Sunday, June 14, 2009

Seed - Bad Memory (Eye Witness Reports)

Interesting article. Nothing new here for people in the field, but a nice article for those not aware of the research on memory and its' unreliability.

Bad Memories

Frontier / by Molika Ashford / June 13, 2009

Eyewitness testimony is both fallible and irreplaceable. How can we know when to trust it?

This February, 10 years after his death in prison, Timothy Cole was posthumously exonerated for a rape he did not commit. Before his trial, a victim picked him out of a series of photographs, but her memory may have been skewed by the fact that his image was the only one in color. Cole’s case is not an isolated one. The Innocence Project, a legal advocacy group that worked on his behalf, has cleared the names of more than 175 people who were wrongly convicted due to the unreliability of human memory.

Psychological research continues to undermine the trust given to eyewitnesses’ ability to accurately remember the details of a crime, and we’re becoming increasingly aware of how often their memories are unconsciously manipulated. Paired with a growing interest in the field of neurolaw, which examines the intersection of neuroscience and legal systems, the desire for tools that can objectively assess the accuracy of memories is palpable. But is it possible?

Recently, the stakes for answering that question have been raised. Last fall in the Indian city of Pune, a woman was convicted of murder on the basis of a brain scan that purported to show that she remembered putting arsenic in her husband’s food. This controversial case piqued the interest of Hank Greely, a law professor at Stanford University who specializes in the legal implications of neuroscience. “We want to figure out how plausible it actually is that you could tell if someone has a memory or not,” he says. “It’s certainly plausible enough to explore it.”

Greely is exploring that question and many others as the codirector of Stanford University’s Law and Neuroscience Project. Awarded a $10 million MacArthur Foundation grant in late 2007, this group of about 40 neuroscientists, lawyers, and philosophers is making the first concerted effort to examine neuroscience’s impact on the law. The team will convert its research into guides for judges and law schools, detailing what current neuroscience technologies—particularly functional magnetic resonance imaging (fMRI)—can and cannot reveal about the information in a person’s mind.

One reason that such technologies are so alluring is that people tend to be poor judges of the quality of their memories. The circumstances in which eyewitnesses find themselves might render their recollections particularly untrustworthy, says Elizabeth Loftus, a professor of psychology at the University of California and a pioneer in the field of eyewitness research. “When it comes to brief episodes of memory, like in many criminal cases, poor lighting, passage of time, biased or suggestive questioning all can produce an erroneous memory,” she says.

Lisa Hasel, a psychology researcher at Iowa State University, recently added to this list of confounders in a paper published in the January issue of Psychological Science. In Hasel’s experiments, hearing that a specific suspect has confessed significantly increased the likelihood that eyewitnesses would choose that person out of a lineup, even if they were initially confident that another person had committed the staged crime they saw. Most troubling, however, was the fact that none of the people in the lineup had committed the crime. “Without outside evidence, a person really can’t tell if their memory is real or false,” says Hasel.

Mistakenly remembering a stranger’s face was what led to the conviction of Cole and so many others; Anthony Wagner, a neuroscientist at Stanford and a member of its neurolaw project, is now using fMRI to examine that phenomenon. Wagner shows subjects two series of faces while they are in a scanner, then compares the patterns of activation in the subjects’ brains as they answer questions about which faces from the second group they remember from the first. By establishing a baseline through multiple trials, a computer program can tell what “remembering” looks like for a particular subject.

However, establishing that baseline may prove more difficult in real-life scenarios, especially if subjects aren’t willing to fully cooperate. “There’s no hard and fast criteria for determining if a memory is real, as the parts of the brain that deal with visual perception and mental imagery have a lot of overlap,” says Ken Norman, a professor of psychology at Princeton University who uses the same memory assessment technique as Wagner. If someone really remembers an event, or just remembers picturing it—after hearing about it from police, say—it could look much the same to the program.

Even though Wagner and Norman have had promising results, neither thinks this technique will produce admissible evidence anytime soon. “We can tell surprisingly well what people are thinking from brain scans,” says Norman. “But if you’re going to convict someone, ‘better than chance’ is not good enough.” Greely also believes that neuroscientific information may be unduly persuasive to people. Indeed, neuroscientists are no exception, judging by the controversy swelling in their ranks over a recent paper that alleges the preponderance of untenable “voodoo correlations” made by fMRI researchers.

Considering these limitations, fMRI might be better used to determine how—rather than what— a person remembers. “Imagine you’re trying to remember what you wrote on your shopping list,” says Norman. “You can try to put yourself back in the mind-set of when you wrote the list, or you can try to think of every vegetable you know until you hit on one you remember. The problem with the second method is that there’s a chance you’re going to come home with some vegetables that weren’t on your list,” he says. Buying a tomato instead of a potato may be no big deal, but this kind of constructive remembering, called “semantic generation,” could lead a witness to remember something that never happened.

Norman’s experiments have been able to distinguish between different recollection methods to the program; “mind-set” memory patterns look more like a person experiencing the event for the first time. If we can better tell when a witness is relying on semantic generation, they could be coached into the mind-set method, or potentially have their testimony weighed differently. Helping judges and juries improve such evaluations is an important part of the Stanford neurolaw project, says Greely. This critical eye could be extended toward the law enforcement practices that happen before a trial begins. Psychological studies have shown the benefits of sequential double-blind lineups—by evaluating each person individually, witnesses are less likely to settle on an incorrect proxy—but few places have adopted them. Showing neurological mechanisms behind the improved accuracy could help tip the balance.

Though they won’t take the number of wrongful convictions down to zero, those kinds of reforms would be a start. Eyewitness testimony isn’t going anywhere, no matter how clear its flaws become in the bore of an fMRI machine. And while brain scans may not be ready to fill in those holes yet, they can help tell us how to avoid falling into them.


2 comments:

Anonymous said...

The Law and Neuroscience Project is based at the University of California, Santa Barbara; not Stanford.

william harryman said...

Cool - thanks for letting us know - the Seed author had it wrong.

Peace,
Bill