Problems with fMRI as a tool of lie detection

by Zachary Shapiro

Functional magnetic resonance imaging (fMRI) evidence of lie detection has, appropriately, faced difficulty gaining evidentiary acceptance in criminal courts. While a comprehensive discussion of the case law is beyond the scope of this post, it is important to note that courts have repeatedly refused to admit such evidence, both under a Daubert test, using Federal Rule of Evidence (FRE) 702, as well as under FRE 403.

Under Daubert, which governs the admissibility of expert testimony, courts have found that fMRI lie detection falls short in meeting the necessary standards, including the identification of error rates and maintenance of uniform testing standards. Courts have also pointed out that the motivation to lie may be different in research v. real-world settings.[1] In a laboratory experiment, one can assume that the participant is complying with investigator directions. However, if the scan is to be used in the courtroom, the subject will have a personal interest in the outcome, and may try to employ counter measures, or disregard instructions, in order to “fool” the scanner. Recent research shows that this task may not be hard, at least not for those who know how to effectively “trick” the scanner.

Judges have highlighted that while there are peer-reviewed studies of fMRI lie detection, said studies have very small patient bases (all N<60), and included a range of participants who were not representative of the general population. Courts recognize that neuroimaging, for the purposes of lie detection, is still not generally accepted by the scientific community.[2] Both of these factors limit the applicability of the results to the general population, and to any individual defendant in particular.

Courts have also held, pursuant to FRE 403, that any probative value related to the fMRI lie detection results would be outweighed by the potential for unfair prejudice of the jury.[3] Courts have analogized this effect to polygraph cases, pointing out that lie detection evidence has been found to be highly prejudicial. There is a great concern that neuroimaging, with its veneer of legitimate science, will be even more influential on juror decisions.[4] This potential for prejudice is even more worrisome given the questionable probative value of the testing, due to the nascent science of the field, and the issues inherently related to fMRI experiments (discussed in earlier blog posts).

Finally, there is a concern that deception is far from a homogeneous behavior, and is still poorly understood by neuroscience. Indeed, there are many types, gradations, and motivations for not telling the absolute truth.[5] It is possible that different types of lying have different neural correlates, which could confound anybody trying to generalize fMRI results to create a specific image of what a “lying” brain looks like.[6] It is also possible that the level of stress, degree of rehearsal, and other factors could influence neuroimaging techniques in ways we do not yet fully understand.[7]

Taken together, these concerns highlight that fMRI-based evidence of truth detection should not yet be accepted in criminal courts. Clearly, the science must advance well past the current state of knowledge, so that the evidence can withstand Daubert challenges, while offering enough probative value to pass FRE 403 challenges. It is only through developing greater scientific understanding and further standardizing of testing methodology, that one day, fMRI based evidence of truth detection may withstand evidentiary challenges, and hopefully provide illumination of truth telling in a criminal context.

[1] Wilson v. Corestaff Services L.P., 900 N.Y.S.2d 639 (N.Y. Sup. Ct. 2010); U.S. v. Semrau, 2010 WL 6845092 at 11-12 (W.D.Tenn.,2010)

[2] Id. at 13.

[3] Id. at 13-16.

[4] Id. at 11.

[5] Bok, S. Lying: Moral Choice in Public and Private Life (Random House Digital, 2011); Barnes, J. A. A Pack of Lies: Towards a Sociology of Lying (Cambridge Univ. Press, 1994).

[6] For example, it is unknown whether different types of lying, such as lying for personal gain, lying to spare the feelings of another, lies that we tell to ourselves, lies we consider trivial, and lies of omission, to name but a few, have the same type of neural correlates.

[7]Kozel, F.A., Padgett, T.M. & George, M.S. (2004). A Replication Study of the Neural Correlates of Deception. Behavioral Neuroscience, Vol. 118, No. 4, pp. 852-856.