Even a short-lived battle with anorexia can be difficult to overcome, but the longer a person goes untreated, the harder it is to recover. Among those who are hospitalized for anorexia and discharged at a normal weight, about 35 percent will relapse within two years, according to a 2005 study in Psychological Medicine. Eating disorder specialists recognize that food restriction becomes entrenched over time, and even patients who genuinely want to get better struggle to eat normally.

Now, a new study offers a possible neurobiological explanation. The study in Nature Neuroscience suggests that, unlike healthy-weight women, anorexic women make decisions about what to eat using a brain region involved in habit learning—the behaviors that become entrenched, inflexible, and automatic over time.

When I think of my eating disorder, I tend to think of this line from Austin Powers: "You're the Diet Coke of evil. Just one calorie, not evil enough."

There was one year, maybe a year and a half, where things got really bad. My weight bottomed out of the healthy BMI range. I counted calories, barely saw friends (because doing so would involve eating and/or drinking, or fielding their questions about why I was doing neither), and I was always cold and miserable. By the time I started outpatient treatment, I was binging and purging regularly, and was diagnosed with bulimia.

I was never hospitalized for my condition and suffered no long-term physical or reproductive consequences, as many women do (eating disorder patients are often, but not always, women). When I terminated therapy, I was considered more or less recovered. Until recently—about seven years later—I never relapsed. Mine was the Diet Coke of eating disorders—just dangerous enough.

Even so, it took nearly a year of treatment to actually start eating like a normal person—to not choose a bowl of carrots over a sandwich or binge wildly and blindly, the moment a bit of fat or sugar passed my lips.

This new study may offer a neurological explanation for why eating disorders remains entrenched and difficult to treat.

"This is a paradigm shift for anorexia," study author Dr. Joanna E. Steinglass, an associate professor of psychiatry at the New York State Psychiatric Institute at Columbia University Medical Center, told ATTN:. "It tells us that there are actually different neural mechanisms at play for the anorexia patients during food choice than for the healthy controls."

The study involved 21 women who had checked themselves into the eating disorders unit at the New York State Psychiatric Institute, along with 21 women of normal weight. They were placed in a functional magnetic resonance imagining (fMRI) scanner to measure brain activity while they chose from a list of foods on a computer screen—a bowl of strawberries, say, or a donut. They were told that after the experiment, they would be given one of their selected foods as a snack. During this task, the women with anorexia were more likely to choose low-fat, low-calorie foods than the healthy women. The next day, the women were offered a buffet-style lunch, and the researchers recorded the caloric value and nutritional profile of the foods they chose. Again, the anorexic women typically chose low-calorie foods and consumed fewer calories overall.

The fMRI revealed a marked difference in brain activity between the groups. The ventral striatum, part of the brain's "reward center," was activated during the food choice task in all the women. But another brain region kicked into gear in the anorexic women: the dorsal striatum, believed to be involved in habitual learning. What's more, the researchers found that activity in a brain circuit linked to the dorsal striatum could predict how many calories the anorexic women would eat the next day.

"This model may help explain why early interventions for anorexia are so important, and why they work better," Dr. Steinglass says. "If you interrupt behavior before it gets over-trained, maybe that's an easier time to intervene and treat."

Dr. Steinglass emphasized that this is still just a hypothesis. But from a behavioral standpoint, she says, the findings make sense. Anorexics are often perceived as having extraordinary self-control—the ability to avoid calorie-laden foods, even to the point that it kills them. But willpower wouldn't explain why those who want to get better continue to eat in a way that keeps them sick. "The patients start dieting for some reason in the beginning, and then they just keep doing it and keep doing it," Dr. Steinglass says. "The behavior sure sounds stuck, like a habit."

Anorexia is the deadliest psychiatric disorder, with an overall mortality rate of around 4 percent, according to the National Eating Disorders Association. (But a Cleveland Clinic estimate puts it as high as 20 percent, mostly due to cardiac arrest and suicide.) And that makes developing better treatments a crucial undertaking. Food restriction is just one aspect of anorexia, even if it is one of the most universal features of the disease. But understanding the mechanisms at work in patients' brains could lead to more effective forms of therapy, Dr. Steinglass says. She and her colleagues are now testing some new therapeutic models at their institute. "Maybe we should be focusing more on what sets off a chain of events and less on trying to tell people what they should be doing differently," she says—"to focus on trying to make what has become automatic less automatic."