When Elizabeth Loftus began to plant false memories to test their therapeutic benefits, the memories seemed innocuous. They weren't about families or politics. They were just about food. You went into the experiment thinking you'd always loved strawberry ice cream, and you came out thinking it had made you sick. Big deal.

But successful ideas have a way of spreading. Soon, Loftus was trying the same technique on alcohol. And she was starting to think about tampering with the kinds of memories whose authenticity had once anchored her career: memories of crime and abuse.

A decade earlier, when she first wrote and testified against recovered memories of sexual abuse, her concern had been that such memories might be false. They could send innocent people to jail. But the more she looked into these cases, the more she noticed a second problem: Regardless of their truth or falsehood, memories of incestuous abuse were hurting the women who had recovered them. Their families were torn apart. Many were losing their jobs, ending their marriages, contemplating suicide, or mutilating themselves.

Her objections to recovered-memory therapy grew. The likely falsity of the memories was just the first problem. The second problem was that even if some of the memories were true, they could still be harmful. And even if they were true and helpful, the therapy might still be harmful on balance. Recovered-memory therapy, like any drug or medical procedure, should be judged by its total costs and benefits across the population, Loftus argued. If it hurt too many patients, its use might be unjustified "even if the benefits for actual victims can be shown." Perhaps, despite the truth, it was better not to remember.

By the end of the 1990s, science was developing new ways to deaden such memories. A study published in 2000 showed that in rats, fearful recollection of an electric shock could be blocked by injection of a drug called anisomycin. Loftus was intrigued. "We're on the brink of being able to figure out how you might accomplish something like memory engineering," she suggested. Patients could be prompted to recall traumatic incidents, she speculated, and drugs could be administered to prevent the memories' reconsolidation.

In 2002 and 2003, studies indicated that another drug, propranolol, could prevent or reduce post-traumatic stress disorder in humans. Adam Kolber, a law professor at the University of San Diego, monitored the research and talked to Loftus about its legal and moral implications. She was fascinated. She went back to her lab determined to get involved. She started with attitudinal research, asking people whether they would take a memory-dampening drug after being mugged and beaten. Nearly half wanted the right to take the drug, but only 14 percent said they would do it. She was surprised. If she had endured such an assault, she decided, she would take the drug.

She understood propranolol's legal implications. Its main effect was to dampen the emotional content of traumatic memories. But to a lesser extent, it also dampened their factual content. A victim who took the drug might lose her ability to testify convincingly against her assailant. Even in this circumstance, Loftus concluded, memory dampening should be permitted. The overriding principle, she argued, was freedom of choice.

But would freedom of choice survive therapeutic deception? Would it survive social pressure to fix unhealthy memories and habits? Loftus and her colleagues were already presenting memory therapy as an alternative to coercion. In their article on fattening food, they warned that unless behavioral scientists stemmed the obesity crisis, laws might be imposed to induce healthier eating, just as "seatbelt laws were imposed upon us when people were not using them on their own." In their article on alcohol, they offered memory doctoring as an alternative to electric shock and other "invasive" aversion therapies.

Society had an obvious interest in purging traumatic memories. These memories, Loftus and her coauthors noted, caused "significant costs to sufferers, their families, and society," such as traffic deaths and reduced productivity.   A similar case could be made for erasing bigotry, an idea that had interested Loftus for years. Memory doctors might be "useful for curing societal ills such as social prejudice," she suggested in 2001. Prejudice might be based "on a few incidents involving a unique group of people, so the memory doctor could wipe out or alter memory of these incidents."

Loftus never endorsed such treatments without the patient's consent. But she was stretching the definition of consent to fit memory therapy. To plant false food memories without violating informed consent, she and a colleague proposed in 2009 that

a therapist might ask for blanket permission in an early session to use various techniques to bring about a positive outcome—a list that might include the planting of false beliefs. If permission is granted, then, much later, when the permission session is long forgotten, the suggestive technique might be attempted.

The proposal sounded like one of Loftus' deception experiments. But it wasn't an experiment. It was the procedure to obtain consent to the deception. Using her expertise in memory's vulnerabilities, she was figuring out how to manipulate people into authorizing their manipulation. In this way, a memory doctor could justify herself.

Loftus answering a question at the Center for Inquiry's World Congress, April 2009

Maybe the doctor would stop at fixing the patient's eating behavior. Maybe she would move on to trauma or prejudice. The patient, having forgotten the initial mention of memory manipulation, would be none the wiser. And once memories were altered, there was no going back. In this respect, Loftus theorized, the brain worked like a computer: "You call up a file, edit it and then put the revised file back. The original is lost."  

Worse, there was no way to distinguish altered memories from originals. For three decades, Loftus looked for telltale signatures that might help judges and juries. She tried everything: confidence, vividness, emotionality, brain scans. She found some differences on average, but nothing that reliably identified a false memory. In 2009, she and her colleagues concluded that "it might be virtually impossible to tell reliably if a particular memory is true or false without independent corroboration."

And where would the corroboration come from? Documents? Photographs? Video? With digital technology, such evidence could easily be altered or fabricated. Two weeks ago, Slate did just that, editing several images to plant false political memories. (The experiment ran here; you can read the results here  and view a slide show of the altered images here.) Even DNA could be faked: In 2009, scientists reported that they had manufactured a blood sample sufficiently incriminating to fool a forensics lab. Altered evidence, in turn, could alter memories, as  Loftus herself had proved. False evidence and false memories would corroborate each other.

This wasn't the future Loftus had envisioned when she first fantasized about doctoring memories. But seeing the future was never her forte. She didn't foresee the memory doctors of the 1990s, either. In 2002, she wrote,

While musing about the hypothetical memory doctor in 1980, I could not have known that a version of the memory specialist was in the making. These "repressed memory therapists" would go out and prospect for early childhood memories of trauma, and in the process they inadvertently created false memories of the most unimaginable kind. The memory doctors I had speculated about in 1980 were supposed to use their talents to help people. The memory doctors of the 1990s went in the wrong direction.

But the memory doctors of the 1990s were trying to help people. They just didn't see that they were going in the wrong direction. That's the danger of doctoring memories: The future is as cloudy as the past.

What happened to the Elizabeth Loftus who warned the world against memory tampering? The one who exposed the memory doctors of the 1990s? The one who risked her career to uncover the truth about Jane Doe? The one who denounced the brainwashing of Tiananmen Square?

She's still there. She still testifies in court. She still teaches the dangers of recovered memories. She still worries about governments and photographic manipulators rewriting history.

No one knows the perils of memory planting better than Loftus does. She ran the experiments. She proved the damage. She literally wrote the book.

Now she has tasted the power of memory therapy. She sees before her all the good things a memory doctor might do. She thinks she can avoid the hazards. But to see where you're going, you have to understand how you got where you are. And that's the irony of her story. To me, her experiments and papers are more than a body of work. They're the trail of bread crumbs that fell out of her pocket on her way into the forest of memory doctoring. That trail is the way in. It's the way out. And it's a warning to anyone who fears the forest.

Today, Loftus struggles with her impulses and the implications of her work. She's preparing a report on a new experiment involving U.S. military personnel in survival training. First, the trainees were brutally interrogated. Then, through bogus photographs, their memories were altered so that they could no longer identify their interrogators. The experiment is an inadvertent blueprint for getting away with war crimes. Loftus worries about who will use it. At the same time, she continues to seek evidence that people are willing to take memory-dampening drugs. The subjects in her experiments keep saying no. She keeps looking for scenarios or information that might persuade them to say yes.

In this 2009 speech, Loftus worried that military enemies would exploit her research on altering memories of interrogations

The last mystery I tried to resolve in my conversations with Loftus is a paper she published with two colleagues in 2007. It describes an experiment aimed at getting people to drink less caffeine. Imagination exercises did the job, but they also created a problem: People who imagined drinking less caffeine came to remember themselves as having consumed less in the past than they really had. This made it hard to detect real reductions. It also created a problem for health research. People with altered memories, the authors noted, would underreport their intake, thereby obscuring caffeine's effects.

These are the same concerns Loftus had raised earlier in her career, before she took up therapeutic tampering with food memories. When I saw the 2007 paper, I thought she might be backtracking. I was mistaken. The initial experiment, it turns out, was done a dozen years ago.   It's a ghost from her past. And if Loftus wrote the warning about health research, she has no memory of having done so. The bread crumb has blown away.

How does her story end? Does she go deeper into the forest? Does she come out? Does she find a safe way through it? And what about the rest of us? What is the future of our past?

None of us can say for certain. But Loftus has left us a warning:

When we have mastered the false memory recipes, we will need to worry about who controls them. What brakes should be imposed on police, lawyers, advertisers? More than ever, we'll need to constantly keep in mind that memory, like liberty, is a fragile thing.

Wise words. Let's not forget them.

Like Slate on Facebook. Follow us on Twitter. William Saletan's latest short takes on the news, via Twitter: