Have we got here a failure to communicate? Or the beginning of one?
A new article by Aharoni, Gazzaniga and Kiehl (eg al) titled Neuroprediction of Future Rearrest (www.pnas.org/cgi/doi/10.1073/pnas.1219302110) has been excitedly discussed in the pop neuroscience press (and just the press in general – see here and here and here) as a showing us the way to catch criminals before they commit crime.
Aharoni and co. had prison inmates do a Go-NoGo task (the kind where you must perform some rote activity as quickly as possible until you get a stimulus telling you to stop, and then must suddenly withhold that activity; it’s a test of impulse control, more or less) in the scanner. They then correlated the number of mistakes the subject made (going when they should have nogoed) with ACC activity. The ACC is a brain region involved in inhibiting primed responses (eg, suppressing impulsive or automatic behavior – its part of what “lights up” when you turn down that piece of chocolate cake). Then they released the inmates and waited for four years to see who got re-arrested.
As an analogy, it would be like catching a bucket of fish from a stream, doing a test in which you saw which ones had trouble resisting eating flies off of plastic (harmless) hooks, tagging them, and then releasing them back into the stream. You’d then go fishing every day for four years and at the end count up which fish got caught again. The hypothesis is the ones who couldn’t resist the plastic flies would get into trouble in the stream again too.
The key finding that’s been trumpetted is that prisoners with relatively lower ACC activity were 2x as likely to be rearrested as those who were not. Some of the pop neuroscience articles have described this as “predicting” recidivism, but of course that’s just like saying that smoking predicts lung cancer, or eating a cheeseburger predicts heart attacks.
The reason everyone in the media has gotten so excited is that it the authors have reminded them of Tom Cruise in Minority Report, eg, seem to have on their hands a way to stop crime before it happens. The only difference is that where Spielberg’s take on precrime in that movie was that it was immoral and could be used enethically, the attitude of our popular press (maybe because it didn’t involve Tom Cruise) has been: cool! Of course, we’re quite a ways from Tom Cruise because all of the subjects were already in prison, and with recidivism rates up near 50% last time I checked, there was already a pretty good chance these guys were going to reoffend. So clearly this study does *not* predict first arrest. It only predicts second arrest. The authors offer no theory on this, and I’d be comfortable saying that doing a go-nogo task on high school students never arrested for anything would fail, in any meaningful way, to pick up who was going to end up in jail in later life. There are way, waay, waaaaaaaaay more people with impulse control problems than prison inmates. Most of them just end up eating the marshmallow – and that’s the end of it.
So, here’s my critique.
First of all, I’m glad they did this. There’s no doubt that asking these questions is a good thing. One thing you always have to ask with these studies is “what did the neuroimaging really add?” I was not clear on whether the go-nogo task alone could have predicted recidivism results, or if it was the ACC, controlling for behavioral results, that did it. This is a subtle point lost on some casual readers. Bascially, if the task results (more errors) predicts the recividism, then you don’t need the imaging, and prisons could just do this task on inmates for around five cents each rather than $1000. The only way the imaging is needed (from a policy standpoint) is if the go-nogo tasks does NOT predict a difference in recidivism, but the ACC results DO. Frankly I tried to read the paper a few times to figure out the answer to this question and missed it. This is almost certainly my mistake and if you catch the answer please write in. But never forget that it is often the case in fMRI that the neural results are a red herring because the images are driven by the task, and the task – done cheaply, outside a scanner – is just as good as predicting the behavioral results. In these cases the neuroscience is just a “huh! interesting!” moment that has no autonomous predictive empirical power.
Second, the authors rightly say all the right things about group results not being individual results. But of course release decisions are made one at a time, and so in the end this technology would have to apply to individuals, just as they are whenever sociological data is applied to individuals.
Third, the authors appear somewhat naive in the way they think about how their findings might be used by unsympathetic prison staff (see the Cool Hand Luke skit above. I once toured, in 1991, a juvenile prison in Louisiana with a white warden like the fellow above, and met not a single non-African American inmate over the course of three hours. Not an encouraging experience on any level). And having not only toured but worked in prisons myself for many years (before going to medical school) I feel confident saying prisons tend not to be run by reformers and liberals. Moreover, prisons have an economic incentive to find reasons to keep inmates in. In particular, many prisons in America are privately run, and their boards, making $40,000/year or more per inmate have reasons to keep people in. The inmates are in a sense their product, and the public is the customer; they’re paid to keep stocks high. So when they write that “this pattern of results raises the possibility that brain activity in regions such as the ACC, elicited by a simple experimental task, may lend incremental utility to existing behavioral risk factors in the ability to predict rearrest” one worries about how, exactly, this information might be used.
The naive idea (hopefully someday actualized) is you could brain scan prisoners prior to release to identify their risks – and tell them so they could guard themselves against themselves! – for recommitting. In the same way you might tell someone with – to use a made up example – an alcoholism brain pattern to avoid alcohol, you might tell someone with a criminality brain pattern to avoid bad parts of town. In a sense you’d be doing what doctors always do: giving them a head’s up on what environmental challenges, given their anatomy, they were up against.
But the more sinister possibility is that parole decisions could be based on findings, with subjects with “bad scans” being held in prison longer than those with good scans. Given the easy manipulation of fMRI results, and the huge margins of error, this seems potentially unfair. One workaround would be to only administer the scan after the parole decision is made, and the information given to the parole officer and inmate to help them make better decisions.
Putting policy considerations aside, there was a psychiatric howler – typical of what happens when academics who don’t work with the population in question begin to think about treatment suggestions. The authors note “finally, this work highlights potential neuronal systems that could be targeted for treatment intervention. One plausible hypothesis is that interventions that modulate ACC activity may help to increase cognitive control systems and thereby reduce future recidivism.” This view seems to “boil down” antisocial behavior to an impulse control problem. But of course much antisocial behavior is deliberate, not impulseive, and most impulsive people are not antisocial. Lack of anxiety, empathy, guilt, and waaaaay too much ego-syntonic aggression (“he deserves to die!” “If you leave your car unlocked you deserve to have it coming!” “if you get drunk in a bar you’re asking to have sex” – all phrases I’ve heard loads when working with sociopaths) aren’t in the ACC. The neural signature of antisocial inmates is likely to be far more complex than revealed in this study. You’d need a far more elaborate battery of tests. The way forward is to do more and more of this work, but for the press to remain cautious.
Medicalization of Deviance
Panning back from all the details, this is the beginning of something that I’ve both hoped and feared would happen for a long time. When I worked in prisons myself, doing research on prison inmates (see my “ The Myth of Inmate Lawlessness: The Perceived Contradiction between Self and Other in Inmates’ Support for Criminal Justice Sanctioning Norms“ in Law and Society Review, publication here), I hoped that someday society could medicalize antisocial behavior. My personal prediction, back in 1990, was that prison inmates would have far less anxiety than normals. In the variant tested here, they were hypothesized to have less impulse control. This would deprive them of the ability to contextualize the crime they are about to commit, decide that all in all it’s not worth it, and stop themselves. I do think that going down this medicalization road will, in the long run, lead to better public policy. But there are clearly risks. Medicalization is a slippery slope. First of all it’s often “wrong.” Abandoning an ancient moral principle – “don’t steal” – for a newfangled concept about ACC dysfunction could turn out to be foolhardy when later research shows that the PCC, or some other part of the brain, explains the problem even better. Second of all, it tends to ignore the social forces at play. There are a lot of rich white women with bad impulse control, and who couldn’t pass a go-nogo task if you paid them, but they aren’t in prison. Their brain scans would make it look like they should be. We wouldn’t want to medicalize crime as a way of disguising or obscuring hidden racial and socioeconomic forces at play in incarceration decisions. And finally, medicalization potentially removes the moral element from thinking about crime. There is an ancient tension between “is versus ought” in philosophy, and science’s ability to explain is not to be confused with morality’s mandate to judge. We’ll want to make sure – as the authors seem keen to – that our moral thinking about this work keeps pace with the scientific studies.
Still, it is my hope that given that being antisocial is such a problem for both the antisocial person and the rest of society, that learning more about antisociality, and giving society tools to manage it – albeit in an ethical framework – is a worthy goal. I know that many of my antisocial patients (and I have many of them – most of them women, intriguingly, a subpopulation left out of this study) would be relieved, not insulted, to learn that there may be a neural basis for the behavior that, all too often, leads them to come into my office complaining that they appear to be ruining their own lives.