Columbia Law School Columbia Law School

Scholarship Archive Scholarship Archive

Faculty Scholarship Faculty Publications

2016

fMRI and Lie Detection fMRI and Lie Detection

Anthony D. Wagner awagner@stanford.edu

Richard J. Bonnie rjb6f@virginia.edu

BJ Casey bjc2002@med.cornell.edu

Andre Davis tonya_allen@ca4.uscourts.gov

David L. Faigman

See next page for additional authors

Follow this and additional works at: https://scholarship.law.columbia.edu/faculty_scholarship

Part of the Criminal Law Commons, and the Science and Technology Law Commons

Recommended Citation Recommended Citation Anthony D. Wagner, Richard J. Bonnie, BJ Casey, Andre Davis, David L. Faigman, Morris B. Hoffman, Owen D. Jones, Read Montague, Stephen J. Morse, Marcus E. Raichle, Jennifer A. Richeson, Elizabeth S. Scott, Laurence Steinberg, Kim Taylor-Thompson & Gideon Yaffe, fMRI and Lie Detection, MACARTHUR FOUNDATION RESEARCH NETWORK ON LAW & NEUROSCIENCE, FEBRUARY 2016; VANDERBILT LAW RESEARCH PAPER NO. 17-10 (2016). Available at: https://scholarship.law.columbia.edu/faculty_scholarship/2015

This Working Paper is brought to you for free and open access by the Faculty Publications at Scholarship Archive. It has been accepted for inclusion in Faculty Scholarship by an authorized administrator of Scholarship Archive. For more information, please contact scholarshiparchive@law.columbia.edu.

Authors Authors Anthony D. Wagner, Richard J. Bonnie, BJ Casey, Andre Davis, David L. Faigman, Morris B. Hoffman, Owen D. Jones, Read Montague, Stephen J. Morse, Marcus E. Raichle, Jennifer A. Richeson, Elizabeth S. Scott, Laurence Steinberg, Kim Taylor-Thompson, and Gideon Yaffe

This working paper is available at Scholarship Archive: https://scholarship.law.columbia.edu/faculty_scholarship/2015

Electronic copy available at: https://ssrn.com/abstract=2881586

Attached hereto:

fMRI and Lie Detection A Knowledge Brief of the MacArthur Foundation Research Network on Law and Neuroscience

Cite as:

Anthony Wagner, Richard J. Bonnie, BJ Casey, Andre Davis, David L. Faigman, Morris B. Hoffman, Owen D. Jones, Read Montague, Stephen J. Morse, Marcus E. Raichle, Jennifer A. Richeson, Elizabeth S. Scott, Laurence Steinberg, Kim Taylor-Thompson, Gideon Yaffe, fMRI and Lie Detection: A Knowledge Brief of the MacArthur Foundation Research Network on Law and Neuroscience (2016).

 

Owen D. Jones Director, MacArthur Foundation Research Network on Law and Neuroscience

New York Alumni Chancellor's Chair in Law & Professor of Biological Sciences, Vanderbilt University

Richard J. Bonnie Harrison Foundation Professor of Medicine and Law Professor of Psychiatry and Neurobehavioral Sciences Director, Institute of Law, Psychiatry and Public Policy Professor of Public Policy Frank Batten School of Leadership and Public Policy University of Virginia  

Stephen J. Morse Ferdinand Wakeman Hubbell Professor of Law Professor of Psychology and Law in Psychiatry University of Pennsylvania Marcus E. Raichle Professor of Radiology, Neurology, Neurobiology, Biomedical Engineering, and Psychology, Washington University  

BJ Casey Professor of Psychology, Yale University Adjunct Professor, Weill Cornell Medical College, New York City Director, Fundamentals of the Adolescent Brain Lab  Andre Davis Judge, United States Court of Appeals for the Fourth Circuit

Jennifer A. Richeson Philip R. Allen Professor of Psychology, Yale University Director, Social Perception and Communication Lab Elizabeth S Scott Harold R. Medina Professor of Law, Columbia University  

David L. Faigman John F. Digardi Distinguished Professor of Law Director of the UCSF/UC Hastings Consortium on Law Science & Health Policy University of California, Hastings  

Laurence Steinberg Distinguished University Professor Laura H. Carnell Professor of Psychology, Temple University Kim A. Taylor-Thompson Professor of Clinical Law, New York University 

Morris B. Hoffman District Judge, Second Judicial District, State of Colorado  Read Montague Professor of Physics Professor of Psychiatry and Behavioral Medicine Director, Human Neuroimaging Laboratory Director, Computational Psychiatry Unit, Virginia Tech & University College, London 

Anthony Wagner Professor of Psychology Director, Stanford Memory Lab Associate Director of the Cognitive & Neurobiological Imaging Center Stanford University  Gideon Yaffe Professor of Law Professor of Philosophy & Psychology, Yale University

 The MacArthur Foundation Research Network on Law and Neuroscience

Vanderbilt Law School, 131 21st Avenue South, Nashville, TN 37203

Electronic copy available at: https://ssrn.com/abstract=2881586 The MacArthur Foundation Research Network on Law and Neuroscience / fMRI AND LIE DETECTION / 02.23.16 1

Law

Neuro+

THE MACARTHUR FOUNDATION RESEARCH NETWORK ON LAW AND NEUROSCIENCE

WHAT IS fMRI?For more than a decade now, scientists have been explor-ing the potential of functional magnetic resonance imaging, or fMRI, to assess increased activity in brain regions asso-ciated with the cognitive processes required for lying.

fMRI does not measure neural activity directly. Instead, it measures small and variable changes in the ratio of oxygenated to deoxygenated blood in the brain when a particular task is performed or stimulus presented—the so-called BOLD, or blood oxygen level-dependent, re-sponse. Firing neurons, like working muscles, require oxygen; follow the trail of oxygenated hemoglobin, and you find neural activity.

LIES, DAMNED LIES, AND BEING COOPERATIVEThe most fundamental question scientists raise when reviewing fMRI lie detection research is this: Do these experiments actually examine lies?

The typical experimental paradigm involves “instructed” lies: a subject is given detailed instructions about how and when to lie, then placed in a scanner. Does conscien-tiously following those instructions constitute lying? Many researchers worry that the answer is no, rendering the experimental results irrelevant.

A distinct but related question arises from the poorly defined nature of the real-world lie. Two equally false statements—“Of course I remember you” and “No, I didn’t kill him”—may be as distinct neurally as they are morally. Similarly, an often-repeated lie or one first told many years ago might look markedly different from an unpracticed or recent lie.

A statement based on faulty memory (“I never said that”) may not trigger any neural activity associated with decep-tion at all. There is some evidence to suggest that fMRI scanning will detect the subject’s belief, even if that belief

fMRI AND LIE DETECTION

In September 2012, the Sixth Circuit Court of Appeals, citing Federal Rule of Evidence 702 and Rule 403, agreed with the trial court’s exclusion of fMRI-based lie detection evidence in the fraud case of United States v Semrau.

A scant month earlier, Judge Eric M. Johnson of the Maryland Sixth Judicial Circuit, Montgomery County, had refused to admit potentially exculpatory fMRI lie detection evidence in the murder trial of State v Gary Smith. Citing the Frye standard, Johnson wrote, “It is clear to the Court that the use of fMRI to detect deception and verify truth in an individual’s brain has not achieved general acceptance in the scientific community.”

While research on fMRI-based lie detection has continued, the general consensus in the scientific community regarding its probative value remains the same. This brief explores why.

The MacArthur Foundation Research Network on Law and Neuroscience / fMRI AND LIE DETECTION / 02.23.16 2

isn’t borne out by the objective truth. In a 2010 memory experiment supported by the Research Network and con-ducted by neuroscientist Jesse Rissman and colleagues, the brain activity observed when subjects recognized a face was comparable to that observed when subjects believed they had seen a face before but hadn’t.

PROBLEMS OF INFERENCEIt is impossible to infer a specific mental process solely on the basis of brain activity in a particular region, or even in a particular set of brain regions. A single brain region is often involved in a number of mental processes, and a mental process often involves multiple areas of the brain.

In 2014, neuroscientist Martha J. Farah and colleagues published a meta-analysis of the fMRI-based lie detection literature to date. Like the meta-analysis performed by neuroscientist Shawn Christ and colleagues in 2009, the study reveals both variability in the particular brain regions activated across experiments and some notable consisten-cy. The regions that consistently showed deception-related activity were the ventrolateral and dorsolateral prefrontal cortex, inferior parietal lobe, anterior insula, and medial superior frontal cortex. Predictably, those regions are activated during other cognitive processes, as well, in particular, those processes that form part of what we call “executive control,” e.g., planning, working memory (the system that provides for temporary storage and manip-ulation of information), inhibitory control (the ability to suppress actions and resist interference from irrelevant stimuli), and attention. Even the instructed lie is cogni-tively complex: a subject must remember a set of circum-stances, attend to stimuli that vary in their significance or salience, decide to lie, suppress the truth, and choose among relevant and plausible details.

CONFOUNDS: DO WE KNOW WHAT WE’RE MEASURING?Even if instructed lies are lies, and there is some com-mon physiological ground shared by all lies, experimental confounds in most of the studies to date make it impos-sible for researchers to know whether the neural activity measured is associated with lying or with something else.

A 2008 experiment by neuroscientist Jonathan Hakun and colleagues, for example, included the following finding: Brain activation was observed whenever the target or “lie” stimulus was presented, independent of whether the

subjects were actually lying about the stimulus at the time. Was the brain activation a result of deception, then, or attention, that is, the salience of the stimulus? This study calls into question many prior published reports that used a similar paradigm, as the brain activity in those studies may not reflect neural responses to deception.

Neuroscientist F. Andrew Kozel and colleagues analyzed data from three independent “mock theft” experiments in which subjects were instructed to look at two objects, select one, take it from a drawer, hide it in a locker con-taining the subject’s personal belongings, and then deny having taken either object. Accuracy rates for those mock theft experiments range from 71 to 90 percent. But when subjects have more and richer memories of one object than another, how much of what’s being detected is deception and how much memory? A subsequent 2012 study by Mathias Gamer and colleagues suggests that memory may be a critical confound in many prior studies.

Variables that can prejudice results aren’t limited to those inadvertently introduced in research studies. Blood flow itself is influenced by a variety of factors independen of neural activity, including age, vascular capacity, and medication. The fMRI results offered in the Semrau case included a confound likely to be unavoidable in civil or criminal applications of the technology: the length of time between the fMRI and the event in question. Relatively little research has been done on how such variables as subject fatigue, anxiety, fear, the presence of a perceived threat, or practice affect fMRI results.

Nature Reviews | Neuroscience

IFG–insula

IPL

MFG

m/SFG

Lateral

Medial

(to our knowledge) report data relevant to detecting deception at the individual-event level. Specifically, Langleben et al.16 and Davatzikos et al.17 focused on whether instructed lie and truth events could be discriminated in the same dataset, using either logistic regression16 or non-linear machine-learning analyses17. Although these event-level analyses showed accuracy rates of 78% and 88%, respectively, the impact of these early findings is limited because of the above-noted response-frequency confound in the experimental design.

Other studies focused on examining whether fMRI BOLD activity differs when an individual is lying compared to telling the truth (pooling data across events). Various statistical approaches have been implemented, including single-subject univariate analy-ses2,7,19,20, univariate analyses combined with the counting of above-threshold voxels in targeted regions of interest8,22,25 and machine-learning classification2. The reported accura-cies in these individual-subject-level analyses have ranged from 69% to 100%, suggesting that these statistical approaches have prom-ise. However, here again, the noted concerns about attention and memory confounds undermine data interpretation7,12,26. Indeed, so long as studies use tasks in which the effects of deception cannot be separated from the effects of attention and memory, the prob-lem of confounds will remain, regardless of

whether the neural correlates of deception are sought using univariate or multivariate analy-ses and regardless of whether the correlates are discovered by simple regression analysis or machine-learning algorithms. Furthermore, even high accuracy rates may decline precipi-tously when subjects use countermeasures in an attempt to conceal their ‘deception’ (REF. 2).

The laboratory studies assessing the accu-racy of fMRI-based lie detection on the indi-vidual-subject level assess the sensitivity and specificity within an individual by differenti-ating trials on which the individual is decep-tive or truthful. However, determining the accuracy of a test in a general population also requires an assessment of the test’s sensitivity and specificity across individuals within that population: that is, what is the likelihood of detecting deception when it is present in a member of the population (sensitivity), and what is the likelihood of correctly indicating when deception is absent (specificity)? To date, fMRI-based lie-detection tests examin-ing accuracy within individuals have gener-ally not assessed the sensitivity or specificity of the test across individuals. One exception is a study by Kozel and colleagues22, which tested participants who had been success-fully classified using the researchers’ method as ‘lying’ on a prior mock crime task (25 out of 36 participants). These pre-selected participants were then examined on a sec-ondary mock crime task. On this secondary

task, some of the participants committed the mock crime and others did not, but all were instructed to indicate that they did not. The authors were able to correctly detect deception, using fMRI, in 100% of the par-ticipants in the ‘mock crime present’ condi-tion. However, they also mistakenly detected deception in 67% of the participants in the ‘mock crime absent’ condition. In the lan-guage of diagnostic testing, the sensitivity of this test was high but the specificity was low.

Determining the real-world accuracy of a detection test also depends on a critical third factor, which is the probability of the event occurring within the population — the base rate. Indeed, the risks associated with a lie-detection test with low specificity will depend on the base rate of lying in the population assessed28,29. Imagine that the test described in REF. 22 was given to 101 people, 100 of them truthful and 1 deceptive. On the basis of the false-positive rates of Kozel and colleagues22, the test would identify 68 participants as ‘lying’ — the 1 participant who lied about the mock crime and 67 who did not. In other words, given a positive result, the probability of the test accurately indicating someone as lying is 1 in 68, or less than 1.5%, and the likelihood of incorrectly indicating deception when it is not present is over 98%. As this example illustrates, even in an ideal circumstance in which a laboratory lie-detection test is developed and used in identical situations, and is sensitive to decep-tion within an individual 100% of the time, its accuracy in a larger population may still be unacceptably low if the specificity of the test is low and the base rate of lying is low.

The real-world validity of fMRI-based lie detection will also depend on the generaliz-ability of the findings obtained in laboratory studies (which typically involve undergradu-ate students — that is, healthy, educated young adult subjects) to the individuals whose veracity is to be assessed by these methods. Consider the differences between criminal offenders, a group that is likely to be subjected to lie-detection methods, and the undergraduate students on which these methods have so far been tested. A relatively high proportion of criminals meet the crite-ria for psychopathy, a condition that is asso-ciated with frequent acts of deception and with alterations in both structural MRI and fMRI studies27. A study of fMRI-based lie detection in criminal offenders with a diag-nosis related to psychopathy — specifically, antisocial personality disorder — found that a large proportion of these participants did not show typical prefrontal BOLD response patterns during instructed deception29.

Figure 1 | Results of the ALE analysis of the functional MRI ‘deception’ literature. Overlay of map of activation likelihood estimation (ALE) values (orange) on the lateral (top) and medial (bottom) inflated PALS surface76, revealing regions that are consistently implicated in deception across studies. The detection thresholded was set at P < 0.05, and the false-discovery rate was corrected as per the method described in REF. 13. IFG, inferior frontal gyrus; IPL, inferior parietal lobule; MFG, middle frontal gyrus; m/SFG, medial and superior frontal gyrus.

P E R S P E C T I V E S

126 | FEBRUARY 2014 | VOLUME 15 www.nature.com/reviews/neuro

© 2014 Macmillan Publishers Limited. All rights reserved

FIGURE 1. Results of the ALE analysis of the functional MRI “deception” literature revealing regions consistently implicated in deception across studies. The meta- analysis was performed over 321 foci from 28 independent statistical contrasts between lie and truth conditions reported in 23 different studies. As noted by others, no region was active in all, or nearly all, studies.

The MacArthur Foundation Research Network on Law and Neuroscience / fMRI AND LIE DETECTION / 02.23.16 3

COUNTERMEASURES: DETECTING A LIAR AND A CHEATAnother serious obstacle to using fMRI for lie detection in the real world is that very little research exists on countermeasures, actions taken to make test results misleading or unusable. Moving during scanning or not following instructions can ruin a test, but they’re also likely to be spotted. More worrisome is whether unnoticeable physical or mental strategies could nonetheless effectively interfere with patterns of neural activity or signal strength.

One study that looked at countermeasures with respect to fMRI lie detection, conducted by Giorgio Ganis and colleagues in 2011, featured prominently in the Gary Smith murder trial. Study participants were instructed to use a series of covert actions, such as imperceptibly moving a left index finger or left toe, just before pressing the response button each time they saw irrelevant dates in a series. In trials without the countermeasure, researchers were able to detect deception with 100 percent accuracy. When the countermeasure was employed, detection accuracy fell to just 33 percent.

A 2015 study by Melina Uncapher and colleagues, this one of memory, showed that participants could successfully conceal or feign memory for faces. Interestingly, and some might say discouragingly, the study showed that both the

magnitude of hippocampal activity—a region long known to be important for memory—and distributed neural patterns could be manipulated by retrieval strategies.

POTENTIAL PROBLEMS OF VALIDITYMany scientists argue that the conclusions drawn from fMRI ”lie detection” experiments conducted to date are only valid within the context of the experimental data.

Group data might not be able to tell us what we need to know about an individual. The holy grail of lie detection is to distinguish truth from lie reliably at the level of the individual subject and at the level of the individual question. But most of the studies conducted on deception to date focus on truth vs. lie differences averaged over multiple subjects and trials.

A sufficient amount of group-averaged data can indicate that a certain pattern of neural activity is frequently associated with a particular experimental condition. However, they cannot tell us whether the pattern of activation is not also common to other experimental conditions (or mental processes). Nor, for the moment, can they shed much light on whether fMRI can reli-ably detect lies at the level of the individual subject or question. In his testimony during the Semrau trial, Cephos Corporation CEO Steven Laken, who conduct-ed the fMRI lie detection tests submitted as evidence,

In contrast to nearly all other studies to date, one fMRI data set shows brain activity during genuine dishonesty—that is, dishonesty related to a freely exercised choice to lie. It was the result of an ingenious experiment published in 2009 by neuroscientists Joshua Greene and Joseph Paxton.

The pair asked participants to predict the outcome of random computerized coin flips while undergoing fMRI. The experiment was presented as an inquiry into paranormal ability to predict the future; the supposed hypothesis was that predictive ability improved when predictions were not made public in advance and were associated with financial gain or loss.

It was a cover story that both encouraged participant honesty (to test the hypothesis adequately required them to tell the truth) and gave them the opportunity to lie (in some trials, they believed they would be self-reporting their success at prediction). In reality, the study was an attempt to deter-mine what makes people behave honestly when they are confronted with an opportuni-ty for dishonest gain.

Throughout a series of “opportunity” (the “opportunity” being to lie) and “no opportunity” trials, participants made their predictions, believing them to be either private or public, depending on the trial. Researchers then classified the participants as honest, dishonest, or ambiguous based on the probability of their self-reported

percentage of wins in the opportunity trials. Subsequent fMRI data analysis revealed that increased activity in the prefrontal cortex—anterior cingulate and dorsolateral and ven-trolateral prefrontal cortices—was associated with the decision to lie in the dishonest group. Interestingly, an even greater increase in prefrontal cortex activity in this group was observed in connection with the decision to refrain from lying. In other words, when individuals who had shown themselves willing to lie passed up the opportunity and instead reported a loss, prefrontal cortical activity was even higher than when they lied. (Whether this increase is due to considering deception, resisting temptation, or something else is currently unknown.) In the honest group, no significant effects were observed when choosing not to lie.

The uninstructed lie

The MacArthur Foundation Research Network on Law and Neuroscience / fMRI AND LIE DETECTION / 02.23.16 4

confirmed that they did not indicate whether Dr. Semrau responded truthfully as to any specific question and that it was “certainly possible” that Dr. Semrau was lying on some of the particularly significant questions.

Experimental conditions often poorly approximate the real world. To date, fMRI studies have focused on detecting lies about an event that just occurred. The event often has no personal relevance and no consequences. Real-world fMRI lie detection focuses on events or facts that are likely to have occurred months or even years before, are deeply relevant to the subject, and have serious consequences. Little is known about whether real-world and experimental conditions yield similar results.

The sensitivity and specificity of fMRI lie detection

have not been established. No diagnostic tool is perfectly accurate. Antiviral software sometimes detects threats that aren’t there; mammograms miss tumors. The probative value of fMRI-based evidence depends on knowing how many lies the tool misses and how often it identifies the truth as a lie; few research studies to date have reported such data.

Findings may not be generalizable to other populations. fMRI studies typically are conducted on undergraduates and other healthy younger adults. Even if we know that there is neural activity in particular regions under the condition of lying when subjects are younger and healthy—a matter of debate, as already discussed—do we know anything at all about what to expect from a woman of 70, or someone with a mental illness?

PRINCIPLED OBJECTIONSAt present, many of the issues that concern the scientific community with respect to the use of fMRI for lie detec-tion are likely to be problematic for the legal community, at least in most contexts. In fact, much of the existing research on deception has no bearing on the question that matters most to judges, lawyers, defendants, and juries, i.e., “Can fMRI-based lie detection methods provide a legally relevant answer to a specific question?”

Most scientists—including many who have reported detecting lies in the laboratory with a high degree of accuracy—agree that more and different research will need to be conducted before fMRI-based lie detection is ready for its day in court.

TO LEARN MORE Brain Imaging for Legal Thinkers: A Guide for the Perplexed Through a Scanner Darkly: Functional Neuroimaging as Evidence of a Criminal Defendant’s Past Mental States Detecting Individual Memories through the Neural Decoding of Memory States and Past Experience Brain Scans as Evidence: Truths, Proofs, Lies, and Lessons Law and Neuroscience in the United States

To download these and other publications exploring the intersection of law and neuroscience, please visit the MacArthur Research Network on Law and Neuroscience, at www.lawneuro.org.

This brief is produced by the MacArthur Research Network on Law and Neuroscience. Supported by the John D. and Catherine T. MacArthur Foundation, the network addresses a focused set of closely-related problems at the intersection of neuroscience and criminal justice: 1) investigating law-relevant mental states of, and decision-making processes in, defendants, witnesses, jurors, and judges; 2) investigating in adolescents the relationship between brain development and cognitive capacities; and 3) assessing how best to draw inferences about individuals from group-based neuroscientific data.

Christ, S. E., Van Essen, D. C., Watson, J. M., Brubaker, L. E., & McDermott, K. B. (2009) The contributions of pre-frontal cortex and executive control to deception: Evidence from activation likelihood estimate meta-analyses. Cerebral Cortex, 19, 1557–66.

Farah, M. J., Hutchinson, J. B., Phelps, E. A., & Wagner, A. D. (2014) Functional MRI-based lie detection: Scientific and societal challenges. Nature Reviews Neuroscience, 15, 123-31.

Gamer, M., Klimecki, O., Bauermann, T., Stoeter, P., & Vossel, G. (2012) fMRI-activation patterns in the detection of concealed information rely on memory-related effects. Social Cognitive and Affective Neuroscience, 7(5), 506–15.

Ganis, G., Rosenfeld, J. P., Meixner, J., Kievit, R. A., Schendan, H. E. (2011) Lying in the scanner: Covert countermeasures disrupt deception detection by functional magnetic resonance imaging. Neuroimage, 55, 312–19.

Greely, H. T., and Illes, J. (2007) Neuroscience-Based Lie Detection: The Urgent Need for Regulation. American Journal of Law and Medicine, 33, 377–431.

Greene, J. D., and Paxton, J. M. (2009) Patterns of neural activity associated with honest and dishonest moral decisions. Proceedings of the National Academy of Sciences, 106:30, 12506–12511.

Hakun, J. G., Seelig, D., Ruparel, K., Loughead, J. W., Busch, E., Gur, R. C., & Langleben, D. D. (2008) fMRI investigation of the cognitive structure of the concealed information test. Neurocase, 14, 59–67.

Kozel, F. A., Johnson, K. A., Mu, Q., Grenesko, E. L., Laken, S. J., & George, M.S. (2005) Detecting deception using functional magnetic resonance imaging. Biological Psychiatry, 58:8, 605–613.

Rissman, J., Greely, H. T., & Wagner, A. D. (2010) Detecting individual memories through the neural decoding of memory states and past experience. Proceedings of the National Academy of Sciences, 107:21, 9849–9854.

Uncapher, M. R., Boyd-Meredith, J. T., Chow, T. E., Rissman, J., & Wagner, A. D. (2015) Goal-directed modulation of neural memory patterns: Implications for fMRI-based memory detection. Journal of Neuroscience, 35(22), 8531-45.

REFERENCES