Imaging for truth in Munchausen's syndrome by proxy

Spence et al. wanted to test the use of fMRI for lie detection. In order to do so, they found a subject who had been convicted for child molestation because she has Munchausen's syndrome by proxy.

There are two important parts of background to this piece.

First, Munchausen's syndrome by proxy (MBP) is a disease (or syndrome or whatever you want to call it) where the individual -- nearly always a mother -- deliberately does things to their child in order to get attention from medical professionals. This contrasts with plain old (if anything in this can be called plain old) Munchausen's where the individual fakes illnesses in themselves to get attention. MBP is one of the only things in psychiatry that makes me shudder. These people will do horrible things to their children like injecting them with feces to make them ill. Then when they get in the hospital, it all becomes about them not their child. They court nurses and doctors and often try and play individuals off against one another. Stories of this, fortunately rare, syndrome destroying the dynamics of entire hospital wards abound. (For more information, this site is pretty good. Also, Robert Sapolsky wrote an essay on MBP in his book Monkeyluv, and he is always eminently readable.)

Second, there has been a lot of interest in recent years about using imaging technology, particularly functional MRI, as a tool for lie detection. Some parts of the brain (including the ventrolateral prefrontal cortex and the anterior cingulate) are more active when an individual is lying suggesting that knowingly lying is an active process. Some researchers believe that we should be able to reliably test for this activation. (There is a lot of controversy over this. It has been sort of hyped in the media, but there is a huge question of accuracy in these studies. This technology is by no means ready to be used diagnostically much less in criminal proceedings.)

So, hey, it's Monday. Why not take these two strange and interesting tags and mix them. I'd like my science shaken not stirred, please.

Spence et al. took an individual convicted for child molestation because she has MBP and scanned her brain. The individual, X, denies that she committed the crimes for which she has been imprisoned. The paper describes X:

X is a 42-year-old married woman who was convicted of a crime against a child in her care (4 years prior to the study). Found guilty by a majority verdict (in another jurisdiction) she served a longer prison term than she might have as she continued to profess her innocence. Her family and friends believed her and there is currently [2007] a campaign to clear her name. Medical opinion obtained by the court initially described a psychologically normal subject [pre-verdict], then features of MSBP [post-verdict]. The grounds for MSBP comprised X's continued denial of guilt. X is right-handed, with a projected verbal IQ of 109 [22]. Prior to her conviction she had no history of neuropsychiatric, forensic or substance misuse disorders, exhibiting a stable work and family background (remarked upon in the judge's summation of her case). The child concerned had an undisputed, though complex, congenital condition (precise diagnosis unknown), necessitating repeated hospital admissions throughout early life. An older child in X's care had never been a focus of concern.

Let's set aside for the moment the issue of whether it is ethical to scan someone who appears to be so cleanly separated from reality. (The authors explained the procedure to the subject and her family. She consented presumably with the intention of clearing her name.)

The authors want to address the following experiment question: if we ask her a question about her crimes what will the activity in her brain look like? When she professes innocence will the brain scan suggest that she is lying? This is a much more interesting subject in the literature of lie detection because most of the other studies are asking people to lie about stuff that they don't care about. Usually the subjects are university students recruited for a study; they have little to lose if their lies are found out. This subject is much better because she is really invested in proving that she is right.

What should we expect from the activation in her brain? There are several scenarios. The question boils down to whether you think she knows that she committed the crimes. If she committed the crimes but for some reason she does not remember or has convinced herself that she did not do them, then we would expect the activity in her brain to correspond to someone telling the truth. If, on the other hand, she committed the crimes and knows she did it, then we would expect the activation in her brain to suggest that she is lying -- assuming fMRI lie detection even works. Or finally, you could take her brain suggesting she is innocent as evidence for a miscarriage of justice.

The authors put her in the scanner, and they use a protocol that I think should be pretty robust. Basically what they ask her to do is both deny and affirm her previous statements about whether she committed the crimes. Then they compare the activation in those two conditions. The idea is that you get a baseline for lying in both cases because she either believes the affirmation or the denial is true. The two conditions should divide into an activated (lying) and non-activated (telling the truth) state.

Here is a snapshot of the data:

The left shows the relative activation when X affirms her version of the story. The right shows the relative activation when X denies her version of the story. The red indicates areas of increased activation. (For those who know about brain anatomy, they include ventrolateral prefrontal cortex and the anterior cingulate.)

The data suggests that X believes that she is telling the truth.

(Some brief experimental details for the interested. 1) The subject was on the antidepressant venlafaxine and gabapentin for peripheral neuralgia. The authors argue, however, that since you are performing a within subject comparison these drugs should not change the data. 2) The variance for activation in the lying situation -- where she denies her story -- was much higher. 3) The authors confirm that the subject was not moving around in order to game the test.)

Now for a moment, let's assume that this technology does in fact work. (I still remain highly skeptical about that.) Further, MBP patients are accomplished liars, and we cannot exclude the possibility that she has gotten so good at lying that she can fool the scanner. Also, let's set aside the possibility that over repeated tellings, X has convinced herself that something that she once knew was a lie is a truth.

With those three large caveats, the philosophical implications of this study are pretty staggering.

Is this patient innocent?

Well, yes, in a sense they are innocent because they don't even know that they committed the crime. They fulfill the criterion for criminal insanity as I understand it. On the other hand, she was convicted of the crime, and though I have no access to those criminal proceedings I assume they had some physical evidence with which to convict her. This patient appears to be cognitively innocent while being guilty in reality. This study is significant because it could in theory allow us to determine whether someone was insane. If we ask them whether they remember the crime and they can't, well, then they have fulfilled one criterion for been innocent by reason of insanity. This is why I think this paper is really cool.

That being said, we still need to be highly skeptical of fMRI lie detection. It is still highly unclear whether an individual can consciously modify the activity in these brain areas. Further, this subject is about the most compelling example of a hostile witness that I can think of. She has no doubt attempted to game the system before and will again. If she could find a way to beat the fMRI, she no doubt would. I guess that is why this is so interesting: because you are working with a subject that is really, really invested in malingering.

This study also reveals one of the limits of fMRI lie detection. Testing the veracity of a statement with a subject in the scanner cannot be accomplished by someone who doesn't know the answer. If, for whatever reason, the subject is unaware of what you want them to confirm then this test is pretty useless.

Finally, there is the very real question about what to do with a subject like this. She clearly cannot be allowed into society. She has committed horrible crimes, and she may well do so again. However, should she be put in a normal prison setting? My gut instinct says no. Psychiatric hospitals are where we put people who do horrible things but don't understand that they did them. We don't expect to ever let them out, but we do recognize a difference in the nature of the crime than those of a willful murderer.