspect. 

QUESTION: What about the MERMER?

DR. FARWELL: MERMER is an acronym for memory and encoding related multifaceted electroencephalographic response.  It's a more extensive response than the P300, and it contains the P300.  When you have a MERMER, you always have a P300, because the P300 is part of the MERMER.  The P300 is simpler to measure and to understand, and it’s a more thoroughly established scientific phenomenon.  As a cutting-edge scientific discovery, the MERMER is of scientific interest, but we don't need anything but the P300 to arrive at a clear determination, with a high statistical confidence, in the Harrington case.  If we want to stick to the tried and true basics, we can use the term P300; if we want to be a little more up-to-date and comprehensive, we can use the term MERMER. 

QUESTION: Is the system you just described the one you used on Harrington? Is it the same as the system used in other studies?

DR. FARWELL: Yes. The system I used on Harrington is essentially the same system used by Dr. Drew Richardson of the FBI and myself in our research on FBI agents, in the studies we did for the CIA, in the research Dr. Rene Hernandez and I conducted for the US Navy, and in the research I am publishing with Sharon Smith of the FBI.  It’s essentially the same as the system Emanuel Donchin and I used in our published research, and as far as the fundamentals and scientific foundations, it's essentially the same as the system used in replications in other laboratories.

The scientific procedure

QUESTION: Exactly what happens during the Brain Fingerprinting test? Can you give more details?

DR. FARWELL: A subject views a series of words or short phrases on a computer. (We can also use pictures, but we usually use words and did so with Harrington.)  Each phrase is flashed for just long enough to read it and see what it means, in Harrington's case, four tenths of a second,.  There are three different kinds of phrases.  Some of the stimuli are things that the subject knows and will recognize, and we make sure he knows these things.  We call these Targets.  To make sure the subject recognizes the Targets, we give him a list of them, and we instruct the subject to press a special button with one thumb when any one of them appears.  So, we know the subject will take note of the Targets. 

QUESTION: What happens when someone takes note of the Targets?

DR. FARWELL: When someone takes note of something, the brain engages in a specific pattern of electrical brain activity that we can measure from the scalp, a MERMER that contains a P300.  Again, I will not get into the technical differences in terminology at this point, because they are not relevant to the basic point or the Harrington results.  It's not necessary to deal with all the technical details to understand the result we obtained in the research and the test with Harrington.  In other words, the fundamental science and the results are the same however you deal with the technical fine points.  I'm going to talk in terms of the P300 for the time being, because that is sufficient to make the necessary scientific points, and also sufficient to obtain the results we obtained with Harrington.

QUESTION: OK, so the person recognizes and notices these things that he knows, and you get a specific brain response called a P300 or a MERMER.  How does this identify a perpetrator?

DR. FARWELL: We need to describe two other kinds of stimuli to get to that point.  The second type of stimuli is phrases that are irrelevant to the subject. These are called Irrelevants. Since these items are irrelevant, that is, they are not significant or noteworthy to the subject in the present context, they do not elicit a P300.

            The stimuli of the third type are the most important ones.  These stimuli are details about the crime that would be known to the subject if he had participated in the crime, but that he would have no way of knowing otherwise. These are called Probes.  These Probes are mixed in with the Irrelevant stimuli, so that a subject who has not participated in the crime will not even know which ones they are.  A subject who participated in the crime, however, will recognize the Probes because they refer to things involved in the crime that he knows about.

            The brain responses to the Probes distinguish whether the person participated in the crime or not.

            If the suspect participated in the crime, he recognizes these crime-relevant stimuli, and his brain produces a P300.  If he is did not participate in the crime and consequently does not know the details about it, he does not recognize these Probe stimuli -- for him, they're indistinguishable from the Irrelevants -- and his brain does not produce a P300.

            The Irrelevants are designed to be equally plausible for a subject who did not participate in the crime, so a subject lacking this crime-relevant knowledge won't even know which stimuli are relevant to the crime.  His is brain will not produce a P300 in response to the Probes.

            Let me give you an example.  Say that the investigated situation involved an espionage crime in which the subject had to find a contact with a blue coat.  "Blue coat" could be a Probe stimulus.  "Red scarf" could be an Irrelevant stimulus.  If someone had not participated in the event, he would not know the difference.  If someone had participated, however, he would recognize the significance of "blue coat."   For a subject who knows about the crime, the brain response to "blue coat" would contain a P300.

            Thus, we have three types of stimuli.  The Targets are things the subject knows, and we know he knows them, and they will produce a brain response with a P300.  The Irrelevants are irrelevant to the subject, and will not produce a P300.  The Probes are relevant to the crime and known only to the perpetrator and investigators, but not to others.  So for a person lacking the crime-relevant knowledge we are testing for, the Probes will not produce a P300 -- like the Irrelevants.  For a person who has the crime-relevant knowledge we are testing for, the Probes will produce a P300 -- like the Targets.

QUESTION:  Could a person beat this system by controlling his response?

DR. FARWELL:  No. Remember the elephant coming into the room?  The first thing we do is to notice the elephant.  After we've noticed him, we can choose to do any number of things -- feed him, run away, etc. -- but we don't have any choice about noticing him and recognizing that this is an elephant.  When we flash stimuli relevant to a crime on the screen, a person who committed that crime first recognizes them, then decides what to do about it.  Maybe he'll try to look innocent, or look like they mean nothing to him, or whatever.  It doesn't matter.  We pick up our information when  he recognizes the stimulus, before he decides what to do about it.

            There are all kinds of things a person could try to do in an attempt to trick the system if he thought he understood how it works, but such manipulations would be easy to pick out in the data analysis, because we have all of the data, and we can analyze responses to individual stimuli or groups of stimuli.  The bottom line is, we've had experts at the Human Brain Research Laboratory -- experts who know exactly how the system works and even people who wrote the software for the data analysis algorithms -- try to beat the system in detecting real-life information, and they can't.  I can't beat it myself.  The information-processing brain function that Brain Fingerprinting detects is a process that takes place automatically when you recognize and take note of a significant stimulus, and you don't have a choice about doing that when such a stimulus is presented.

            A person could refuse to even look at the screen, and then of course we would not get any data, but this would be obvious.  If he pretended to look at the screen but didn't, he could not push the right buttons, and we would know this because we record the button presses.  So, we can't physically force a person to take the test, but if he sits in front of the computer screen and actually pushes the buttons when the Target and Irrelevant stimuli come up, he has to recognize the Probes as well if he has the crime-relevant information, and we'll detect the response.

Limitations on the applicability of Brain Fingerprinting Testing

QUESTION: Is Brain Fingerprinting applicable in every case?

DR. FARWELL: No. Like every other scientific technique, there are limitations on the applicability of Brain Fingerprinting.  DNA and conventional fingerprints, which are very effective technologies, are applicable in only about 1% of cases.  In the vast majority of cases, there are no DNA and no fingerprints at the crime scene.  With Brain Fingerprinting, the brain is always there, planning, executing, and recording the crime.  We can apply Brain Fingerprinting whenever we have a suspect and we have specific details about the crime that would be known only to the perpetrator and investigators, and not to an innocent suspect.  This certainly provides a much higher level of applicability than fingerprints or DNA, but it is not applicable in every single case.  When we don't have and can't get the necessary information, then Brain Fingerprinting is not applicable.  When it's not applicable, obviously we don't apply it. 

            Note that this is a limitation on the applicability, not the reliability or the accuracy of the system.  It's like fingerprints -- where there are none, it's not that the technique will be used and will be inaccurate: the technique simply does not apply and we do not use it.  With Brain Fingerprinting, where we cannot find specific details of the crime that are known only to the perpetrator and investigators, then the technique is not applicable, and we do not use it.

QUESTION: What if the person knows about the crime because he was there, but he was a witness and not a perpetrator?

DR. FARWELL: Brain Fingerprinting will detect what information is in the brain, but will not tell us how it got there.  It's like having fingerprints at the crime scene.  Someone's fingerprints could be there because he was there witnessing the crime and not because he committed it.  In a case where there are two people at a crime scene and only one committed the crime, all Brain Fingerprinting can do is to narrow the search down to two suspects. It can not be used to distinguish why a person was at the crime scene.  This, however, is relatively rare.  In almost every case when a person says he is innocent, he also says he was not there. 

            Like DNA and fingerprints, Brain Fingerprinting matches evidence at a crime scene with evidence on the person of the perpetrator.  It can place a person at the crime scene or exonerate someone who was not there, but in some cases there will be some other reason someone was there other than committing  a crime.  In such cases, we can apply Brain Fingerprinting to narrow the range of suspects to the people who were present at the crime, but we cannot use it to determine which one committed the crime.

QUESTION: What if a suspect read about the crime in the newspaper, or what if he's a suspect and the police told him all about it in interrogation?

DR. FARWELL: If a suspect knows everything we know or could discover about the crime, from some other means other than from committing the crime, then we cannot apply Brain Fingerprinting.  To apply the test, we need some information that the subject will have only if he committed the crime.  This is why it is important to educate law enforcement personnel so that they will not reveal everything they know about a crime to a suspect before a Brain Fingerprinting test can be administered.

QUESTION: What if some of the stimuli are relevant to the subject for some other reason that has nothing to do with the crime?

DR. FARWELL: We go over a list of all the stimuli just before we run the test.  This reinforces the context in which the test is run.  It also gives the subject a chance to tell us if some stimulus is relevant to him for any reason.  If so, we eliminate the stimulus.   Whether he's innocent or guilty, he can be expected to  deny knowing the details of the crime that would be known only to the perpetrator.  If a Probe is significant to a suspect because of the crime, we don't expect him to reveal it to us by saying so.  So a guilty suspect is expected to say he does not recognize the crime-relevant Probes as significant. What he says, however, makes no difference to the outcome of the Brain Fingerprinting test.  If we test him and his brain responses reveal that he does recognize the crime-relevant Probes, there's only one reason why: that he participated in the crime.

Application of Brain Fingerprinting Testing in the Harrington case

QUESTION: Harrington was tried and convicted.  In the course of the trial he heard all about the crime, and that information was stored in his brain.  If you detected that information, it wouldn't prove he participated in the crime, only the trial.  So how did you use Brain Fingerprinting in this case?

DR. FARWELL: We couldn't use any of the information Harrington heard at the trial, because showing that he had this information stored in his brain would not prove anything about the crime.  He would have this information stored in his brain from the trial, not from committing the crime. So we had to find information about the crime that he was never told in the trial.

QUESTION: Was this difficult?

DR. FARWELL: The Harrington case is about as difficult a case as we would ever expect to find, but it was still quite possible to solve it.  If Brain Fingerprinting had been invented in 1977, it would have been a very easy case to solve, but Brain Fingerprinting was not invented until the mid-1980s.  On the day he was picked up for the crime, there were hundreds of details about the crime that Harrington did not know, and that the investigators did know.  We could have easily used these to test Harrington's brain and demonstrate that he did not know a thing about the crime.  But he heard voluminous information about the crime in the interrogations and the trial, so we couldn't use any of that information in a test.  What we had to do was examine the available evidence and develop a test that would include only things he would know about the crime if he committed it, but that he had never been told at the trial.

QUESTION: Were you able to do that?

DR. FARWELL: Yes.

QUESTION: Can you give me an example?

DR. FARWELL:  Yes.  The only alleged witness to the crime was Kevin Hughes, a 16-year-old black who was himself accused of the crime. Hughes told a detailed story about Harrington committing the crime, which he has now recanted.  Hughes' account of the crime was essentially the only evidence against Harrington.  I couldn't test Harrington directly on what Hughes said at the trial -- because Harrington knew this from the trial, not from the crime.  I could, however, use Hughes' testimony, along with police reports, crime scene photos, and the crime scene itself, to determine what the perpetrators must have encountered in committing the crime that Hughes described.  Some of these things that the perpetrators would have had to experience in committing the crime were never mentioned specifically by Hughes or otherwise revealed at trial, so Harrington would have no way of knowing them unless he committed the crime. 

            Fortunately, the basic lay of the land at the crime scene, in terms of major buildings, streets, railroad tracks, etc., is the same now as it was at the time of the crime.

            Here is the account that Hughes gave of the crime at the trial.  Hughes testified that he drove to the crime scene with Harrington and the other alleged perpetrator, Curtis McGhee.  He said they parked the car at a certain place near the lot, Harrington took a shotgun out of his trunk and wrapped it in a jacket, then he saw Harrington and McGhee disappear around the corner onto the lot where they were going to steal a particular car. The victim was shot, not at that car lot, but on some railroad tracks about a block away, and the perpetrators had to cross the street to get to the place he was shot.  Ballistic evidence showed the direction from which he was shot, so I could determine from that and crime scene photos what was behind him when he was shot.  There was actually another car lot behind him when he was shot, with parked cars immediately behind him.  Hughes said that after he heard a shot, he saw Harrington and McGhee come running out from behind a particular building, up onto the street and over to the getaway car.  There was a large drainage ditch next to the road that the perpetrators had to get through to get to the getaway car. The car was parked next to some large trees, but McGhee did not say that at the trial.  Then, according to Hughes, they drove away.  He specified the parking place on an aerial photo, and named the streets they drove on.  I was able to determine from Hughes description and crime scene photos which way they would have had to drive the car immediately upon entering it to follow the route he specified, although he never said this explicitly at trial.

            There are several significant things here that never came up explicitly in the trial, but the perpetrators must have encountered to commit the crime.  Some of these things are easy to figure out if you have the crime scene photos and the testimony, and can actually walk around the crime scene and see what is the lay of the land, what is next to what, where the parts of the scene are in relation to each other and in relation to the photos, etc.

            For example, to get from the murder scene to the getaway car, Hughes testified that Harrington ran behind a specific building.  At that time, I was able to piece together from crime scene photos and visiting the crime scene that there were waist-high weeds and grass behind the building, but you couldn’t tell that from what you would have heard and seen at the trial.  So I asked Harrington, "Did you shoot Schweer?"

"No."

"Were you at the crime scene?"

"No."

" Did you run behind that building next to the tracks?"

"No."

"So you don't know what was behind the building?"

"No."

"You don't know whether it was cement and blacktop, sand and gravel, or weeds and grass?"

"No."

            Running from a murder scene in the dark, one could not fail to notice that one had to run through waist-high weeds and grass, which would have been a significant impediment.  He claimed not to know that.

            There were several major facts about the crime that the perpetrator knows but Harrington claimed not to know.  Another example is, the perpetrators had to cross the street to get from the car lot where the car was to be stolen to the scene of the shooting.  I asked Harrington,

"Do you know where the perpetrators had to go to get from the car that was to be stolen to the scene of the shooting?"

"No."

"Do you know if they had to go under an underpass, across the street, or over a bridge?"

"No."

            So Harrington was claiming not even to know where the perpetrators went to commit the crime. 

            To get to the getaway car, the perpetrators had to negotiate a deep drainage ditch.  Harrington claimed not to know what kind of obstacle they had to negotiate -- drainage ditch, wire fence, or concrete wall.

            He also claimed not to know what was behind the victim when he was shot (parked cars), or what the getaway car was parked next to (large trees), or what direction the car was driven initially for the getaway (straight ahead). (Hughes testified that Harrington drove the car.)

            Clearly, if Harrington committed the crime, this is information he would have.

QUESTION: This was 23 years ago. What if he has forgotten what happened that night?

DR. FARWELL:  We ran a second test to eliminate this possibility. Harrington had several alibi witnesses who testified that at the time of the crime he was in Omaha, not Council Bluffs, at a concert and later driving around town with friends.  I examined the transcripts and spoke to one of Harrington's major alibi witnesses, Mr. Karl Wright, who was his football coach at the time.  I found out details of the events of that same evening, as recounted by the alibi witnesses, events that took place far from the crime scene at the time of the crime.  I tested Harrington's brain for information relevant to the events of the evening, as recounted by the alibi witnesses.  For example, Harrington's football coach said he had a long conversation with Harrington that evening about football, and the fact that he had benched Harrington.  They met at a music concert where the football coach had gone to pick up his daughter. They sat on some white bricks near the gate to the concert at the time. 

Harrington's brain responses

QUESTION: What were Harrington's brain responses?

DR. FARWELL:  Please see Appendix 2, Figure 1: Harrington's Brain-Wave Responses to Crime-Scene Information. 

            These are plots of Harrington's brain responses.  To clarify the data and to isolate the data of interest, we have included only the time range from 600 to 1500 milliseconds (0.6 to 1.5 seconds) after the stimulus. (The full waveforms are included in the original report.)  Before this range, the subject has not yet recognized and processed the differences in meaning of the stimuli of different types, so the brain responses to the different stimulus types are the same.  After this time range, the response of interest is over.  So we're focusing on this time range.

QUESTION: What do Harrington's brain responses reveal?

DR. FARWELL:  First let's look at Figure 1, Harrington's Brain-Wave Responses to Crime Stimuli.  The red line is the response to the Targets.  This is the voltage across this time range at the parietal area of the scalp, a standard scalp site known as Pz, where the P300 is of maximum amplitude.  We can see that there is a positive peak here.  That's the P300.  This is followed by a negative deflection.  The two of these together constitute a MERMER.

QUESTION: So the MERMER includes the P300 plus a subsequent negative deflection?  Is that all?

Dr. FARWELL: There is more to it than that, but the additional scientific details have no bearing on this case, because I did not use them in the data analysis.  They are of scientific interest, but we can safely ignore them for the purposes of the findings on Harrington.  For those who are interested in the technicalities,  I am convinced that there are phasic changes in the frequency-domain brain response that do not appear in the time-domain averages, because they are not phase-locked to the stimulus.  I have some data to support that, and I maintain that this is a scientifically interesting and potentially important distinction. The negative deflection also has a different scalp distribution than the P300, with a prominent frontal component.  These distinctions are not, however, necessary for understanding what we measured in the Harrington case, and they were not included in the data analysis algorithm we applied to Harrington's data.  So, for the purposes of this discussion we can say that the MERMER contains the P300 positive peak and a subsequent negative deflection.  We can isolate our discussion and analysis to the P300 without changing in any way the scientific conclusions warranted by the data.  This allows us to remain within the realm of scientific phenomena that have been extensively tested, peer-reviewed, and published, are known to be accurately detectable, and are extremely well accepted in the scientific community. 

QUESTION: Why do we get a large P300 in response to the Targets?

DR. FARWELL: Because they are noteworthy to the subject.  Remember the elephant?  The subject knows these details about the crime, we have discussed them with him, and he is required to press a special button when they appear.  So he recognizes and takes note of them, and we get a P300.

QUESTION: What about the other brain responses?

DR. FARWELL: The green line represents the response to the Irrelevant stimuli.  These are things that are irrelevant to the crime and irrelevant to the subject.  They are items that would be equally plausible as details of the crime for a subject who did not know about the crime.  As you can see here, the Irrelevants do not produce a large P300, nor do they produce a MERMER.

QUESTION: What about the crime-relevant stimuli?

DR. FARWELL: Recall that the crime-relevant stimuli -- that is, the ones that the subject will know if he participated in the crime but has no way of knowing if he did not -- are called Probes.  The response to the Probes is represented by the blue line.  Note that with the crime stimuli, the response to the Probes is just like the response to the Irrelevants.  In the time range of interest, there are no significant peaks or troughs; there is no P300 and no MERMER.  This indicates that Harrington does not recognize the Probes as any different from the Irrelevants.  That is, his brain does not contain the relevant information about the crime. 

QUESTION: What about the alibi stimuli?

DR. FARWELL: See Appendix 2, Figure 2: Harrington's Brain-Wave Responses to Alibi Information. For the alibi stimuli, as with the crime stimuli, we get a large P300 and a large MERMER to the Targets.  Recall that these are things we know that he knows.  Also as before, we do not get a large P300 or a large MERMER to the Irrelevants. 

            Here, however, we do get a large P300 and a large MERMER to the Probes.  These are stimuli relevant to the alibi that we did not identify to Harrington during the test. 

QUESTION: So what does this show?

DR. FARWELL: The Probe stimuli contain information about the events of the evening of the crime, as described by Harrington's alibi witnesses, who placed him in a different city, at a concert and later driving around town with friends at the time of the crime.  For example, according to an alibi witness who was said he saw Harrington at the concert, he and Harrington sat on a white brick wall and talked about football during the concert.  "White bricks" and "football" were two of the Probes.

            Brain Fingerprinting showed that the information stored in Harrington's brain regarding the events of that evening matched the alibi, and did not match the crime scene. So Harrington did have a record of the evening of the murder stored in his brain.  It was a record of the events that he actually did participate in, according to alibi witnesses -- attending a concert with friends. It was not a record of the crime scene.

            The scientific fact here is that the record of the evening of the murder stored in Harrington's brain matches the alibi and does not match the crime.  This is essentially the same as if his fingerprints or DNA match the fingerprints or DNA at the scene of the alibi and not the fingerprints or DNA at the scene of the crime. How do we interpret that fact?  Well, one reasonable conclusion is that he was not at the crime scene, but rather was at a concert in another city with friends.  I'll discuss this in more detail later.

Data analysis and statistical confidence

QUESTION: How certain is this result?

DR. FARWELL:  You can clearly see the result here on the plots, but we don't go by visual examination, or by your impression or mine of the graphs.  We compare the brain responses mathematically, and come up with a mathematical determination of "information present" or "information absent" and a statistical confidence for this.  For the crime scene information, the determination was "information absent" -- the details of the crime were not stored in Harrington's brain.  For the alibi information, the determination was "information present" -- Harrington's brain did have a record of the events of his alibi stored in his brain.  The statistical confidence for this determination depends on how much data we include in the analysis.  When we include only the P300, we get a confidence of 99% for both determinations -- "information absent" regarding the crime, and "information present" regarding the alibi.   When we use the full MERMER, we get exactly the same results, but with a higher statistical confidence, 99.99% for both determinations.

            I analyzed the data using the MERMER, and I also analyzed the data using only the P300, just to be on the conservative side.  When I leave out all of the refinements, discoveries, and innovations I have come up with recently, and use only the most traditional, well-established scientific phenomenon in the entire field of cognitive psychophysiology over the last 35 years -- that is, the P300 -- I get exactly the same result, with a confidence of 99%.  Using the full MERMER doesn't change the results; it only gives us more data to work with so we get a higher statistical confidence.

QUESTION: How did you analyze the data?

DR. FARWELL: I have described the algorithms in detail in my publications and patents. (See Appendices 6 and 7 of the Report.) I used the same standard algorithms on the Harrington data as my colleagues and I used previously.  I used the statistical procedure of bootstrapping on the correlations between the Probe and Target waveforms, compared with the correlations between the Probe and Irrelevant waveforms.  Essentially, the question we are asking mathematically is, "Are the Probe brain responses like the Target responses or are the Probe responses like the Irrelevant responses?"  If the Probe responses are like the responses to the Targets, this indicates the subject recognizes the Probes as significant.  If the Probe responses are like the responses to the Irrelevants, this indicates that the subject does not recognize the Probes as being any different from the Irrelevants.

QUESTION: Is this the standard data-analysis procedure for this application?

DR. FARWELL: Yes.  It is the same procedure we used in Farwell and Donchin 1991 and Farwell and Smith 2001.  Essentially this same procedure has been used by Iacono as well. 

QUESTION: What is the difference between the P300 data analysis you conducted and the MERMER data analysis you conducted?

DR. FARWELL:  They are identical except for the time window.  In both cases I used bootstrapping on correlations.  For the P300, I used a time window from 600 to 1000 msec.  For the MERMER, I used 600 to 1600 msec.  The results I got are also identical, except that the MERMER gave a higher statistical confidence.

QUESTION: Why do we look at this time range, and not the early part of the brain response?

DR. FARWELL: There is a pattern to the response for the early period before this time range also, but it is not relevant to what we are measuring.  Initially, immediately after the stimulus, the brain is processing the stimulus at the sensory level.  We see event-related potentials here, but they are the same for all three stimulus types, Targets, Probes, and Irrelevants. This is because on a sensory level, these three types of stimuli are not different.  They cause the same sensory processing and the same exogenous or sensory event-related potentials.  A little later, the subject recognizes the cognitive meaning of the stimuli, and at this point he differentiates between the different stimulus types -- Targets, etc.. At this point cognitive or endogenous event-related potentials begin. 

            The brain responses for the different stimulus types diverge at the point where the subject begins processing the different cognitive meanings of the Target, Probe, and Irrelevant stimuli for the subject.  If we were studying sensory processing, the early event-related potentials would be of interest.  For our purposes, however, we are only interested in the cognitive event-related potentials that manifest information processing of the meaning of the stimuli.  That is why we isolate this time range in analyzing the brain responses.

Accuracy rate of Brain Fingerprinting Testing

QUESTION: Is the 99% or 99.99% confidence you yielded by your data analysis algorithm an accuracy rate?

DR. FARWELL: No. It is a statistical confidence for this specific result. It means, what is the mathematical probability that this is a genuine phenomenon, and did not just take place by chance?  What is the probability that these differences we see here on the graph are actually real? This is our statistical confidence for the result.  Accuracy rate is how often we get the correct result.

QUESTION: What has been the accuracy rate of Brain Fingerprinting in the past?

DR. FARWELL: The accuracy rate has been 100% in all the studies I have done.  Here we have to distinguish the different possible outcomes of a test.  The outcome can be one of three: information present, information absent, or indeterminate.  An indeterminate outcome means that the mathematical algorithm determines that we do not have enough data to make a clear determination.  In simple language, the outcomes are "this information is stored in the brain," "this information is not stored in the brain," and "no clear distinction can be made based on the available data."

            For the "information present" and "information absent" outcomes, they can in theory be either correct or incorrect.  In the work I have done, they have always been correct.  "Indeterminate" does not mean incorrect, it just means that the system could not make a clear determination, so it does not make either determination.

            Incorrect determinations can be of two types, false positives -- people who lack the crime-relevant knowledge falsely found to be "information present," -- and false negatives -- people who have the crime-relevant knowledge falsely found to be "information absent." My colleagues and I have never had either one of these.

            In the study we published originally in 1986, in the three CIA studies including the Navy study, and in the two FBI studies including the one in the Journal of Forensic Sciences, we made a definite determination in every case.  There were no indeterminates. 

            In the Farwell and Donchin study published in 1991, there were 12.5% indeterminates.  That is, we made a determination in 87.5% of the cases.  In every case where we made a determination, 100% of these determinations were correct.  There were no false positives and no false negatives.  So the system never made any inaccurate determinations, but there were 12.5% of the cases where no determination was made.

Standard technical and scientific procedures, techniques, and phenomena

QUESTION: Did you follow standard technical and scientific procedures for measuring brain waves in your test of Harrington?

DR. FARWELL: Yes.

QUESTION: Could you specify these standard technical and scientific procedures, techniques, and the relevant scientific phenomena?

DR. FARWELL: Yes.  Table 1 summarizes the basic standard technical and scientific procedures we used and phenomena we measured.  The experimental design features are described in more detail in the original report, along with a description of the most salient scientific and technical details.

Table 1

Scientific and Technical Procedures and Phenomena

1.      Measurement of electrical brain signals (electroencephalography or EEG) non-invasively from the scalp

2.      Signal averaging for detecting event-related brain potentials (ERPs)

3.      The International 10-20 System for electrode placement

4.      The P300 component of the event-related potential

5.      The measurement of P300 from Fz, Cz, and Pz scalp sites

6.      Analysis of P300 using data from the Pz scalp site where it is maximal

7.      The statistical technique of bootstrapping as described in Wasserman and Bockenholt 1989,  Farwell and Donchin 1991, and Efron 1979.

8.      Optimal digital filters as described in Farwell, Martinerie, Bashore, Rapp, and Goddard 1991 with a passband cutoff frequency of 6 Hz and a stopband cutoff frequency of 8 Hz

9.      Measurement of eye movements (EOG) through their electrical signals

10.  Artifact rejection of trials with EOG range greater than 117 microvolts

11. Grass P5 amps set to a gain of 50,000 for EEG, 10,000 for EOG, 0.1 Hz 1/2 amplitude low frequency analog filter, 30 Hz 1/2 amplitude high frequency analog filter, 60 Hz notch filter

12. Silver-silver chloride electrodes (disposable)

13. Linked ears as a reference

14.  Visually presented word stimuli consisting of short phrases for the elicitation of the event-related brain potentials

15.  Button-press responses with the left and right thumbs to different categories of stimuli

16.  A 400 msec stimulus duration for visually presented word stimuli consisting of multiple words in a phrase

17.  Blocks consisting of 72 trials each, averaging and analyzing across 24 blocks for the crime stimuli and 16 blocks for the alibi stimuli

18.  Presentation of visual word stimuli in a random sequence

19.  Digitizing rate of 100 Hz

20.  Digitizing with a Scientific Solutions AD signal processing board.

QUESTION: Are these procedures, techniques, and phenomena standard and accepted in the scientific community for cognitive psychophysiology?

DR. FARWELL: Yes.

QUESTION: Does this mean that all laboratories will use exactly the same procedures, parameters, and techniques?

DR. FARWELL:  No, not exactly. In any scientific field, the standard procedures are not a straight jacket, but provide a range and a set of guidelines for what will produce optimal results.  For many of these parameters, there is an acceptable or optimal range, and different scientists make choices within the range based on the specific experimental design, the scientific phenomena investigated, and personal preferences.   For example, in the Farwell et al. paper on Optimal Digital Filters for Long-Latency Event-Related Potentials published in Psychophysiology we conducted research on the best features and settings for the several parameters involved in a digital filters used to minimize noise in the data and increase the signal-to-noise ratio for data analysis.  We found that a range of parameters for digital and analog filters provided acceptable performance, and there are some tradeoffs with the different parameters depending on which features are important in a specific experimental design. I am not saying that every successful laboratory uses exactly the same parameters we use.  What I am saying is that we used parameters, procedures, phenomena, and techniques that are well known, well accepted, and within the range of what other respected, experienced, and successful laboratories in the field use. 

Dr. Farwell's research and results

QUESTION: Could you describe your scientific studies on Brain Fingerprinting, or detection of concealed information through the use of brain waves, and their results?

DR. FARWELL: Table 2 summarizes the studies and results my colleagues and I have conducted on this scientific phenomenon.  (Complete references are contained in the original report, in the references section of Appendix 6, Farwell and Smith 2001.) Note that the terminology used in the various studies varies.  Not all of these used the term "Brain Fingerprinting," but what is important is the phenomenon and not what name we may give to it.  Terminology tends to evolve over time, and this is no exception.

Table 2

Scientific Studies on Brain Fingerprinting by Farwell and Colleagues