The Tortured Brain

Image via Wikipedia

A new paper (PDF) in Trends in Cognitive Science, “Torturing the Brain,” claims to show that torture is not only evil — it’s based on bad science.

Professor Shane O’Mara, of the Trinity College Institute of Neuroscience in Dublin, Ireland, did not examine the brains of victims of American torture. But he analyzes the likely effects of torture on the brain, based on the vast existing literature on the effects of extreme stress on motivation, mood and memory, using both animals and humans.

He concludes, essentially, that torture is likely to destroy the very memories it’s trying to extract.

O’Mara starts out by trying to describe the “folk” model that underlies the belief that torture is effective in extracting information (basically, what laymen think about how humans work). The American government has never released information on the scientific underpinnings of its program — if there are any — but O’Mara dissects it as such:

From reading the [CIA torture] memos, the neuropsychobiological model seems to be the following: a person possesses information (by definition, this information is in their long-term memory – the enduring personal register of experience, events, and facts that lasts at least for minutes and may extend to decades); they intentionally withhold this information under questioning; applying certain non-verbal techniques … over prolonged periods of time (press and other reports indicate up to six months or more) will facilitate release of this information from long-term memory by the captive. The memos do not fully articulate the mechanisms by which coercion makes captives reveal information they hold in memory. Nevertheless, they seem based on the idea that repeatedly inducing shock, stress, anxiety, disorientation and lack of control is more effective than standard interrogatory techniques in making suspects reveal information. Information retrieved from memory in this way is assumed to be reliable and veridical, as suspects will be motivated to end the interrogation by revealing this information. No supporting data for this model are provided; in fact, the model is utterly unsupported by scientific evidence.

The problem, according to O’Mara, is that the mechanisms of long-term memory are compromised by the effect torture is certain to have on the brain:

There is a vast literature on the effects of extreme stress on motivation, mood and memory, using both animals and humans. To briefly summarise a complex literature: chronic, prolonged and extreme stress: (i) inhibits long-term potentiation  (LTP; the biological process believed to underlie memory formation in the brain) and facilitates long-term depression (the inverse of LTP); (ii) causes hippocampal atrophy and hence impairs learning in humans and animals; (iv) is implicated in many neuropsychiatric disorders (especially anxiety, depression and post- traumatic stress disorder. Notably, repeated, chronic exposure to uncontrollable pain (e.g. electric shocks) causes many similar effects to those found under severe but non-painful stress.

…

Extreme stress studies in Special Operations Soldiers have found impaired visuo-spatial capacity and impaired recall of previously-learned information in stressed soldiers (who undergo stress, including food and sleep deprivation, during training modelled on the experiences of American prisoners-of-war).

What’s more, false memories are likely to take root under torture:

It is difficult or impossible to determine during interrogation whether the information a suspect reveals is true: information presented by the captor to elicit responses during interrogation may inadvertently become part of the suspect’s memory, especially since suspects are under extreme stress and are required to tell and retell the same events which may have happened over a period of years. Other factors exacerbate this problem. Confabulation, the pathological production of false memories, is a common consequence of frontal lobe disorders and, as already noted, prolonged and extreme stress has a deleterious effect on frontal lobe function. Thus, distinguishing between confabulations and what is true in the verbal statements of tortured suspects will be very difficult.

And these false memories will serve not just the psychological needs of the captive, but of the captor as well:

In a torture situation, the captor and the captive have different motivations. The captor wants the captive to speak and reveal key information from long-term memory. The captive wants to escape the extreme stress while not revealing key information. In classical conditioning, circumstances signaling escape from stressful or noxious/aversive events are known as conditioned safety signals. Here, the detainee’s own words provide the safety signal: ‘while I’m talking, I’m not being waterboarded’. The truth of what the detainee says does not provide the safety signal, just the fact that s/he is talking. In other words, speech acts signal periods of safety. Equally, when the captive is talking, the captor’s objective has been obtained. Finally, and presumably, subjecting a fellow human being to torture is very stressful for all but the most psychopathic. In fact, the historical literature is replete with accounts of alcohol or drug abuse by torturers. Thus, the fact that the captive is speaking also provides a safety signal to the captor; making the captive talk may become the end (not the truth of what the captive is revealing)—so long as the captive is talking the captor can avoid using torture.

That is, torturing someone is unpleasant (for a non-psychopath), so both torturer and torture-ee are happy just so long as the captive is talking.

The caveats? I’m sure plenty of readers will be skeptical of this study — and I am, too.

1) As we’ve discussed before, social science has a clear anti-torture bias. You may think that’s understandable, even fantastic. But it’s not a recipe for objective science.

2) How useful are these models of long-term memory to the kind of information one might want to obtain (via torture) from a terror suspect? Say we want to know who someone’s contact was in Germany during the planning of an attack. Or we want to know which subway line was the target. Or we want to know who provided the explosives. Are the negative effects of stress on long-term memory really going to wipe out this information? It’s possible, but I’m not confident lab experiments on stress and long-term memory really shed enough light on such real-world situations.

Those caveats laid out, I do think the incentives to give (and listen to) false information are very high — and the possibility of a captive beginning to believe his own BS is also very significant, given what we know about the malleability of memory.

I’m on record as saying that “torture doesn’t work” is a terrible political argument. But, from a policy perspective, we have to evaluate the science.

UPDATE (9/23/09): After conducting an Internet-based gender test, corrected gender pronouns related to Mr. O’Mara.