At first, everything about L.’s baby boy seemed normal. He met every developmental milestone and delighted in every discovery. But at around 12 months, B. seemed to regress, and by age 2, he had fully retreated into his own world. He no longer made eye contact, no longer seemed to hear, no longer seemed to understand the random words he sometimes spoke. His easygoing manner gave way to tantrums and head-banging.

At first, everything about L.’s baby boy seemed normal. He met every developmental milestone and delighted in every discovery. But at around 12 months, B. seemed to regress, and by age 2, he had fully retreated into his own world. He no longer made eye contact, no longer seemed to hear, no longer seemed to understand the random words he sometimes spoke. His easygoing manner gave way to tantrums and head-banging. “He had been this happy, happy little guy,” L. said. “All of a sudden, he was just fading away, falling apart. I can’t even describe my sadness. It was unbearable.” More than anything in the world, L. wanted her warm and exuberant boy back.

A few months later, B. received a diagnosis of autism. His parents were devastated. Soon after, L. attended a conference in Newport, R.I., filled with autism clinicians, researchers and a few desperate parents. At lunch, L. (who asked me to use initials to protect her son’s privacy) sat across from a woman named Jackie, who recounted the disappearance of her own boy. She said the speech therapist had waved it off, blaming ear infections and predicting that Jackie’s son, Matthew, would be fine. She was wrong. Within months, Matthew acknowledged no one, not even his parents. The last word he had was “Mama,” and by the time Jackie met L., even that was gone.

In the months and years that followed, the two women spent hours on the phone and at each other’s homes on the East Coast, sharing their fears and frustrations and swapping treatment ideas, comforted to be going through each step with someone who experienced the same terror and confusion. When I met with them in February, they told me about all the treatments they had tried in the 1990s: sensory integration, megadose vitamins, therapeutic horseback riding, a vile-tasting powder from a psychologist who claimed that supplements treated autism. None of it helped either boy.

Together the women considered applied behavior analysis, or A.B.A. — a therapy, much debated at the time, that broke down every quotidian action into tiny, learnable steps, acquired through memorization and endless repetition; they rejected it, afraid it would turn their sons into robots. But just before B. turned 3, L. and her husband read a new book by a mother claiming that she used A.B.A. on her two children and that they “recovered” from autism. The day after L. finished it, she tried the exercises in the book’s appendix: Give an instruction, prompt the child to follow it, reward him when he does. “Clap your hands,” she’d say to B. and then take his hands in hers and clap them. Then she would tickle him or give him an M&M and cheer, “Good boy!” Though she barely knew what she was doing, she said, “he still made amazing progress compared with anything he’d gotten before.”

Impressed with B.’s improvement, both families hired A.B.A. specialists from the University of California, Los Angeles (where A.B.A. was developed), for three days of training. The cost was enormous, between $10,000 and $15,000, covering not only the specialists’ fees but also their airfare and hotel stays. The specialists spent hours watching each boy, identifying his idiosyncrasies and creating a detailed set of responses for his parents to use. The trainers returned every couple of months to work on a new phase, seeking to teach the boys not just how to use language but also how to modulate their voices, how to engage in imaginative play, how to gesture and interpret the gestures of others. The families also recruited and trained people to provide A.B.A. to their sons, so each boy received 35 hours a week of one-on-one therapy.

The specialists taught the parents that if their child wanted something, they should hand it to him — but should not let go until he looked at them. Within a month, B. was looking at people when he asked them for something, having learned it was the only way to get what he wanted. Within four months, he was looking at people even when he wasn’t soliciting help. Soon he learned to point to things he desired, a skill that required weeks of lessons. Once B. understood the power of pointing, he no longer pulled his mother to the refrigerator and howled till she happened upon the food he wanted; now he could point to grapes and get grapes. “Between the time he was age 1 and almost 3,” L. said, “I remember only darkness, only fear. But as soon as I figured out how to teach him, the darkness lifted. It was thrilling. I couldn’t wait to get up each morning and teach him something new. It wasn’t work at all. It was a huge, huge relief.” Soon B. began to use language to communicate, albeit inventively at first. One time when B. pointed to the grapes in the fridge, L. took them out, plucked them off the stem and handed them to him — at which point he started screaming. He threw himself on the ground, flailing in misery. L. was baffled. He had clearly pointed to the grapes. What had she misunderstood? Why were his tantrums so frustratingly arbitrary?

Suddenly, B. pleaded: “Tree! Tree!” It hit her: He wanted the grapes still attached to the stem. He wanted to pull them off himself! “It was like, Oh, my god, how many times have I thought his tantrums were random, when they weren’t random at all? I felt so bad for him: What other things have you wanted that you couldn’t tell me?”

After that, B.’s language blossomed quickly. By the time he finished kindergarten, he was chatty and amiable, though he remained socially awkward, hyperactive and unyieldingly obsessed with the animal kingdom — he knew every kind of dinosaur, every kind of fish. Whatever his preoccupation of the moment, he would talk about it incessantly to anyone who would, or wouldn’t, listen. L. made three small laminated coupons, and each morning, she’d tuck them into B.’s front pocket and remind him that whenever he talked about his favorite animal or noticed kids walking away or changing the subject, he should move a coupon to his other pocket. Once he ran out of coupons, she told him, he had to find other things to talk about for the rest of the day. Whether because of the coupons or maturation or something else, B.’s monologues stopped by second grade. Around the same time, his fixations eased. B.’s doctor concluded that the last vestiges of his autism were gone; he no longer met the criteria, even in its mildest form.

L. was ecstatic, but she was also plagued by guilt. Though Jackie’s son received the same treatments as B., he had made no such progress. Matthew still could not talk. He remained uninterested in other children and most toys. And despite efforts to teach him, Matthew’s communication remained extremely limited: When he squealed loudly, he was happy. When he threw up — which for a year he did daily — his parents concluded that he was distressed, after a doctor assured them that there wasn’t anything physically wrong with him.

“Jackie did everything for him,” L. told me, her voice filled with angst. “Everything. She tried just as hard as I did. She hired the same people, did the same work. . . . ” Her voice trailed off. She was sure that the behavioral therapy had allowed her to reclaim her son, but she could not understand why it had not done the same for Matthew.

Autism is considered a lifelong developmental disorder, but its diagnosis is based on a constellation of behavioral symptoms — social difficulties, fixated interests, obsessive or repetitive actions and unusually intense or dulled reactions to sensory stimulation — because no reliable bio-markers exist. Though the symptoms of autism frequently become less severe by adulthood, the consensus has always been that its core symptoms remain. Most doctors have long dismissed as wishful thinking the idea that someone can recover from autism. Supposed cures have been promoted on the Internet — vitamin shots, nutritional supplements, detoxifiers, special diets, pressurized rooms filled with pure oxygen and even chelation, the potentially dangerous removal of heavy metals from the body. But no evidence indicates that any of them can alleviate any of the core symptoms of autism, let alone eradicate it.

The idea that autistic people could recover first took hold in 1987, after O. Ivar Lovaas, the pioneer of A.B.A., published a study in which he provided 19 autistic preschoolers with more than 40 hours a week of one-on-one A.B.A., using its highly structured regimen of prompts, rewards and punishments to reinforce certain behaviors and “extinguish” others. (An equal number of children, a control group, received 10 or fewer hours a week of A.B.A.) Lovaas claimed that nearly half the children receiving the more frequent treatment recovered; none in the control group did. His study was greeted with skepticism because of several methodological problems, including his low threshold for recovery — completing first grade in a “normal” classroom and displaying at least an average I.Q. The therapy itself was also criticized, because it relied, in part, on “aversives”: sharp noises, slaps and even electric shocks. By the 1990s, after a public outcry, Lovaas and most of his followers abandoned aversives.

While subsequent studies did not reproduce Lovaas’s findings, researchers did find that early, intensive behavioral therapy could improve language, cognition and social functioning at least somewhat in most autistic children, and a lot in some. A few studies claimed that occasionally children actually stopped being autistic, but these were waved off: Surely, either the child received a misdiagnosis to begin with or the recovery wasn’t as complete as claimed.

In the last 18 months, however, two research groups have released rigorous, systematic studies, providing the best evidence yet that in fact a small but reliable subset of children really do overcome autism. The first, led by Deborah Fein, a clinical neuropsychologist who teaches at the University of Connecticut, looked at 34 young people, including B. She confirmed that all had early medical records solidly documenting autism and that they now no longer met autism’s criteria, a trajectory she called “optimal outcome.” She compared them with 44 young people who still had autism and were evaluated as “high functioning,” as well as 34 typically developing peers.

In May, another set of researchers published a study that tracked 85 children from their autism diagnosis (at age 2) for nearly two decades and found that about 9 percent of them no longer met the criteria for the disorder. The research, led by Catherine Lord, a renowned leader in the diagnosis and evaluation of autism who directs a large autism center and teaches at Weill Cornell Medical College, referred to those who were no longer autistic as “very positive outcome.”

Autism specialists hailed the reports. “Those of us who work closely with children with autism,” says Geraldine Dawson, a psychologist and researcher at Duke University’s department of psychiatry and the Institute for Brain Sciences, “have known clinically that there is this subgroup of kids who start out having autism and then, through the course of development, fully lose those symptoms — and yet people always questioned it. This work, in a very careful and systematic way, shows these kids exist.” She told me that she and many of her colleagues estimated that 10 percent or more of their autistic patients no longer had symptoms.

The findings come at a time when the number of autism cases nationwide appears to be climbing rapidly. No nationally representative study of autism’s prevalence exists, but the Centers for Disease Control and Prevention’s most recent study of 11 communities in the United States found that one in 68 children has autism, up from one in 88 two years earlier. Experts attribute much of that increase to greater awareness of the disease and its symptoms, as well as to broader diagnostic criteria. Some researchers say additional factors — among them toxic substances and older parental age — may contribute to the rise as well. Scientists suspect that what is called autism may actually be an array of distinct conditions that have different genetic and environmental etiologies but happen to produce similar symptoms. If true, it could help explain why some children progress so much while others don’t.

The research by Fein and Lord doesn’t try to determine what causes autism or what exactly makes it go away — only that it sometimes disappears. There do, however, seem to be some clues, like the role of I.Q.: The children in Lord’s study who had a nonverbal I.Q. of less than 70 at age 2 all remained autistic. But among those with a nonverbal I.Q. of at least 70, one-quarter eventually became nonautistic, even though their symptoms at diagnosis were as severe as those of children with a comparable I.Q. who remained autistic (Fein’s study, by design, included only people with at least an average I.Q.) Other research has shown that autistic children with better motor skills, better receptive language skills and more willingness to imitate others also tend to progress more swiftly, even if they don’t stop being autistic. So do children who make striking improvements early on, especially in the first year of treatment — perhaps a sign that something about their brains or their kind of autism enables them to learn more readily. Researchers also say that parental involvement — acting as a child’s advocate, pushing for services, working with the child at home — seems to correlate with more improvements in symptoms. Financial resources, no doubt, help too.

For now, though, the findings are simply hints. “I’ve been studying autistic kids for 40 years,” Fein says, “and I’m pretty good at what I do. But I can’t predict who is going to get better and who’s not based on what they look like when I first see them. In fact, I not only can’t predict who is going to turn out with optimal outcome, but I can’t even predict who will have high-functioning autism and who will be low-functioning. There’s so much we still don’t understand.”

Mark Macluskie, an animated 16-year-old, is another of the children in Fein’s study who no longer has autism. He spends his spare time playing video games, building robots, writing computer code and hanging out with friends at the local park near his home in a Phoenix suburb. He co-hosts a weekly Internet radio show called “Tech Team,” which has 32,000 listeners. On the program, he and a buddy review apps, discuss tech news, tell (very) corny jokes and produce regular features like “Gadget on a Budget.”