Autistic children struggle with many obstacles, including learning to speak. And, experts have noted, if these children learn verbal skills by age five, they tend to become happier and higher-functioning adults than do their nonverbal peers. Thirty years ago, psychiatrists expected only half of all autistic children would gain speaking abilities. Recent studies, however, indicate that as many as 80 percent of children with autism can learn to talk. One such study in 2006 showed that toddlers who received intensive therapy aimed at developing foundational oral language skills made significant gains in their ability to communicate verbally. Now researchers have followed up with a number of those kids and found that most of them continued to reap the benefits of that therapy years after it had ended.

Several early behaviors build a foundation for language. These abilities have also been linked to whether a child can anticipate another person's mental state and use that understanding to explain and predict behavior. Developing this "theory of mind" may be a central difficulty for children with autism. Kasari's team targeted two of the early behaviors in their work: The first is the ability to engage in symbolic play, in which one object represents another—a child pretending a doll is his parent, for instance. The second is joint attention, wherein a child divides focus between an object and another person. This behavior can be thought of as "sharing looks." For example, when a child points to show a playmate a toy train, she looks at the moving train and checks to see if her playmate is engaged.

In the initial study, Connie Kasari of the University of California, Los Angeles, and her colleagues evaluated 58 children between three and four years old in a randomized controlled study. The children played with trained graduate students for 30 minutes each day over a period of five to six weeks. The time-intensive interventions focused on either symbolic play or joint attention. A third group, serving as a control, participated in playtime but was not directed to complete tasks and goals.

Independent clinical testers assessed the children before and after the intervention. They measured language and cognitive skills with standard tests, evaluated play level and diversity, and interaction with a caregiver. The initial study, published in 2006, showed that the joint-attention group was better at showing and pointing behavior whereas the symbolic play group showed more symbolic behavior, both in terms of play level and diversity. Twelve months after the therapy period, Kasari's group assessed the kids' language skills. On a standard language test, the two intervention groups showed spoken language improvement that corresponded to 15 to 17 months of development; the control group had only made a nine-month gain during the same period. Younger children and children at the lowest language levels prior to the intervention made the largest improvements. Kasari was initially surprised the groups showed such progress. The most important aspect of both interventions, she says, was "engaging the child for periods of time with a social partner."

In the new study, Kasari's team revisited 40 of the children five years later. The researchers found that 80 percent of them, who were by then eight to nine years old, still had "useful, functional spoken language." A small number of children remained nonverbal, which Kasari says is typical for studies of children with autism. Some children do not seem to be able to learn useful language by age five, but studies suggest it is possible to acquire language later. The new studies show a method for teaching preschool-aged children basic skills that will help them develop language by five and continue to make improvements years later. The researchers detail their findings in the May issue of the Journal of the American Academy of Child and Adolescent Psychiatry.

Previous studies have targeted skills important to language development, but many only looked at small groups of children or infrequent treatment sessions, Kasari notes. Understanding what makes a treatment successful or not is vital. "We need to distill down the active ingredients in early intervention," she says, "then take these elements and match them to programs."

This kind of long-term follow-up is rare. "The study is important in terms of raising expectations of what can be accomplished, and in raising awareness of how much work it takes," says Sally J. Rogers, a psychiatry professor with the MIND Institute at the University of California, Davis. Rogers, who was not involved in the research, emphasized that because the subjects were very young, the study builds on evidence indicating that the earlier the intervention the better—and children even younger than the toddlers in the original study could benefit. This has important pubic policy implications, she says, because there is little funding for children younger than three.

Finding a one-size-fits-all approach to helping autistic kids talk may be tricky, however: Autism affects each child differently, Rogers observes, and even the best interventions will have varied outcomes.