The following is a transcript of the presentation video, edited for clarity.
The overview of my talk is I’ll start by introducing Autism Spectrum Disorder and then drill down into, you know, two words about language and Autism. I’m then going to talk about studying high-risk infants in the first year of life and these are infants at high risk for Autism. I’ll then talk about what we see about early concerns and language milestones that are drawn from the data of our project but very much reflect what is coming out of other work in this area. I’ll focus then on mechanisms which is what I largely wanted to address. These are sort of risk mechanisms, and we’ll talk about the neural underpinnings of domain general functioning, speech perception and language, and then also potentially maternal contributions to later language and/or Autism outcomes. Then I’ll talk about comparing the studies of early mechanisms and risk factors in Autism to those in the extent literature on SLI and finally come to some conclusions.
Autism Spectrum Disorder
Okay, let me start by talking about Autism Spectrum Disorder and I’m going to focus on DSM5. We now have two broad domains in which we identify impairments. Social communication and secondly repetitive behavior and restricted interests. The communication impairments, the examples that are included in the symptom list in DSM5 include failure in back and forth conversation, deficits in nonverbal communicative behaviors, and difficulties adjusting behavior to suit various social contexts and here we could think about communicative or language within social context. In addition, in DSM5, we now need to specify whether ASD occurs with or without accompanying language impairment. So it’s clearly acknowledged that not all kids with Autism have what I’d call formal or structural language impairments, though they all have social communicational pragmatic deficits. You also have to specific other co-occurring conditions, particularly intellectual disability.
So this is kind of a very fluid graph. Don’t write down these numbers. What you can see here in blue, about 25 percent of the population don’t have language impairment, 45 percent are language impaired but they do speak, and about 30 percent, 25 to 30 percent, are what we call minimally verbal, they have no or extremely limited functional language. So you can see why we need to do greater specification, and I think we need to learn a great deal more about the characteristics of the language and underlying biological mechanisms that may be similar or different across these different, broadly-defined subgroups.
Just as an aside, to touch on epidemiology, I think everyone knows the rates of Autism in the population are soaring, and we see the same kind of huge geographical variation across states even within the United States. I will point out to Sean that the rates in Utah are extremely high. And everywhere you see very high rates of Autism, that proportion of no language impairment is what’s driving the increase today.
Developmental profiles. What do we know about this? About 90 percent, although again I don’t think we have very good data on this, have delayed language milestones. About 20 percent show language regression. They lose words or phrases in the second year of life and this regression path is quite unique to Autism. We don’t see this in, for example, specific language impairment. About 25 percent of children who are quite delayed in language through to the third birthday may still have no words to speak of or phrases show accelerated language growth in the preschool years and they can catch up to peers. Every single one of those children is enrolled in some form of intensive early intervention. This doesn’t happen spontaneously.
The most significant influence of early intervention, which is designed, I mean most of the kids with an Autism diagnosis go into these specialty programs and we have a variety of intervention programs that have been developed, but basically the common finding is that what early intervention does is to promote language development and it’s actually pretty unclear that you change very much else. Everything else seems to travel with increasing language development. So what this tells us is that early intervention and later language intervention is absolutely critical for Autism and this is true whether we’re talking about comprehensive behavioral programs such as ABA programs or the Early Start Denver model or whether we’re talking about more focused, targeted interventions that drill down to core ingredients like joint attention and here I’m just showing you data from two such studies. On the left, this is Gerry Dawson’s data from Early Start Denver. The two year outcomes. Here, visual reception or nonverbal IQ and ADOS or social symptoms we see actually, well, this is nonsignificant change but both receptive and expressive language we see huge gains compared to a community control group who are also making gains as you can see. And then on the right, this is data from a study by Connie Kasari. Also longer term outcomes, joint attention focusing in on training joint attention leads to extremely efficacious changes in particularly expressive language.
Studying High Risk Infants
Okay, let me turn now to studying high-risk infants and what we mean by that. And the rest of my talk is going to focus in on this population. So there’s been a surge in interest in the Autism world in the past decade to study infants at risk for neurodevelopmental disorders. I mean, almost all the work is on Autism. There are more than 20 research teams now from around the world. We actually all sort of get together. We form a consortium under the auspices of Autism Speaks called The Baby Sibling Research Consortium, and the primary goal of the consortium now is to pool our data so that we have a whole bunch of common measures and they are now over 4000 infants in this database so we can get very good data on what’s happening with infants who are high risk. And we define high risk in the baby sibling consortium to infants who have an older sibling with Autism. And every one right now in the consortium is comparing these infants to infants to what we’d call a low-risk control. For most of us that means babies who have an older sibling but they’re squeaky clean. We’ve screened them for Autism and other neurodevelopmental disorders.
Now I just added this line this morning because I realized I needed to say this. Our grant was designed to also compare infants at high risk to Autism to infants at high risk for SLI and we failed. Okay. We absolutely failed. We were able, over a five-year period, to recruit 25 families. Every one of these infants who we did recruit who was at high risk for SLI because they had an older sibling with the disorder, every one of them turned out perfectly fine, no language problems, no nothing. So it’s like a dud group for us. And we can talk more about that later in the discussion as to why we failed so miserably.
Okay, so one of the things we know, why do we study these babies, well, we know that this is a highly inheritable disorder and we know that the recurrence risk if you’re a younger sibling is way higher. And based on the BSRC database, the figure is about 18.7 percent overall but it really divides up depending whether you have a y chromosome or whether you have a double x. And in white, what you see, these are infants who have more than one older sibling with a diagnosis of Autism so that what you can see here at the highest group is if you have two older siblings and the baby is a male, that baby has almost a 50 percent chance. Okay, but it’s 18.7 percent overall.
Okay, so what do we know about demographic risk factors? As I’ve said, males more than females, family history, and the more loaded your family history the higher the risk pointing to genetics. Another demographic risk factor which is parental age which I think also does come out in epidemiological studies of SLI but I’ll be comparing the risk factors later.
General Findings from Infant Sibling Studies
Okay, so what are some of the general findings, and I’m really just skimming over what is now well over 100 papers, the first of which was published just about ten years ago, and that is that Autism manifests a change in developmental trajectories. The onset when we look at behavior is very much in the second year of life and what we see are declines in social engagement which are very reminiscent then of language regression. So its regression, we could say, extends to the social domain as a defining characteristic, a change in developmental trajectory, so infants who had good eye contact at six and 12 months, we see a decline in the second year. The same is true for social smiles and vocalizations don’t show the increases that are expected. We see slowed cognitive development and especially delays in language and gesture at 12 months but also patters of unusual vocalizations even before that, at nine months when we look at babbling, for those groups that are actually seeing babies somewhere between six and 12 months.
So I think this is an important point to keep in mind because we all started out thinking we’d see something different about these babies very early on in the first year of life, and the answer is you don’t. They look like lovely babies just like typically developing, like the low-risk controls. And it’s only the changes in the behavior in the Autism outcome infants that in the second year of life that really defines the disorder.
Our Infant Sibling Project
So we jumped into this venture about seven or eight years ago and my main collaborator is Chuck Nelson from Children’s Hospital Boston and in our project we see babies multiple times in the lab up until 36 months and for all of us 36 months is our major outcome time period. That’s when we’ll give a diagnosis at 36 months based on a bunch of things including expert clinical judgement that this is a child with autism. Okay. We have a lot of behavioral and observational measures. Between six and 18 months in our study, we actually send the families home with a camera and we ask them to send us videos, and they also do online written diaries and I’ll talk a little about that. We do some eye tracking stuff with the kids and we do brain measures and the focus of much of the talk today will be on how we can use our brain measures to investigate what might be some of the underlying mechanisms. These are measures of brain function including electrophysiology and neuro infrared spectroscopy and we have different paradigms for the infants which includes both giving them faces to look at of different sorts and also language and just parenthetically, of course, I bring the language and the Autism expertise to the project and Chuck Nelson brings the faces and the infant brain development expertise to the project. We have completely non-overlapping skills.
Early Concerns and Language Milestones
So let me talk a little bit about what we see in early concerns and language milestones. So this is work of my former, I have to say former, doctoral student who is now Dr. Tolbert. She is a post-doc at The Mind Institute and what she did was she went through all the written diaries and we looked at– we ask in the written diaries lots of different questions. Did your kid do something cool this week? What happened? Tell me something nice. You know, just a bunch of general questions and embedded in there is did you have any concerns about your child this week?
So at six months, here’s what we found. Okay. And a lot of the slides, you can kind of read them the same. The blue is our low-risk control group. The red is the group of infants who are high risk but did not have Autism at 36 months. And the pink are the Autism outcome kids. And what you can see is we’ve divided the concerns into the sort of domains, the social communication domain and the repetitive behavior and restricted interest domain that are part of the defining characteristics and language. And you can see that for every one, they’re worried about language at six months. Are you kidding? Okay. But what is important to see here, it’s the high risk who don’t have Autism, those are the moms who have the most concerns. And say this, no relationship to reality here. This is just what moms are feeling. Okay. Ditto at nine months. Okay. And it’s still the mothers of the infants who don’t end up with Autism who are expressing the same concerns. At 12 months, we see a flip and now all of a sudden we do see that mothers of infants who later turn out with Autism– so you can see, these are the biggest change– are now expressing concerns but again look where their main concerns are. It’s always language. Okay. Now we sort of knew this from retrospective studies. We know this from, you know, anecdotes and so forth but this is a prospective study so I think we see a lot about who’s got concerns at an almost 60 percent level.
We also looked at their gesture use and here this is tokens and in the light of pink are types. These are our low-risk controls. These are our high-risk who don’t have Autism at the end and they’re very comparable but we can see far reduced gesture use at 12 months in those infants who later are diagnosed with Autism. And then looking at early vocalizations and this is ongoing work soon to be submitted by my wonderful doctoral student, Karen Chenausky who’s sitting right here. She looked at early vocalizations at 12 months, and this is just to cut to the chase. These are the low-risk controls. These are the Autism negative outcomes. And these are the kinds who end up with Autism. So we can see that they are vocalizing less and if you have questions to drill down into these data, you can address them to Karen.
What I’m showing you here are some growth curves that we did of Mullen language scores and on the left is expressive language and on the right is receptive language and this is work from Catarina Stamoulis at Children’s Hospital and she just mapped out the data for us. And what you can see here is in gray are the low-risk controls. This is kind of the aggregate data but you can see all the variability. And in pink are the high-risk infants. And what you can see here is that there’s something happening. They’re kind of traveling together quite nicely during between 18 and 24 months but it’s in the third year of life that we see the language kind of peeling apart a little bit. They’re still in the normal range, but they’re really separated out and I will say that the groups are extremely well-matched for socioeconomic and other demographic variables like that. So these infants as a group begin to kind of fall off where they should be on their developmental curves. We could see that this gray curve is what they should be traveling on and they’re peeling apart a little bit. And we actually have some data from– which I don’t have time to show– from eye tracking which suggests the same thing, that language processing by 36 months we’re seeing differences in the high-risk infants even when they don’t end up with Autism.
I’m going to start by talking about looking at gamma. Basically, you can take the raw EEG signal, and you transform it to obtain the power spectrum. And at the low end we see, you know, sort of alpha, beta, etc. Anyway, gamma is here in the 30 to 50 hertz range. We’re actually now very interested in looking at very high gamma in the over 100 hertz range because there we’re beginning to see some quite interesting things except that we didn’t filter our system to capture the very high frequency range for a lot.
So why are people interested in gamma? Well, because it is associated, it’s electrical activity in that power spectrum that reflects general aspects of cognition broadly, attention, working memory, learning, and also language so it’s just one of those kind of pretty important power spectra. And it’s involved in integrating information in different brain networks which is required for complex skills, more broadly defined. The gamma frequencies that we’re picking up are actually picking up local circuitry. This is not long-range, whole-brain gamma frequency.
So this is how we collect the data. Well, this is Chuck Nelson blowing bubbles to the baby. Let me tell you, this was just done for the photo shoot. I don’t think he’s ever done this for a real baby in this study. But you can see the baby is sitting on the mom’s lap. You want to try to keep them still. You have to do something. So people refer to this as resting state. They’re not resting. It’s a baseline EEG and we collect these two minutes and it’s over the frontal regions and we looked at this between six and 24 months.
And the analyses of this were done by a postdoc, actually it was for her doctoral dissertation, Adrienne Tierney, and what you can see here is this is the low-risk controls and we can see that they start out quite high at six months and there’s a change, sort of a decline, we don’t really understand the significance other than, you know, you can throw out nice terms like more efficiency, blah, blah, blah. Really, we don’t know. And then these are the high-risk infants and what you can see is that at six months, a point in time when we look at the baby’s behavior, we see absolutely nothing different about them, and these are high-risk infants, most of whom don’t have an Autism outcome and the data look the same with or without the ones with an Autism outcome. These babies are different and they’re different fairly early on. Okay. But by the time we see them at 24 months, they’re pretty much at the same point. So again I think the important point here is it’s all in the developmental trajectory, not just taking cross sectional time points.
Okay, let me turn now, so that’s it for looking at domain general mechanisms in this population. Let me turn now, I’m going to focus much more on speech in the first year of life since this is ASHA. So we started out thinking that what would be really interesting to look at in high-risk infants is the phenomenon of perceptual narrowing which takes place at around nine to 12 months which according to Pat Kuhl is related to attending to the linguistic environment in social context. So the idea here is that, of course, early on we can discriminate all speech sounds and then by 12 months we’re only able to discriminate speech sounds that are forming making the language to which we’re being exposed in our social contexts.
So we asked the question, and we got an awful lot of money, all our grant proposals in that first round were built on this fabulous hypothesis. Do infants who are later diagnosed with Autism fail to show perceptual narrowing? I mean it’s a no brainer, right? If these infants are attending less to the social context, they’re not going to show perceptual narrowing at 12 months, maybe they will a little bit later.
We also know that there’s a lot of work showing that children and adults with Autism and their first-degree relatives who are unaffected show atypical brain asymmetry. They’re generally not left lateralized for language. Of course, there’s enormous variability, but they’re far less likely to show standard left lateralization. They may show reduced or equal or even in some cases reversed lateralization. And so we could use the data that we collected in to look at perceptual narrowing taking a later part of the wave form to ask were the infants at risk or who are later diagnosed with Autism show atypical lateralization to speech.
So this was worked by another one of my former doctoral students, Seery, and we took one of Pat Kuhl’s research paradigms and it’s a double oddball procedure and basically the baby, still sitting on mom’s lap, is hearing da, da, da 80 percent of the time. Ten percent of the time, they hear a native contrast ta which we can all hear, and then ten percent of the time they’re hearing a non-native contrast which I’ll attempt to produce, da, okay which you shouldn’t hear as being different from da unless you’re Hindi-speaking. So as I say, at six months we expect all the infants to differentiate the standard and the non-native contrast, the ta and the non-native da, but by 12 months not to do so.
And here are the data six months, nine months, 12 months low risk, high risk and basically what you see here is at six and nine months we see faster peaks. I’m showing you the actual peaks here. In the P150, that’s the part of the waveform that’s most sensitive to these consonant differences, to both the deviants, both the da and the ta relative to the da which is in the solid blue. We see that in both groups. By 12 months, we see that only to the native deviant, to the ta, are they showing a differentiation. And again, we see this in both groups. There is no interaction with group at any age, and we’ve looked at this in replicated samples.
So that hypothesis went down the tubes, the perceptual narrowing one. But that’s always so much more interesting because that’s what gets you to look at more things in your data. So what Annie went on to look at was the late slow wave so that occurs kind of, in this region, 300 to 700 milliseconds after six, nine, and 12 months. Here in the low-risk controls, what you can see is at six months the amplitude is the same in both the left and the right hemispheres. Different by nine months and continuing to be different, so beginning to show some lateralization. The emergence of lateralization to speech sounds by nine and 12 months.
In contrast, we don’t see any developmental change in the high-risk infants. So we’re seeing here now our first peek at something. We could say it’s at nine months even, and we didn’t follow this out with the same paradigm to see what happens over time, but I would suspect that this atypical lateralization continues because we know studies of older children or parents of kids with Autism who themselves don’t have autism show this atypical lateralization effect, and I think it’s really interesting to see how early it emerges.
Well, we then separated, once the samples got larger, Annie went back and divided that high-risk group into high-risk infants with an Autism outcome. High-risk infants who are just like low-risk controls — their behavior on all our measures is perfectly, perfectly normal. The typical kids. And then we’ve got the high-risk infants who show atypical behavior. They don’t have Autism but they’re showing elements of what’s called in the literature, the broader Autism phenotype, so their language may be a little bit off, their social scores not in the Autism range later but at 36 months they’ve got lower language and/or some social impairments and that’s are atypical group.
And these are the lateralization data. What you can see as we see this nine, sort of lateralization, this kind of higher amplitude in the left hemisphere for the lowest controls, all three of our groups show this atypical lateralization but it seems to be most extreme in the Autism outcome where they’re showing quite distinct reversed lateralization. What’s remarkable, I think, is that this isn’t traveling with behavior at all. So even in this group, we’re not seeing the lateralization effect, so it’s one cautionary point and I’ll come to another which is not everything that’s different in the brains of a high-risk infant who comes from a family and is carrying genetic burden for Autism, not everything that’s different is abnormal. Normality should be defined on the basis of behavior not on what your brain is doing.
Annie then went back and analyzed the amplitude of the P150 to the da, the repeated da stimulus. We were sort of interested initially in habituation effects but we really didn’t find it in any group. We looked at this at nine months and what we found was, if you look here these are the low-risk controls and these are the high-risk and you can see that in general the P150 is larger in amplitude. So we’re seeing an atypical response. This is the repeated. They hear this over and over and every time they hear it, their brain is saying wow, another da. And these infants are like bored with all those das because they hear too much of them. What you see here is the amplitude to the deviant and you can see so it’s not a question of their brain is, you know, saying wow to everything, but they’re actually the same on those tas here.
Then asked the question, well, you know, what is this enhanced P150 all about? So she looked at the correlation between the amplitude at nine months and looked at expressive language outcomes at 18 months and low and behold what she found was it doesn’t matter what the amplitude of your P150 is if you are a low-risk control. It means absolutely nothing. There is not a significant correlation. But we find a highly significant correlation for the high-risk infants whereby the more atypical your P150, the larger the amplitude, the better your expressive language outcome is. So again this is another example where being different does not mean being impaired and in fact for these infants this may be some form of compensatory mechanism, we don’t quite know how to interpret this. You know, let’s be honest, it’s a bit of a fishing expedition. But we did find this and I think it’s really interesting that there is something different from the start about how the brains of high-risk infants process language and I will say that all our kids did acquire spoken language, even the ones with Autism. We don’t have any minimally verbal kids in our sample.
Brain Connectivity in ASD
We then went to look at connectivity in the brain. This is a hot topic, you know, if you look up what’s the general hypothesis about what’s different in the brains of people with Autism, well, they have different connectivity between brain regions. So we know anatomically there seem to be fewer long-range connections that includes, say, the arcuate fasciculus so between regions, particularly in children and adults with Autism and when we look at functional connectivity, and you can do this in a variety of ways depending on whether you’re doing MRI or EEG, we also find reduced connectivity, both intra- and inter-hemispheric connectivity. This is a quite replicable finding across studies.
So we asked, how early do these connectivity differences emerge, and are they specific to Autism outcome infants and this was worked on by two postdocs. Well, this is Adrienne Tierney, now a postdoc, and then Giulia Righi and they looked at connectivity and their measure was event-related coherence in gamma between frontal and parietal regions of interest. And it was during the speech task, so basically they took that window right after the onset of the speech sound, the sort of 150 to 300, the same window in which we’re seeing the P150, and asking how synchronized or similar are the signals in the different regions of the brain reflecting the strength of the functional connection. And in blue I’m not sure how well it stands out here. These were our regions of interest. These were the electrodes we used in the left and the right hemisphere.
And these are what the data looked like, the coherence at six and 12 months. On the left, we see six months and we can see there’s no group differences, and on the right we see 12 months and what you see is for the low-risk controls, as we’d expect, coherence is increasing, and we see the opposite pattern, the opposite developmental pattern in the high-risk infants. And we see that here there’s a statistically significant difference between the groups.
When we break out that 12-month data by outcome who does have Autism and who doesn’t, what you can see here is a pattern that emerges across a lot of studies of high-risk infants. The lowest level of connectivity are in the Autism outcome infants here. We have an intermediate pattern for the high-risk but who don’t have Autism and the greatest at 12 months in the low-risk controls.
We then looked at connectivity using a different technology and a somewhat different paradigm and this was work that’s done by another postdoc, Brandon Keehn and what the infants were doing, and we started this work when the infants were three months, it was a paradigm that I think we borrowed from Janet Werker who’s done a lot of functional neuro infrared spectroscopy work with typically developing infants. The babies are listening. They’re wearing that cute, little blue cap I showed you and they’re listing to different syllables, artificial words, either a b b patterns or a b c patterns so penana, baloti, etc. And we compared them at these different ages and what Brandon did was to investigate the connectivity and neuro infrared spectroscopy is often called the poor man’s functional MRI and basically what it is is you are like in FMRI you’re picking up blood flow but you’re using these light electrodes and you look at both oxygenated and deoxygenated blood separately.
For this particular paradigm, we used the following regions of interest in the left and the right hemisphere, frontal, sort of temporoparietal, left and right hemispheres. And these were the findings. And he’s basically again using a statistical coherence measure, how well-correlated the signals were in the different regions of the brain. And let me walk you through these. So this is regional connectivity and we’ve got here left parietal, right anterior, right parietal, and here is left anterior and you see all the different things and this is three months, six months, nine months, 12 months. No difference — the kind of richer the color, you know, it’s like FMRI, the more striking the difference between the groups. Okay. And what you see here is that at three months, the regional connectivity is significantly higher in our high-risk infants compared to the low-risk infants. And then we see the reverse by 12 months of age. And in particular, we see far stronger intraregional connectivity in the left hemisphere for the low-risk controls compared to the high-risk controls. But we also see somewhat significant inter-he– so left and right frontal, left and right parietal regions. And then when we looked at global connectivity, you can see that pattern where here it’s high-risk infants are far greater than the low-risk controls and we see these different developmental trajectories. By 12 months, we’re seeing the reverse.
So that’s it for brain mechanisms and now let me talk about maternal behavior. Between six and 18 months, our moms recorded these videos. We gave them some vignettes we wanted them to do with the babies, playing with novel toys, etc., etc. And the data complement what we collect in the lab but provides us with a pretty fairly naturalistic mother-child interaction from which we can ask the question, and I’m just going to go through these slides quickly and not go into the details of all the different work here, are moms of high-risk infants interacting differently with their babies, particularly around language and gesture, and are the mothers of the infants who end up with Autism interacting differently to the other two groups?
So this is looking at maternal responses to infant vocalizations at nine months and suffice it to say all three groups are doing exactly what they’re supposed to be doing. They’re responding positively. They’re giving much more reinforcing feedback when the baby produces a lovely consonant vowel syllable than when they’re just doing vowels, and so we see no group differences. The moms are all doing the right thing.
This is during the toy drop. The mother drops the toe and then we see what does the baby do and how does the mother respond to the baby trying to get the mom to get the toy back for him or her? And, you know, do they request, do they ask a question, are they commenting, blah, blah, blah? No group differences.
And then we looked at maternal gesture at 12 months. And here, I mean there’s a lot I also have outlier dots on my box plot so you can see there’s a lot of variability, but we see that the low-risk controls are pretty much equivalent but there’s a lot more variance about the mothers of the infants who end up with Autism. The group that’s actually gesturing the most are the mothers of the infants who don’t end up with Autism. Okay. But again, we’re not seeing that these mothers are doing the wrong thing with their babies.
Comparisons to SLI
So let me turn now and compare this to SLI and as you know, based on what I said at the beginning, this is not my own data, this is coming from the literature. And what I want to ask here, what we’re picking up here is we haven’t found anything that’s particularly definitive and predicting to Autism. When we want to predict to Autism based on the work that’s been done with this high-risk infants, we’re going to need to develop a risk model that incorporates a whole bunch of features, none of which can stand alone in telling us that at 36 months this is a baby with Autism. And I think we’re currently working on pulling out data together to develop these kinds of risk models and our expectation is that demographic, behavioral, and most especially neural response patterns in the first year of life are going to be important components of this multifactorial kind of model.
So what do we see in the literature on SLI? If you’re a male, you’re more likely to have SLI than females. Family history. That actually carries a huge proportion of the variance where we see it. Parental concerns. Delay in early gesture use and language milestones and slowed growth in language during the preschool years. Identical to the data that I showed you on Autism.
What do we know from neural and cognitive mechanisms? Well, going through the literature, and this was in the reading that I assigned to you, we see lower resting frontal gamma. There’s some differences here. Much of this work comes from April Benasich’s lab and there’s also a group in Finland who’s done some work but mostly from her work. But in her findings, she found lower between the 16 and 36 months and that was not the age range that we found the biggest differences. Higher rapid auditory processing threshold to tones at seven months. Atypical lateralization of response to tone pairs at six to 12 months. Delayed mismatch responses to changes in syllable length at two months and that comes out of a study in Holland. Delayed mismatch response to changes in word stress at four to five months. And reduced mismatch responses to tones at six months.
So, what can we draw conclusions from here? Well, this doesn’t quite capture the kinds of things that we looked at in the kids with Autism. Okay. We’re not really doing the same things.
Gaps and Future Research
So to conclude, what are the gaps and where do we need to go if we really want to understand what are the similarities and differences in early risk markers and developmental trajectories in these two populations? So I think there are many parallels in the risk factors and mechanisms for language impairment, particularly the atypical lateralization and perhaps reduced gamma, a more general marker of brain integrity. But our studies of brain development have employed different paradigms and different measures. So we just can’t mix our apples and camels here at all.
What we really need, I think, are longitudinal studies that do highlight developmental trajectories. I think this has been enormously productive. It’s extremely challenging, I get it. You know, I’ve aged a lot more than the seven years since I got started on this. But I think we really want to know from longitudinal studies, maybe there’s something to the difference in April’s finding of differences in gamma at older ages that we didn’t see in the high-risk Autism infants. We don’t really know. We need longitudinal studies that may actually tell us whether there are these different developmental trajectories. We already know that Autism is a disorder with an onset during the second year of life, that we don’t see those patterns in kids with SLI. So it would be fascinating to have brain studies that really did much more parallel work.
And finally, my one tiny, little clinical implication since I’m not a clinician, is that studies of risk I think do contribute to the development and implementation of targeted early or preventive interventions because we see that across these two populations there’s a lot of vulnerabilities, particularly in the domain of language and we’re really not doing enough and certainly the literature on Autism and early intervention suggests that we can make an amazing difference in the lives of infants who are captured very early on. And if we can pick out those infants who are at the highest risk, we may be able to do far more improve the outcomes and perhaps begin to slow the ever-rising prevalence rates in our society.
And so then just to thank everybody, my funders, my collaborators, our labs over time, and most especially the children and families who participate in the projects that I talked about today. And thank you all for listening.
Questions and Discussion
Thank you. I’m Uma Soman at Vanderbilt University and I predominantly work with children who have a hearing loss. In your assigned reading, you talk in the conclusions about auditory processing differences in children, with Autism Spectrum Disorder as well as SLI, and my question was related to Brandon Keehn’s work. When you did the auditory processing paradigm, did you consider using words that did not follow the phonotactic rules of the child’s native language given that there were some differences in perceptual narrowing with da, da, and ta?
No. We didn’t. There’s only so many things that you can do. Rhea wants to comment on this.
I did that. And there was no difference. I felt the same way . I was heartbroken.
Oh I wasn’t heartbroken. If I was heartbroken, I would have quit so many years ago because I’ve always been wrong. Okay.
I’m Tracy Centanni. I’m a postdoc at MGH Institute for Health Professions and my work is in neuromechanism and genetics of dyslexia and so I was actually curious about a couple of different things. The first is, you were looking at the differences in connectivity across hemispheres and I was wondering if you had any thoughts about whether or not that could be due to a difference in neural pruning post birth and if that could have any implications for connectivity.
And actually is a nice transition into my second question. So I’ve actually worked in rat models of autism in my doctoral work and so I was wondering if you’ve thought about doing ABR recordings, auditory brainstem responses, looking at basic sensory responses versus higher order language and if you think those two thinks are related in the, of course, human model of autism versus the animal models.
Yeah. So there’s kind of a long history of looking at ABR in autism and I think the most recent review has basically come out and said, guys there’s nothing in it. In my new life, along with babies, we are now studying minimally verbal kids. And one of the primary things we are looking at there is auditory processing and we’re using EEG, ERP to look at auditory processing and I work with these amazing engineers. You know, engineers can do stuff with crappy signals that the rest of us would have no idea how to do and so they’ve managed to extract, because these minimally verbal kids have challenging behaviors and extremely challenging brains, and when we look at their cortical response just to tones, you know, it’s a huge mess. There’s just, you know, it looks like garbage. But these engineers have been extracting out the signal from different components that reflect processing at different levels of the auditory pathway. And based on the initial work, this is work of my colleague, Barbara Shinn-Cunningham and her postdoc, we are not seeing anything like a problem at the auditory brainstem. Nothing there. Basically, we are seeing differences in their auditory seen segregation but it seems to be exclusively at the cortical level. It’s not lower down the pathway which I think is really interesting and, you know, that yeah there may be differences in terms of how we look at animal models. My main beef there with animal models is, if you can’t replicate the developmental trajectory, alright, yes the seeds of autism are laid down prenatally. Presumably, brain development goes awry sometimes during the second or third trimester of pregnancy and as you can see when we look at the brains at three months they’re majorly different but which does not necessarily travel with behavior. And I think we’re going to have to grapple with these developmental changes in the brain as well as behavior when we refine our animal models.
Karen Chenowski, Boston University. Helen, the enhanced P150 amplitude that you found in the– okay. To what extent do you think that could be a protective factor and functioning as like the subtractive effect that Sean was talking about?
You know, I’ll say two things. First, it could be for those infants and it would be quite interesting, you know, unfortunately, well, I should fortunately. That’s not a nice thing to have said. We don’t have kids with very poor language outcomes so I can’t tell you whether, you know, that would be kind of cool to look at, right? So it may be that we’d see different things there. It looks like it’s a protective factor. I’m not quite sure what one does with something like that in terms of developing an intervention. You know, that’s for you guys, the clinicians here, to figure out what you do with that. But it’s certainly, I think, interesting and it looks that.
But the second thing I’m going to say is, replication, replication, replication. Alright? Until somebody else shows something similar, you know, take what I have said with a grain of salt. I feel more confident in the atypical lateralization and the reduced connectivity, partly because we’ve got replication at least within our sample of cross paradigms, but it also tells the right store that fits with everything else we know from the existing literature on older kids. So I feel a little less worried about that. The enhanced P150, I think, is an intriguing finding, but I’d love to see somebody replicate it.
I am Eliane Lazar-Wesley. I am scientific review officer at the Institute on Deafness and Communication Disorder. I was intrigued by the fact that, unless I misunderstood, that the outcome among the autistic children, they all had language development?
Yes, in our sample. And that’s fairly typical when you study these high-risk infants.
So how do you explain that because the outcome is 25 percent of the autistic children don’t develop language?
Right. So, you know, I think that’s a really good question. I think one of the answers to your questions is that we run our research in a highly ethical manner. And we all do, okay? All the people studying these high-risk infants. Many of them, even all the autism outcome infants, enter early intervention early, okay. If we see significant developmental problems at 12 months or at 18 months or at 24 months, they’re shipped off. You know, they get a fast track diagnosis and they enter early intervention. And in Massachusetts, specialty early intervention services on average is 20 or more hours a week of intensive early intervention. So the fact that we don’t, I would like to, it’s not a randomized control trial in any way. There’s nothing controlled here, but I think the fact that across the BSRC we’re seeing sort of more optimal language outcomes in these infants. It’s because we’re all affiliated with, you know, clinical facilities and early intervention and so we’re doing the right thing by them. You know, so, I’d love someone to take a look at this more systematically, so that’s how I would explain it.
And what is the spectrum of language that they acquire? Are you talking about knowing ten words or are you talking about learning to speak full sentences?
They all have at least phrase speech by 36 months.
They should all come to your program.
That’s not my, I mean, there’s other places that are doing, because we’re not doing the clinical lab, we’re not doing the intervention. There are other places that are that really are doing a great job, too.
But they start about 15, 18 months or even, when do they start with this treatment?
Early intervention, some of our, listen, I think you have to sort of step back and say, who the heck signs up for a study like this, alright? It’s a high SES sample. Highly neurotic mothers. I showed you the data. Six months they’re very worried about everything. They’re very knowledgeable. These mothers are really knowledgeable because they already have a child with autism. They know the lingo, you know. They’re asking at six months. Today I’m worried because this week I haven’t seen any joint attention episodes. Okay. Joint attention. Six months. It’s not happening, right? Not for anyone. So it’s a very special sample of kids and who can afford to take off the time and, you know, schlep to Children’s Hospital at three months and six months and they’ve got all these kinds in tow. It’s not a population-based sample and, you know, I think Sean mentioned this point in his talk related to other variables, I think you have to ask who enters into the research. So the good language outcomes, I think, are related to that fact.
I think you just spoke to the importance of early diagnosis and of course we know it’s a holy grail and there are many reasons in addition to our belief that early intervention is going yield better outcomes, we also are concerned with our neurotic mothers’ concerns and worries and, you know, it is just part of motherhood that something kicks in and you just start worrying about wars that are barely on this planet. You’re so obsessed with how your kid is going to do in life. So in part, it’s very intriguing to me that you sort of mentioned in passing that there might be some indications about early vocalizations that might be yielding at least some predictive validity, of course yet to be fully met. And I was wondering if we could return to that topic and get a better sense of what it is that is special perhaps about these vocalizations.
Karen’s going to have answer that question. This is, I mean truly I am just, you know, I’m just, the information flows through me. It’s really all these people who have done the hard work of the project and I just have the privilege of pulling it together for you. So Karen, why don’t you talk more about the early vocalizations in our sample.
I would love to. So what I looked at as part of my dissertation was the distribution of speech-like, non-speech-like and proto vowel utterances in the three groups that Helen mentioned at 12, 18, and 24 months and one of the most striking results is that the HRA-positive kids, the high-risk kids who go on to get a diagnosis at 36 months, vocalize significantly less and this is actually consistent with things that Amy Weatherby has found and Kim Oller and something that I was really intrigued by that again I didn’t expect, I hypothesized exactly the opposite. I thought that the high-risk kids would have a greater proportion of proto vowels because I thought, you know, if you’re delayed in speech or language then you’re not going to have a really nice adult-like vowels, you’re going to have these sort of very centralized schwa-like vowels a lot and that’s what you’re going to rely on and so maybe that’s a thing we could look for as a marker. Exactly the opposite. The kids who eventually went on to get a diagnosis had a lot fewer of these proto vowels which I now, you know, have to like lie down with cucumbers over my eyes and think about for a while.
And before you go, you know, I always hate measures that end up with this population’s going to do more or less of something than that population because that’s very hard to translate for an individual. How do I know with this kid? And of course, there could be norms. I’m wondering if you could give us a sense, you talked about prototypes not being as plentiful. Was there anything else about their vocal behavior, pitch modulation, repetitives, those sorts of things that might be qualitative and maybe inference at this point but might be since you’ve looked at it so closely. ?
Well, it’s a good question. I didn’t look at any of the atypical vocalizations in detail but I have to say nothing really jumped out at me. You know, there are some kids who kind of like to get into repeating pop, pop, pop during the ADOS with the bubbles but it doesn’t seem to go by group. I did discover, though, that consistent with Rhea Paul’s work that the middle eight consonants, especially at 12 months, I think is what I found, are a really interesting differentiator between especially the high-risk autism negative kids and the high-risk autism positive kids in that at 12 months the HRA minus kids has like the lowest number of these middle eight consonants. By 24 months, they had the highest number. So that’s another indication of these really interesting, exactly, developmental changes and possibly because of protective factors. Nothing was really predictive, though. There were a lot of things that kind of played in together but nothing was really like oh, this is going to tell you whether he gets a diagnosis.
Thank you. And my last question concerns the recent changes in the DSM-5. We’ve had a lot of discussion, actually internationally, going on about whether this is working better, whether it’s working worse, and of course depending on what kind of stakeholder you are, you may fall one way or another on this. But my concern and specifically question to you is given that the social communication and repetitive behaviors tend to be late-appearing problems for these children, what do you think of the consequences of the new DSM-5 criteria for early identification and early intervention?
Well, actually I think we certainly used to think that the repetitive behaviors were later-appearing, but now that it also includes sensory sensitivities so you can pick up and it turns out there’s a couple of papers now in the literature on high-risk infants to suggest that you can pick up at 12 months risk that cuts across these behavior domains.
I think altogether what this line or research, if I was to step back and sort of think what have we learned a decade into this, and I actually pulled back and do that last year at the BSRC meeting, is this is a subtle change in development, that that’s what signals autism, that we need to develop these models or risk early on. Now we need to start pulling together our different risk markers, and I think we need to be shifting from the idea that you provide intervention only after there’s a full-blown diagnosis of SLI which is not going to happen until three or four or autism which may also not happen until the child is two or three even in the best of circumstances. We need to pull back and argue that intervention, preventative intervention is the way to go, just like we have done in cardiovascular disease where we know that high blood pressure — there’s nothing wrong with having high blood pressure. What’s wrong with having high blood pressure is it’s a risk marker for cardiovascular disease. And in others of medicine, we’ve been very, we’ve invested, you know, 40-50 years in developing these kinds of risk models, and we haven’t done that and I think that’s where this whole field needs to move.
And I don’t know whether the best way to do that is in our little silos where I do it for autism and you do it for SLI and Sean does it for ADHD or whether we may in fact be far better off thinking about looking at risk markers and risk models that may be slightly different but nevertheless there’s going to be a lot of shared, common factors there, and that we do this across neuro developmental disorders from the perspective of at least let’s get them into early intervention, that birth to three program, and enrich those programs from our point of view here at ASHA, let’s really infuse early intervention with a targeted focus on how to promote language in very young kids who are right at the right period of doing that. That’s where I think we need to be going with this focus because nothing that we found, we didn’t find that magic bullet that says if you do this you’re going to have autism down the line and I don’t think it’s there anymore. So that’s where I stand on that issue and I don’t think DSM-5 makes a difference one way or another.