Neuroscience, mental privacy, and the law.

• Author: Shen, Francis X.
• Position: III. Mind Reading with Neuroimaging: What We Can (and Cannot - Privacy, Security, and Human Dignity in the Digital Age

3. MIND READING WITH NEUROIMAGING: WHAT WE CAN (AND CANNOT) DO

   Having established in Part II a working definition of neuroimaging mind reading, the Article now briefly discusses several recent legal applications of such technology. Part III reviews: (A) fMRI-based lie detection; (B) fMRI-based memory detection; (C) EEG-based memory detection; and (D) fMRI-based decoding and reconstruction of visual stimuli.

   1. Lie Detection with fMRI (127)

      Neurons, the cells of greatest interest in the brain and nervous system, need oxygen to live. This oxygen is supplied to them via blood flow. fMRI is premised on the logic that tracking relative blood flow to different parts of the brain will reveal relative oxygen uptake, and thus show which neurons are more active (at a given moment in time). (128) Changes in blood oxygen levels in the brain at different moments during a given experimental task allow for inferences about brain-activation patterns.

      Different protocols have been used in fMRI lie detection, most of which rely on a paradigm known as the "Concealed Information Test" (CIT) (also known as the "Guilty Knowledge Test" (GKT)). (129) This paradigm is different than the Control Question Test typically used by professional polygraphers. (130)

      fMRI lie detection evidence has been proffered in several U.S. cases, has been the topic of much neuroscience research, and has drawn the attention of many commentators. (131) There are a large number of conceptual and technical problems with this approach. Conceptually, one major challenge with neuroimaging lie detection is defining a "lie." (132) In practice, neuroscience lie detection has utilized an "instructed lie" experimental paradigm, (133) in which subjects are told to lie under certain conditions in the experiment. Critics point out that this may limit the inferences we can make about "lying," because an instructed lie in the lab may not involve the same brain activity as a high-stakes lie in real life outside the lab. (134) Additionally, technical issues include general concerns about using fMRI techniques to study higher-order cognitive functions. (135)

      Of particular note here is the "reverse inference" fallacy. The reverse inference fallacy is the idea that just because a particular part of the brain is more active during a certain cognitive state, it does not necessarily follow that whenever that brain area is more active, a person is in that cognitive state. (136) The reverse inference fallacy is acute in the lie detection case, as "it is not lying per se that is being decoded from these brain areas but rather the cognitive and emotional processes that are associated with lying." (137)

      Despite these limitations, two for-profit fMRI-based lie detection companies are now in operation, (138) and both have proffered evidence in criminal trials on behalf of defendants. (139) So far, the evidence has been ruled inadmissible under both the Daubert standard in federal court (140) and the Frye standard in state court. (141) However, the judge overseeing the evidentiary hearing in the federal case suggested that such evidence may one day become admissible:

      [I]n the future, should fMRI-based lie detection undergo further testing, development, and peer review, improve upon standards controlling the technique's operation, and gain acceptance by the scientific community for use in the real world, this methodology may be found to be admissible even if the error rate is not able to be quantified in a real world setting. (142) For purposes of the Fourth Amendment and Fifth Amendment analysis in Part IV, it is important to note that all of these experimental paradigms involve researcher-subject interaction such as requesting a response to a visual stimulus or question. (143) Although fMRI may be used in what is known as "resting state" analyses (in which the subject just lies in the scanner), such resting-state approaches have not been employed in the lie detection context. (144)

   2. Memory Detection with fMRI

      Scientists have also made intriguing progress in detecting memories. Neuroscientists Jesse Rissman and Anthony Wagner were able to use fMRI, combined with an advanced data analysis methodology, (145) to identify with great accuracy the subjective memory states of subjects, such as whether the subject thought he had seen a particular face before. (146) Subjects were initially shown a battery of faces, and then, while in the scanner, were shown both previously seen and new faces. (147) The researchers could tell with great accuracy whether a subject remembered seeing a particular face. (148) Further, "neural signatures associated with subjective memory states were sufficiently consistent across individuals to allow one participant's mnemonic experiences to be decoded using a classifier trained exclusively on brain data from other participants." (149)

      At the same time, the researchers were much more limited in their ability to determine from brain signals alone whether a subject had actually seen the face before--the subject's objective memory state. The distinction between subjective and objective memory states, as the authors noted, has very important legal implications. (150) The law generally is interested in objective memory states, such as whether a witness actually saw the alleged criminal.

      As with fMRI-based lie detection, current memory-detection techniques with fMRI require both subject-researcher interaction (such as pressing a button to indicate when a face is remembered).

   3. Memory Recognition with EEG

      Distinct from the fMRI-based approaches just described are memory-recognition approaches using electroencephalography (EEG). These techniques are not lie detection per se, though they are typically used to improve assessment of an individual's veracity. For instance, if a defendant's alibi is that he was never at the scene of the crime, an EEG memory-recognition test could theoretically help the fact finder or investigator's assessment of the defendant's credibility. (151)

      As discussed earlier, EEG is a method of measuring the electrical activity produced by the brain. (152) Electrodes are placed on the subject's scalp, and electrical activity is recorded. (153) As with fMRI lie detection studies, EEG memory-recognition paradigms use a version of the Concealed Information Test. (154) The logic is that the brain will react differently to a stimulus (such as a photo of a particular aspect of a crime scene) if that person recognizes the stimulus. (155)

      A measurement of electrical activity called the P300 wave specifically is of note. (156) "The P300 is a special ERP [event-related potential] component that results whenever a meaningful piece of information is rarely presented among a random series of more frequently presented, non-meaningful stimuli often of the same category as the meaningful stimulus." (157) The theory is that if a series of objects are shown to a subject, the brain will automatically respond in a different way to items that have been seen before and are thus recognizable. Starting in the 1980s, research confirmed this to be the case. "The P300 would not represent a lie per se but only a recognition of a familiar item of information, the verbal denial of which would then imply deception." (158)

      In a recognition task with EEG, subjects are exposed to three types of stimuli: probes (the stimuli that only the guilty party would know); irrelevant stimuli (the stimuli that have nothing to do with the crime scene); and target stimuli (the stimuli that are related to the crime scene but that everyone knows). (159) The legal system has seen a particular version of this approach--the "brain fingerprinting" approach developed by scientist Lawrence Farwell. (160) Farwell presented his evidence in two cases, (161) but his approach has not been admitted into evidence since, and has been heavily criticized. (162)

      There are many scientific challenges to the brain fingerprinting approach. As some critics have pointed out:

      [T]here is no simple one-to-one relationship between the P300 and memory. Even though information stored in memory may very well cause some events to be identified as distinct and therefore elicit a P300, reducing the P300 to a simple "Aha!" response driven by 'recognition of the relevant information contained in the probes as significant in the context of the crime' is quite at variance with what is known about the P300. (163) Moreover, "laboratory research on brain fingerprinting published in peer-reviewed journals amounts to a single study containing 20 participants." (164)

      Setting aside the scientific shortcomings, and thus its admissibility on the merits, two features of the brain fingerprinting approach are particularly relevant to subsequent legal analysis discussed in the next Part. Like the fMRI studies just reviewed, every brain fingerprinting study conducted to date requires substantial subject-researcher interaction. Here, the researcher instructs the subject to press a button to indicate that an image is recognized. As Farwell writes, "A subject neither lies nor tells the truth during a brain fingerprinting test. He simply observes the stimuli and pushes the buttons as instructed." (165)

      Farwell's claim is that "[b]rain [f]ingerprinting testing has nothing to do with lie detection. Rather, it detects information stored in the brain by measuring brain responses." (166) The critical word is responses, as the testing relies on the researcher's questions and the subject's response. Even if the subject's response was not required via pushing a button, it is difficult to see how the protocol could work without requiring the subject to look at the screen in front of him or her. Farwell is clear that the protocol must "[r]equire an overt behavioral task that requires the subject to recognize and process every stimulus, specifically including the probe stimuli." (167)

      In addition to response during the testing itself, the brain...