Brain fingerprinting is a method that detects concealed information stored in the brain by measuring event-related brain potentials (ERPs) in response to stimuli. It works by presenting crime-relevant words or images and measuring the brain's P300-MERMER ERP, which indicates recognition. Tests on the FBI, CIA, and US Navy achieved 100% accuracy with no false positives or negatives, though 3% of results were inconclusive. Brain fingerprinting provides an objective means to access information stored in the brain beyond what traditional forensic methods allow.
Brain fingerprinting is a method that detects concealed information stored in the brain by measuring event-related brain potentials (ERPs) in response to stimuli. It works by presenting crime-relevant words or images and measuring the brain's P300-MERMER ERP, which indicates recognition. Tests on the FBI, CIA, and US Navy achieved 100% accuracy with no false positives or negatives, though 3% of results were inconclusive. Brain fingerprinting provides an objective means to access information stored in the brain beyond what traditional forensic methods allow.
Brain fingerprinting is a method that detects concealed information stored in the brain by measuring event-related brain potentials (ERPs) in response to stimuli. It works by presenting crime-relevant words or images and measuring the brain's P300-MERMER ERP, which indicates recognition. Tests on the FBI, CIA, and US Navy achieved 100% accuracy with no false positives or negatives, though 3% of results were inconclusive. Brain fingerprinting provides an objective means to access information stored in the brain beyond what traditional forensic methods allow.
Brain fingerprinting is a method that detects concealed information stored in the brain by measuring event-related brain potentials (ERPs) in response to stimuli. It works by presenting crime-relevant words or images and measuring the brain's P300-MERMER ERP, which indicates recognition. Tests on the FBI, CIA, and US Navy achieved 100% accuracy with no false positives or negatives, though 3% of results were inconclusive. Brain fingerprinting provides an objective means to access information stored in the brain beyond what traditional forensic methods allow.
Brain ngerprinting: a comprehensive tutorial review of detection
of concealed information with event-related brain potentials Lawrence A. Farwell Received: 17 March 2011 / Revised: 26 November 2011 / Accepted: 30 January 2012 / Published online: 17 February 2012 Springer Science+Business Media B.V. 2012 Abstract Brain ngerprinting (BF) detects concealed information stored in the brain by measuring brainwaves. A specic EEG event-related potential, a P300-MERMER, is elicited by stimuli that are signicant in the present con- text. BF detects P300-MERMER responses to words/ pictures relevant to a crime scene, terrorist training, bomb- making knowledge, etc. BF detects information by mea- suring cognitive information processing. BF does not detect lies, stress, or emotion. BF computes a determination of information present or information absent and a statistical condence for each individual determination. Laboratory and eld tests at the FBI, CIA, US Navy and elsewhere have resulted in 0% errors: no false positives and no false negatives. 100% of determinations made were correct. 3% of results have been indeterminate. BF has been applied in criminal cases and ruled admissible in court. Scientic standards for BF tests are discussed. Meeting the BF scientic standards is necessary for accu- racy and validity. Alternative techniques that failed to meet the BF scientic standards produced low accuracy and susceptibility to countermeasures. BF is highly resistant to countermeasures. No one has beaten a BF test with coun- termeasures, despite a $100,000 reward for doing so. Principles of applying BF in the laboratory and the eld are discussed. Keywords Brain ngerprinting P300-MERMER P300 Event-related potential Detection of concealed information Introduction and background The state of the art prior to brain ngerprinting Brain ngerprinting is an objective, scientic method to detect concealed information stored in the brain by mea- suring electroencephalographic (EEG) brain responses, or brainwaves, non-invasively by sensors placed on the scalp. The technique involves presenting words, phrases, or pic- tures containing salient details about a crime or investi- gated situation on a computer screen, in a series with other, irrelevant stimuli. Brain responses to the stimuli are mea- sured. When the brain processes information in specic ways, characteristic brainwave patterns can be detected through computer analysis of the brain responses. When an individual recognizes something as signicant in the cur- rent context, he experiences an Aha! response. This response is characterized by a specic brainwave pattern known as a P300-MERMER. Brainwave responses are analyzed to determine whether or not the specic infor- mation tested is stored in the brain of the subject or not. Brain ngerprinting computes a determination of infor- mation presentthe subject knows the critical informa- tion, or information absenthe does not. The system also computes a statistical condence for each individual determination, e.g., information present, 99.9% con- dence indicates that there is a 99.9% probability that the subject knows the relevant information tested. If the sta- tistics computed do not provide a statistical condence high enough to meet a predetermined criterion for either a determination of information present or information absent, then no determination is made: the outcome is indeterminate. This tutorial review discusses the science of brain n- gerprinting in light of the current state of the art in forensic L. A. Farwell (&) Brain Fingerprinting Laboratories, Inc., 14220 37th Ave NE, Seattle, WA 98125, USA e-mail: [email protected] 1 3 Cogn Neurodyn (2012) 6:115154 DOI 10.1007/s11571-012-9192-2 science, the scientic principles on which the technique is based, published scientic data, successful eld applica- tions, applications in the judicial system and legal admis- sibility in court, scientic methods and standards for brain ngerprinting tests, correcting misconceptions and avoid- ing errors in applying the science, and the role of brain ngerprinting in criminal investigations and security. Prior to the invention of brain ngerprinting, the state of the art in forensic science, investigations, and criminal justice was as follows. The goal is to reconstruct the crime and accurately identify the perpetrator. This is accom- plished by connecting features of the crime scene with features associated with the perpetrator. The crime involves participants and a crime scene. The participants include one or more perpetrators, and may include one or more victims and/or witnesses. The crime scene has two types of features that may be of use: permanent features and changes wrought by the crime. Permanent features pre-exist the crime and persist after the crime. These include buildings, streets, the lay of the land, etc. Changes that took place at the time of the crime include such things as the positioning of the body in the case of a murder, ngerprints, blood at the crime scene, etc. The participants in the crime also have permanent and changed features. Permanent features of the participants include, for example, DNA and ngerprints. Changed features of the perpetrator may include, for example, wounds sustained in the course of the crime. Changes to the victim wrought by the crime can be considered along with the crime scene. The fundamental task of investigations is to establish accurate and reliable connections between features of the crime scene and victim on the one hand and features of the perpetrator and witnesses on the other. The perpetrator may leave traces of a permanent feature of the perpetrator, such as DNA or ngerprints, at the crime scene. The perpetrator may take with him from the crime scene traces of a per- manent feature of the crime scene, such as an unusual kind of soil on his shoes. The perpetrator may take with him from the crime scene traces of changed features of the crime scene, such as the blood of the victim. In some cases, investigators may establish a connection directly between a feature of the crime scene and a per- manent feature of the perpetrator himself. This can be the case with DNA and ngerprints. In other cases, investi- gators may establish a connection between the crime scene and something associated with the perpetrator. For exam- ple, the perpetrator may leave behind ber samples at the crime scene that can be matched to clothing he owns. The crime scene generally has many features that can be established with certainty. Every suspect also has many features that can be established with certainty. When and to the extent that specic features common to both are available, forensic science can establish objective, deni- tive connections between the crime and the perpetrator. Generally, however, in the past there have been only very few specic features of the crime scene that can be den- itively and uniquely matched to features of a particular suspect. Generally the most major change that takes place in the perpetrator and witnesses is something that has not been directly detectable through forensic science techniques. The perpetrator and witnesses to a crime virtually always know that they have observed and/or participated in the crime. They virtually always know more than a few small details about what took place and who was involved. The record stored in the brains of the witness and perpetrator is often a much more comprehensive account of the crime than what can be pieced together from connecting a few specic features of the crime scene with a few specic features of the perpetrator. The record stored in the brains of witnesses and perpe- trators, however, has not been accessible to scientic scrutiny. The only way to access this record has been through interrogation and testimony. The record of the crime stored in the brain of witnesses and revealed through testimony has the advantage of often providing by far the most comprehensive account of the crime available. Except in rare cases where a crime is recorded on video, witness testimony often provides much more information than is available through any other means. Witness testimony, however, has several disadvantages. Even when the witness is deemed to be truthful, witness testimony is not a complete and accurate account of the crime. Witness testimony is a subjective report of the con- tents of memory. It is well known that human memory is not a perfect record of events. Memory is known to be limited and imperfect in a number of ways. It is approximate, sometimes distorted, selective, and subject to numerous inuences. Memory is known to be affected by mental or physical ill- ness, injury, passage of time, drugs, and many other factors. The two primary disadvantages of witness testimony are the following: (1) Human memory is imperfect; and (2) The witness may lie. In weighing the advantages and disadvantages of wit- ness testimony, which constitutes subjective reports of the contents of human memory, every judicial system throughout history has reached the same three conclusions: 1. The brain is a sufciently accurate recorder of events that testimony constituting subjective reports of the contents of memory is universally admitted and considered as evidence. 2. The brain is a sufciently imperfect recorder of events that witness testimony is never taken to be absolute fact. Any proceeding that includes witness testimony 116 Cogn Neurodyn (2012) 6:115154 1 3 must weigh the testimony in light of the well known limitations of human memory. 3. Any proceeding that includes witness testimony must consider the veracity of the witness. The same principles that apply to the subjective reports of witnesses of the contents of their memory also apply to suspects, with the added disadvantage that perpetrators have a greater motivation to lie. Even in the case of con- fessions, the same limitations that apply to witness testi- mony apply to testimony by a suspect. In an attempt to eliminate one of the three major limi- tations of testimony, investigators have developed psy- chophysiological methods to attempt to detect deception (Farwell 1995a, 2013). The fundamental premise of the various techniques for detection of deception is that lying produces emotional stress and other psychological effects, which in turn produce physiological arousal and other physiological changes. These can be measured through changes in perspiration, blood pressure, breathing, etc. The commonly used control question test (CQT) in conven- tional detection of deception employs direct, relevant questions regarding participation in the crime, such as Did you shoot Mr. Jones? Lykken (1959, 1960) originated a new technique for discovering more of the details of the record of the crime stored in the brain. It is known as the guilty knowledge test (GKT) or concealed information test (CIT) (Iacono 2007, 2008; Iacono and Lykken 1997; Iacono and Patrick 2006; Verschuere et al. 2011). Rather than asking direct, ultimate questions about participation or non-participation in the crime, the subject is asked about a series of alternative details of the crime, only one of which is correct. For example, the interrogator may ask, Regarding the murder weapon, do you know it was a pistol? Do you know it was a shotgun? Do you know it was a knife? All subjects are expected to answer No in response to each question. Asubject who does not knowthe details about the crime will be telling the truth in every case, and will not know which item is crime-relevant. A subject who partici- pated in the crime will recognize the guilty knowledge or concealed information contained in the correct, relevant item, and is expected to lie in response to the relevant question. The theory is that the resulting psychophysiologi- cal responses will reveal this lie and the subjects possession of the associated concealed information. As such, the con- ventional concealed information test is an adjunct to inter- rogation and testimony. It is a method not of directly detecting evidence of a crime, but of determining the veracity of a subject who is testifying regarding the evidence. Methods for detection of deception or credibility assessment have met with limited success. They have been used primarily to guide investigations rather than to denitively establish the relevant facts. They have not generally been admissible in court. Overview of brain ngerprinting science and technology In summary, the state of the art in investigations, forensic science, and criminal justice has been a combination of two modes of establishing a connection between the crime scene and the perpetrator: 1. Objective data collection and forensic science: estab- lishing objective, scientic connections between a few specic features of the crime scene and a few specic features of the perpetrator. 2. Interrogation and testimony: obtaining a subjective account of the contents of memory, and attempting to determine whether or not the subject is lying. What is generally the most comprehensive repository of information about the crime, the record stored in the brains of participants, has not been available to scientic scrutiny or objective investigation. There has been no objective, scientic way to detect the record stored in the brain and thereby to connect the perpetrator with the crime scene. Brain ngerprinting addresses this fundamental lack. Brain ngerprinting was developed to provide an objective, scientic way to connect the established features of the crime scene with the record of the crime stored in the brain of the perpetrator (Farwell 1992a, 1994, 1995a, b, 2010, Farwell and Donchin 1991). The major benet of brain ngerprinting is that it brings the record stored in the brains of participants within the realm of scientic scrutiny and objective investigation (Farwell and Smith 2001; Iacono 2007, 2008; Iacono and Lykken 1997; Iacono and Patrick 2006). Brain ngerprinting provides an objective method to detect features of the crime that are stored in the brain of the suspect. This is accomplished by measuring the sub- jects brain response to stimuli in the form of words or pictures presented briey on a computer screen. During a brain ngerprinting test, electroencephalograph (EEG) signals are recorded non-invasively from the scalp. When a subject recognizes and takes note of something signicant in the present context, the brain emits an Aha! response. This involves the ring of neurons in a specic, identiable pattern known as a P300-MERMER that can be detected by computer analysis of the EEG signals. When a subject recognizes a specic feature of the crime scene, such as the murder weapon, the brain ngerprinting system detects the Aha! response and its corresponding EEG pattern (Farwell 1992a, 1995a; Farwell and Donchin 1991; Farwell and Smith 2001). This reveals that the subject knows the Cogn Neurodyn (2012) 6:115154 117 1 3 relevant information. If the subject does not possess the relevant knowledge, the tell-tale brain response is absent. In a brain ngerprinting test, words or pictures relevant to a crime, terrorist act, terrorist training, or other investigated situation are presented on a computer screen, in a series with other, irrelevant words or pictures. (For brevity, the inves- tigated situation will generally be referred to herein as the crime, although other situations can of course also be investigated.) A subjects brainwave responses to these stimuli are measured non-invasively using a headband equipped with EEG sensors. A computer program then analyzes the data to determine if the crime-relevant infor- mation is stored in the brain. The specic, measurable brain response known as the P300-MERMER is emitted by the brain of a subject who has the details of a crime stored in his brain, but not by a subject lacking this record in his brain. The P300 response has been extensively researched and widely published in scientic journals for more than 30 years. It has gained broad acceptance in the scientic eld of psychophysiology (Harrington v. State 2001; Johnson 1988). The discovery of the P300-MERMER has allowed the results of brain ngerprinting testing to be more accurate, and has produced a higher statistical con- dence in the outcomes (Farwell 1994, 2008, 2010; Farwell and Richardson 2006a, b; Farwell et al. 2011, in press; Farwell and Smith 2001). Brainwave (EEG) measurements and event-related brain potentials (ERPs) The neurons in the brain re electrically, forming a vast network of electrical potential conduits. Electroencepha- lography (EEG) involves the measurement of these pat- terns of electrical voltage changes that originate in the brain. These measurements are made non-invasively from the scalp. When the brain conducts certain tasks, specic patterns of EEG (or brainwave) activity are produced. An example of such a specic task is noticing, recognizing, and processing the information contained in a signicant stimulus such as a murder weapon presented on a screen in a brain ngerprinting test. These specic patterns of brainwave activity are known as event-related brain potentials, or ERPs. Brain ngerprinting technique uses event-related brain potentials to determine what informa- tion is stored in a persons brain. This is based on how the brain processes specic information such as the features of a crime that are presented on a computer screen. At the same time the brain is engaging in the informa- tion processing of interest in a scientic experiment, the brain is also engaging in many other activities. The result is that the brainwaves measured at any time are a mixture of the relevant (event-related) activity and other brainwave activity. In order to isolate the activity of interest, the standard procedure in event-related potential research is to present a stimulus many times and average the responses (Donchin et al. 1978, 1986; Farwell and Donchin 1988a; Miller et al. 1987; Picton 1988). All of the brainwave activity that is not specically related to processing this specic stimulus averages out to zero, since its timing is unrelated to the event of the appearance of the stimulus on the screen and its processing by the subject. High-fre- quency noise is also removed using analog and digital l- ters (Farwell et al. 1993). What is left in the average response is the event-related potential: the brainwave pat- tern that is specically related to the event of interest. Each stimulus presentation and associated response is referred to as a trial. The larger the number of trials in each average brainwave response, the more extraneous brainwave activity is eliminated by the averaging proce- dure (Farwell and Donchin 1988a). In order to obtain valid and reliable results, a minimum number of trials is required. Experimenters usually run several tests, each containing about 100 trials. Each separate test is referred to as a block. Successive blocks may use the same stimuli or different stimuli relevant to the same event. The P300 is a very well known event-related potential that is utilized in brain ngerprinting testing. Several thousand studies have been published in the scientic lit- erature on the P300 brain response. This response takes place when the brain recognizes and processes a stimulus that is signicant in a particular context (Fabiani et al. 1987; Farwell and Donchin 1988a, 1991; Miller et al. 1987). For example, the murder weapon is signicant in the context of a brain ngerprinting test about a specic murder. The rst event-related potentials discovered were related to sensory processing of stimuli. These responses are referred to exogenous event-related potentials or evoked potentials. They are driven purely by sensory processing. They have nothing to do with the meaning of the stimulus or with any cognitive activity the subject may undertake. Initially, scientists recorded and analyzed only a very short epoch after the stimulus was presented, under the assumption that anything later than a few milliseconds did not have to do with the processing of the stimulus. In early evoked potential research scientists used evoked potentials occurring in the rst 100 ms after a stimulus. These were used primarily to study sensory processing. In the 1960s, scientists began to look at late potentials in excess of 100 ms after the stimulus. They discovered that there were patterns of neural activity reliably mea- surable from the scalp that manifested cognitive processing rather than sensory processing. This cognitive processing was independent of which sense delivered the information to be processed. In their seminal research on the P300, Sutton et al. (1965) showed that the response depended not on the physical characteristics of the stimulus but rather on 118 Cogn Neurodyn (2012) 6:115154 1 3 how that stimulus was processed cognitively. The same stimulus would result in different brain responses depending on its signicance in the context of the experi- ment and on the mental information-processing task per- formed by the subject. Sutton et al. used auditory stimuli to elicit these responses. Subsequent research has often uti- lized visually presented stimuli. Brain responses that manifest cognitive activity are referred to as endogenous event-related potentials (Donchin et al. 1978, 1986; Miller et al. 1987; Gaillard and Ritter 1983; Picton 1988). Before they could see these endogenous potentials, scientists had to begin looking in the time range beyond the rst 100 ms after the stimulus. The discovery of endogenous event-related potentials opened the door to a host of potential applications not only in discovering how the brain works but also in applying these discoveries to practical situations in the real world. The P300 is an electrically positive potential that occurs at 300800 ms after the stimulus (Sutton et al. 1965; Donchin et al. 1978; Long et al. 2011). The name refers to the fact that the response is electrically positive (P) and has a latency of at least 300 ms (300). The P300 occurs when a subject recognizes a stimulus as signicant in the context in which it is presented. It may be called an Aha response. In the early P300 research, the responses were elicited by very simple stimuli such as clicks or tones. These were made signicant in context by the experimental instruc- tions. When the stimulus and the task are simple, the P300 peak occurs at about 300 ms after the stimulus. As research progressed in the 1970s through the early 1990s, scientists began to use more complex stimuli such as a word ashed on a computer screen (Fabiani et al. 1987; Johnson 1988; Farwell 1992a). This had the advantage of providing the subject with more meaningful information. This enhanced the ability of event-related potential research to reveal cognitive brain processes. When the stimulus becomes more rich and complex, it takes longer for the subject to discern what the stimulus is and evaluate its sig- nicance. Thus the response is delayed. As scientists used more meaningful, rich, and complex stimuli, the P300 latency (elapsed time from the stimulus onset to the brain response) increased. By the 1990s it was not uncommon to measure a P300 that peaked 600 or 700 ms after the stimulus. The name P300 still remained in use, however, because it was the same response, just delayed by the increased com- plexity of the cognitive processing taking place. The discovery of the P300-MERMER In the initial brain ngerprinting research, Farwell and Donchin used the P300 event-related brain potential (Far- well and Donchin 1986, 1988b, 1991; Farwell 1992a). Later Farwell discovered that the P300 can be considered to be part of a larger response he called a memory and encoding related multifaceted electroencephalographic response or P300-MERMER. The discovery of the P300-MERMER was one more step in the ongoing progression from very short latency evoked potentials to longer and longer latency event-rela- ted potentials as the stimuli and the processing demanded by the experimental task become more rich and complex. In the 1990s when Farwell and FBI scientist Drew Rich- ardson were conducting the brain ngerprinting research on FBI agents, P300 latencies of 600 to 700 ms were typically found in experiments where the stimuli were information rich and the cognitive processing required was substantial (Farwell and Richardson 2006a, b; Farwell et al. 2011, in press). At that time, in such research a new stimulus was typically presented every 1,0001,500 ms (11.5 s). In the rst brain ngerprinting study, for example, Farwell and Donchin (1991) presented a stimulus every 1,500 ms. In dealing with real-life situations, Farwell and Rich- ardson (2006b), Farwell et al. (2011, in press) found it necessary to use longer and more complex stimuli to accurately communicate the necessary information to the subject. In order to present realistic stimuli that accurately represented knowledge unique to FBI agents, they found it necessary to use stimuli consisting of several words, sometimes several words of several syllables each. It took the subjects longer to read the words and evaluate their signicance than in previous experiments with simpler stimuli. To give the subjects time to process the stimuli and respond appropriately, Farwell and Richardson lengthened the interval between stimuli from 1,500 to 3,000 ms. They recorded a longer segment of brainwave data in each trial. Recall that in the 1960s when scientists looked farther out in time after the stimulus, they found previously unseen responses such as the P300 (Sutton et al. 1965). The same thing happened to Farwell and Richardson. They were looking for the P300 response, and indeed the brain responses contained a clear P300 peak at about 500800 ms. Surprisingly, however, this positive peak was followed by a negative peak with a latency as long as 1,200 ms. This unexpected late negative potential consis- tently followed the positive P300 peak. It was reliably elicited by the same Aha response that elicited the P300. This more complex P300-MERMER response included both the P300 and a late negative peak (the Late Negative Potential or LNP). Farwell (Farwell 1994, 2010; Farwell and Smith 2001) called this a memory and encoding related multifaceted electroencephalographic response (MER- MER), or P300-MERMER. The P300 is maximal in the parietal area. The late negative potential (LNP) that con- stitutes the latter part of the P300-MERMER is parietally maximal yet also frontally prominent (Farwell 1994, 1995b, 2010). Cogn Neurodyn (2012) 6:115154 119 1 3 Experimentation (including recording without analog lters), scalp distribution (the relative amplitude at different scalp sites), and morphology (the latency and shape of the waveforms) proved that the LNP was not an artifact of the signal-detection or noise-reduction procedures or equip- ment, such as digital and analog lters (Farwell 1994, 1995b; Farwell et al. in press). The recording equipment is identical for all scalp sites and all subjects. If the LNP were an artifact of the equipment, the identical equipment would produce the same effects in different instances. On the contrary, Farwell et al. found that the relative latency and amplitude of the P300 and the LNP are very different for different subjects and for different scalp sites in the same subject. Like any new discovery, the P300-MERMER raises questions both of nomenclature and substance. The clas- sical P300 is also known by various other names, including the P3, N2-P3 complex, P3a and P3b, late positive com- plex, and LPC. There has been considerable discussion as to whether the P300 is a unitary response or in fact a constellation of several responses (Johnson 1989; Spencer et al. 2001). There has also been discussion over whether the various names refer to the same or slightly different phenomena (Spencer et al. 2001). No doubt there will be considerable discussion as to whether the MERMER or P300-MERMER is a unitary phenomenon inclusive of the P300 and the late negative potential (LNP), or whether the late negative potential is a separate component from the component or components that make up the P300. The answers to these questions are empirical, to be settled by further research. Differences in nomenclature also exist. Over a thousand published studies have associated the name P300 with a positive peak. Farwell rst reported the P300-MERMER, including the positive peak of the P300 and the late neg- ative potential (LNP), in 1994 (Farwell 1994). By 2001, not only Farwell and Smith (2001) but also almost all other researchers brainwave-based in detection of concealed information were using the full P300-MERMER in data analysis. Such research now almost always includes both the positive peak of the P300 and the late negative potential (LNP) of the P300-MERMER. For example, in the algo- rithms that use amplitude computations, the amplitude of the response is typically computed as the sum of the amplitudes of the positive and negative peaks (that is, the difference between the highest and lowest points). Some authors (Soskins et al. 2001; Rosenfeld et al. 2004) how- ever, have used the name P300 to refer to the entire P300-MERMER response, including not only the tradi- tional P300 peak but also the late negative potential (LNP) of the P300-MERMER. In view of the fact that over a thousand previous publications have used the term P300 to refer only to the positive peak, such usage is confusing and ambiguous. Authors may avoid confusion by using the term P300 to refer to the positive peak alone and P300- MERMER to refer to the complex consisting of the positive peak followed by the negative peak (LNP). Farwell and colleagues (Farwell 1994, 1995b, 2010; Farwell and Smith 2001) reported additional facets in the P300-MERMER brainwave response that occur simulta- neously with the positive and negative peaks. These are detected through different signal-processing algorithms than the signal-averaging algorithm typically used to detect event-related potentials (Farwell 1994; Rapp et al. 1993). These include an event-related, short-term shift in the frequency of the EEG signal. The nature of these additional facets and their relationship to the more readily visible positive and negative peaks is also an empirical question to be resolved by further research (see Guntekin and Basar 2010). The discovery of the P300-MERMER has allowed the brain ngerprinting results to be more accurate than the results obtained with the P300 alone. The P300-MERMER consists of a positive peak followed by a negative peak. The P300 includes only the positive peak. Thus, the full P300-MERMER contains more information and more distinctive features, and can be more reliably and accu- rately detected by a mathematical signal-detection algo- rithm than the P300 alone. In all brain ngerprinting research using either the P300- MERMER or the P300 alone, there have been no false negatives and no false positives. 100% of determinations made have been correct. (See, however, the discussion below regarding the term 100% accurate.) When Farwell and colleagues have included the full P300-MERMER in the data analysis algorithm, there have also been no inde- terminates. In brain ngerprinting research using the P300 alone, results have been indeterminate in 3% of cases overall, consisting of 12.5% in one experiment. As dis- cussed below, an indeterminate response is not an incorrect response, but rather the determination that insufcient data are available to make a determination in either direction with high statistical condence. Neurodynamics, physiological mechanism, and signal characteristics of the P300-MERMER Many important phenomena in science are not easily described in terms of characteristics and categories that existed before their discovery. For example, Heisenberg (Heisenberg 1958), Neils Bohr, and others showed that elementary particles appear as particles when measured with apparatus designed to detect particles, and as waves when measured with apparatus designed to detect waves (see Farwell and Farwell 1995; Farwell 1999 for a detailed discussion). Like elementary particles, the characteristics of the P300-MERMER that are detected depend on how it 120 Cogn Neurodyn (2012) 6:115154 1 3 is measured. Farwell and Smith (2001; see also Farwell 1994) described the process of discovery of the charac- teristics of the P300-MERMER as reminiscent of the story of the blind men and the elephant. Touching the trunk, one may conclude the elephant is like a snake; touching the tail, one may conclude it is like a rope. Touching the leg, the tusk, or the abdomen, etc. lead to different impressions. Originally, the P300 was discovered as an event-related potential, detected through signal averaging of scalp- recorded voltage patterns. As discussed above, event-rela- ted potentials are voltage changes that take place in the time domain that are measured from the scalp. They manifest specic information-processing activities in the brain. Much of the subsequent research has been directed towards discovering what structures and processes in the brain cre- ate the event-related potentials recorded at the scalp. The phenomenon that manifests at the scalp as the P300, however, is much more complex and multifaceted than a simple average time-domain pattern at the scalp can reveal or characterize. Farwell and Smith (2001; see also Farwell 1994) described the phenomenon manifested as the P300- MERMER and the P300 as a multifaceted electroen- cephalographic response (MER). The different facets of the phenomenon, like those of elephants and the electrons, are discovered through different methods of data acquisi- tion and analysis. The P300 phenomenon has been explored through sev- eral different methodologies. Measurements of EEG elec- trical voltage patterns made non-invasively at the scalp have been analyzed in terms of time-domain analysis through signal averaging, frequency-domain analysis, relationships between voltage patterns at different areas of the scalp (e.g., event-related coherence), and various characteristics of the signal (e.g., dynamical systems/chaos/ fractal analysissee Rapp et al. 1993; Farwell et al. 1993). Mathematical analysis of multichannel scalp recordings of EEG has been used to explore the source of these patterns in the brain (Li et al. 2011). Intracranial recordings have been analyzed in both the time and frequency domains, and in some cases have been analyzed with respect to relationships between sites (e.g., coherence). Animal models have allowed for intra-cranial recordings in all areas of the brain, whereas such record- ings in humans are restricted to areas of interest during surgery. Functional imaging (e.g., functional magnetic resonance imaging, fMRI) and magnetoencephalography have served to further explore which areas of the brain are activated during the process. Through a combination of these techniques, a reason- ably clear and comprehensive picture of the phenomenon has emerged. Like elephants and elementary particles, however, the P300 pattern is not without paradox and even apparent contradiction. To begin with, signal averaging makes the event-related potential visible as a pattern in the time domain, but the same process eliminates any fre- quency-domain signals that are a part of the phenomenon. A full characterization of the phenomenon, even at the scalp, must include both time-domain and frequency- domain characteristics (Basar-Eroglu et al. 1992). Intra-cranial recordings have clearly revealed patterns in several areas of the brain, both in the time and frequency domains, that respond to the same experimental manipu- lations and appear at the same time as the P300. Some of these undoubtedly contribute in large measure to the scalp- recorded potentials. Clear patterns of activation in some specic areas, however, appear to be an integral part of the phenomenon, but due to the physical positioning of the structures in which they occur, they apparently do not contribute signicantly to the scalp-recorded potentials. Lesion studies combined with scalp recording have served to isolate the structures that contribute to the scalp-recor- ded potentials, and in some cases to show that deep structures that are activated in the phenomenon do not contribute to its manifestation at the surface of the scalp. To further complicate the picture, the P300 is not a unitary phenomenon. As discussed above, it is a multifaceted elec- troencephalographic response. Different facets correspond to different sub-tasks and modality-specic tasks in the information-processing of which the scalp recorded poten- tial is a manifestation. Some of the same structures, and some different structures, are involved in these various different sub-tasks. The P300 includes information processing related to both the novelty and the signicance of the stimulus, which involve somewhat different neural processes and structures. Information processing, and the structures involved, also varies somewhat depending on the modality (usually auditory or visual) of the stimulus. Combining the ndings from the various methods of discovery reveals the following pattern (Baudena et al. 1995; Huster et al. 2010; Halgren et al.1995; Johnson 1993; Kiss et al. 1989; Sabeti et al. 2011; Smith et al. 1990; Wang et al. 2005; for a review, see Linden 2005.) Signicance-related (or target-related) contributions to the response appear to be generated largely in the parietal cortex and the cingulate. Novelty-related contributions to the response appear to be generated mainly in the inferior parietal and prefrontal regions. Contributions specic to the visual modality appear to be generated in the inferior temporal and superior parietal cortex. Contributions specic to the auditory modality appear to be generated in the superior temporal cortex. A number of subcortical structures appear to be inte- grally involved in the phenomenon, but do not contribute signicantly to the scalp-recorded potentials. Both fre- quency-domain and time-domain signals, and polarity reversals indicating a local source, have been recorded in temporal structures, most signicantly in the hippocampus Cogn Neurodyn (2012) 6:115154 121 1 3 and the amygdala (Halgren et al. 1986; Sochurkova et al. 2006; Stapleton and Halgren 1987). Initially, temporal structures were thought to be major contributors to the scalp-recorded potentials. Lesion studies, along with the limited degree to which such electrical potentials from such deep sources are propagated to the scalp, however, have cast doubt on this conclusion. Lesions of the medial tem- poral lobe have little effect on the scalp-recorded potentials (Johnson 1988). The early research on the phenomenon of interest gen- erally focused on the positive peak of the P300. In experi- ments involving meaningful stimuli and relatively long inter-stimulus intervals such as those applied in detection of concealed information, the P300-MERMER has a triphasic shape. The positive P300 peak is preceded by a negative peak, the N200, and followed by another negative peak, the LNP. This triphasic negative-positivenegative pattern was observed at the scalp by Farwell and others (Farwell 1994; Farwell and Smith 2001) as a dening characteristic of the P300-MERMER. The same negative-positivenegative pattern was observed in intracranial recordings in various structures (Halgren et al. 1998), including the inferior pari- etal lobe/supramarginal gyrus, superior temporal sulcus (Halgren et al. 1995), the amygdala and hippocampus (Halgren et al. 1986; Stapleton and Halgren 1987), dorso- lateral and orbital frontal cortices, and the anterior cingulate (Baudena et al. 1995). By nowvirtually all of the researchers involved in detection of concealed information with brain- waves include in their computational algorithms both the positive peak of the P300 and the late negative potential (LNP) that constitutes the other major facet of the P300- MERMER. (As discussed above, however, differences in nomenclature still exist; some have called the entire multi- faceted P300-MERMER phenomenon P3 or P300.) One hypothesis that may explain the simultaneous involvement of numerous cortical and sub-cortical struc- tures in the P300-MERMER is that this phenomenon reects phasic activity of the neuromodulatory locus coe- ruleus-norepinephrine (LCNE) system (Murphy et al. 2011; Nieuwenhuis et al. 2005; Pineda et al. 1989). The hypothesis is that the properties of the P300 reect the function of the LCNE system to potentiate the infor- mation-processing activities undertaken in response to signicant events. According to this theory, the LCNE system affects or is affected by each structure in the entire constellation of structures that are activated in the P300, and the activation of large pyramidal neurons in the cortex by the LCNE system is the primary source of scalp-recorded potentials of the P300. Evidence converges from several types of studies. The LC responds to experimental manip- ulations similar to those that produce the P300. Lesion studies show that normal functioning of the LCNE system is necessary for the generation of the P300. The projections of the LC are also consistent with this hypothesis. The LC is the sole source of NE input to the neocortex and hippo- campus, and also projects to the amygdala and the thalamus. The LC receives afferent input from structures known to be involved in decision making, in signaling novelty, and in representing task goals and signicance of stimuli, includ- ing the prefrontal cortex, the anterior cingulate cortex, and the orbitofrontal cortex. This is consistent with the activa- tion of the LC in response to stimuli that are evaluated as novel or signicant in the current context, in accord with the antecedent conditions for the P300. The timing of the LC NE responses to events is also consistent with the timing of the P300 response to similar events. Brain ngerprinting scientic protocol Experimental design Brain ngerprinting tests are conducted according to the following scientic protocols. In a brain ngerprinting test, stimuli are presented to the subject in the form of words, phrases, or pictures on a computer screen. (Auditory stimuli may also be presented.) Brain responses are mea- sured non-invasively from the scalp, digitized, and ana- lyzed to determine the presence or absence of information stored in the brain. Figure 1 outlines the stages of data acquisition and analysis in brain ngerprinting. Three types of stimuli are presented: probes, targets, and irrelevants. (Farwell and Donchin 1986, 1991; Farwell and Smith 2001). Probes contain information that is relevant to the crime or other investigated situation. Probes have three necessary attributes (Farwell 1994, 1995a, b; Farwell and Smith 2001; Iacono 2008): 1. Probes contain features of the crime that in the judgment of the criminal investigator the perpetrators would have experienced in committing the crime; 2. Probes contain information that the subject has no way of knowing if he did not participate in the crime; and 3. Probes contain information that the subject claims not to know or to recognize as signicant for any reason. For example, if a subject claims not to have been at the murder scene and not to know what the murder weapon was, a probe stimulus could be the murder weapon, such as a knife. Brain ngerprinting experimental protocols ensure that probes do not contain information that the subject knows from the news media, interrogations, etc. The scientic question addressed by a brain nger- printing test is whether or not the subject is knowledgeable regarding the crime or investigated situation. Specically, the critical variable is his recognition of the information contained in the probes as signicant in the context of the 122 Cogn Neurodyn (2012) 6:115154 1 3 crime (or lack thereof). If, and only if, this is present, it is predicted that the probes will elicit a P300-MERMER. The amplitude, morphology and latency will be characteristic of the individual subjects response to such stimuli when the subject knows the relevant information. For a subject who is knowledgeable or information present, the probes contain information describing known features of the crime. For a subject who is information absent, the probes contain information describing plau- sible features of the crime that are not known to be correct. To objectively classify the probe responses into one of these two categories, it is necessary to isolate the critical variable. To accomplish this, two standards are required: a standard for the response of this subject to stimuli con- taining known features of the crime, and a standard for the response of this subject to stimuli containing plausible but unknown (or incorrect) features of the crime. Target stimuli are details about the crime that the experimenter is certain the subject knows, whether or not he committed the crime. They may have been previously revealed through news accounts, interrogation, etc. In any case, the experimenter tells the subject the target stimuli and explains their signicance in terms of the crime. Because they are signicant in the context of the crime for all subjects, targets elicit an Aha response in all subjects. Targets elicit a P300-MERMER whether the subject knows the other salient features of the crime contained in the probes or not. For example, a target stimulus might be the name of the victim, which is revealed to the subject in the course of test instructions (and may be already known from news reports, etc.). Irrelevant stimuli contain information that is not rele- vant to the crime and not relevant to the subject. They consist of incorrect but plausible crime features. Irrelevant stimuli are designed to be indistinguishable from correct crime-relevant features (probes) to someone who does not know the features of the crime. Since the irrelevant stimuli are not signicant in context, they do not elicit a P300- MERMER. If a probe stimulus is the murder weapon, a knife, then irrelevant stimuli could be other plausible (but incorrect) murder weapons such as a pistol, a rie, and a baseball bat. Thus, the targets and irrelevants both provide standard responses. Thetargets provide a standardfor thesubjects brain response to relevant, signicant information about the crime in question. The irrelevants provide a standard for the subjects brain responseor rather lack of a responseto irrelevant information that is plausible as being crime-relevant. It is vital in any science to isolate the critical variable. This three-stimulus design accomplishes this purpose. Targets and irrelevants differ only in whether or not they contain the critical feature being tested, that is, whether they contain known crime-relevant information. For an infor- mation-present subject, probe stimuli are virtually identical to target stimuli: both contain known features of the crime. The only difference is which button is pressed when the stimuli appear. For an information-absent subject, probes are indistinguishable from irrelevants. The probe responses are classied as being more similar to targets or irrelevants. Everything except the critical variable, namely the subjects recognition of the probes as crime-relevant, is controlled. The subject is given a list of the targets and instructed to press a specic button when a target appears and another button when any other stimulus appears. Since the subject does not know which of the three types of stimulus will occur on each trial, he must read and evaluate each stim- ulus, and demonstrate behaviorally in each trial that he has done so by pressing the appropriate button. Table 1 outlines the types of stimuli and predicted brain responses in brain ngerprinting. For a subject who knows the relevant details about the crime, the probes, like the targets, are signicant and rele- vant. Thus, the probes produce an Aha response when presented in the context of the crime. This manifests as a Montage: Fz, Cz, Pz, EOG EEG Amplification: 50,000 Analog Filtering: .1 to 30 Hz Signal Averaging: Probes, Targets, Irrelevants Bootstrapping Correlations: Probe Target vs. Probe Irrelevant Determination: Information Present / Information Absent; Statistical Confidence Digitizing: 100 Hz Digital Filtering: 0 to 6 Hz Fig. 1 Stages of data acquisition and analysis Cogn Neurodyn (2012) 6:115154 123 1 3 P300-MERMER in the brainwaves that will be virtually iden- tical to the target response. For a subject who lacks the knowledge contained in the probes, the probes are indistin- guishable fromthe irrelevants. Probes donot produce anAha response or the corresponding P300-MERMER: the probe response will be virtually identical to the irrelevant response. The brain ngerprinting computerized data analysis algorithm computes a mathematical determination as to whether the probe response is more similar to that of the targets or that of the irrelevants. The former yields a determination of information present; the latter, infor- mation absent. The information that is either present or absent in the brain of the subject is the information con- tained in the probes. The brain ngerprinting system also computes a statistical condence for each individual determination, e.g., information present, 99.9% con- dence. If there is insufcient data to reach either an information present or an information absent deter- mination with a high statistical condence, the algorithm returns the outcome of indeterminate. Since the inclusion of the P300-MERMER in the brainwave data analysis algorithm, brain ngerprinting testing has made a denite determination in every case. All determinations have been correct. There have been no false negatives, no false positives, and no indeterminates. Error rate has been 0% in all studies and eld applications. Accuracy has been 100%. (As discussed below, these are usually represented as less than 1% and over 99% respectively.) When brain ngerprinting data analysis has been conducted using the P300 alone, there have been no false positives or false negatives, and about 3% of the results have been indeterminate. All of these were in a single experiment (Farwell and Donchin 1991), wherein indeterminates comprised 12.5% of the results. Note that an indeterminate result is not incorrect. It is not an error. It is neither a false negative nor a false positive. Rather, it is a determination that the data analysis algorithm has insufcient data to make a determination in either direction with a high statistical condence. Before conducting a brain ngerprinting test, the subject is interviewed to nd out what he knows about the crime from any non-incriminating source such as news reports or prior interrogations. Any such information is excluded from the probes. (Such information may be contained in targets, since the targets are known to contain information that the subject knows.) The experimenter describes to the subject the signicance of each probe in the context of the crime. The experimenter does not tell the subject which stimulus is the probe and which are similar, irrelevant stimuli. Only information that the subject denies knowing is used for probe stimuli. Also, the experimenter shows the subject a list of all the stimuli including the probes, without of course identifying which ones are probes. As an extra precaution, the subject is asked if any of the stimuli are signicant to him for any reason at all. Any stimuli that are signicant to the subject for reasons unrelated to the crime are eliminated. If for example, a potential probe is the name of a known accomplice, and coincidentally it is also the name of the suspects brother-in-law, it is not used. Things are signicant to a person in context. The context of the probe stimuli in relation to the crime is established in the interview prior to the brain ngerprinting test. Imme- diately before the test, the experimenter describes the sig- nicance of each probe in the context of the crime. Before the test, the subject has explicitly stated that he does not know which stimulus is the probe containing the correct information. Under these circumstances, a large P300-MERMER in response to the probes provides evidence that the subject recognizes the probes as signicant in the context of the crime. If the experimenter has followed the proper Table 1 Types of stimuli and predicted brain responses Stimulus type Relative frequency Description Instructions Subjects stimulus evaluation Predicted brain response Target 1/6 Relevant to investigated situation; known to all subjects Press left button Relevant, rare for all subjects P300-MERMER Irrelevant 2/3 Irrelevant Press right button Irrelevant, frequent No P300-MERMER Probe 1/6 Relevant to investigated situation; known only to investigators and subjects who have the specic knowledge tested Press right button (treat like irrelevants) Informationabsent subjects: Irrelevant, frequent (Indistinguishable from irrelevants) No P300-MERMER Informationpresent subjects: Relevant, rare P300-MERMER 124 Cogn Neurodyn (2012) 6:115154 1 3 protocols, the subject has eliminated all plausible non- incriminating explanations for this knowledge by his own account prior to the test. Therefore, an information-present response can provide evidence that a judge and jury may reasonably evaluate as being probative regarding the subjects involvement in the crime. Note, however, that the question of whether or not the subject participated in the crime is in the domain of the judge and jury. The brain ngerprinting sci- entist provides evidence and testimony only regarding whe- ther the crime-relevant information contained in the specic probes tested is stored in the brain of the subject. Electroencephalograph (EEG) data are collected from midline frontal, central, and parietal scalp sites (Fz, Cz, and Pz respectively). Electrooculograph (EOG) data are col- lected from the forehead to monitor artifacts generated by eye movements. Data are amplied, digitized, ltered, and analyzed. Stimuli are presented for a duration of 300 ms at an inter-stimulus interval of 3,000 ms. A xation point is presented for 1,000 ms prior to each stimulus presentation. Figure 2 outlines the timing parameters for stimulus pre- sentation, data acquisition, and data analysis. The analysis epoch is 300900 ms post-stimulus for the P300 analysis, and 3001,800 ms post-stimulus for the P300-MERMER analysis. Data analysis The purpose of data analysis in brain ngerprinting studies is to determine whether the brain responses to the probe stimuli are more similar to the responses to the target stimuli or to the responses to the irrelevant stimuli, and to provide a statistical condence for this determination. The determination and statistical condence must be computed for each individual subject. If the probe responses are mathematically more similar to the same subjects responses to target stimuli containing known features of the crime, the subject is determined to be information present. If the probe responses are mathe- matically more similar to the same subjects responses to irrelevant stimuli containing plausible but unknown (or incorrect) features of the crime, the subject is determined to be information absent. To be valid, the statistical condence for an individual determination of information present or information absent must take into account the level of variability in the individual brain responses that are aggregated in the average response. Farwell and Donchin (1988b, 1991) and their colleagues Wasserman and Bockenholt (1989) applied bootstrapping to compute a statistical condence for each individual determination that takes this variability into account. Bootstrapping is described in detail in these publications. It has now become a standard procedure in event-related brain potential research. Bootstrapping is a method to compute probability and statistical condence regardless of the shape of the distribution of the data. It also provides a means to re-introduce the variability across single trials present in the original data, while preserving the feature of a smooth average that is necessary for comparing the waveforms of the three types. The algorithm is as follows. Conduct the following procedure twice, once using the time epoch characteristic of the P300-MERMER (typically 3001,800 ms after the stimulus) and once using the time epoch characteristic of the P300 alone (typically 300900 ms post-stimulus). 1. Sample randomly with replacement T target trials, P probe trials, and I irrelevant trials, where T, P, and I are equal to the total number of trials in the data set of the respective types. A trial consists of one stimulus presentation and the associated brain response data. 2. Average the trials by trial type, yielding three average waveforms: probe, target, and irrelevant. Compare the average waveforms to determine if the probe average is more similar to the target average or to the irrelevant average. 3. Repeat the above procedure 1,000 times. Each iteration yields a new set of 3 averages containing probe, target, and irrelevant trials respectively. Keep a tally of the number of times the probe average is more like the irrelevant average than like the target average. 4. For each iteration, compare the probe, target, and irrelevant waveforms according to the following algorithm: (a) subtract the grand mean of all trials, or grand average waveform, from each of the 3 averages, yielding 3 adjusted averages; (b) compute the corre- lation between the adjusted probe average and the adjusted irrelevant average; (c) compute the correla- tion between the adjusted probe average and the adjusted target average; (d) compare the probe-irrel- evant correlation with the probe-target correlation: if the probe-irrelevant correlation is greater, then incre- ment the information present tally by one; other- wise, increment the information absent tally by one. 5. Compute the percentage of times that the probe target correlation is higher than the probeirrelevant Fig. 2 Timing parameters. Timing of events for each individual trial in brain ngerprinting Cogn Neurodyn (2012) 6:115154 125 1 3 correlation. This the percentage of times that the probe waveformis more similar to the target waveformthan to the irrelevant waveform. This provides the probability or statistical condence for an information present result. 100% minus this gure the provides the proba- bility that the probe response is more similar to the irrelevant response, which provides the statistical con- dence that for an information absent result. 6. Compare the computed statistical condence to a decision criterion. If the statistical condence for an information present result is greater than 90%, classify subject as information present. If the statistical condence for an information absent response is greater than 70%, classify the subject as information absent. If neither criterion is met, no determination is made: the subject is not classied as either information present or information absent; this is an indeterminate outcome. In other words, if the bootstrapping procedure produces a high statistical condence that the probe response is more similar to the target response than to the irrelevant response, then the determination is information present. If the bootstrapping procedure produces a high statistical con- dence that the probe response is more similar to the irrelevant response, then the determination is information absent. If neither the statistical condence for information pres- ent nor the condence for information absent is high enoughtomeet establishedcriteria, the subject is not classied ineither category, and the result is indeterminate. Typically a condence of 90% is required for an information present determination. A lower criterion, typically 70%, is generally required for an information absent determination. The outcome of brain ngerprinting data analysis con- sists of two determinations, each of the form information present/absent, x% condence, e.g., information present, 99.9% condence. One determination is computed using the full P300-MERMER, and one using the P300 alone. This allows us to report one result with the method that applies the most well established science and is most cer- tain to meet the standard of general acceptance in the scientic community, and one with the method that applies the state of the art and generally produces the highest statistical condence. By computing bootstrapped correlation as described above, the brain ngerprinting data analysis algorithm takes into account the amplitude, latency, and morphology (shape and time course) of the brain response. This maxi- mizes the information extracted from the data and also controls for individual differences in brain responses from different subjects. Before applying the bootstrapping technique on corre- lations between waveforms, noise in the form of high-fre- quency activity is eliminated by the use of digital lters. Farwell et al. (1993) have shown that a specic type of lters known as optimal digital lters are highly effective for eliminating this high-frequency noise while preserving the brainwave pattern of interest in event-related brain potential research. These lters are optimal in the precise mathematical denition of the word. In evaluating the error rate/accuracy of any technique, it is important to establish ground truth as objectively and cer- tainly as possible. In any scientic test, ground truth is the true state of exactly what the test seeks to determine. In a DNA test, ground truth is not whether the suspect is a rapist or a murderer; ground truth is whether sample A (putatively from the subject) matches Sample B (putatively from the crime scene). In a ngerprint test, ground truth is not whether the suspect is a burglar, but whether the prints at the crime scene match the prints on the suspects ngers. The same principle applies to brain ngerprinting. A brain nger- printing test detects the presence or absence of specic information stored in the brain. Ground truth is the true state of that which is tested. Ground truth is whether or not the specic information embodied in the probe stimuli is in fact stored in the specic brain tested. Ground truth is not whether the suspect is a murderer, or a liar, or whether the experi- mental subject participated in a knowledge-imparting pro- cedure consisting of a mock crime, or whether the experimenter thinks the subject should knowthe information tested. In evaluating the accuracy of a brain ngerprinting study, it is important to establish ground truth as objectively and certainly as possible. When subjects are cooperative (or eventually become cooperative), this is accomplished by post-test interviews to determine whether the subject knew the information tested at the time of the test. In eld cases with any forensic science, establishing ground truth can be challenging. In brain ngerprinting eld cases with unco- operative subjects, ground truth can never be known abso- lutely. It is established with as much certainty as possible through indirect means such as testimony, converging evi- dence, judicial outcome, and confessions. Brain ngerprinting and other techniques Advantages and disadvantages of brain ngerprinting Brain ngerprinting has advantages and disadvantages with respect to other forensic science and investigative methods. Compared to previously available scientic methods for matching features of a crime scene with features of a suspect, the primary advantage of brain ngerprinting is that in most crimes very few such features can be found. In some crimes none are available. The record stored in the brain of the perpetrator is often a rich source of information that can be connected to the crime scene. Except in rare 126 Cogn Neurodyn (2012) 6:115154 1 3 cases where the crime has been recorded on video, the record stored in the brain is generally the most compre- hensive available record of the crime, even though it is not perfect. Brain ngerprinting also has advantages in comparison to witness testimony. It provides an objective, scientic way to detect the record of the crime stored in the brain directly. Witness testimony provides an indirect, subjective account of this record. Witnesses may lie. The brain never lies. If the information is stored in the brain, it can be objectively detected regardless of the honesty or dishonesty of the subject. Brain ngerprinting thus eliminates one of the two major disadvantages of witness testimony, that of deception on the part of the witness. The primary disadvantage of brain ngerprinting in comparison to other forensic science methods of connect- ing features of a perpetrator to features of a crime scene is the same as the primary disadvantage of witness testimony: human memory is imperfect and limited. Just as all pro- ceedings involving witness testimony must weigh the evi- dence obtained thereby in light of the limitations of human memory, all proceedings involving brain ngerprinting evidence must do the same. Brain ngerprinting and lie detection Brain ngerprinting and the guilty knowledge test or concealed information test Brain ngerprinting has some the same features, and all of the advantages, of the conventional guilty knowledge test or con- cealed information test described above. It can be considered a type of guilty knowledge test (Farwell 2007; Iacono 2007, 2008; Iacono and Lykken 1997; Iacono and Patrick 2006). Brainngerprintingtakes advantage of thefeatures of theguilty knowledge test that have made it well accepted in the relevant scientic community (Iacono 2008). Brain ngerprinting, however, is fundamentally different from the conventional guilty knowledge test in several important ways. These dif- ferences provide signicant advantages over the conventional guilty knowledge test (Farwell 1994, 1995a, 2007). Both brain ngerprinting and the conventional guilty knowledge test are concerned with the relevant features of the crime that are known to the perpetrator and not to an innocent suspect. Brain ngerprinting directly detects information stored in the brain based on information-pro- cessing brain activity. The conventional guilty knowledge test questions the subject, detects a stress-related response in an attempt to detect lies, and makes indirect inferences about what the subject knows on that basis. A conventional guilty knowledge test asks two types of questions, relevant and irrelevant. The data analysis attempts to determine whether the stress-related response to the relevant questions is larger than the response to the irrelevant questions. If so, the subject is found to be deceptive. The determination of a conventional guilty knowledge test is deceptive or non-deceptive. A brain ngerprinting test, as described above, presents three types of stimuli. Two of these are relevant to the crime. Targets contain crime-relevant information that is known to all subjects. Probes contain information that is known only to the perpetrator and investigators. The information-processing responses to the probes are classi- ed as being more similar to the irrelevant responses or to the target responses. The latter indicates that the probes, like the targets, contain information that the subject knows and recognizes as signicant in the context of the crime. Since brain ngerprinting measures an information- processing brain response rather than an emotional stress response, it does not depend on the emotional responses of the subject. It does not seek to assess the veracity of the subject. A subject neither lies nor tells the truth during a brain ngerprinting test. He simply observes the stimuli and pushes the buttons as instructed. The determination of a brain ngerprinting test is the same whether the subject tells the truth or lies about any subject at any time. The determination of a brain ngerprinting test is infor- mation present or information absent. An information present determination means that the subject possesses the specic knowledge tested. An information absent deter- mination means that the subject does not possess this infor- mation. This is entirely independent of whether the subject tells the truth or lies about this information or anything else. Conventional CQT polygraphy and fMRI Conventional polygraphy involves questioning the subject, measuring a physiological response, and thereby attempt- ing to determine if he is lying. A conventional control question polygraph test (CQT) measures peripheral responses, usually skin conductance (related to perspira- tion), cardiovascular activity/blood pressure, and breathing. Conventional polygraphy measures this peripheral physi- ological response in an attempt to detect changes consid- ered to accompany lying (Farwell 2013; Iacono 2007, 2008; Iacono and Lykken 1997; Iacono and Patrick 2006; National Research Council 2003; Vrij 2008). Scientists have recently introduced measurements of cerebral blood ow with functional magnetic resonance imaging (fMRI) in an attempt to detect lies. Like conven- tional polygraphy, fMRI does not directly measure lying. Since lying is not a unitary phenomenon, there is no unique lie response. The underlying theory of fMRI detection is similar to the theory of conventional polygraphy. Instead of measuring stress, however, fMRI measures brain processes Cogn Neurodyn (2012) 6:115154 127 1 3 putatively connected with conict or other processes considered to accompany lying. As discussed above, there are two fundamental ways to attempt to obtain information regarding a suspects par- ticipation in a crime: (1) collect objective data linking the subject to the crime, and (2) question the subject or a witness about the crime and attempt to discern if he or she is lying. Conventional CQT polygraphy and fMRI-based lie detection methods both fall into the latter category. Their purpose is lie detection. They serve as adjuncts to interrogation and testimony. Brain ngerprinting is fundamentally different from lie detection. Mistakenly classifying brain ngerprinting as a form of lie detection (e.g., Verschuere et al. 2009; Ro- senfeld 2005) arises from a fundamental misunderstanding of the science and technology (Farwell 2013; Farwell and Makeig 2005; Farwell and Smith 2001). Brain nger- printing detects information stored in the brain, not lies (Farwell 1992a, 1994, 1995b, 2013; Farwell and Donchin 1991; Iacono 2008). Like ngerprinting and DNA, brain ngerprinting is a method to collect objective data linking the subject to the crime. Origin of the term brain ngerprinting Brain ngerprinting is so named based on the following analogy (Farwell 1994). Fingerprinting establishes an objective, scientic connection between ngerprints at a crime scene and the ngers of a suspect. DNA nger- printing, as it is sometimes called, establishes an objec- tive, scientic connection between biological samples from the crime scene and biological samples from the suspect. Brain ngerprinting was so named because like nger- printing it detects a match between evidence from the crime scene and evidence on the person of the suspect. It establishes an objective, scientic connection between features of the crime scene and the record stored in the brain of a suspect. Principles of applying brain ngerprinting in the laboratory and the eld The purpose of brain ngerprinting is to determine whether or not specic relevant knowledge is stored in the brain of the subject (Farwell 1994; Farwell and Smith 2001; Iacono 2008). In eld cases, the relevant knowledge generally is information that an investigator thinks represent the details of a crime. Alternatively, it may be information that is known only to a particular group of people, such as FBI agents (Farwell and Richardson 2006b; Farwell et al. 2011, in press), skilled bomb makers (Farwell 2009; Farwell et al. 2011, in press), trainees of an Al-Qaeda training camp, or members of a terrorist cell. The primary example used herein will be the case where the relevant knowledge constitutes information that an investigator believes con- stitutes salient features of a crime that the perpetrator experienced in the course of committing the crime. The relevant knowledge is provided by the criminal investigator to the brain ngerprinting scientist. The goal of brain n- gerprinting is to determine whether or not the relevant knowledge is known to the subject. Note that brain ngerprinting does not evaluate whether or not the investigators account of the crime is accurate, or whether the putatively relevant knowledge actually cor- rectly represents the crime. Brain ngerprinting does not detect guilt or innocence. The determination of whether the subject is guilty is a legal determination that is made by a judge and/or jury, not by a scientist or a computer. Brain ngerprinting does not detect whether or not the subject committed the crime. It only detects whether or not the subject knows the relevant knowledge contained in the probes. The prosecution may argue that the best explanation for an information present determination is that the sub- ject learned the relevant knowledge while committing the crime. (In a properly executed brain ngerprinting test, plausible alternative hypotheses such as the subject being told the information after the crime have been eliminated before the test.) The defense may argue that an information absent determination introduces a reasonable doubt that the subject is guilty of committing the crime, and provides support for his claims of innocence. The defense may argue, for example, that a subject should or would knowthe relevant knowledge if he had committed the crime. Brain ngerprinting does not evaluate whether the sub- ject should, could, or would know the information, and under what circumstances. It only determines whether or not the subject actually does know the relevant knowledge. The interpretation of the results of a brain ngerprinting test in terms of guilt or innocence, participation or non- participation in a crime, goes beyond the science and is outside the realm of expert testimony by a brain nger- printing scientist. Brain ngerprinting is similar to other forensic sciences in this regard. A DNA expert testies that Sample A, which the investigators say came from the crime scene, matches Sample B, which the investigators say came from the subject. Similarly, an expert may testify that two nger- prints match. He does not testify, report, or attempt to scientically determine Therefore, the subject committed the murder. A brain ngerprinting scientist testies regarding only one specic fact: the subject does or does not know the specic relevant knowledge tested (Har- rington v. State 2001). The degree to which this fact is probative regarding the subjects participation in a crime is 128 Cogn Neurodyn (2012) 6:115154 1 3 outside the realm of science. That is a matter to be debated by the prosecution and defense and decided by a judge and/ or jury based on their non-scientic, common sense judg- ment and life experience. In a laboratory setting, the relevant knowledge is fabri- cated by the experimenter. One additional step is necessary before a test can be implemented to test whether or not the subject knows the relevant knowledge. The experimenter designs and implements a knowledge-imparting procedure to impart the relevant knowledge to the subject. The knowledge-imparting procedure generally constitutes a training session, a mock crime, or some combination thereof. The purpose of the knowledge-imparting procedure is to make certain that the subject knows the relevant knowledge. The accuracy of a method to detect the relevant knowledge can only be evaluated when the relevant knowledge is actually there to be detected. If the knowledge-imparting procedure fails to impart the knowledge to the subject, then the knowledge is not there to be detected. No method, no matter howperfect, can detect knowledge that is not there. As discussed above in the context of ground truth, in order to conduct a valid test of a knowledge-detection procedure in a laboratory study, the experimenter must independently assess whether the knowledge-imparting procedure actually succeeded in imparting the knowledge so it was there to be detected. This is accomplished by post-test interviews. In a eld case, the brain ngerprinting procedure begins after the criminal investigator has provided the relevant knowledge to the scientist. In a laboratory case, the brain ngerprinting procedure begins after the experimenter has fabricated the relevant knowledge and successfully imple- mented the knowledge-imparting procedure. The relevant knowledge generally comprises 1230 short phrases or pictures, along with an explanation of the signif- icance of each in the context of the crime. The investigator also provides the scientist with a detailed account of which items in the relevant knowledge are or may be already known to the subject for any known reason. For example, the investigator notes any specic features of the crime that have been published in the newspaper or revealed to the subject in interrogation or previous legal proceedings. The relevant knowledge generally contains six to nine or more items that have never been revealed to the subject. These will constitute the probe stimuli. If there is an insufcient number of features that are known only to the perpetrator and investigators (probes), a brain ngerprint- ing test cannot be conducted. Generally there are also six or more items that have already been revealed to the subject or are commonly known. These will constitute the target stimuli. The test requires an equal number of targets and probes. If there are too few features already known to the subject for non-incriminating reasons (potential targets), the experimenter may request additional information about the crime from the criminal investigator to use for target stimuli. Alternatively, if there are ample available features of the crime that are not commonly known and have not been revealed to the subject (potential probes), the exper- imenter may elect to inform the subject about some of these features and use these as targets instead of probes. Scientic standards for brain ngerprinting tests The following procedures comprise the scientic standards for a brain ngerprinting test (Farwell 1992a, 1994, 1995a, b; Farwell and Donchin 1991; Farwell and Smith 2001). 1. Use equipment and methods for stimulus presentation, data acquisition, and data recording that are within the standards for the eld of cognitive psychophysiology and event-related brain potential research. These standards are well documented elsewhere (Donchin et al. 1978, 1986; Fabiani et al. 1987). For example, the standard procedures Farwell introduced as evidence in the Harrington case were accepted by the court, the scientic journals, and the other expert witnesses in the case (Farwell and Donchin 1991; Farwell and Makeig 2005; Farwell and Smith 2001; Harrington v. State 2001). Use a recording epoch long enough to include the full P300-MERMER. For pictorial stimuli or realistic word stimuli, use at least a 1,800 ms recording epoch. (Shorter epochs may be appropriate for very simple stimuli.) 2. Use correct electrode placement. The P300 and P300- MERMER are universally known to be maximal at the midline parietal scalp site (Fabiani et al. 1987; Farwell 1994), Pz in the standard International 1020 system. 3. Apply brain ngerprinting tests only when there is sufcient information that is known only to the perpetrator and investigators. Use a minimum of six probes and six targets. 4. Obtain the relevant knowledge from the criminal investigator (or for laboratory studies from the knowledge-imparting procedure). Use stimuli that isolate the critical variable. Divide the relevant knowledge into probe stimuli and target stimuli. Probe stimuli constitute information that has not been revealed to the subject. Target stimuli contain information that has been revealed to the subject after the crime. 5. If initially there are fewer targets than probes, create more targets. Ideally, this is done by seeking additional known information from the investigators. Note that targets may contain information that has been publicly disclosed. Alternatively, some Cogn Neurodyn (2012) 6:115154 129 1 3 potential probe stimuli can be used as targets by disclosing to the subject the specic items and their signicance in the context of the crime. 6. For each probe and each target, fabricate several stimuli of the same type that are unrelated to the crime. These become the irrelevant stimuli. Use stimuli that isolate the critical variable. For irrelevant stimuli, select items that would be equally plausible for a non-knowledgeable subject. The stimulus ratio is approximately one-sixth probes, one-sixth targets, and two-thirds irrelevants. 7. Ascertain that the probes contain information that the subject has no known way of knowing, other than participation in the crime. This information is provided by the investigator for eld studies, and results from proper information control in laboratory studies. 8. Make certain that the subject understands the signif- icance of the probes, and ascertain that the probes constitute only information that the subject denies knowing, as follows. Describe the signicance of each probe to the subject. Show him the probe and the corresponding irrelevants, without revealing which is the probe. Ask the subject if he knows (for any non-crime-related reason) which stimulus in each group is crime-relevant. Describe the signi- cance of the probes and targets that will appear in each test block immediately before the block. 9. If a subject has knowledge of any probes for a reason unrelated to the crime, eliminate these from the stimulus set. This provides the subject with an opportunity to disclose any knowledge of the probes that he may have for any innocent reason previously unknown to the scientist. This will prevent any non- incriminating knowledge from being included in the test. 10. Ascertain that the subject knows the targets and their signicance in the context of the crime. Show him a list of the targets. Describe the signicance of each target to the subject. 11. Require an overt behavioral task that requires the subject to recognize and process every stimulus, specically including the probe stimuli. Detect the resulting brain responses. Do not depend on detecting brain responses to assigned tasks that the subject can covertly avoid doing while performing the necessary overt responses. 12. Instruct the subjects to press one button in response to targets, and another button in response to all other stimuli. Do not instruct the subjects to lie or tell the truth in response to stimuli. Do not assign different behavioral responses or mental tasks for probe and irrelevant stimuli. 13. In order to obtain statistically robust results for each individual case, present a sufcient number of trials of each type to obtain adequate signal-to-noise enhancement through signal averaging. Use robust signal-processing and noise-reduction techniques, including appropriate digital lters and artifact- detection algorithms (Farwell et al. 1993). The number of trials required will vary depending on the complexity of the stimuli, and is generally more for a eld case. In their seminal study, Farwell and Donchin (1991) used 144 probe trials. In the Harrington eld case, Farwell used 288 probe trials (Harrington v. State 2001). In any case, use at least 100 probe trials and an equal number of targets. Present three to six unique probes in each block. 14. Use appropriate mathematical and statistical proce- dures to analyze the data (Farwell 1994; Farwell and Donchin 1991). Do not classify the responses according to subjective judgments. Use statistical procedures properly and reasonably. At a minimum, do not classify subjects in a category where the statistics applied show that the classication is more likely than not to be incorrect. 15. Use a mathematical classication algorithm, such as bootstrapping on correlations, that isolates the critical variable by classifying the responses to the probe stimuli as being either more similar to the target responses or to the irrelevant responses (Farwell and Donchin 1991; Farwell 1994; Wasserman and Boc- kenholt 1989). In a forensic setting, conduct two analyses: one using only the P300 (to be more certain of meeting the standard of general acceptance in the scientic community), and one using the P300- MERMER (to provide the current state of the art). 16. Use a mathematical data-analysis algorithm that takes into account the variability across single trials, such as bootstrapping (Farwell 1994; Farwell and Donchin 1991; Wasserman and Bockenholt 1989). 17. Set a specic, reasonable statistical criterion for an information-present determination and a separate specic, reasonable statistical criterion for an infor- mation-absent determination (Farwell 1994; Farwell and Donchin 1991; Wasserman and Bockenholt 1989). Classify results that do not meet either criterion as indeterminate. Recognize that indetermi- nate outcome is not an error, neither a false positive nor a false negative. 18. Restrict scientic conclusions to a determination as to whether or not a subject has the specic crime- relevant knowledge embodied in the probes stored in his brain (Farwell and Makeig 2005; Farwell and Smith 2001; Harrington v. State 2001). Recognize that brain ngerprinting detects only presence or 130 Cogn Neurodyn (2012) 6:115154 1 3 absence of informationnot guilt, honesty, lying, or any action or non-action. Do not offer scientic opinions on whether the subject is lying or whether he committed a crime or other act. Recognize that the question of guilt or innocence is a legal determination to be made by a judge and jury, not a scientic determination to be made by a scientist or computer. 19. Evaluate accuracy based on actual ground truth (Farwell and Donchin 1991; Farwell et al. 2011, in press). Ground truth is the true state of what a scientic test seeks to detect. Brain ngerprinting is a method to detect information stored in a subjects brain. Ground truth is whether the specic informa- tion tested is in fact stored in the subjects brain. Establish ground truth with certainty through post- test interviews in laboratory experiments and in eld experiments wherein subjects are cooperative. Estab- lish ground truth insofar as possible through second- ary means in real-life forensic applications with uncooperative subjects. Recognize that ground truth what the subject in fact knows, not what the experimenter thinks the subject should know, not what the subject has done or not done, and not whether the subject is guilty, or deceptive. 20. Make scientic determinations based on brain responses. Do not attempt to make scientic deter- minations based on overt behavior that can be manipulated, such as reaction time. Error rate/accuracy standards for eld applications In the United States and many other jurisdictions, the error rate of a scientic technique is critical for admissibility as scientic evidence in court. The error rate is the percentage of determinations made that are either false negatives or false positives. In brain ngerprinting, this is the percent- age of information present and information absent determinations that are false positives and false negatives respectively. 1 In our view, in order to be viable for eld use or any application with non-trivial consequences, a technique must have an error rate of less than 1% overall, and less than 5% in each and every individual study. As discussed in the section below on common errors in research, failure to meet the brain ngerprinting scientic standards generally produces error rates ten times higher than this standard. Brain ngerprinting exceeds this standard. In all labora- tory and eld research and eld applications to date, brain ngerprinting has had an error rate of less than 1%. In each individual study, brain ngerprinting has also had an error rate of less than 1%. In fact, to date brain ngerprinting has never produced an error, neither a false positive nor a false negative, in any research or eld applications. Accuracy is 100% minus the error rate. In reporting results, it is important to report the error rate directly, or to report the accuracy rate in such a way that the true error rate can be computed 2 (see the section entitled Is Brain Fingerprinting 100% Accurate?). Brain ngerprinting in criminal cases and in court In addition to laboratory and eld studies conducted by the author at the CIA, the FBI, the US Navy, and elsewhere as well as replications in independent laboratories, brain n- gerprinting has been applied in real-world criminal cases and has been ruled admissible in court. According to courtroom testimony by expert witnesses on both sides of the issue, the fundamental science underlying brain ngerprinting testing has been established by hundreds of studies published in the peer-reviewed scientic literature and is well accepted in the relevant scientic community (Farwell and Makeig 2005; Harrington v. State 2001; Iacono 2008). The James B. Grinder case On August 5, 1999, Dr. Lawrence Farwell administered a brain ngerprinting test to murder suspect J. B. Grinder (Fig. 3). The test was designed to determine if Grinders brain contained specic details of the rape and murder of Julie Helton. Drew Richardson, then a scientist in the FBI Laboratory, was the criminal investigator who developed the probe stimuli. The brain ngerprinting test found that the specic details of the crime were recorded in Grinders brain (Fig. 4). The result was information present, with a statistical condence of 99.9%. Considering the brain ngerprinting test results and other evidence, Grinder faced an almost certain conviction and highly probable death sentence. One week after the brain ngerprinting test, Grinder pled guilty to the rape and 1 Note that an indeterminate outcome is neither a false positive nor a false negative error. Rather, it is a determination that there was insufcient data to draw a conclusion with a high statistical condence in either direction. False negatives and false positives are errors that provide false evidence, to the detriment of the judicial process. An indeterminate provides no evidence, and has no legal impact. 2 Accuracy is correctly reported as 100% minus the error rate. This allows the reader to compute the true error rate. Reports of accuracy that confound false positive and false negative errors with indeterminate outcomes have the effect of hiding the true error rate, and thus make comparison with correctly reported studies problematic. Cogn Neurodyn (2012) 6:115154 131 1 3 murder of Julie Helton in exchange for a sentence of life in prison without the possibility of being released. He is currently serving that sentence. In addition, Grinder con- fessed and later pled guilty to the murders of three other young women. The Terry Harrington case In 1977 John Schweer, a retired police captain, was mur- dered near the car dealership in Council Bluffs, Iowa where he was working as a security guard. Terry Harrington was arrested for the murder. An alleged witness claimed that he had accompanied Harrington to the crime scene and wit- nessed Harrington committing the crime. A jury found Harrington guilty in Iowa District Court in 1978. He was sentenced to life in prison without the pos- sibility of being released. On April 18 and 25, 2000, Dr. Lawrence Farwell administered a brain ngerprinting test to Harrington. The test results demonstrated that Harring- tons brain did not have a record of certain specic salient features of the crime. Another test showed that he did recognize salient details of his alibi. The result was information absent with respect to the crime, and information present with respect to the alibi. In both cases the statistical condence was over 99%. When Farwell confronted the only alleged witness to the crime with the brain ngerprinting test results, he recanted his testimony. He admitted that he had lied in the original trial, falsely accusing Harrington to avoid being prosecuted for the murder himself. In Harrington v. State (2001), Terry Harrington sought to overturn his murder conviction on several grounds, including an allegation that newly discovered evidence in the form of brain ngerprinting entitled him to a new trial (Erickson 2007; Farwell and Makeig 2005; Roberts 2007). Standard of review To obtain relief, the petitioner Harrington had to show that the newly discovered evidence was unavailable at the ori- ginal trial, and that the new evidence, if introduced at the trial, would probably change the verdict. Additionally, in view of the fact that the proffered evidence consisted of a novel forensic application of psychophysiological tech- niques, the court was required to determine whether this scientic evidence was sufciently reliable to merit admission into evidence and, if admitted, whether the weight of the scientic evidence was sufciently compel- ling to change the verdict. Fig. 3 Brain ngerprinting test on a serial killer. Dr. Lawrence Farwell conducts a brain ngerprinting test on serial killer J. B. Grinder, then a suspect in the murder of Julie Helton. The test showed that Grinders brain contained a record of certain salient features of the crime. He then pled guilty and was sentenced to life in prison (Photo: Brain Fingerprinting Laboratories, Inc.) Fig. 4 Information present brain response of a serial killer. Brain response of serial killer J. B. Grinder to information relevant to the murder of Julie Helton. There is a clear P300-MERMER in response to the known targets. The P300 is the positive voltage peak at the upper left. The P300-MERMER contains both the P300 peak and the late negative potential (LNP) at the lower right. There is no P300- MERMER in response to the irrelevants. Grinders brain response to the crime-relevant probes clearly contains a P300-MERMER. This shows that the record in the brain of J. B. Grinder contains salient details of the murder. Determination: information present. Statis- tical condence: 99.9% 132 Cogn Neurodyn (2012) 6:115154 1 3 In the Daubert case (Daubert v. Merrell Dow Pharma- ceuticals, Inc. 1993; Erickson 2007) the US Supreme Court has held that the standard for admissibility of novel sci- entic evidence is a showing of reliability based on (1) whether a theory or technique can be (and has been) tested; (2) whether it has been subjected to peer review and pub- lication; (3) whether, in respect to a particular technique, there is a known or potential rate of error, and whether there are standards controlling the techniques operation; and (4) whether the theory or technique enjoys general acceptance within a relevant scientic community. The Iowa Supreme Court has not formally endorsed this federal evidentiary standard, but in Leaf v. Goodyear Tire & Rubber Co. (1999) it announced that the Iowa courts may use the Daubert factors in assessing the admissibility of novel scientic evidence. Moenssens (2002), Erickson (2007), and Roberts (2007) discuss the issues involved in admissibility of brain n- gerprinting in some detail. The brain ngerprinting assessment of Harrington In the Harrington case, Farwell developed a series of probes for the crime scene, and a separate series of probes for the petitioners alibi, from investigations, witness interviews, and previously undisclosed police les. Farwell administered the test to Harrington in May 2000. In October 2000, he rendered a report to the Iowa District Court analyzing the P300-MERMER responses. He sup- plemented the report with a separate analysis based solely on P300 responses on November 10, 2000. Both analyses produced a result of information absent regarding the crime scene probes and information present regarding the alibi probes, with a high degree of statistical condence (over 99%). Figure 5 presents Harringtons brain responses to specic crime-relevant information to which he had not been exposed prior to the test. Proceedings in the Iowa District Court The District Court held a 1-day hearing on the brain n- gerprinting evidence on November 14, 2000. The court took preliminary testimony on Farwells credentials, the efcacy of the test, and the reliability of the underlying science. The court also examined the test results, subject to a later determination whether this scientic evidence was sufciently reliable to be admissible. At the November 14 session, Dr. Farwell testied and was cross-examined on the basis of his test reports. Addi- tionally, two other psychophysiologists with EEG exper- tise, Dr. William Iacono of the University of Minnesota and Dr. Emanuel Donchin of the University of Illinois at Champaign/Urbana, testied on Dr. Farwells credentials, his test reports and the science underlying the brain n- gerprinting test. Iacono testied at Harringtons request, and Donchin was called by the state. Both experts validated the science underlying brain n- gerprinting and acknowledged Dr. Farwells credentials. However, while Iacono validated the forensic application of P300 science based on his own research, Donchin asserted that the testers selection and presentation of the specic probes is the point at which science ends and art begins. The investigative phase of preparing the brain nger- printing test discovers the salient features of the crime that are used as probe stimuli. It depends on the skill and judgment of the criminal investigator. This is not a scientic process. The scientic phase of brain ngerprinting testing begins after the investigation has identied appropriate probes. The science of brain ngerprinting testing determines how the subjects brain responded to the probes, providing an Fig. 5 Information absent brain response of an innocent convict. Brain response of Terry Harrington to information relevant to the murder of which he had been convicted. There is a clear P300- MERMER in response to the known targets. P300 is the positive peak in the top center. The P300-MERMER is the P300 plus the late negative potential (LNP) in the lower right. The response to irrelevant stimuli lacks a P300-MERMER. There is no P300-MERMER in response to the crime-relevant probes. This shows that Harringtons brain does not have a record of these specic features of the crime. He was exonerated and released. Determination: information absent. Statistical condence: 99.9% Cogn Neurodyn (2012) 6:115154 133 1 3 objective result: information present or information absent. This result does not depend on the subjective judgment of the scientist conducting the test. The test result is then presented to the trier of fact to assist in the determination of guilt and innocence. The brain ngerprinting scientist does not opine on guilt or innocence, or whether the suspect committed the crime, but only on the presence or absence in the brain of the suspect of a record of the specic crime- relevant information contained in the probe stimuli. Donchin contended that the selection of probes in brain ngerprinting is the end of science and the beginning of art. Farwell noted that the selection of probes is a feature of the skilled investigation and not of the scientic brain nger- printing testing. Farwell agreed that the selection of probes is a subjective element depending on the skill and judgment of the criminal investigator. He asserted, however, that this subjective element is the kind of evidence that judges and juries are competent to evaluate. A non-scientist is well equipped with common sense and life experience to evaluate all the facts and circumstances of the case and determine whether a nding that the specic probes in questionreturned a scientic result of information present or information absent helps to establish the subjects guilt. Farwell, Iacono, and Donchin agreed that brain nger- printing as practiced using the P300 and published by Far- well and Donchin (1991) in both the laboratory and the real- life cases was generally accepted in the relevant scientic community (Harrington v. State 2001). They also agreed that the additional analysis using the MERMER did not yet have the same level of acceptance, and was not necessary to reach the scientic conclusions relevant to the case. The District Courts ruling The court determined that brain ngerprinting was new evi- dence not available at the original trial, and that it was suf- ciently reliable to merit admission of the evidence (Erickson 2007; Farwell and Makeig 2005; Harrington v. State 2001; Roberts 2007). However, the court didnot regardits weight as sufciently compelling in light of the record as a whole as meeting its exacting standard, and thus it denied a newtrial on this and the other grounds asserted by Harrington. The court stated the following: In the spring of 2000, Harrington was given a test by Dr. Lawrence Farwell. The test is based on a P300 effect. The P-300 effect has been recognized for nearly 20 years. The P-300 effect has been subject to testing and peer review in the scientic community. The consensus in the community of psycho-physiolo- gists is that the P300 effect is valid. The evidence resulting from Harringtons brain n- gerprinting test was discovered after the verdict. It is newly discovered. The court admitted only the brain ngerprinting evidence using the P300. The additional analysis using the P300- MERMER was superuous, and not necessary to establish the brain ngerprinting results. It reached the same statistical and scientic conclusions as the P300 analysis, with essentially the same extremely high statistical con- dence. At that time, Farwell had not yet published the P300-MERMER in peer-reviewed journals. It has now been peer reviewed and published. Appeal The Iowa Supreme Court reversed the trial court and granted Harrington a new trial (Harrington v. State 2003). The Supreme Court did not reach the brain ngerprinting issue, and decided the case on other grounds. Due to a constitutional rights violation, Harrington was accorded a new trial. The only alleged witness to the crime, Kevin Hughes, had recanted when Farwell confronted himwith the information absent results of the brain ngerprinting test on Harrington. Without its star witness, the state subsequently dismissed the murder prosecution without prejudice for lack of evidence due to witness recantations and the passage of time. Resolution and vindication of Harrington In his recantation, Hughes stated under oath under ques- tioning by Farwell that the detectives and prosecutors had told him he would go to prison for life if he didnt implicate Harrington. He stated that when he agreed to falsely accuse Harrington of the murder, they coached him in fabricating the story to which he later testied in the trial. He stated that when he said something that contradicted known facts such as identifying the wrong murder weaponthey cor- rected him, and he changed his story accordingly. Harrington sued the prosecutors and the State of Iowa for framing him. The prosecutors did not deny the accu- sations brought by Hughes and Harrington. Their defense was that they enjoyed absolute immunity due to their professional position. The US Supreme Court agreed to hear the case. Before the Supreme Court heard the case, however, the State of Iowa settled with Harrington and another man falsely convicted of the same crime. The state paid them a $12 million settlement. The Jimmy Ray Slaughter case In 2004, brain ngerprinting testing was offered in support of the Oklahoma petition for post-conviction relief led by 134 Cogn Neurodyn (2012) 6:115154 1 3 death-row inmate Jimmy Ray Slaughter (Slaughter v. State 2004). The Oklahoma court of nal resort in post-convic- tion matters declined to order an evidentiary hearing on numerous issues raised by the petitioner. These included not only an information-absent result for crime-scene items returned by a brain ngerprinting test, but also exculpatory DNA evidence; the sworn testimony of the original lead investigator on the case in which he stated that he had come to believe that Slaughter was innocent, and that others involved in the investigation had falsied reports and fabricated evidence against Slaughter; and other exculpatory evidence. Slaughter was subsequently executed. The Oklahoma court declined to return the case to the trial court where it could reach the merits of the brain ngerprinting challenge, based on procedural grounds and on the appeal courts view that the petitioners brief af- davit contained insufcient evidence of the efcacy of the test and salience of the probes, and that the newly dis- covered evidence was presented in an untimely manner. [B]ased on the evidence presented, we nd the brain ngerprinting evidence is procedurally barred, What we have are some interesting, indeed startling, claims that are not backed up with enough information for us to act on them. Published research on brain ngerprinting science and technology Overview of scientic research Farwell and colleagues have tested brain ngerprinting technology in over 200 cases, including over a dozen sci- entic studies as well as individual forensic cases involving real-life crimes and other events. Numerous other scientists have conducted similar research on the P300 brain response and have replicated Farwells brain ngerprinting research. The scientic studies conducted on brain ngerprinting testing have included both eld/real-life and laboratory studies. Real-life studies involve using brain ngerprinting technology to detect information stored in the brain regarding real-life events that took place in the course of actual life experience. Laboratory studies involve detecting information that was acquired by subjects in the course of a knowledge-imparting procedure such as a mock crime. Of the approximately 200 cases where the author has tested brain ngerprinting technology, about half were real-life situations and half were laboratory experiments. Brain ngerprinting testing has been used to detect information stored in the brain regarding two different types of situation: 1. Specic issue tests detect information regarding a specic incident or a particular crime. 2. Specic screening or focused screening tests detect information relevant to a specic type of training or inside knowledge of a specic eld or organization, such as FBI agent training or knowledge of bomb making. Brain ngerprinting technology detects information stored in the brain. Therefore brain ngerprinting testing is not applicable for general screening or interrogation where the investigators do not know what specic information they seek to detect. General screening includes most pre- employment screening and periodic general security screening of employees. Federal Bureau of Investigation (FBI) studies Farwell conducted FBI Experiment 1, the FBI agent study, (Farwell and Richardson 2006a; Farwell et al. 2011) with Drew Richardson, then a scientist at the FBI Laboratory. Brain ngerprinting produced 100% accurate results in detecting FBI-relevant knowledge in 17 FBI agents and lack of this knowledge in four non-FBI agents. In this experiment, the information detected was specic knowledge that is known to FBI agents and not to the public. The detection of FBI agents indicates that the system could detect knowledge specic to members of specic organization such as a terrorist, criminal, or intel- ligence organization as well as perpetrators of a specic crime. For example, members of a particular terrorist organization or terrorist cell share a particular body of knowledge that is unknown to the public, and the detection of such knowledge could assist in the identication of a suspect as a member of such an organization. Stimuli were words, phrases, and acronyms presented on a computer screen. Analysis using the P300-MERMER resulted in correct determinations in every case, with a high statistical condence in every case. There were no false positives, no false negatives, and no indeterminates. Analysis using the P300 alone resulted in the same deter- minations, with somewhat lower statistical condence in some cases. The FBI agent study included the following innovation that proved useful in making more accurate determinations in future studies. The scientists used the usual probe and irrelevant stimuli. Probes consisted of FBI-relevant knowledge. Irrelevants consisted of irrelevant, unknown items. In initial pilot studies, they used targets consisting of irrelevant items that had been disclosed to the subjects. Some of the stimuli were acronyms wherein the probes were known to the subjects, and both targets and irrelevants were random strings of letters. Cogn Neurodyn (2012) 6:115154 135 1 3 For an FBI-knowledgeable subject, the probes were easily recognizable combinations of letters, and both tar- gets and irrelevants were nonsense letter strings. The sub- jects recognized the probes more quickly than targets and irrelevants, resulting in shorter latency P300-MERMER responses only to the probes (see Fig. 6; contrast this with Fig. 4, wherein probe and target brain responses have identical time course.). When computing correlations, this discrepancy in latency between targets and probes reduced the correlations between these response types. This resulted in decreased statistical condence in detection of infor- mation-present pilot subjects. To correct this, the authors used targets that were FBI- relevant acronyms (like the probes) in the blocks where stimuli were acronyms. The only difference between tar- gets and probes was that the targets had been disclosed to the subjects immediately before the test. Subjects were instructed to push a different button for targets. With this algorithm, targets were more similar to probes for FBI- knowledgeable subjects. Both were quickly recognizable as known acronyms. Both resulted in short-latency P300- MERMER responses. The probe-target correlations were higher, resulting in higher statistical condence. (For non- FBI-knowledgeable subjects, all three types of stimuli were random letter strings, so using FBI-relevant targets made no difference in latency or correlations.) Farwell and colleagues generalized the procedure of using situation-relevant targets in this and subsequent studies. Brain ngerprinting targets now consist of features of the crime or investigated situation, like probes. For a knowledgeable subject, this makes the targets more similar to the probes than targets that are not relevant to the investigated situation. The only difference is that targets have been disclosed to the subject, and the subject pushes a special button only for targets. This tends to increase the accuracy of the algorithm for classifying the probes as being more similar to targets for a knowledgeable subject. Farwell conducted FBI Experiment 2, the real-life FBI study, (Farwell and Smith 2001) at the FBI with Sharon Smith of the FBI Laboratory. Brain ngerprinting tech- nology correctly detected whether or not subjects who were FBI agents had participated in specic, real-life events. All determinations were correct, with a high statistical con- dence in every case. Central Intelligence Agency (CIA) and US Navy studies Farwell and colleagues studies at the CIA and the US Navy (Farwell and Richardson 2006a, b; Farwell et al. 2011) showed that brain ngerprinting could accurately and reliably detect individuals possessing information regarding both mock crimes and real-life activities, including some actual major crimes. In CIA Experiment 1, the picture study, (Farwell and Richardson 2006a) the information detected was relevant to a mock espionage scenario enacted by some of the subjects. The stimuli that elicited the brain responses were relevant pictures presented on a computer screen. Fifteen subjects were correctly determined to be information present, and 13 were correctly determined to be information absent. An example of the stimuli was a picture of the subjects contact person in the mock espionage scenario. CIA Experiment 2 was a collaboration between the CIA and the US Navy (Farwell and Richardson 2006b). In this experiment words and phrases relevant to knowledge of military medicine were presented on a computer screen, Fig. 6 FBI agent brain response to FBI-relevant acronyms. Brain response of an FBI agent to acronyms known to FBI agents and random letter sequences. There is a clear P300-MERMER in response to the known targets. The P300 is the positive voltage peak at the upper left. The P300-MERMER contains both the P300 peak and the late negative potential (LNP) at the lower right. There is no P300- MERMER in response to the irrelevants. The FBI agents brain response to the FBI-relevant probes, like the target response, clearly contains a P300-MERMER. This shows that FBI agent knows the FBI-relevant acronyms. Note, however, that the response to the probes has a shorter latency than the response to the targets and irrelevants. This is because only probes, and not targets, were FBI- relevant acronyms. Subsequent research uses targets that are relevant to the investigated situation, like probes. Subjects are informed of the targets, but are not told which stimuli are probes 136 Cogn Neurodyn (2012) 6:115154 1 3 and subjects were determined to be information present or information absent with respect to knowledge of military medicine. Brain ngerprinting technology resulted in the correct determination in every case. Sixteen subjects were correctly classied as information present, and 14 subjects were correctly classied as information absent. In CIA Experiment 3, the real-life CIA study, (Far- well and Richardson 2006b; Farwell et al. 2011) the information detected was relevant to real-life events, including a number of felony crimes. This study used visually presented words and phrases as stimuli. Seventeen subjects were correctly classied as information present, and three control subjects were correctly classied as information absent. An example of the stimuli was the type of automatic pistol used in one of the crimes. In the CIAand Navy studies there were no false negatives, no false positives, and no indeterminates. All determinations were correct, with a high statistical condence in every case. Two separate analyses were conducted, one using the P300- MERMER and one using only the P300. Both analyses resulted in the same correct determinations. The analysis using the P300-MERMER produced a somewhat higher statistical condence for some of the determinations than the analysis using only the P300 (Farwell et al. 2011). In the three CIA and Navy experiments, the statistical condence for the information present results was 99.9% for each of 44 of the 48 individual determinations. The lowest condence for any information present determination was 98.8%, well above the 90% criterion for information present determinations. Information absent determinations for the real-life experiment were also at least 99.9% in every case. Information absent determinations for the Navy and picture studies were lower on average, but all met the criterion of 70% necessary for an information absent determination. Other studies by Farwell and colleagues Farwell and colleagues more recent studies have focused on real-life/eld applications. One study successfully detected knowledge of improvised explosive devices in bomb makers (Farwell 2009; Farwell et al. 2011). In other eld studies, Farwell and colleagues (Farwell 2008; Farwell et al. 2011) detected information regarding real crimes. To ensure that the brain ngerprinting tests potentially had a major, life-changing impact on subjects regardless of judicial consequences, they offered a $100,000 cash reward to any subject who could beat the test. They also taught subjects countermeasures that been effective in defeating alternative, non-brain ngerprinting tests (Mertens and Allen 2008; Rosenfeld et al. 2004). Brain ngerprinting correctly detected all subjects, with no false positives, no false negatives, and no indeterminates. Summary of results of research and eld applications by Farwell and colleagues In over 200 test cases by Farwell and colleagues, brain n- gerprinting resulted on no false positives and no false neg- atives. Accuracy rate for determinations made was 100%; error rate was 0%. Determinations made were information present or information absent with a criterion statistical condence for each individual determination. Since the introduction of the P300-MERMER in the data analysis algorithm, there have been no indeterminates. In 3% of cases, all of them using the P300 alone prior to the discovery of the P300-MERMER, the data analysis algorithm returned a result that insufcient data were available to make a determination with a strong statistical condence in either direction. No determination was made: the result was indeterminate. All of the indeterminates were in one study (Farwell and Donchin 1991), wherein they comprised 12.5% or results. For all brain ngerprinting studies by Farwell and col- leagues, Grier A (Grier 1971) values are 1.0. Table 2 outlines the laboratory studies on brain nger- printing testing conducted by Farwell and colleagues. Table 3 outlines the eld/real-life studies on brain n- gerprinting testing conducted by Farwell and colleagues. Replications of brain ngerprinting science in other, independent laboratories Others who have followed similar or comparable scientic procedures to those of Farwell and colleagues have pub- lished similar accuracy results in the peer-reviewed liter- ature. For example, Iacono and colleagues have published studies reporting similar procedures and similar results to those achieved by Farwell and colleagues. Iacono and colleagues (Allen et al. 1992) used P300 event-related potentials to detect learned information in a three-stimulus experimental design similar to Farwells technique. The authors achieved 94% accuracy in detecting learned material as learned, and 96% accuracy in identi- fying unknown material as unknown. Like Farwell and colleagues, the authors used an algorithm that included a method for arriving at one of two different determinations, a determination that the subject knew the information or a determination that the subject did not. Also like Farwell and colleagues, they computed a statistical condence for whichever determination was achieved for each individual subject. The authors used a Bayesian algorithm for com- puting a determination and statistical condence for each individual subject. Although the mathematical algorithm was not identical to the bootstrapping algorithm used by Farwell and colleagues, the results showed a relatively high level of accuracy. Cogn Neurodyn (2012) 6:115154 137 1 3 In another study, they (Allen and Iacono 1997) replicated Farwell and colleagues brain ngerprinting technique, and compared their Bayesian algorithmwith the bootstrapping of the brain ngerprinting technique and with a simplied application of bootstrapping. The authors replicated the high accuracy of the brain ngerprinting technique. Like Farwell and colleagues, they reported no false positives using this method. They also found that increased motivation to beat the test increased the accuracy of Farwells brain nger- printing technique. This may be one of the reasons for the extremely high accuracy achieved by Farwell and colleagues using brain ngerprinting in eld situations. The authors Table 2 Brain ngerprinting laboratory studies by Farwell and colleagues Study Authors (year) Type of information detected Type of study Number of subjects Accuracy rate (%) a Indeterminates b Bootstrapping study Farwell and Donchin (1988b) Mock crime; bootstrapping analysis Specic issue 4 100 0 Mock espionage study Experiment 1 Farwell and Donchin (1991), Farwell (1992a) Mock crime/espionage; word stimuli Specic issue 40 100 5 CIA study 1 Farwell and Richardson (2006a) Mock espionage; picture stimuli Specic issue 29 100 0 a Percent correct in all cases wherein a determination was made b Number of cases where no determination was made. In all indeterminate cases, analysis was with P300 alone, not P300-MERMER Table 3 Brain ngerprinting eld/real-life studies by Farwell and colleagues Study Authors (year) Type of information detected Type of study Subjects Accuracy rate (%) a Indeterminates b Crime detection Farwell and Donchin (1986) Real-life minor crimes Specic issue 8 100 0 Real-life Experiment 2 Farwell and Donchin (1986, 1991), Farwell (1992a) Real-life minor crimes Specic issue 8 100 1 Occupation study Farwell (1992b) Occupation-specic knowledge Specic screening 4 100 0 FBI study 1FBI agents Farwell and Richardson (2006b), Farwell et al. (2011) FBI-relevant knowledge, FBI agents Specic screening 21 100 0 CIA/US Navy study 2 Farwell and Richardson (2006b), Farwell et al. (2011) Expertise in military medicine Specic screening 30 100 0 CIA study 3: Real-life CIA study Farwell and Richardson (2006b), Farwell et al. (2011) Real-life events (some crimes) Specic issue 20 100 0 FBI study 2: Real-life Farwell and Smith (2001) Real-life events in FBI agents lives Specic issue 6 100 0 Field tests on suspects Farwell and Richardson (2006b), Farwell et al. 2011) Information on crimes in brains of suspects Specic issue 7 100 0 Brain ngerprinting in counterterrorism Farwell (2009), Farwell et al. (2011) Bomb-making knowledge Specic screening 20 100 0 Real crime $100,000 reward test Farwell (2008), Farwell et al. (2011) Knowledge of actual crimes; $100 k reward for beating test Specic issue 8 100 0 a Percent correct in all cases wherein a determination was made b Number of cases where no determination was made. In all indeterminate cases, analysis was with P300 alone, not P300-MERMER 138 Cogn Neurodyn (2012) 6:115154 1 3 theorized that the basis of this difference was cognitive rather than emotional: that the difference resulted from increased cognitive salience of stimuli in the more motivated condition. As reported in Iacono and colleagues previous study (Allen et al. 1992) the Bayesian procedure achieved an accuracy rate nearly as high as that achieved by Farwell and colleagues using the brain ngerprinting technique, although unlike brain ngerprinting the Bayesian proce- dure returned some false positive/negative errors. The accuracy rate achieved by the authors using Farwell and colleagues bootstrapping technique was comparable. The alternative, simplied bootstrapping procedure achieved a slightly lower accuracy rate. Other experimenters detected concealed information with event-related brain potentials by applying methods that are in some ways similar to, and in some ways dif- ferent from, brain ngerprinting. Some studies have used mock crimes or virtual mock crimes (Abootalebi et al. 2006; Hahm et al. 2009). Some have applied various other knowledge-imparting procedures (Gamer and Berti 2009; Lefebvre et al. 2007, 2009; Meijer et al. 2007). Some have detected recognition of well-known personal information such as pictures of known individuals (Meijer et al. 2007, 2009). These studies have met some but not all of the brain ngerprinting scientic standards. Accuracy rates have in some cases been quite high. Accuracy has varied consid- erably based on the methods used. A number of researchers in Japan (Hira and Furumitsu 2002; Miyake et al. 1993; Neshige et al. 1991) used a variety of procedures applying event-related brain poten- tials in the detection of concealed information. Results varied considerably according to the methods applied. Kakigi and colleagues (Neshige et al. 1991) achieved similar results to Farwells CIA picture study (Farwell and Richardson 2006a). All of these studies serve to further establish the valid- ity, reliability, and general acceptance in the scientic community of the fundamental science on which brain ngerprinting is based. A number of studies have been conducted attempting to detect simulated malingering relevant to brain injury and memory loss. These studies are not directly comparable to brain ngerprinting, and are not reviewed herein. Limitations of brain ngerprinting Is brain ngerprinting 100% accurate? As described in detail above, brain ngerprinting technique using the P300-MERMER has resulted in no false posi- tives, no false negatives, and no indeterminates. All of the determinations have been correct. Overall including studies using the P300 alone, there have been 3% indeterminates. Whether using the P300 alone or the P300-MERMER, 100% of determinations in brain ngerprinting tests by Farwell and colleagues have been correct. Error rate to date has been 0%. Does this mean that brain ngerprinting is 100% accu- rate? No. In science, there is no such thing as 100% accurate. There is always a margin of uncertainty, a margin of error. In reporting on a specic series of laboratory or real- life cases wherein there were no errors, however, it is correct to say (and incorrect not to say), In these specic cases, brain ngerprinting testing produced 100% accurate results. This is simply a statement of the observed facts. To state that brain ngerprinting (or any science) is 100% accurate, without qualication or reference to a specic, existing data set, however, would never be correct. Such a statement contains an implicit prediction about the future. A technology that is 100% accurate never makes an error, now or ever. There is no guarantee that any technology ever can meet that standard throughout all future applications. A technology may have produced 100% accurate results in a particular set of tests already completed, as brain ngerprinting has. Nevertheless, even a technology such as brain ngerprinting technology that has achieved this standard in the past cannot be characterized as 100% accurate without qualication. In short, neither brain ngerprinting nor any other sci- ence or technology can be unqualiedly characterized as 100% accurate. Limits to the applicability of brain ngerprinting testing The brain of the perpetrator plays a prominent role in every crime. Perpetrators virtually always know of their partici- pation in the crime, and often know the features of the crime in considerable detail. Nevertheless, brain ngerprinting is not applicable in every case for every suspect. Probes must contain infor- mation that, in the judgment of the criminal investigators, was experienced by the perpetrator in the course of com- mitting the crime. Probes must contain information that the subject claims not to know. Consequently, there are some circumstances where no probes can be developed for a particular crime for a particular subject. Obviously, in such cases a brain ngerprinting test cannot be conducted. If the investigators have no idea what took place in the perpetration of a crime, for example, when a person simply disappears and foul play is suspected, they cannot develop any probes. No brain ngerprinting test can be conducted. A subject may claim that he was at the crime scene as a witness and not a perpetrator. In such a case, there would be no information that the subject claimed not to know. Cogn Neurodyn (2012) 6:115154 139 1 3 Thus there could be no probes, and a brain ngerprinting test could not be structured. Similarly, a brain ngerprint- ing test is not applicable when the subject knows absolutely everything about the crime that investigators know because he has been told this information after the crime. This may occur when a subject has already gone through a trial and has been convicted of the crime. If there is no information known to investigators that the subject claims not to know, there is no material for probe stimuli, and one cannot structure a brain ngerprinting test. In some cases, however, such as the Terry Harrington case (Harrington v. State 2001), it is possible to nd salient features of the crime to which the subject was never exposed in the trial or investigation, and which he claims not to know. Under these circumstances, a brain nger- printing test can be conducted using these salient features of the crime as probe stimuli. Brain ngerprinting and the limitations of human memory Human memory is not perfect. It is affected by myriad factors, including mental and physical illness, trauma, injury, drugs, aging, passage of time, and many other well known factors. The limitations on human memory already gure prominently in all judicial proceedings that include testi- mony by witnesses or suspects, whether they involve brain ngerprinting evidence or not. A witness, even if he is truthful, does not testify to the absolute truth. He testies only to the contents of his memory. To perform their evaluation of witness testimony ade- quately, judges and juries must already be aware of the well established limitations on human memory and take them into account. Judges and juries must apply these exact same considerations and standards when weighing brain ngerprinting evidence. The argument that brain ngerprinting evidence should not be admitted or considered due to the limitations of human memory is without merit in any forum that admits witness testimony of any kind. Witness testimony consti- tutes a subjective report of the contents of memory. Brain ngerprinting constitutes objective, scientic evidence of the contents of memory. In any forum where subjective reports of the contents of memory are considered, objective evidence of the contents of human memory warrant at least the same treatment. For brain ngerprinting, witness tes- timony, and confessions, the well-known limitations of human memory go to the weight of the evidence, not to admissibility or applicability. When the brain ngerprinting determination is informa- tion present, the limitations of human memory play a minor role. Despite these limitations, the technology has shown that the suspect knows details about a crime that he has previously claimed not to know and has no reasonable explanation for knowing other than having participated in the crime. With brain ngerprinting science, as with all science, negative ndings must be interpreted with caution. When the brain ngerprinting determination is information absent, then the judge and jury must take into account the limitations on human memory in the same way as they do when weighing witness testimony. Conducting a brain nger- printing test on the alibi as well as the crime can help to minimize the possibility that the subjects lack of knowledge of the crime was due to a catastrophic memory failure. (Note, however, that an information present determination with respect to the alibi does not prove that the alibi is true, only that the subjects memory of the alibi is intact.) The effect of the imperfections of human memory and perception on brain ngerprinting evidence is identical to the effect of these factors on the testimony of a witness. The evidence provided by a brain ngerprinting test is limited to a specic determination as to whether certain information is stored in the subjects brain or not. (See above discussion of brain ngerprinting scientic standard 18.) The brain ngerprinting determination is informationpresent or informationabsent withrespect tothe specic probe stimuli provided by the criminal investigators, which in the criminal investigators judgment are salient features of the crime. Neither brain ngerprinting science nor any other sci- ence tells us directly what took place at the crime scene, or whether a particular individual is guilty of a particular crime. Like DNA, ngerprints, and every other forensic science, brain ngerprinting science does not provide a determination of guilty or innocent, or a determina- tion that this suspect did or did not do specic actions. The value of brain ngerprinting science is that it can provide evidence that the triers of fact use in their decisions regarding what took place and who was involved. Brain ngerprinting science does not determine what the facts are, other than the one fact of presence or absence of specic information stored in a specic brain. Brain n- gerprinting expert witnesses testify only to this fact and to the validity and reliability of the science that establishes this fact. They do not opine regarding whether or not the suspect committed the crime; this is to be decided by the judge and jury. The role of brain ngerprinting science in judicial pro- ceedings is to provide evidence that the judge and jury can utilize in reaching their verdict. This evidence must be considered along with all other available evidence. Like other evidence and witness testimony, it must be consid- ered in light of the known limitations on human memory. For a forensic scientist, the import of all discussions about human memory is simply the following: Draw sci- entic conclusions only regarding what the subject knows 140 Cogn Neurodyn (2012) 6:115154 1 3 at the time of the brain ngerprinting test. This is one of the brain ngerprinting scientic standards discussed above (standard 18). For the trier of fact, memory considerations can be summarized as follows. The contents of human memory are revealed subjectively (and not always truthfully) by witness testimony, and objectively by brain ngerprinting. In weighing the evidence and extrapolating from facts regarding the contents of human memory to facts regarding crimes or guilt, use common sense and take into account the well known limitations of human memory. Countermeasures Brain ngerprinting has proven to be highly resistant to countermeasures. No one has ever beaten a brain nger- printing test with countermeasures. Farwell (2008), Farwell et al. (2011) tested countermeasures in a series of brain ngerprinting tests on actual crimes. In order to produce life-changing effects regardless of judicial outcomes, Far- well offered perpetrators of actual crimes a $100,000 cash reward for beating the brain ngerprinting test. The per- petrators were trained in countermeasures that had previ- ously reduced the accuracy of other techniques, but not of brain ngerprinting (Rosenfeld et al. 2004; Mertens and Allen 2008). No one succeeded in beating the brain n- gerprinting test. Brain ngerprinting accurately detected the crime-relevant knowledge in all such subjects, with no false positives, no false negatives, and no indeterminates. Other countermeasure experiments (Sasaki et al. 2002) found a simple mental-task distraction countermeasure to be ineffective. Rosenfeld et al. (2004) report several different coun- termeasure experiments and several different data analysis and statistical methods. In every case, they used funda- mentally different subject instructions, subject tasks, sta- tistics, data acquisition procedures, and methods (or lack thereof) for establishing ground truth than those of brain ngerprinting. Their methods failed to meet 15 of the 20 brain ngerprinting scientic standards. As a result of the fundamental differences between their methods and those of brain ngerprinting, Rosenfeld et al. did not achieve accuracy rates as high as the accuracy rates consistently achieved by brain ngerprinting. For some of Rosenfelds methods in some studies, accuracy was as low as 54%, no better than chance. All of their methods are very different from brain ngerprinting. Even their one method that they correctly characterize as most similar to brain ngerprinting lacks some of the most essential fea- tures of brain ngerprinting methods. All of Rosenfelds (2004) alternative, non-brain ngerprinting methods were found to be susceptible to countermeasures. The countermeasure taught in Rosenfeld et al. (2004) was to perform covert actions such as wiggling the toe in response to each irrelevant stimulus. This was predicted to increase the P300 amplitude to irrelevants, thus lessening the difference between probe and irrelevant brainwave responses. Some of the same subjects had slower reaction times to the stimuli. Reaction times, however, are easily manipu- lated and therefore not suitable for detection in real-life situations with real consequences. Rosenfeld et al. (2008) conducted a second series of studies that showed that their non-brain ngerprinting complex trial protocol method is susceptible to coun- termeasures. Accuracy was 92% without countermeasures and 83% when subjects practiced Rosenfeld et al.s (2004) countermeasure described above. As discussed below, the complex trial protocol has three fundamental characteris- tics that render it unusable in the eld: high and variable error rates, failure to isolate the critical variable along with invalid statistics, and procedures that are ineffective when used with motivated subjects. Another non-brain ngerprinting study, Mertens and Allen (2008), found similar countermeasures to be effec- tive against their procedure. As discussed below, that procedure failed to meet the brain ngerprinting scientic standards, resulting not only in susceptibility to counter- measures but also in very low accuracy even without countermeasures. In discussing countermeasures, it is important to avoid over generalizing the susceptibility to countermeasures of non-brain ngerprinting techniques that fail to meet even the most essential of the brain ngerprinting standards. Some authors (e.g., Rosenfeld 2005; Rosenfeld et al. 2004, Mertens and Allen 2008) have mistakenly generalized the inaccuracy and susceptibility to countermeasures of the non-brain ngerprinting techniques they studied to apply to brain ngerprinting, whereas the actual data on studies that meet the brain ngerprinting scientic standards demon- strate denitively that this generalization does not apply. All available actual data have shown that although these other, non-brain ngerprinting techniques are inaccurate and susceptible to countermeasures, brain ngerprinting is highly accurate and highly resistant to countermeasures. Criticisms of brain ngerprinting Critics have advanced the following criticisms of brain ngerprinting. 1. Criticism: Brain ngerprinting is inaccurate. The relevant facts: Purported support for this criticism comes solely from citing the inaccuracy of non-brain Cogn Neurodyn (2012) 6:115154 141 1 3 ngerprinting studies that used alternative techniques and did not meet the brain ngerprinting scientic standards outlined above. All studies that have met these scientic standards have had extremely high accuracy. Brain n- gerprinting has never produced a false positive or false negative error. 2. Criticism: Brain ngerprinting could be used to detect knowledge that was acquired innocently. The relevant facts: Any forensic science could be used to detect evidence that arose through innocent means. For example, a suspect may have left ngerprints at a crime scene innocently before the crime. Observing common sense, and specically observing the brain ngerprinting standards summarized above, will ensure that innocently acquired information is eliminated in advance from the test (Farwell 1994; Iacono 2008). The standards require the experimenter to establish in advance that the probes con- tain only information that the subject has no known way of knowing, that the subject denies knowing, and that the subject states are not signicant to him and are indistin- guishable from the irrelevant stimuli. This criticism does not apply to brain ngerprinting, but only to non-brain ngerprinting methods that fail to meet the brain ngerprinting scientic standards, particularly standards 4 and 710. 3. Criticism: Brain ngerprinting detects the contents of human memory, and human memory is imperfect. The relevant facts: Human memory is indeed imperfect. Brain ngerprinting is not the only evidence commonly admitted in court that depends on human memory, how- ever. All witness testimony depends critically on human memory. In every trial involving witness testimony, judges and juries must already be aware of and take into account the limitations of human memory. Witness testimony constitutes a subjective report of the contents of human memory. Brain ngerprinting is an objective account of the contents of human memory. In both cases, the trier of fact must evaluate the facts in light of common sense, life experience, and the known limitations of human memory. Such considerations go the weight of the evidence, not to admissibility. This is discussed in more detail above in the section entitled Brain ngerprinting and the limitations of human memory. For a forensic scientist, the import of all discussions about human memory can be stated in its entirety in one sentence, as follows. Observe brain ngerprinting scientic standard # 18: draw scientic conclusions only regarding what the subject knows at the time of the brain ngerprinting test. Critiques of human memory in the context of a discus- sion of brain ngerprinting (Allen 2008; Allen and Mertens 2009; Meegan 2008) amount to a criticism not of brain ngerprinting but rather of any non-brain ngerprinting technique that fails to follow brain ngerprinting scientic standard #18. 4. Criticism: Brain ngerprinting is art not science, subjective not objective. The relevant facts: There are three different processes involved in the application of brain ngerprinting science in a judicial case. These are (1) the investigation that precedes the science; (2) the objective, scientic procedure of brain ngerprinting; and (3) the legal interpretation that may follow later. Before a brain ngerprinting test, a criminal investigator investigates the crime. He formulates an account of the features of the crime. These are the probe stimuli to be tested (and the targets). This criminal investigation is outside the realmof science. This process is based on the skill, expertise, and subjective judgment of the criminal investigator. The criminal investigator provides the scientist with the probe stimuli that in the criminal investigators judgment represent the actual events at the time of the crime. The scientist applies the scientic procedure of brain ngerprinting to determine objectively whether or not the subject knows the crime-relevant information contained in the probes. Brain ngerprinting determines only the pres- ence or absence of this information stored in the subjects brain. The brain ngerprinting scientist opines only on the presence or absence in the subjects brain of the specic knowledge embodied in the probes that were provided by the criminal investigator. Here the science ends. The sci- ence and the scientist do not address the question of whether the results are probative of the subjects guilt or innocence. The science does not even address whether the probes provided by the investigator have anything to do with the crime, or whether a crime took place. The judge and/or jury weigh the brain ngerprinting evidence along with other evidence to reach a non-scientic, common-sense judgment regarding the suspects participa- tion in the crime. This process is outside the realmof science. They may reach a legal determination of guilty or not guilty. The role of the scientically produced brain ngerprinting evidence is only to inform the trier of fact, not to render a scientic conclusion regarding guilt or innocence. In short, brain ngerprinting is an objective, scientic process that is preceded by a process outside the realm of science and followed by another process outside the realm of science. Criticisms of brain ngerprinting as being unscientic result from mistakenly lumping brain ngerprinting with the preceding and/or subsequent non-scientic processes. In effect, all such criticisms amount to a criticism not of brain ngerprinting but of any non-brain ngerprinting 142 Cogn Neurodyn (2012) 6:115154 1 3 technique that fails to observe brain ngerprinting scien- tic standards, particularly standards 4 and 18. 5. Criticism: Brain ngerprinting does not prove that a subject is innocent or guilty, and it would go beyond the science for a brain ngerprinting expert to opine on the guilt or innocence of a subject based on test results. The relevant facts: This is a limitation that brain nger- printing shares with all other forensic sciences. As descri- bed with reference to the preceding criticism, brain ngerprinting accurately and objectively detects whether certain specic information is or is not stored in a subjects brain. Brain ngerprinting standard procedures do not allow a brain ngerprinting expert to draw any conclusions beyond the presence or absence of specic information stored in the brain. It is up to the court to weigh the pro- bative value of these scientically established ndings (Erickson 2007; Harrington v. State 2001; Iacono 2008; Roberts 2007). DNA, ngerprints, and all other forensic sciences also do not prove a subject guilty or innocent. Like brain n- gerprinting experts, experts in these other forensic sciences testify only to what the science actually shows. For example, an expert may testify that DNA putatively from the crime scene matches DNA putatively from the subject. As discussed above, it is up to the judge and jury, not the scientist, to decide if brain ngerprinting evidence, taken along with all the other evidence, warrants a legal deter- mination of guilty or not. This criticism amounts to a criticism not of brain n- gerprinting but rather of any non-brain ngerprinting technique that fails to observe brain ngerprinting scien- tic standard 18. 6. Criticism: Brain ngerprinting is subject to counter- measures. The relevant facts: All evidence cited in support of this contention arises solely from research showing only that non-brain ngerprinting studies that did not meet the scien- tic standards for brain ngerprinting were susceptible to countermeasures. The one study cited as evidence for this contention (Rosenfeld et al. 2004) purported to be a repli- cation of Farwell and Donchin (1991), but in fact failed to meet over half of the brain ngerprinting standards that were met in the original Farwell and Donchin brain ngerprinting research and all other brain ngerprinting research. Farwell (Farwell and Richardson 2006b) offered a $100,000 reward to any subject who could beat a brain ngerprinting eld test using the countermeasures that proved effective against non-brain ngerprinting tests. Brain ngerprinting correctly detected all subjects prac- ticing countermeasures. This is discussed in more detail above in the section entitled Countermeasures. 7. Criticism: A brain ngerprinting test requires that the investigators have specic information regarding the features of the investigated situation. Therefore it is not applicable for general screening when the investigators have no idea what undesirable activities the subject may have undertaken. The relevant facts: This is a limitation that brain nger- printing shares with other evidentiary forensic sciences. To use evidence to connect a suspect to a crime, there must be evidence of the crime. Brain ngerprinting is indeed inapplicable for general screening purposes such as pre- employment screening wherein the investigators have no knowledge of what information they are seeking. Brain ngerprinting can be applied, however, in specic or focused screening for a specic type of knowledge. For example, brain ngerprinting has been successfully used to detect knowledge unique to FBI agents, to US Navy mili- tary medical experts, and to bomb experts (Farwell and Richardson 2006b, Farwell 2009; Farwell et al. 2011). 8. Criticism: A 2001 report by the US General Accounting Ofce (GAO) quoted representatives of several federal agencies as stating that they did not see a role for brain ngerprinting in their current operations at that time. The relevant facts: The GAO report was entitled Fed- eral Agency Views on the Potential Application of Brain Fingerprinting. It was essentially a sampling of opinions of individuals associated with the polygraphy in the federal government prior to 911. (It was completed before 911 2001 and issued shortly thereafter.) It reported that most such individuals did not see the need for brain nger- printing in their pre-911 operations over a decade ago. The report stated: Ofcials representing CIA, DOD, Secret Service, and FBI do not foresee using the brain ngerprinting technique for their operations because of its limited application. For example, CIA and DOD ofcials indicated that their counterintelligence operations and criminal investigations do not usually lend themselves to a technique such as brain ngerprinting because use of the technique requires a unique level of detail and information that would be known only to the perpe- trator and the investigators. These ofcials indicated that they need a tool to screen current and prospective employees, which as indicated above, involves ques- tioning a subject about events unknown to the inves- tigator. Further, a Secret Service ofcial indicated that the agency has had a high success rate with the poly- graph as an interrogative and screening tool and therefore saw limited use for brain ngerprinting. The report noted, however, that the only two US govern- ment scientists interviewed who had conducted research on Cogn Neurodyn (2012) 6:115154 143 1 3 brain ngerprinting both were convinced that it would be useful in FBI investigations. The report did not include an account of the peer- reviewed scientic research on brain ngerprinting or its successful use as scientic evidence in court. The report did not discuss the value of brain ngerprinting for other applications other than general screening, for which it does not apply as discussed above. The GAO did not evaluate or opine on the effectiveness, accuracy, or validity of brain ngerprinting. The report stated: we did not independently assess the hardware, software, or other components of the technology nor did we attempt to determine independently whether brain ngerprinting is a valid technique. The report concluded that a number of federal ofcials did not see an immediate application for brain ngerprinting in their general screening operations before 911. The report constituted a reasonably accurate opinion poll of federal employees associated with the polygraph a decade ago, before the 911 terrorist attacks. This is not relevant to the validity, value, accuracy, or current applicability of brain ngerprinting. Consequently Senator Grassley, who commissioned the original GAO report, has asked the GAO to develop a new report. He asked the GAO to discuss the potential applica- tions of brain ngerprinting in criminal investigations and counterterrorism in the post-911 world. He also asked the GAO to include the views of experts well versed in brain ngerprinting and MERMER technology, and to include the successful brain ngerprinting research at the FBI, CIA, and US Navy. The new report is currently being prepared. Non-brain ngerprinting research on brainwave-based concealed information tests Common errors in research on brainwave-based concealed information tests The seminal papers on brain ngerprinting, Farwell and Donchin (1991), Farwell (1992a, 1994, 1995a, b), and Farwell and Smith (2001) described the scientic standards for brain ngerprinting outlined herein. By meeting these standards, the authors achieved error rates of less than 1% in every study. In our view, this is the level of accuracy that is required for eld use. Subsequent experimenters whose research met the brain ngerprinting standards, such as the replication of Farwell and Donchins (1991) seminal brain ngerprinting research by Iacono and colleagues (e.g., Allen and Iacono 1997), achieved similar accuracy levels to those of brain ngerprinting. Virtually all subsequent researchers adopted some of the original experimental design introduced in the original brain ngerprinting studies. Many of the subsequent experiment- ers, however, did not followthe scientic standards for brain ngerprinting, and consequently did not achieve results comparable to those of brain ngerprinting. Methods that failed to meet the brain ngerprinting standards have gen- erally produced error rates at least ten times higher than the error rates necessary for eld useerror rates of over 10%, and in some cases as high as 50%, no better than chance. In their seminal research and publications on the subject, Farwell and Donchin (1986, 1991) made it clear that brain ngerprinting detects information, not lies, guilt, or actions. Unfortunately, some commentators and even some subsequent researchers have mistakenly considered brain ngerprinting to be a method to detect lies, guilt, or past actions, rather than information stored in the brain. Many errors by subsequent commentators and researchers have resulted from this fundamental misunderstanding. Most of the criticisms of brain ngerprinting (see above) have arisen from the mistaken understanding that brain nger- printing is supposed to detect truthfulness/lies or past actions, rather than information stored in the brain. The following is a summary of the most common errors and the errors that have produced the greatest decrements in accuracy or validity. (A) Failure to recognize that brain ngerprinting detects only the presence or absence of certain specic knowledge stored in the brain. This fundamental misunderstanding or misrepresentation of what is detected by the brainwave measurements is expressed in several ways: Failure to dis- tinguish between what the experimenter knows, or thinks the subject should know, and what the subject actually knows; confounding the knowledge-imparting procedure with the knowledge-detection procedure; failure to establish ground truth; failure to distinguish between subjects actions and knowledge; drawing conclusions that are not warranted by the data, such as that the subject is innocent or guilty, rather than that the subject does or does not possess the specic knowledge tested. (Brain Fingerprinting standards 18 and 19. This error is often combined with failure to meet other standards, particularly 4, 7, 8, and 9.) In several studies that reported low accuracy rates, for example, Rosenfeld et al. (2007), the experimenters con- founded the knowledge-imparting procedure with the knowledge-detection procedure. This study failed to meet 13 of the 20 brain ngerprinting standards, numbers 4, 5, 7, 8, 9, 10, 12, 13, 14, 15, 17, 18, and 19. This resulted in failure to establish valid ground truth, failure to distinguish between knowledge and actions, and unwarranted conclu- sions. The result was very low accuracy rates, ranging from 33% to 62% in different conditions and averaging 51% (no better than chance) overall. 144 Cogn Neurodyn (2012) 6:115154 1 3 The experimenters implemented a knowledge-imparting procedure consisting of a mock crime. They tested the subjects on probe stimuli that the experimenter (but not necessarily the subject) knew were associated with the mock crime. They did not conduct post-test interviews to determine whether or not the knowledge-imparting proce- dure had been effective in imparting the relevant knowl- edge to the subjects. They interpreted the lack of a large P300 to the probes to indicate that the test had failed to detect information pos- sessed by the subjects. This result may, however, have been simply a failure of the knowledge-imparting procedure to impart the relevant knowledge. No knowledge-detection procedure, no matter how perfect, can detect knowledge that the subject does not have. The knowledge-detection procedure may have correctly detected that the subjects lacked the relevant knowledge, because the knowledge- imparting procedure had failed to impart it. Since ground truth was unknown, true accuracy was unknown. Confounding the knowledge-imparting proce- dure with the knowledge-detection procedure produces results that cannot be meaningfully interpreted. One study that obtained very low accuracy rates, Mer- tens et al. (2003), combined this fundamental error with several other errors. This study failed to meet 10 of the 20 Brain Fingerprinting standards, numbers 4, 5, 8, 9, 10, 11, 12, 13, 18, and 19. (B) Confounding or confusing lying with knowing the relevant information. Instructing laboratory subjects to lie by pushing a specic button. (Brain Fingerprinting standard #12.) This is generally done in the laboratory without any real intent or attempt to deceive. Brain n- gerprinting includes no such instruction. Subjects neither lie nor tell the truth in a brain ngerprinting test; they simply observe stimuli and push buttons as instructed. One fundamental error that Rosenfeld and a number of other researchers have made is to instruct subjects to lie in response to probe items in a laboratory experiment (see Farwell 2011a, b, 2012; Rosenfeld et al. 1987, 1988, 2004; Rosenfeld 2002). Recall that in brain ngerprinting the subject is instructed to press one button in response to targets, and another button in response to all other stimuli. Rosenfeld et al. also instructed their subjects to press one button for targets and another for all other stimuli. They told the subjects, however, that their instructed button presses meant yes, I recognize the stimulus and no, I dont recognize the stimulus respectively. They told the subjects they would be lying when they pressed the no button as instructed for probes. Telling subjects to press a button and then telling them that pressing that button constitutes a lie, however, does not create a lie. The Rosenfeld studies and other similar labo- ratory studies involved no intent to deceive and no attempt to deceive. Subjects did not seriously intend to deceive the experimenter into believing they did not recognize well known information by simply pressing the no button as instructed by the same experimenter. The P300 obviously is not a lie response. If a large P300 in response to probes is interpreted as indicating a lie, then a large P300 to targets logically would indicate the same thing. In one study Rosenfeld and colleagues (Verschuere et al. 2009) attempted to validate the procedure wherein subjects are told they will be lying when they press the instructed button in response to probes. The experimental instructions emphasized the salience of the probes in the lie condi- tion, and not in the control condition. The large P300 to the probes in the lie condition was predictably produced by the experimental instructions emphasizing their salience, rather than by the act of pretending to lie. (The lie was pushing a no button as instructed in response to the subjects own nameclearly subjects did not actually intend to deceive the experimenter into believing they did not recognize their own name.) Kubo and Nittono (2009) showed that enhanced P300s in deception conditions are caused not by a deception- specic process but by increased signicance due to additional processing. One reason that there is no lie response is that lying is not a unitary phenomenon. Many different cognitive and emotional processes can be involved in a lie. No one set of cognitive and emotional processes uniquely denes a lie. Instructing subjects to lie and interpreting a large P300- MERMER as being due in part to a lie response is contrary not only to brain ngerprinting standards but also to logic and common sense. Such a ction unnecessarily confounds and complicates the phenomenon being mea- sured and makes the interpretation of results problematic. (C) Failure to inform subjects of the signicance of the probes and to describe the signicance of the probes and targets that will appear in each block immediately before the block (brain ngerprinting standard 8). Some experimenters failed to describe the signicance of the probes in the context of the crime to the subjects. The brainwave responses depend on the subject recog- nizing the signicance of the probe stimuli in the context of the crime. Simply presenting probe stimuli that the experimenter interprets as being meaningful in the context of the crime is insufcient to ensure an appropriate brain response from the subject. For example, Rosenfeld et al. (2007) had subjects pretend to steal an item from a desk drawer lined with blue paper. Then they used blue as a probe, with no indication to the subjects why that might be signicant. They failed to describe the signicance of the probes in the context of the crime to the subjects, and also failed to determine ground truth through post-test inter- views. The consequence of this and the dozen or so other Cogn Neurodyn (2012) 6:115154 145 1 3 errors common to all the Rosenfeld studies was detection rates that were extremely low (33% to 62%, averaging no better than chance) for this study (see Farwell 2011a, b, 2012). To ensure that a knowledgeable subject recognizes the probes as signicant in the context of the investigated sit- uation, the experimental instructions must explicitly inform the subject of this signicance. Of course, the instructions must not inform the subject which is the probe and which are the similar irrelevants constituting incorrect details with the same possible signicance. Failure to follow this procedure tends to markedly decrease accuracy (Meijer et al. 2009). (D) Failure to run a sufcient number of trials for ade- quate signal-to-noise enhancement, or failure to apply adequate signal-processing and noise-reduction techniques such as digital lters and artifact detection algorithms (brain ngerprinting standard #13). This is a very common error. All studies with extremely low accuracy rates have failed to meet this standard, including Mertens and Allen (2008), Mertens et al. (2003), Miyake et al. (1993), Rosenfeld et al. (2007), and some conditions of Rosenfeld et al. (2006) and Rosenfeld et al. (2004). (For discussion see also Roberts 2007). Each of these studies also failed to meet at least several other standards. (E) Teaching subjects countermeasures or instruc- tions that transparently accentuate the probe stimuli, or that require the subject to read and attend to the probe stimuli when motivated subjects otherwise would not do so (Rosenfeld et al. 1987, 2008 and subsequent complex trial protocol studies; see discussion below). Such counter- measures actually increase the salience of the probes and enhance the responses to the probes, making the test appear to be more effective and/or resistant to countermeasures. Motivated subjects who understand how the test works would not be coerced or tricked into following such instructions. (See Farwell, 2012 for a review.) (F) Failure to require an overt behavioral task that requires the subject to recognize and process every stim- ulus, specically including the probe stimuli (brain n- gerprinting standard 11). Studies that fail to meet this standard are unusable in the eld, where there is no guar- antee that the subject can be trusted to do what the experimenter would like him to do except insofar as required by overt actions. Rosenfeld and colleagues have published several studies using a complex trial protocol that fails to meet this and other standards and consequently is unusable in the eld (Meixner et al. 2009; Meixner and Rosenfeld 2010; Meixner and Rosenfeld in press; Rosenfeld et al. 2008, 2009; Winograd and Rosenfeld in press; for a review, see Farwell 2011a, b; 2012). In addition, all of these studies failed to meet 17 of the 20 Brain Fingerprinting standards. This is further discussed below. (G) Failure to isolate the critical variable, generally combined with improper and invalid use of statistics. Failure to establish separate determinations and reasonable statistical condence criteria for both information-present and information-absent results. Failure to include an indeterminate category. Classifying some results in a cat- egory where there is up to an 89% probability that the classication is an error. (Brain ngerprinting standards 47, 14, 15, and 17.) This error and its consequences are illustrated by a series of experiments conducted by Rosenfeld and colleagues (see Farwell 2011a, b, 2012; Lui and Rosenfeld 2008; Johnson and Rosenfeld 1992; Rosenfeld et al. 1987, 1988, 1991, 2004, 2006). All of Rosenfeld and colleagues studies failed to meet 1217 of the 20 brain ngerprinting standards, including standards 14, 15, and 17 in every case (see Farwell 2011a, b, 2012). Although they used some of the major features of the probe-target-irrelevant stimulus design introduced by Farwell and Donchin, their methods were signicantly different (Farwell 2011a, b, 2012; Far- well and Smith 2001; Rosenfeld 2002). Their subject instructions, subject tasks, data acquisition procedures, data analysis procedures, and even stimulus types all failed to meet the brain ngerprinting standards outlined above. They conducted a fundamentally different procedure from brain ngerprinting, and not surprisingly obtained different and less accurate results. Their methods and results varied in their various attempts (see Farwell 2011a, b; 2012). In some cases results were over 80% or 90% accurate (Johnson and Rosenfeld 1992; Lui and Rosenfeld 2008; Rosenfeld et al. 1991, 2004, 2006). In some cases Rosen- feld et al.s results were no better than chance or less than chance (Rosenfeld et al. 2004, 2006, 2007). None of their techniques approached the requirement of less than 1% error rate across all studies and less than 5% error rate in every individual study that is necessary in our view for a technique to be viable for eld use. These studies generally failed to meet the same 14 brain ngerprinting scientic standards, 36, 810, 1215, and 1719. In some cases they also failed to meet additional standards. For example, in their rst attempt Rosenfeld et al. (1987) also measured P300 from the wrong location on the head (standard #2) and did not require any overt responses (standard #11). They made their determinations on the basis of subjective judgment based on looking at the plots of the waveforms, and did not compute any statistics on individual subjects (standards 1417) (see Bashore and Rapp 1993; Rosenfeld 1995). Moreover, they included instructions to the subject to engage in covert behavior that introduced a confound that made results uninterpretable. In their second attempt Rosenfeld et al. (1988) corrected the 146 Cogn Neurodyn (2012) 6:115154 1 3 confound and the scalp location, but repeated all of the other same errors. In all of Rosenfeld et al.s subsequent studies (e.g., Rosenfeld et al. 2004) they failed to isolate the critical variable, and failed to properly use the bootstrapping algorithm introduced by Farwell and Donchin (1988b, 1991). They ignored target responses, and simply attemp- ted to determine if probe responses were larger than irrelevant responses. This resulted in invalid statistical procedures that classied subjects in a category where the statistics computed had determined that there was up to an 89% probability that the classication was incorrect. This is discussed below in reference to the complex trial pro- tocol, which combines this error with numerous others. Failure to establish separate, reasonable criteria for both information-present and information-absent determinations virtually guarantees classication errors. This obviously reduces accuracy. It also makes results uninterpretable and difcult to compare to results obtained in studies that have used statistics properly and reasonably. The complex trial protocol (Rosenfeld et al. 2008 and subsequent studies cited above) fails to isolate the critical variable; it lacks the necessary standard to do so. The statistical procedures based on (and necessitated by) this failure are not valid or viable for eld use. In addition, it fails to meet 17 of the 20 brain ngerprinting scientic standards, specically standards 36 and 820. It has not been used in the eld or in any real-world situation with non-trivial consequences, and is not viable for such use. The following three fundamental characteristics make the complex trial protocol fundamentally unusable in the eld (for details see Farwell 2011a, b, 2012). First, complex trial protocol published error rates are far too high for eld use. In our view, to be viable for eld use a technique must have false positive/negative error rates of less than 1% overall and less than 5% in each and every individual study. Complex trial protocol error rates in published studies average over ten times higher than these criteria. Even in highly contrived laboratory conditions with accommodating subjects and no real consequences to the outcome, published accuracy in some studies is as low as 53%, no better than chance. Error rates in all published studies average 15% without countermeasures and 29% with countermeasures. Second, the complex trial protocol does not isolate the critical variable. It lacks a standard for an information- present response. Without a standard for comparison, it is impossible to compute valid and meaningful statistics. The statistics applied are invalid and result in extremely low statistical condences, on average no better than chance for information-absent determinations. Targets are simply meaningless number strings. They are not comparable to probes and do not provide a standard for an information- present response. Data analysis ignores target responses, and simply seeks to determine whether the probe responses are larger than the irrelevant responses (with larger variously dened). Bootstrapping computes the probability that the probe responses are larger than the irrelevant responses. 3 It classies subjects as information present (or guilty) if this probability is 90% or greater, and other- wise classies them as innocent. Thus, the statistics applied classify subjects as innocent when there is up to an 89% probability that they are in fact guilty, that is, an 11% probability that the determination is correct. In pub- lished research subjects have been classied as innocent when the statistics used returned as high as an 86% prob- ability that the opposite classication would be correct, 4 or 14% probability that the classication returned by the procedure is correct. Moreover, in the published data, the average statistical condence for an innocent classi- cation is 50%, no better than a coin ip. This is in accord with the predictions of the statistical model used. This is a fundamental aw in the statistical procedures, due to the lack of a standard for an information-present response, that cannot be corrected by simply changing the criterion. Setting a higher statistical condence criterion for an innocent determination will simply increase the already unacceptably high error rate (see Farwell 2011a, b, 2012 for a detailed explanation). For example, a criterion could be established that subjects are classied as innocent when there is a 50% or greater probability that the probe response is larger than the irrelevant response (that is, a higher than chance probability that this classication is correct). Applying this criterion to the actual reported data results in an error rate of 50% or more (less than chance accuracy) in the published studies. In other words, according to both published data and the predictions of the statistical model used, any criterion that results in greater than chance statistical condence results in lower than chance accuracy. Obviously, the procedures that produce such low statistical condences and high error rates are unsuitable for eld use, where high statistical condence is a necessity and errors have real consequences for life and freedom. Third, on the stimulus presentations that are included in the data analysis (probes and irrelevants), the subjects are 3 Note that in brain ngerprinting, both targets and irrelevants provide standards. Bootstrapping computes the probability that the probe responses are more similar to the target responses, or more similar to the irrelevant responses. This results in high statistical condences for both information-present and information-absent determinations. 4 For example, in Meixner and Rosenfeld (2010, p. 63, Table 2, subject 10), GNO condition was classied innocent when there was an 86% probability that the probe response was larger than the irrelevant response (i.e., 86% probability that guilty was the correct classication). Cogn Neurodyn (2012) 6:115154 147 1 3 not required to distinguish behaviorally between stimulus types. They simply push a button indicating that something appeared in a general area on the screen. Probes/irrelevants occur at totally predictable times that are known to the subjects. Thus, subjects are not required to read and pro- cess probe and irrelevant stimuli and prove behaviorally that they have done so on each trial. Motivated subjects with something to hide do not read and process these stimuli, and consequently their brain responses do not reveal their concealed knowledge. The accuracy of the complex trial protocol in detecting real-world information in motivated subjects is 0% (Farwell et al. 2011). A sug- gested means to coerce a subject into accommodatingly revealing her concealed knowledge (Rosenfeld et al. 2008) was to periodically ask the subject to recall what stimuli she had seen, and threaten to le a false report that the subject had failed the brainwave test if she did not recall the stimuli accurately. Such a report would be false, since in truth she had passed the brainwave test: her brain responses contained no evidence of any relevant knowl- edge. Clearly, this would be both unethical and ineffective in any real-world situation with motivated subjects and real consequences. Subjects could easily discern that neither their brain responses nor such a false report would provide any scientic evidence of their knowledge of the relevant information, and the false report would carry no weight in any real-world judicial proceeding. This method differs from brain ngerprinting in several other ways. The complex trial protocol has never been independently replicated. It has never been applied in the eld or in any real-world situation with non-trivial conse- quences. It has never been ruled admissible in court, and in light of the above it appears extremely unlikely that it ever will be in the future. In our view, any attempt to apply it in the real world would be scientically, ethically, and legally untenable. Reports in the popular press have sometimes mistakenly considered another method, the brain electrical oscillation system (BEOS) developed in India by Chanpadi Raman Mukundan, to be based on the authors original brain n- gerprinting research. Unlike brain ngerprinting, however, it is not based on established scientic phenomena and published data. Like the complex trial protocol, it has not been independently replicated or published in the peer- reviewed literature. It purports to distinguish between knowledge gained while committing a crime and knowl- edge learned after the fact by an innocent person. No known mechanism or psychophysiological phenomenon has been proposed on the basis of which to make this distinction. Initially it was used in some criminal cases in India, but it later was ruled inadmissible in court there. In addition to the complex trial protocol and the BEOS system, three studies, Miyake et al. (1993), Rosenfeld et al. (2004), and Mertens and Allen (2008) warrant particular mention in the context of scientic and methodological errors. This is because (1) the errors resulted in excep- tionally low accuracy rates; and (2) although the methods are fundamentally different from brain ngerprinting and fail to meet the brain ngerprinting standards, the studies have been mistakenly considered to reect negatively on brain ngerprinting (Harrington v. State 2001; Rosenfeld 2005; Mertens and Allen 2008). In fact, the studies show only that these alternative, non-brain ngerprinting meth- ods, which fall far short of the brain ngerprinting stan- dards in fundamental ways, are inaccurate and susceptible to countermeasures. A study in Japan by Miyake et al. (1993) failed to meet 18 of the 20 brain ngerprinting scientic standards (all but numbers 3 and 20). Moreover, the experimenters failed to implement data collection, artifact rejection, and data analysis procedures that meet the universal standards met by other laboratories in the eld of event-related brain potential research. They measured responses from the wrong scalp location. These errors resulted in an excep- tionally low accuracy rate. Only 65% of their determina- tions were correct. In a study that has been mistakenly (Rosenfeld 2005, Mertens and Allen 2008) considered to be similar to brain ngerprinting, and even purported to be a replication of the original Farwell and Donchin brain ngerprinting research, Rosenfeld et al. (2004) failed to meet 15 of the 20 brain ngerprinting scientic standards, specically numbers 36, 810, 1215, and 1720. They reported a variety of accuracy rates for different methods and analysis proce- dures. None of the accuracy rates were as high as the accuracy rates of brain ngerprinting, and some were as low as chance (54%) even without countermeasures (see Farwell 2011a, b, 2012). This study also showed that Ro- senfelds techniques (but not brain ngerprinting) are susceptible to countermeasures. In a third study that has been mistakenly considered to be similar to brain ngerprinting, Mertens and Allen (2008) failed to meet standards 8, 13, 18, and 19 and consequently achieved very low accuracy rates as well as susceptibility to countermeasures. They used valid statistical procedures and met several other standards that were not met by the Rosenfeld et al. (2004) and Miyake et al. (1993) studies. Nevertheless, failure to meet several vital scientic stan- dards resulted in a major decrement in accuracy and also in susceptibility to countermeasures. The above studies serve to demonstrate that meeting the scientic standards for brain ngerprinting research out- lined herein is important in order to obtain valid, accurate, reliable, and interpretable results. Meeting certain stan- dards is also necessary for establishing procedures that can be applied in eld settings. These studies also highlight the 148 Cogn Neurodyn (2012) 6:115154 1 3 fact that the proven inaccuracy and susceptibility to countermeasures of other, non-brain ngerprinting tech- niques does not imply that brain ngerprinting shares these same shortcomings. Summary The role of brain ngerprinting in forensic science is to bring within the realm of scientic scrutiny the record of a crime, terrorist act, terrorist training, specic crime- or terrorism-related knowledge or expertise, or other relevant information stored in the brain of a suspect or other person of interest, and to develop reliable forensic science evi- dence based on the accurate detection of such information. Brain ngerprinting is a scientic technique to detect concealed information stored in the brain by measuring event-related potential (ERP) brainwave responses. Brain ngerprinting laboratory and eld tests at the CIA, the FBI, the US Navy, and elsewhere have detected the presence or absence of information regarding the following: real-life events including felony crimes; real crimes with substantial consequences, including judicial outcomes such as the death penalty or life in prison; concealed information in real-world cases where sub- jects were offered a $100,000 reward for beating the test; knowledge unique to FBI agents; knowledge unique to explosives (EOD/IED) experts; knowledge unique to US Navy medical military personnel; pictorially represented knowledge; mock crimes and mock espionage scenarios; other laboratory tests and real-world applications. Brain ngerprinting has been successfully applied in eld settings, including actual criminal cases wherein the brain ngerprinting test was ruled admissible in court and/or contributed to bringing the perpetrator to justice or exonerating the innocent. Brain ngerprinting helped to bring a serial killer to justice and to exonerate an innocent man falsely convicted and imprisoned for murder. A brain ngerprinting test measures the subjects brain responses to specic information. The information is embodied in stimuli consisting of words, phrases, or pic- tures presented on a computer screen. Some of the stimuli are probe stimuli. Probes contain information that is rele- vant to the crime or situation under investigation and that the subject has no way of knowing outside of having par- ticipated in the crime. When the subject recognizes the relevant information contained in the probes as signicant in the context of the crime, the brain emits an Aha! response. Brain ngerprinting measures and analyses the brainwaves and detects the corresponding P300-MERMER brain response. Brain ngerprinting computes a determination of information present (the subject possesses the specic information tested) or information absent (he does not) for each individual subject. The brain ngerprinting boot- strapping algorithm also computes a statistical condence for each individual determination. In data analysis, responses to probes are compared with two standards. Target stimuli provide a standard for the subjects brain response to known crime-relevant informa- tion, information which is provided to all subjects. Irrelevant stimuli provide a standard for the subjects response to irrelevant information consisting of plausible but incorrect features of the crime. Data analysis determines whether the response to the probes is more similar to the response to the targets or to the response to the irrelevants, and provides a statistical condence for this determination. In our view, in order to be viable for eld use, a tech- nique must have less than 1% error rate overall, and less than 5% error rate in every individual study. Brain n- gerprinting exceeds this standard. In over 200 cases including all eld and laboratory research so far, brain ngerprinting has not produced a single error, neither a false negative nor a false positive. Error rate has been 0%. 100% of determinations have been correct. (Note, however, that this is simply a report of the actual data to date; no science can be generally characterized as 100% accurate without qualication or reference to a specic data set.) In brain ngerprinting using the P300-MERMER, all tests have resulted in a denite determination with a high sta- tistical condence. There have been no indeterminates. In brain ngerprinting using the P300 alone, in less than 3% of cases, the data analysis algorithm has concluded that insufcient information is available to make a determina- tion in either direction with a high statistical condence, resulting in an indeterminate outcome (not an error). Brain ngerprinting specic issue tests detect information regarding a specic event at a particular time and place, such as a crime or terrorist act. Brain ngerprinting specic screening tests detect a specic type of knowledge or expertise, such as knowledge unique to FBI agents, bomb makers, or Al-Qaeda-trained terrorists. Brain ngerprinting is not applicable for general screening, when the investiga- tors have no idea what information is being sought. Brain ngerprinting is highly resistant to countermea- sures. Subjects have been taught the same countermeasures that have proven effective against other, non-brain nger- printing techniques. Countermeasures had no effect on brain ngerprinting, despite the $100,000 reward offered for beating the test with countermeasures and the motiva- tion to beat the test inherent in real-world applications. All Cogn Neurodyn (2012) 6:115154 149 1 3 subjects, whether practicing countermeasures or not, have been correctly detected. The results of original research by the author and independent replications in other laboratories show that accuracy and validity can be reliably attained by following the established brain ngerprinting scientic standards outlined herein. Studies that have failed to meet the brain ngerprinting scientic standards show that such failure can result in low accuracy, susceptibility to countermea- sures, and in some cases unreliable results and invalid procedures. Studies in the original laboratory and independent rep- lications elsewhere that meet the brain ngerprinting sci- entic standards consistently report extremely accurate results. Studies that fail to meet the brain ngerprinting standards, particularly certain critical standards, have resulted in inconsistent and lower accuracy rates, in some cases no better than chance. In some cases such failure also produced results that are uninterpretable and/or invalid. We have outlined herein the most common errors in brain- wave-based concealed information tests, and the errors that have produced the greatest decrements in accuracy, reli- ability, and validity. This paper reviews all relevant research previously published in English. In view of the published data, some caveats are necessary with respect to reliability, accuracy, practical usefulness in the eld, and generalizations regarding results. The published results indicate that fol- lowing the established brain ngerprinting scientic stan- dards outlined herein is sufcient to ensure accurate, reliable, and valid results in laboratory studies and eld applications. However, the available evidence does not support the notion that any attempt that fails to meet the established brain ngerprinting scientic standards could be expected to obtain accurate, reliable, or valid results. Techniques that fail to meet at least the most essential of the brain ngerprinting scientic standards have generally produced error rates at least ten times higher than the error rates that in our view are necessary for viable eld use. To be meaningful and practically useful, generalizations about brainwave-based concealed information tests must distin- guish between the studies that meet the brain ngerprinting standards and those that fail to meet the standards. Gen- eralizations that fail to recognize this distinction are inad- equate to present a meaningful interpretation of the available data, and can result in drawing erroneous con- clusions about brain ngerprinting that in fact apply only to non-brain ngerprinting tests that fail to meet the stan- dards. For example, the low accuracy and susceptibility to countermeasures characteristic of several non-brain n- gerprinting techniques has sometimes been erroneously generalized to apply to brain ngerprinting, whereas in fact the actual data directly contradict this generalization. In addition to conducting the science according to established brain ngerprinting standards, it is also vitally important to restrict the interpretation of brain nger- printing results to what the science actually shows. Brain ngerprinting detects the presence or absence of specic information stored in the brain. It does not detect guilt, lies, emotions, intentions, or any action, including participation in a crime. Prior to a brain ngerprinting test, a criminal investi- gator develops his account of the crime, based not on science but on his skill and judgment as an investigator. He determines that, in his non-scientic judgment, informa- tion contained in specic probe stimuli is relevant to the crime. The brain ngerprinting scientist tests scientically whether or not this specic information is stored in a specic subjects brain. Scientic reports of brain nger- printing results, and testimony by brain ngerprinting expert witnesses in court, must be conned to an expla- nation of the science and a report of what the science actually shows. The only legitimate scientic conclusion to be reported is that the brain ngerprinting evidence shows, with a particular statistical condence, that the subject either does or does not know the information contained in the probe stimuli in the context of the crime. Any inter- pretation of the results in terms of the subjects guilt or innocence goes beyond the science and is outside the legitimate purview of testimony by a brain ngerprinting scientist. Brain ngerprinting scientists whose testimony on brain ngerprinting has been admitted as scientic evidence in court have adhered closely to this requirement. As discussed herein, brain ngerprinting science does not evaluate whether the criminal investigators account of the crime, and the probe stimuli included therein, accu- rately represent the crime, or whether the suspect com- mitted the crime. Brain ngerprinting, and a brain ngerprinting scientists testimony, do not address what the suspect should know, could know, or would know about a crime under what circumstances (e.g., if he is innocent or guilty). Brain ngerprinting only detects what the subject actually does know about the crime. It is up to the criminal investigator to come up with an account of the crime and the knowledge relevant thereto (the probe stimuli). It is up to the prosecuting and defense attorneys to debate, and the judge and jury to decide, what all the evidence, including the brain ngerprinting evidence, means with respect to what happened, whether a crime was committed, and if so what was the crime, who committed it, and who is guilty or not. Brain ngerprinting is similar to other forensic sciences in this regard. DNA testing, for example, concludes only that Sample A (ostensibly from the crime scene) matches sample B (ostensibly from the subject). DNA science and 150 Cogn Neurodyn (2012) 6:115154 1 3 DNA expert witnesses do not determine or opine on whether the subject is guilty of a murder. That is up to the judge and jury to decide based on all the evidence. Brain ngerprinting does not present a conclusion regarding the subjects guilt or innocence of the crime. Like other forensic sciences, it simply provides evidence that is useful for the the judge and jury in their determinations regarding what took place and who is guilty or not. In weighing the evidence and extrapolating from witness testimony and scientic evidence to the question of whe- ther a suspect committed a crime, judges and juries must use common sense and take into account the well known limitations of human memory. This consideration applies equally to witness testimony, which is a subjective (and not always truthful) account of the contents of memory, and to brain ngerprinting, which provides an objective, scientic account of the contents of human memory. Such consid- erations are evaluated by the judge and jury on the basis of their common sense and life experience, and are outside the realm of the scientic testimony of a brain ngerprinting scientist. The results of published laboratory research and eld applications indicate that brain ngerprinting testing con- sistently provides accurate, reliable, and valid scientic evidence in the detection of concealed information, pro- vided that the science is conducted strictly according to the established brain ngerprinting scientic standards and interpreted according to these same standards such that the scientic interpretation stays strictly within the boundaries of what the science actually demonstrates. Brain nger- printing evidence, when based on science so conducted and interpreted, has proven to be of value in real-world crim- inal justice and national security applications. Brain ngerprinting has been successfully applied in real-world cases and ruled admissible as scientic evidence in court. Scientists in the United States currently continue to apply it successfully in the eld. The results reviewed herein suggest that brain ngerprinting provides a new scientic method to accurately and reliably detect the presence or absence of concealed information that can generate useful forensic evidence in real-world applica- tions in criminal justice and national security. Acknowledgments The author is grateful to former FBI scientists Drew Richardson, PhD and Sharon Smith, PhD for their substantial contributions to some of the research reviewed herein; to attorneys Thomas Makeig, Mary Kennedy, and Thomas Frerichs for their legal expertise in obtaining admissibility in court for brain ngerprinting; to the US Central Intelligence Agency (CIA) for funding several of the studies reviewed herein; and to the US Federal Bureau of Investigation (FBI) and the US Navy for their cooperation and assistance in conducting research with their personnel in their facil- ities. The views expressed herein are the views of the author, and do not necessarily reect the views of any other individual or agency. References Abootalebi V, Moradi MH, Khalilzadeh MA (2006) A comparison of methods for ERP assessment in a P300-based GKT. Int J Psychophysiol 62(2):309320 Allen J (2008) Not devoid of forensic potential, but. Am J Bioethics 8(1):2728 Allen J, Iacono WG (1997) A comparison of methods for the analysis of event-related potentials in deception detection. Psychophys- iology 34:234240 Allen JJ, Mertens R (2009) Limitations to the detection of deception: true and false recollections are poorly distinguished using an event-related potential procedure. Soc Neurosci 4(6):473490 Allen J, Iacono WG, Danielson KD (1992) The identication of concealed memories using the event-related potential and implicit behavioral measures: a methodology for prediction in the face of individual differences. Psychophysiology 29:504522 Basar-Eroglu C, Basar E, Demiralp T, Schurmann M (1992) P300- response: possible psychophysiological correlates in delta and theta frequency channels. A review. Int J Psychophysiol 13(2):161179 Bashore T, Rapp P (1993) Are there alternatives to traditional polygraph procedures? Psychol Bull 113:322 Baudena P, Halgren E, Heit G, Clarke JM (1995) Intracerebral potentials to rare target and distractor auditory and visual stimuli. III. Frontal cortex. Electroencephalogr Clin Neurophys- iol 94:251264 Daubert v. Merrell Dow Pharmaceuticals, Inc. 509 US 579, 594 (1993) Donchin E, Ritter W, McCallum WC (1978) Cognitive psychophys- iology: the endogenous components of the ERP. In: Callaway E, Teuting P, Koslow S (eds) Brain event-related potentials in man. Academic Press, New York, pp 349441 Donchin E, Miller GA, Farwell LA (1986) The endogenous components of the event-related potentiala diagnostic tool? In: Swaab DF, Fliers E, Mirmiran M, Van Gool WA, Van Haaren F (eds) Progress in brain research. Vol 70: Aging of the brain and Alzheimers disease. Elsevier, Amsterdam, pp 87102 Erickson MJ (2007) Dauberts bipolar treatment of scientic expert testimonyfrom Fryes polygraph to Farwells brain nger- printing. Drake Law Rev 55:763812 Fabiani M, Gratton G, Karis D, Donchin E (1987) The denition, identication and reliability of measurement of the P300 component of the event-related brain potential. In: Ackles PK, Jennings JR, Coles MGH (eds) Advances in psychophysiology, vol 2. JAI Press, Greenwich, pp 178 Farwell LA (1992a) The brain-wave information detection (BID) system: a new paradigm for psychophysiological detection of information. Doctoral Dissertation, University of Illinois at Urbana-Champaign, pp 1165 Farwell LA (1992b) Two new twists on the truth detector: brain-wave detection of occupational information. Psychophysiology 29(s4A):S3 Farwell LA (1994) Method and apparatus for multifaceted electro- encephalographic response analysis (MERA). US Patent #5,363,858 Farwell LA (1995a) Method and apparatus for truth detection. US Patent #5,406,956 Farwell LA (1995b) Method for electroencephalographic information detection. US Patent #5,467,777 Farwell, LA (1999) How consciousness commands matter: the new scientic revolution and the evidence that anything is possible. Sunstar. Faireld, IA Cogn Neurodyn (2012) 6:115154 151 1 3 Farwell LA (2007) Apparatus for a classication guilty knowledge test and integrated system for detection of deception and information. UK Patent #GB2421329 Farwell LA (2008) Brain ngerprinting detects real crimes in the eld despite one-hundred-thousand-dollar reward for beating it. Psychophysiology 45(s1):S1 Farwell LA (2009) Brain ngerprinting in global security. Presented at the Global Security Challenge Security Summit, Nov 2009. London Business School, London Farwell LA (2010) Method and apparatus for brain ngerprinting, measurement, assessment and analysis of brain function. US Patent # 7,689,272 Farwell LA (2011a) Brain ngerprinting: corrections to Rosenfeld. Sci Rev Mental Health Practice 8(2):5668. http://www.brain wavescience.com/Farwell_Brain_Fingerprinting_Corrections_ to_Rosenfeld_Scientic_Review_of_Mental_Health_Practice. pdf Farwell LA (2011b) Brain ngerprinting: comprehensive corrections to Rosenfeld in scientic review of mental health practice. Excalibur Scientic Press, Seattle. http://www.brainwavescie nce.com/Scientic_Review_of_Mental_Health_Practice_Farwell_ Corrections_to_Rosenfeld.pdf Farwell LA (2012) Brain ngerprinting: a tutorial review of laboratory research and eld applications. Evergreen Press, Seattle. Available at: http://www.brainwavescience.com/Far well_Brain_Fingerprinting_Tutorial_Review Farwell LA (2013) Lie detection. In: Saukko P (ed) Encyclopedia of forensic sciences, 2nd edn. Elsevier, Oxford Farwell LA, Donchin E (1986) The brain detector: P300 in the detection of deception. Psychophysiology 23(4):434 Farwell LA, Donchin E (1988a) Talking off the top of your head: toward a mental prosthesis utilizing event-related brain poten- tials. Electroencephalogr Clin Neurophysiol 70:510523 Farwell LA, Donchin E (1988b) Event-related brain potentials in interrogative polygraphy: analysis using bootstrapping. Psycho- physiology 25(4):445 Farwell LA, Donchin E (1991) The truth will out: interrogative polygraphy (lie detection) with event-related potentials. Psychophysiology 28(5):531547. http://www.Brainwavescience. com/Farwell_Donchin_1991_Psychophysiology_Brain_Fingerprin ting_The_Truth_Will_Out.pdf Farwell LA, Farwell GW (1995) Quantum-mechanical processes and consciousness. Bull Am Phys Soc 40(2):956957 Farwell LA, Makeig TH (2005) Farwell brain ngerprinting in the case of Harrington v. State. Open Court X [10]:3, 7-10 Indiana State Bar Assoc. Available at: http://www.brainwavescience. com/Farwell_Brain_Fingerprinting_in_Harrington_v_State Farwell LA, Richardson DC (2006a) Brain ngerprinting in labora- tory conditions. Psychophysiology 43(s1):S37S38 Farwell LA, Richardson DC (2006b) Brain ngerprinting in eld conditions. Psychophysiology 43(s1):S38 Farwell LA, Smith SS (2001) Using brain MERMER testing to detect concealed knowledge despite efforts to conceal. J Forensic Sci 46(1):135143. Available at: http://www.Brainwavescience. com/Farwell_Smith_Journal_of_Forensic_Sciences_Brain_Finger printing.pdf Farwell LA, Martinerie JM, Bashore TR, Rapp PE, Goddard PH (1993) Optimal digital lters for long-latency components of the event-related brain potential. Psychophysiology 30(3):306315 Farwell LA, Richardson DC, Richardson G (2011) Brain ngerprint- ing eld studies comparing P300-MERMER and P300 ERPs in the detection of concealed information. Psychophysiology 48: S95S96 (abstract). Available at: http://www.brainwavescience. com/Farwell_Richardson_Richardson_2011_Brain_Fingerprin ting_Field_Studies_Comparing_P300MERMER_and_P300_in_ the_Detection_of_Concealed_Information.pdf Farwell LA, Richardson DC, Richardson GM (in press) Brain ngerprinting eld studies comparing P300-MERMER and P300 brainwave responses in the detection of concealed information. Evergreen Press, Seattle. http://www.brainwave science.com/Farwell_Richardson_Richardson_Brain_Fingerprin ting_Field_Studies_Comparing_P300-MERMER_and_P300_ Brainwave_Responses_in_the_Detection_of_Concealed_Infor mation Gaillard AKW, Ritter W (1983) Tutorials in event-related potential research: endogenous components. North-Holland, Amsterdam Gamer M, Berti S (2009) Task relevance and recognition of concealed information have different inuences on electrodermal activity and event-related brain potentials. Psychophysiology 47(2): 355364 Grier JB (1971) Non-parametric indexes for sensitivity and bias: computing formulas. Psychol Bull 75:424429 Guntekin B, Basar E (2010) A new interpretation of P300 responses upon analysis of coherences. Cogn Neurodyn 4(2):107118 Hahm J, Ji HK, Jeong JY, Oh DH, Kim SH, Sim KB, Lee JH (2009) Detection of concealed information: combining a virtual mock crime with a P300-based Guilty Knowledge Test. Cyberpsychol Behav 12(3):269275 Halgren E, Stapleton JM, Smith ME, Altafullah I (1986) Generators of the human scalp P3(s). In: Cracco RQ, Bodis-Wollner I (eds) Evoked potentials, frontiers of clinical neuroscience, vol 3. Alan R. Liss, New York Halgren E, Baudena P, Clarke JM, Heit G, Liegois C, Chauvel P et al (1995) Intracerebral potentials to rare target and distractor auditory and visual stimuli. I. Superior temporal plane and parietal lobe. Electroencephalogr Clin Neurophysiol 94:191220 Halgren E, Marinkovic K, Chauvel P (1998) Generators of the late cognitive potentials in auditory and visual oddball tasks. Electroencephalogr Clin Neurophysiol 106:156164 Harrington v. State. Case No. PCCV 073247(Iowa District Court for Pottawattamie County, 5 March 2001) Harrington v. State. 659 N.W.2d 509 (Iowa 2003) Heisenberg W (1958) Physics and philosophy, the revolution in modern science. Harper, New York Hira S, Furumitsu I (2002) Polygraphic examinations in Japan: applications of the guilty knowledge test in forensic investiga- tions. Int J Police Sci Manag 4:1627 Huster RJ, Pantev C, Konrad C, Westerhausen R (2010) The role of the cingulate cortex as neural generator of the N200 and P300 in a tactile response inhibition task. Hum Brain Mapp 31(8):12601271 Iacono WG (2007) Detection of deception. In: Cacioppo J, Tassinary L, Berntson G (eds) Handbook of psychophysiology. Cambridge University Press, New York, pp 688703 Iacono WG (2008) The forensic application of Brain Fingerprint- ing: why scientists should encourage the use of P300 memory detection methods. Am J Bioethics 8(1):3032 Iacono WG, Lykken DT (1997) The validity of the lie detector: two surveys of scientic opinion. J Appl Psychol 82:426433 Iacono WG, Patrick CJ (2006) Polygraph (lie detector) testing: current status and emerging trends. In: Weiner IB, Hess AK (eds) The handbook of forensic psychology. Wiley, New York, pp 552588 Johnson R (1988) The amplitude of the P300 component of the event- related potential: review and synthesis. In: Ackles P, Jennings JR, Coles MGH (eds) Advances in psychophysiology: a research annual, vol 3. JAI Press, Greenwich, pp 69137 Johnson RJ (1989) Auditory and visual P300s in temporal lobectomy patients: evidence for modality-dependent generators. Psycho- physiology 26(6):633650 Johnson RJ (1993) On the neural generators of the P300 component of the event-related potential. Psychophysiology 30(1):9097 152 Cogn Neurodyn (2012) 6:115154 1 3 Johnson MM, Rosenfeld JP (1992) Oddball-evoked P300-based method of deception detection in the laboratory II: utilization of non-selective activation of relevant knowledge. Int J Psycho- physiol 12(3):289306 Kiss I, Dashieff RM, Lordeon P (1989) A parieto-occipital generator for P300: evidence from human intracranial recordings. Int J Neurosci 49:12 Kubo K, Nittono H (2009) The role of intention to conceal in the P300-based concealed information test. Appl Psychophysiol Biofeedback 34(3):227235 Leaf v. Goodyear Tire & Rubber Co., 590 N.W.2d 525, 533 (Iowa 1999) Lefebvre CD, Marchand Y, Smith SM, Connolly JF (2007) Deter- mining eyewitness identication accuracy using event-related brain potentials (ERPs). Psychophysiology 44(6):894904 Lefebvre CD, Marchand Y, Smith SM, Connolly JF (2009) Use of event-related brain potentials (ERPs) to assess eyewitness accuracy and deception. Int J Psychophysiol 73(3):218225 Li Y, Hu Y, Liu T, Wu D (2011) Dipole source analysis of auditory P300 response in depressive and anxiety disorders. Cogn Neurodyn 5(2):221229 Linden D (2005) The P300: where in the brain is it produced and what does it tell us? Neuroscientist 11(6):563576 Long J, Gu Z, Li Y, Yu T, Li F, Fu M (2011) Semi-supervised joint spatio-temporal feature selection for P300-based BCI speller. Cogn Neurodyn 5(4):387398 Lui M, Rosenfeld JP (2008) Detection of deception about multiple, concealed, mock crime items, based on a spatial-temporal analysis of ERP amplitude and scalp distribution. Psychophys- iology 45(5):721730 Lykken DT (1959) The GSR in the detection of guilt. J Appl Psychol 43:385388 Lykken DT (1960) The validity of the guilty knowledge technique: the effects of faking. J Appl Psychol 44:258262 Meegan DV (2008) Neuroimaging techniques for memory detection: scientic, ethical and legal issues. Am J Bioethics 8:920 Meijer EH, Smulders FTY, Merckelbach HLGJ, Wolf AG (2007) The P300 is sensitive to face recognition. Int J Psychophysiol 66(3):231237 Meijer EH, Smulders FTY, Wolf A (2009) The contribution of mere recognition to the P300 effect in a concealed information test. Appl Psychophysiol Biofeedback 34(3):221226 Meixner JB, Rosenfeld PJ (2010) Countermeasure mechanisms in a P300-based concealed information test. Psychophysiology 47(1):5765 Meixner JB, Rosenfeld PJ (in press) A mock terrorism application of the P300-based concealed information test. Psychophysiology Meixner JB, Haynes A, Winograd MR, Brown J, Rosenfeld PJ (2009) Assigned versus random, countermeasure-like responses in the p300 based complex trial protocol for detection of deception: task demand effects. Appl Psychophysiol Biofeedback 34(3): 209220 Mertens R, Allen JJB (2008) The role of psychophysiology in forensic assessments: deception detection, ERPs, and virtual reality mock crime scenarios. Psychophysiology 45(2):286 298 Mertens R, Allen J, Culp N, Crawford L (2003) The detection of deception using event-related potentials in a highly realistic mock crime scenario. Psychophysiology 40:S60 Miller GA, Bashore TR, Farwell LA, Donchin E (1987) Research in geriatric psychophysiology. In: Schaie KW, Eisdorfer C (eds) Annual review of gerontology and geriatrics, vol 7. Springer, New York, pp 127 Miyake Y, Mizutanti M, Yamahura T (1993) Event related potentials as an indicator of detecting information in eld polygraph examinations. Polygraph 22:131149 Moenssens AA (2002) Brain ngerprintingcan it be used to detect the innocence of persons charged with a crime? UMKC Law Rev 70:891920 Murphy PR, Robertson IH, Balsters JH, OConnell RG (2011) Pupillometry and P3 index the locus coeruleusnoradrenergic arousal function in humans. Psychophysiology 48:15311542 National Research Council (2003) The polygraph and lie detection. National Academies Press, Washington Neshige R, Kuroda Y, Kakigi R, Fujiyama F, Matoba R, Yarita M, Luders H, Shibasaki H (1991) Event-related brain potentials as indicators of visual recognition and detection of criminals by their use. Forensic Sci Int 51(1):95103 Nieuwenhuis S, Aston-Jones G, Cohen JD (2005) Decision making, the P3, and the locus coeruleusnorepinephrine system. Psychol Bull 131(4):510532 Picton TW (1988) Handbook of electroencephalography and clinical neurophysiology: human event-related potentials, vol 3. Else- vier, Amsterdam Pineda JA, Foote SL, Neville HJ (1989) Effects of locus coeruleus lesions on auditory, long-latency, event-related potentials in monkey. J Neurosci 9:8193 Rapp PE, Albano AM, Schmah TI, Farwell LA (1993) Filtered noise can mimic low dimensional chaotic attractors. Phys Rev E 47(4):22892297 Roberts AJ (2007) Everything new is old again: brain ngerprinting and evidentiary analogy. Yale JL Tech 9:234270. Available at: http://www.yjolt.org/les/roberts-9-YJOLT-234.pdf Rosenfeld JP (1995) Alternative views of Bashore and Rapps (1993) alternatives to traditional polygraphy: a critique. Psychol Bull 117(1):159166 Rosenfeld JP (2002) Event-related potentials in the detection of deception, malingering, and false memories. In: Kleiner M (ed) Handbook of polygraph testing. Academic Press, New York, pp 265286 Rosenfeld JP (2005) Brain ngerprinting: a critical analysis. Sci Rev Mental Health Practice 4:2037 Rosenfeld JP, Labkovsky E (in press) New P300-based protocol to detect concealed information: Resistance to mental countermea- sures against only half the irrelevant stimuli and a possible ERP indicator of countermeasures. Psychophysiology Rosenfeld JP, Nasman VT, Whalen R, Cantwell B, Mazzeri L (1987) Late vertex positivity in event-related potentials as a guilty knowledge indicator: a new method of lie detection. Int J Neurosci 34:125129 Rosenfeld JP, Cantwell G, Nasman VT, Wojdac V, Ivanov S, Mazzeri L (1988) A modied, event-related potential-based guilty knowledge test. Int J Neurosci 42:157161 Rosenfeld JP, Angell A, Johnson M, Qian J (1991) An ERP-based, control-question lie detector analog: algorithms for discriminat- ing effects within individuals average waveforms. Psychophys- iology 28:319335 Rosenfeld JP, Soskins M, Bosh G, Ryan A (2004) Simple effective countermeasures to P300-based tests of detection of concealed information. Psychophysiology 41(2):205219 Rosenfeld JP, Biroschak JR, Furedy JJ (2006) P-300-based detection of concealed autobiographical versus incidentally acquired information target and non-target paradigms. Int J Psychophysiol 60(3):251259 Rosenfeld JP, Shue E, Singer E (2007) Single versus multiple probe blocks of P300-based concealed information tests for autobio- graphical versus incidentally learned information. Biol Psychol 74:396404 Rosenfeld JP, Labkovsky E, Lui MA, Winograd M, Vandenboom C, Chedid K (2008) The complex trial protocol (CTP): a new, countermeasure-resistant, accurate P300-based method for detec- tion of concealed information. Psychophysiology 45:906919 Cogn Neurodyn (2012) 6:115154 153 1 3 Rosenfeld JP, Tang M, Meixner JB, Winograd M, Labkovsky E (2009) The effects of asymmetric vs. symmetric probability of targets following probe and irrelevant stimuli in the complex trial protocol for detection of concealed information with P300. Physiol Behav 98(12):1016 Sabeti M, Moradi E, Katebi S (2011) Analysis of neural sources of p300 event-related potential in normal and schizophrenic participants. Adv Exp Med Biol 696:589597 Sasaki M, Hira H, Matsuda T (2002) Effects of a mental counter- measure on the physiological detection of deception using P3. Stud Humanit Sci 42:7384 Slaughter v. State, No. PCD-2004-277 (Okla. Ct. of Crim. App., 16 Apr 2004) Smith ME, Halgren E, Sokolik M, Baudena P, Musolino A, Liegois- Chauvel C et al (1990) The intracranial topography of the P3 event-related potential elicited during auditory oddball. Electro- encephalogr Clin Neurophysiol 76:235248 Sochurkova D, Brazdil M, Jurak P, Rektor I (2006) P3 and ERD/ERS in a visual oddball paradigm: a depth EEG study from the mesial temporal structures. J Psychophysiol 20(1):3239 Soskins M, Rosenfeld JP, Niendam T (2001) The case for peak to peak measurement of P300 recorded at 3 Hz high pass lter settings in detection of deception. Int J Psychophysiol 40:173180 Spencer KM, Dien J, Donchin E (2001) Spatiotemporal analysis of the late ERP responses to deviant stimuli. Psychophysiology 38:343358 Stapleton JM, Halgren E (1987) Endogenous potentials evoked in simple cognitive tasks: depth components and task correlates. Electroencephalogr Clin Neurophysiol 67:4452 Sutton S, Braren M, Zubin J, John ER (1965) Evoked potential correlates of stimulus uncertainty. Science 150:11871188 Verschuere B, Rosenfeld JP, Winograd M, Labkovksy E, Wiersema JR (2009) The role of deception in P300 memory detection. Legal Criminol Psychol 14(2):253262 Verschuere B, Ben-Shakhar G, Meijer E (2011) Memory detection: theory and application of the concealed information test, 1st edn. Cambridge University Press, Cambridge Vrij A (2008) Detecting lies and deceit: pitfalls and opportunities, 2nd edn. Wiley, Chichester Wang C, Ulbert I, Schomer DL, Marinkovic K, Halgren E (2005) Responses of human anterior cingulate cortex microdomains to error detection, conict monitoring, stimulus-response mapping, familiarity, and orienting. J Neurosci 25:604613 Wasserman S, Bockenholt U (1989) Bootstrapping: applications to psychophysiology. Psychophysiology 26:208221 Winograd MR, Rosenfeld P (in press) Mock crime application of the Complex Trial Protocol (CTP) P300-based concealed informa- tion test. Psychophysiology 154 Cogn Neurodyn (2012) 6:115154 1 3