Filter by Categories
Clinical Practice Research
Ethical Legal and Regulatory Considerations
Planning, Managing, and Publishing Research
Research Design and Method

Identifying Children with Specific Language Impairment

Laurence B. Leonard

DOI: 10.1044/cred-dsa-bts-001

I think, as some British colleagues have said before, children with specific language impairment seem to be a group of children who, the person on the street will never have heard of, and yet, the prevalence of specific language impairment — kids who meet these criteria are every bit as common as dyslexic children, children with ADHD, and much more common than children on the autism spectrum. Because I think so much in these children is intact, that is, this children exhibit normal hearing, they score at age appropriate levels for nonverbal tests of intelligence, and they do not have any frank neurological damage or disease, it’s probably a little easier to pay more attention to children whose deficits are broader, whose deficits have a very direct impact on academics such as children with reading problems.

Challenges of Identification

Many of us who work in this area are very concerned because when we do long-standing longitudinal research in these children, we realize that their outcomes are not so stellar. These children are often having academic difficulties, including reading difficulties. They do have social problems, not so much because their social skills are so bad, but rather their language skills really do restrict their ability to interact. Especially when you’re talking about conversations among two, three people. Coordinating language in that way is really difficult for these kids. And it even affects, later on, their economic functioning. These children eventually when they go through school, tend to be employed in obviously less verbal skills, but it has an impact on their economic wellbeing. So these are lifelong conditions. Even though intervention can most definitely facilitate these children’s language ability, it is nevertheless the case that it is a long standing problem. So a lot of us are committed to making sure that these children don’t get lost in the shuffle. We really do need to research and learn more about them.

I think I see the biggest problem as a disconnect between the research world — even dare I say, the clinical research world — and the clinical practice world. I think there is a huge disconnect, and I’ve always been concerned about it.

A great case in point would be, we as researchers often use the term “specific language impairment” and sometimes we argue among ourselves because, on the one hand, we recognize that many of these children have some subtle weaknesses in other areas, but they tend to be sub-clinical. They would never warrant a term like “intellectual disability” or “autism spectrum disorder” or “frank motor impairment” but they are kind of weak in these areas. Some people when they hear a term like specific language impairment and they know these kids are not so pure, that maybe another term is relevant, so we argue among ourselves a bit. But we can define our kids so we can replicate our own studies.

If you go to the clinical world, for instance in school service delivery systems, no one uses the term “specific language impairment.” Zero. And some of the alternative terms that could be used like primary language impairment — we don’t, no. Some of that can be explained by the fact that, as I mentioned earlier on, children with specific language impairment are very much underserved because school systems and other agencies are hard-pressed, they often have to deal with children of great severity. Right now, many children with autism spectrum disorders are being served. And so therefore children who would be closer to the kinds of kids I study, if they get served, they get referred only if, for instance, their language problems are such that they are leading a child to act out in school. So there is now a little bit of a behavior problem component, along with the language component. Or the child is somehow expressing some real overt emotional frustration. Or the child might have some other kind of problem in combination with the language problem.

So clinicians working in these settings are not seeing anything close to a pure case anyway. And so to them, specific language impairment is like saying, I’ve got a kid from Mars that I’d like you to see. They don’t quite relate to it. And yet, to a large degree, they are seeing some of the same kids, but they aren’t called the same. Across different states, across different regions of the country, the terms vary. And the criteria for service delivery most definitely varies. So you can go from one school district where they’ll say John has to have two scores 1.5 standard deviation below the mean. One county over they have to say, one test with a score two standard deviations below the mean. And they’ll call the kid something slightly different in each case.

I’m disappointed that researchers, even when they publish a lot, and we’re making good grounds, that somehow, especially in the terminology, it’s not getting translated very well. If we were to apply that loose terminology that sometimes one sees in the clinical world, if we were to apply it to research, no one could ever replicate a study. Two people would think they are doing the same study and come up with very different results. Why? Because the children are being selected in wildly different ways, because they weren’t described in a precise enough way.

Emerging Research Trends: Early Identification

Early identification is really critical. Yesterday morning I participated in a panel on late talkers. There were a number of longitudinal, prospective studies on late talkers, these were excellent studies. The problem is, to the field’s surprise, when these late talkers were followed, very few of them turned out to be diagnosed with language impairment. Much lower than the actual prevalence of the disorder. So if you follow children from 24 months until they reach five years of age, the way late talkers have been defined, precious few of them turn out to have a language impairment. And the prevalence of specific language impairment is 7% among five-year-olds. And we (the field) weren’t coming up with those kind of figures at all. Which is difficult, because it says, okay if all these kids at five years have language impairment, how did they get there?

Well, now we’re getting the idea that, okay, we defined the field, defined late talkers maybe in such a pure way, we looked at kids who were late in talking, but in other ways they look so normal. Their birth history was fine. Their cognition was fine. Their hearing of course was fine. Their motor development was fine. They have absolutely nothing wrong with them, it’s just that they weren’t talking, or they were talking quite late.

Well that might be too clean a profile for kids who eventually have a cleaner profile. It could be that these kids had somewhat broader delays. We’ve always had the suspicion that these kids have some subtle weaknesses in other areas. Maybe their fine motor development is not quite as refined as would be expected. Maybe they’ve got messiness in their birth history that’s not enough to cause great alarm, but it’s enough that the pediatrician would have made note of it, yet it didn’t reach a clinical stage, and so on.

One of the things we can do is try to identify future SLI kids by broadening the profile at 24 months when we look at children. There are a few other things too. But what it boils down to — we need earlier identification techniques. Other people in the field are starting to adapt measures that people studying young typically developing infants — their early sentence processing — you may have seen some of these looking tasks. Children will hear a sentence, and there will be two pictures on a screen, and these eye trackers will focus just where the kids are looking at it. Thirty-three millisecond frames. Frame-by-frame, where are these kids looking? How quickly are they reacting? For instance, some researchers have found that, let’s say there’s a ball and a car, and the child hears something like, “Where’s the ball?” Not only does the kid look at the ball, but how quickly in milliseconds does the kid turn his or her attention to the ball? The speed with which the kid — within milliseconds — switches to the, shall we say, the correct object that’s being heard actually has predictive value three or four years later.

We just need to find other ways to identify kids at an early age. Because this population is not syndromic. They do not look differently. They wouldn’t have that kind of diagnoses if there was a huge medical complication at birth, and so on. We have to find other subtle ways of identifying future problems in these kids.

Emerging Research Trends: Early Intervention

The stock answer, and we need a lot of research on this as well, but the stock answer is that we need to be as certain as we can that a child identified early on really is going to be seriously at risk for a language problem. Because we don’t want to over identify, because some of these kids are going to be late bloomers.

Let’s assume that we do, as a field, come up with really accurate early identification. Then, of course, the natural thing would be early intervention. But you don’t want to intervene when so many of these kids are going to be fine without you. But early intervention would be the answer.

Then, of course, that also prompts the question, do we have good early intervention techniques? There are certainly procedures out there. This could bring to the next topic, which you’ve heard a lot about, everyone in the field has, the importance of future randomized controlled trials. Where large numbers of kids are randomly assigned to one group or another, and very important controls are built in. Like the people assessing the child’s progress are blind to which treatment condition the child was in. So the next step will be those kinds of controls will need to be placed in early intervention programs. That’ll be another important thing that’s going to be done. We hope.

Emerging Research Trends: Statistics

And I think it’s also going to be important — in the old days, we used to take statistics that were kind of limited to analysis of variance kinds of designs, and now regression techniques, structural equation modeling is going to be more important. One value about these techniques is that some of these boundaries between typically developing kids and clinically significant kids, some of these boundaries are really gray areas. They shift around a lot. A lot of these statistical techniques are looking at the whole distribution of children, kids who on one continuum are very clearly impaired, and kids on another continuum are above normal, but you have everything in between, rather than these discrete categories. We’re really learning a whole lot about language processing in kids and we’re learning that the difference between low normal, and mildly impaired is not a qualitative difference, it seems to be a quantitative difference. So we need to have analytic techniques that take those things into account. I think young researchers would be really good to — if their mentors are not already steering them in this direction — to go in the direction of seeking out these kinds of analytic tools during their research preparation.

Emerging Research Trends: Genetics and Neurobiology

At some point in the not-distant future, I sure hope we’re going to have genetic and neurobiological measures, both neuroanatomical and electrophysiological measures, and genetic measures and behavioral measures finally starting to be coordinated. Right now it’s tough.

In the genetics world, we used to think for the longest time, that the reason why we couldn’t really move forward in genetics — the field couldn’t — was because, using specific language impairment as a good example, because the phenotype was a little bit inconsistent. We would describe these kids slightly differently. We would recognize the heterogeneity. Geneticists would say, once you can really pinpoint and say these kids and not these kids — these kids have the condition and these kids don’t — then we can work on genotype. Well, as it turns out the genotype is a bit complicated too, to say the least, because it seems to be a multifactorial disorder. There is no single gene with a variant or mutation that would be sufficient to create the symptoms of specific language impairment, just as there isn’t for some of these other conditions that the field has studied. And we’re talking about subtle combinations of genes. Each contributing a very small amount of variance. And probably interacting with the environment in ways we haven’t fully appreciated yet, that are creating these kinds of disorders.

In each of these areas, neuroanatomically speaking, MRIs and functional MRIs for that matter, as well, we’re finding lots of group differences between kids with SLI and typically developing kids. Not damage — they’re not like lesions or brain disease but rather brain configurations that are a little more atypical than one might expect. The problem is that the particular brain configurations that are different will vary from kid to kid, just the genetic contribution seems to vary from kid to kid. So we have a long way to go. But what I’m really starting to see are big collaborative studies where people are bringing together as I said neuroanatomical, electrophysiological, genetic, behavioral work. People from other disciplines, as well. I really think we’re going to be making good headway.

Laurence B. Leonard
Purdue University

The content of this page is based on selected clips from a video interview conducted at the ASHA Convention.
Copyright © 2015 American Speech-Language-Hearing Association