Professor Maggie Snowling on ‘Dyslexia and Language Impairment’
Recorded on 29 September 2017 at the Welsh ACAMH Branch as part of the conference, ‘Dyslexia from assessment to intervention’. The day promoted the notion that the assessment of literacy difficulties should be the first step on a clear pathway to individualised interventions for children or young people.
ACAMH members can now receive a CPD certificate for watching this recorded lecture. Simply email firstname.lastname@example.org with the day and time you watch it, so we can check the analytics, and we’ll email you your certificate.
Bio via St John’s College, Oxford
I came to St. John’s in September 2012 as President and Honorary Professor in the Department of Experimental Psychology. I also hold an honorary contract as a clinical psychologist in the Department of Psychiatry. Before this, I was Professor of Psychology at the University of York where I co-directed the Centre for Reading and Language.
I completed my first degree in Psychology at Bristol and my doctorate at University College London under the supervision of the eminent psychologist, Uta Frith. Later I qualified as a clinical psychologist. I am a Past-President of the Society for the Scientific Study of Reading and was one of the Joint Editors of the Journal of Child Psychology and Psychiatry. I served as a member of Sir Jim Rose’s Expert Advisory Group on provision for Dyslexia in 2009 and as an expert member of the Education for All: Fast Track Initiative group in Washington DC in 2011. I have offered advice to the Government Departments of Education and of International Development, particularly in relation to the early years and specific learning difficulties (SpLD).
I am a Fellow of the British Academy and a Fellow of the Academy of Medical Sciences. I hold honorary doctorates from Goldsmiths London (2010), University College London (2014), Warwick University (2016) and University of Bristol (2019) for contributions to the science of reading and dyslexia and was appointed CBE for services to science and the understanding of dyslexia in 2016.
My research focuses on children’s language and learning and I am specifically interested in the nature and causes of children’s reading difficulties and how best to ameliorate them. In recent years, my work has focused on longitudinal studies of children at high-risk of reading problems, and I have been involved in the implementation of randomised controlled trials assessing the efficacy of reading and language interventions in projects led by Charles Hulme in the Department of Education.
Professor Maggie Snowling – Well, good morning, everybody, and I guess I should also say good morning to the people who are live streaming this event, which is very exciting. So particularly any of my friends out there!
Today, I know that the theme of the conference is dyslexia from assessment to intervention and I am actually talking about dyslexia and development language disorder, or I will do when I get some slides to look at.
OK, the reason I’m doing this is I can’t anymore speak about dyslexia without talking about developmental language disorder, because unlike the position I took in the earlier part of my career of many years, which I thought dyslexia should be considered a very specific phonological disorder, my views have now changed. I think there is much more relationship between dyslexia and developmental language disorder than we had realised before, and one of the reasons that my views have changed is changes in the sort of methodology that have been used for undertaking research on dyslexia. So more of that later.
First of all, some definitions. Here are two classic definitions from DSM five. DSM five categorises dyslexia and developmental disorder, both as neuro developmental disorders. Dyslexia comes under the rubric of a specific learning disorder and is one of a number of specific learning disorders. In the case of dyslexia, it affects fluent word reading, word reading accuracy and spelling abilities, but notably not necessarily comprehension.
In DSM five, language disorder is classified as one of the communication disorders and is characterised by persistent difficulties in understanding or production of language through different modalities.
Now those of you who are avid social media users will perhaps be aware that last Friday was DLD day, and that was a day in which a group of us, led by Dorothy Bishop, presented a consensus view gathered from 50 international professionals that suggested that language disorder could be, or should be, relabelled as developmental language disorder (DLD) and that term should be used for children who have language difficulties that are not associated with any specific biological condition.
Downstairs on the table, I’ve just left some handouts about DLD because I think all the professionals in this room ought to know about it and through this talk, I will try to use the terminology Developmental Language Disorder or DLD, although sometimes because the slides were produced a month ago on the slide it will say language disorder. So just a little bit about terminology. I don’t want to confuse you. I’m going to be talking about the relationship between dyslexia and developmental language disorder.
But of course, that relationship is, by its nature, a confusing and complex one, because language is a very complex system and it comprises a number of different subsystems. As I already mentioned, dyslexia used to be thought to only affect the phonological system of language and not grammar or semantics or pragmatics. But I’m going to be arguing that that is perhaps rather a simplistic view of the condition.
In the literature, people for many years of trying to understand the relationship between dyslexia and DLD and there have been three primary hypotheses about this.
One is that dyslexia might be considered a sort of developmental outcome of an earlier language disorder. The second, and perhaps somewhat similar, is that dyslexia is just a mild form of language disorder, possibly when there’s been a language difficulty that’s resolved. And then the third sort of hypothesis has been that dyslexia and language disorder occur there, they are comorbid disorders. They may have different aetiologies, but you sometimes see them in the same child.
In 2004, Dorothy Bishop and I, feeling rather dissatisfied with each of those explanations, suggested that in order to understand the relationship between dyslexia and developmental language disorder, you have to consider the two dimensions that underpin fluent reading for a different meaning.
First of all, phonological skills which are at the core of decoding, and here in the centre of the slide, they vary in the population from weak to strong and also broader oral language skills sometimes called non phonological language skills, which also vary in the population from weak to strong. Where a child’s profile sits depends upon the interaction between these two dimensions or axes.
So dyslexia essentially can be seen along the bottom here. Dyslexia is characterised by phonological problems, but sometimes there are comprehension difficulties that complicate that. Up in the upper left quadrant is a group of children called poor comprehenders who have language difficulties but essentially don’t have dyslexia.
One thing that at the time I think we under estimated, was the importance of timing in thinking about the interactions between those two dimensions and that’s something I’m going to talk about today.
By timing I mean development itself. Each of these sets of skills, the phonological and the non phonological language skills are developing and one of the things that we need to understand, if we want to understand reading difficulties, is how the development of the two different dimensions comes together.
Now, one of the reasons that I changed my view about the specificity of the deficit underlying dyslexia was because of findings from studies of children at family risk of dyslexia. There’s now some 25 years of research on this topic and around 20 studies published and ongoing.
So a few years ago, a couple of years ago, Monica Melville* and I undertook a meta analysis of these studies to try to understand better what the early precursors of dyslexia are, because if you study children at family risk, you essentially recruit children in the pre-school years before they start to learn to read. So before reading or some diagnosis or some form of instruction could have affected them and you compare them with children who are from families in which there is no affected member with dyslexia.
We know dyslexia is highly heritable. So by studying children at high risk and children at low risk, over time, you can start to unpick what might be precursors of a later, if you like, dyslexia diagnosis.
I’m just going to very briefly tell you about the findings of that study, and I’ve chosen to do this just in two schematic diagrams. So what we have here is development from infancy through to grade two.
So usually these studies recruit children in the pre-school years and then they make an assessment of their reading skills around about year two, grade two. At each point in time, we have a benchmark control group who are at low risk of dyslexia, no history of dyslexia in the family and we compare to them children at family risk of dyslexia. Now, let me see if this works.
So what we see in the pre-school years, if we compare children at family risk of dyslexia with controls, is very early on some evidence for atypical auditory and atypical speech processing.
These findings primarily come from studies using neurophysiological measures, principally EEG and Meg, and then later, say, age three and four, what we see is a pattern of delayed speech and language development. That is poorer speech and language development than the typical control group. That’s true in the family risk group as a whole. But of course, what we really want to know about is how do those patterns relate to the outcomes? And here we have the outcomes of family risk studies.
In these studies, about 45 percent of children reach a threshold which you would be able to say that they have a dyslexic difficulty in year two, and of course, some of them have pure dyslexia and some of them have dyslexia with poor reading comprehension.
The other thing that these studies have revealed is that in these families, there’s another group of children who have some literacy difficulties, but they don’t reach a sort of threshold where you’d want to actually identify them as having a significant reading problem.
And in terms of the characteristics, those children have spelling problems and they have problems with challenging decoding tasks like non word reading and very interestingly, many of those children go on to get some kind of dyslexia diagnosis later on. Perhaps in teenage years, perhaps for those who go on to college, perhaps college level.
So a key question, though, is what determines those outcomes?
So, first of all, that what this also indicates is move from mild to more severe reading problems is that dyslexia is really a continuum. It’s a dimension and where we put the cut-offs is arbitrary and that will depend upon externally agreed criteria.
What differentiates these different outcomes according to this meta analysis or what we kind of inferred was what happens to language during those pre-school years? And it sort of turns out that if an early language difficulty resolves, so that the child’s language skills are now within the normal range, they’re much less likely to get a dyslexia diagnosis than if they have a continuing or persisting language impairment.
So here we have just thinking about that language dimension from OK to sort of affected and this seems to moderate the outcomes that you get in children at family risk of dyslexia.
OK, so just to perhaps sum that up, think what this meta analysis revealed was that a familial dyslexia is associated with difficulties in language and in phonological skills. I haven’t talked about the phonological skills much at the moment, but they were certainly affected, but that it’s more likely that dyslexia is more likely to be diagnosed if you have a persisting language difficulty into school age.
We also know something about children in sort of risk families who don’t have an identified reading problem. They often still have vulnerabilities in the literacy domain and they’re sometimes referred to as having a broader phenotype of dyslexia. I’ve sometimes seen them quoted to have compensator dyslexia, though I think it’s not quite clear what they’re compensated for. If anything, their difficulties are getting greater over time. And thirdly, we see the dyslexia as a dimensional disorder.
But of course, those conclusions come from a meta analysis across a number of studies, and they’re not each, although each study is longitudinal, it isn’t a longitudinal study per say in what’s really needed is longitudinal work that can track how language impacts and then to read. I’m going to talk about data from one such study today and really try to address five aspects.
First of all, I’m going to talk about language as a foundation for reading. I’m then going to look at the profile of dyslexia and development language disorder early in development to try to set up some hypotheses about risk factors for dyslexia and look at who actually becomes dyslexic in a high risk population. That will lead me to answer the question of what the relationship is between dyslexia and DLD, and we’ve seen some of those hypotheses and then hopefully there is time to talk about implications for intervention.
But first of all, language as a foundation for reading. I’ve spoken about different aspects of language and made a broad distinction between phonological and non phonological aspects of language. And we might think about these differences here. If you think about these different language tasks, tasks like articulation, word repetition and non word repetition place a high reliance on speech processing, on phonological processing. But of course, they’re not devoid of language. You couldn’t repeat a word very well the first time if you hadn’t heard the word before, except insofar as you can do the speech processing part.
But when we’re asked to repeat words and even when we’re asked to repeat non words, we do activate our lexical knowledge. In contrast, the tasks in Blue Task, which you’ll be familiar with from a test like the KELF vocabulary, sentence recall, receptive grammar and basic concepts, rely less on speech and more on broader language systems.
So one might expect that these different aspects of language would have different roles in development. And there is some evidence from behaviour genetics that the phonological aspects of language are more highly heritable than the non phonological aspects, which in turn are explained by more environmental variance.
So there are some theoretical reasons for keeping these two aspects of language apart. But when you come to actually look at data, it’s quite clear that all of these skills load on one language factor. And so when we try to model the language foundation of reading, what we need to do is to allow individual differences in each of these language tests to come together to define a common language factor, and that’s what we’ve done here.
So here we have defined a latent factor language and all of these measures load on this factor.
Now, if you can see, and you might be able to see in your hand out the loadings, you’ll see the loadings of these non phonological factors on language are high, but they’re less hard for these phonological measures of language. Which is, perhaps, in line with the idea that these different components are a bit different, but what it also indicates from a sort of statistical point of view is a lot of unexplained variance in these skills from this factor.
So what we did was we had a secondary factor called…which I’m calling Speech. It’s probably some kind of speech motive factor and the residual variance, so things not accounted for by this, now go on to this second speech like facture.
So let’s for a moment just then say that there are two aspects to the Language Foundation. Early on in development, there’s a common language factor and then there’s something else which is sort of speech like. And what we want to know is how those two different components predict later literacy development.
Now, the full model is in your next slide. I’m not going to talk about that, but just to say we do this full model. But just to keep it simple, I’m going to actually take you through step by step how we think about development.
So what we…I have shown you the language foundation at age three and a half. This is one year later, and at this point in time, age four and a half children are really developing prereading skills, and we sometimes refer to this as the foundational skills, the triple foundation.
And there are three skills that form the triple foundation of language across all the European languages. So there has been cross linguistic research to show that these are the three key skills that seem to be predictors of reading. You will be aware of phonological awareness or phoneme awareness. We’ve got here letter sound knowledge and rapid learning. Rapid learning is measured by a test of random objects and colours. Letter sound knowledge here was measured by a letter sound knowledge and letter writing and phoneme awareness by the ability to isolate phonemes at the beginning of the end of words.
So we put these into the model and we look to see whether the three and a half year old language skills predict these four and a half year old language foundations. And what we find is the language is a predictor of all three of these foundational skills. But speech is not a predictor at all. So speech doesn’t account for any of the variance in four and a half year old prereading skills. That’s quite interesting because those of you who work with children of speech sound disorder will know that even though these children have quite significant speech difficulties, they’re actually at low risk of reading problems. They are slightly heightened risk compared to typical controls, but it’s a low risk and certainly nowhere near the risk that we see in children with language disorder.
So what we now know is that the three and a half year old language skills predict prereading skills and they particularly strongly predict phoneme awareness. Wait, here is point seven. What we now need to ask is a year later, at five and a half, these children will be now learning to read. How do these prereading skills predict their ability to decode? Because decoding is, I guess, the core of learning to read at this age. And that’s shown on this slide.
We measured decoding using two measures of single word reading and a measure of spelling and we get a very nice, strong factor there. What you can see here is there are two predictors of decoding, individual differences in decoding a child’s letter sound knowledge and their phoneme awareness.
Rapid naming is not a predictor at this relatively early stage in development, in reading, in English. It is in other alphabetic languages, is predictive of reading within a year of instruction, but not in English. And that’s almost certainly because English is a harder orthography to learn.
OK, so I think just one thing I want to point out about this and this is that in line with the sort of idea that phonological skills are crucial for learning to read. We see that here, we see the alphabetic foundation at age four and a half predicting reading at age five and a half. And that’s in line with much of the literature and indeed much, much reading research has focussed on that particular window of development. But if you go earlier to age three and a half, you can see that the foundation of phoneme awareness is actually oral language.
We also want to ask about outcomes later on in development, but as you might know, reading is a very stable trait. So actually individual differences in reading at five and a half correlate above point nine with individual differences in reading at six and a half, seven and a half and so on, which in itself is a kind of reminder that one should never just wait to see if a child will pick up, because actually their rank ordering at five and a half is likely to be the same at eleven and a half unless you make some kind of intervention. But we also want to know about predicting reading comprehension, because that is the goal of reading.
We were first able to get a really reliable measure of reading comprehension at age eight. Here it is in this final model, and what you can see here is that there are two predictors of reading comprehension, decoding from five and a half, but also a very long range effect of early language on later reading comprehension. That’s really quite striking. Of course, cross-sectional, if you look at language and reading comprehension, aged 8, you would also get a prediction. But here we show that there is still a prediction from very early language to later reading comprehension.
So just to sum up that piece. Language, I hope I have convinced you is the foundation for learning to read, it’s an important precursor of the development of phonological awareness. The effect of language on decoding is actually via phonological awareness and letter knowledge. So we think about mediation here. So if you look at language. It goes through here to decoding and through here to decoding, so that’s known as a mediated relationship.
In terms of reading comprehension we have both direct effect of language and indirect effect via decoding. So if you accept that, then one has to accept that if you’re thinking about children at risk of dyslexia, you have to contend that there are two kinds of risks. There are children who have got phonological problems. Phoneme awareness problems are also children with lurking language difficulties because that language problem will probably have a mediated effect on their ability to decode.
So let’s then look at some children who are at risk of dyslexia and children who are also at risk of developing developmental language disorder.
I’m going to now tell you a bit more about this longitudinal study from which I just showed you some data actually. This was a study which followed the other family risk studies that we had analysed, because what they revealed really was if there is a possible relationship between dyslexia and language disorder, what family risk studies ought to do is include a second group of risk children, children who in the pre-school years have a diagnosed language difficulty. So that’s what we did in the Welcome Language and reading project.
We compared children at family risk of dyslexia across development with children who had in the pre-school years a developmental language difficulty. We recruited these children at three and a half and we assess them at approximately annual intervals until they were eight. Sorry, that’s…that’s there.
Just a word about our recruitment strategy. We asked parents and also speech and language therapists and early years professionals to refer children to the study who were at family risk of dyslexia, who had a pre-school language impairment or who were, as far as everyone knew, not at risk of either dyslexia or language impairment.
So we had three groups recruited. We then did various tests and assessments, questionnaires. Firstly to determine which children were really at family risk of dyslexia, as opposed to parents saying they were. That involved assessing parents as well as assessing children. And then we also assessed each child’s language skills so we could determine which children fulfilled criteria for a pre-school language impairment. And we did that using a criterion score on tests from the KELF also the test of expressive grammar.
This actually led us to end up with four subgroups rather than three. So we had typically developing children. Then we had two groups of children with language impairment. One group with pure language impairment and one group with language impairment who also had a family risk of dyslexia. And interestingly, many of those parents didn’t realise that there was a family risk of dyslexia. They didn’t actually realise that their own literacy skills were very, very poor. And we also had a group of children at family risk of dyslexia who had no pre-school language impairment.
And so our plan was to compare those three different risk groups against the typically developing control group throughout our study, just focussing in on pre-school and the nature of the language problem. Again, I just simplified this here. We’ve got the four groups on the left of the slide and the bars are just indicating the groups that have the deficit is actually tautological to say that the two groups with language impairment have got language deficits that’s simply the way in which we classify the children in our study.
But what’s interesting in this slide is that those two groups, and the group at family risk of dyslexia, only had phonological deficits. So on those more phonological language measures, these children were scoring significantly below the low risk, typically developing group. So they had problems on tests of articulation in terms of percentage of consonants correct, they had problems with word and non word repetition and they also had some problems in assigning tense and third person singular in a test of expressive morphology but that problem went away within 12 months.
So children at family risk of dyslexia who don’t have a language problem in the pre-school years, nonetheless seem to have some phonological processing difficulties, which of course is interesting, given a long tradition of dyslexia being construed as a phonological deficit.
Now, we were also interested in looking at skills beyond the language domain, and we therefore had a battery of tasks, tapping fine motor skills and various executive function tasks, tapping the control and manipulation of attention. I’m just showing you here data from the family risk only group and data here combined from the two language impairment groups with regard to their difficulties in these domains.
So compared to the typically developing control group, 82 percent of children at family risk of dyslexia were free of any comorbidity, but some of them did have problems. In contrast, in the language impaired group, only 54 percent were free of comorbidities, occurring difficulties. In the language impaired group, 18 percent had significant problems with fine motor control, 21 percent significant problems with executive function and seven percent with difficulties in both of these domains.
So I think it’s important to notice that these two groups don’t just differ in language, in the nature of the language problem, they also differ in co occurring problems. And we can just summarise the profile in terms of the effect size of the deficit for children at family risk of dyslexia, who are in the dark green bars, and children who’ve got a pre-school language disorder, in the light green bars, on a number of tasks I’ve put here phonology, vocabulary, language, motor skills and executive function.
Don’t pay too much attention to the variability between the different tasks, that’s really just to do with the nature of the measure. But within each task, what you can see, first of all, for phenology, the effect size for the family risk group is point six four, which is quite a large deficit because much larger for the language impaired. In fact, the language impaired have got big deficits in all of these domains. But here for phenology, we see a similarity with the dyslexic group who are showing significant deficits.
They’re not showing deficits in the other domains, although in language it’s creeping up. Point four two is a medium sized deficit. So I think the Take-Home message here is that language disorder brings with it a whole range of risk factors, many of which will affect learning in the early years classroom. Children with dyslexia as a group bring a phonological vulnerabilities. But of course, some of them do have other problems, which is why the effect size isn’t zero for that group.
So just to summarise, then. Children at family risk of dyslexia and children with pre-school language disorder show phonological deficits, so that’s a kind of shared risk factor between the two groups. They differ, the two groups differ, in vocabulary, grammar and comprehension. That’s a kind of non shared risk factor for reading comprehension. And also the language disorders group have broad deficits affecting executive and motor skills.
So if you think about all of those risk factors, we can now start to predict who might become dyslexic and of course, here I have to define dyslexia.
So in order to avoid any issues around qualitative distinctions or definitions, we simply took the DSM five idea that the core feature of dyslexia is poor decoding and poor spelling. So we formed a composite of decoding and spelling at age eight and we said that’s our outcome measure. And we defined as having dyslexia. Any child who fell one and a half standard deviations below the mean of the typically developing group who are similar on various other factors as being dyslexic. Though that cut-off was the standard score of 88, which some people think is rather liberal, other people think is rather severe.
The actual, as I’m sure you know, the actual average reading in the population has come up a lot in recent years and this is probably around about 108. So this is 20 standard score points below that expectation. So this is a significant reading problem. And some of the kids had much more severe problems.
Who might you think would fall into that group? Well, children at family risk. Children with poor language, children just with poor phonology. Maybe some children who’ve got problems with executive function because classroom learning might actually be affected and that might affect their reading acquisition.
Here are the data and I’ve split them up into children with typical language and children who had a language impairment in pre-school. The family with risk group are in blue I think, yeah, blue. So if you start off, rather counterintuitively, from the right of the slide, you see seven percent of our control group fulfil that criterion for dyslexia.
And that’s an interesting prevalence figure because it’s the same figure that comes out of population studies. You’ll see, regardless of whether or not there’s a language impairment, being at family risk really elevates the risk of a reading problem. So we go with no language difficulty, 26 percent of the family risk group have got a dyslexia level sort of difficulty. If you’ve got a language impairment and you’ve also got family risk of dyslexia, then the prevalence is around 40 percent.
So if you think back to the meta analysis, we found that overall the study’s prevalence of forty five percent, which suggests to me that in some of those studies there were children with language impairment. But the investigators at that point were not actually screening out children who had a language disorder.
So that’s what we find. To me one of the peculiar things was that the prevalence of reading difficulties in the language impaired group was the same as in the family risk group, assuming no comorbidity.
And I thought that was interesting. I wasn’t really sure. I was trying to understand why that might be. Maybe in the family risk group, you’ve got some phonological mechanism that’s impaired, maybe in the language impaired group, because phonology is part of language. Its language is mediating that problem and then I suddenly realised these figures are wrong. At least they’re misleading, and the reason that they’re misleading is because of development. Because essentially because of language development and in study, overall, about 50 percent of children who had a pre-school language disorder resolved their language problems before the end of the study, which was the time at which we were assessing those children to assess whether or not they had a reading problem.
And so development itself was complicating the sort of what we could see of the picture, more than one might like. If you look at the children who fulfil criteria for language impairment, obviously the beginning of our study of the two groups with language impairment, 100 percent of them two and the other groups zero, because that’s what we did. But if you look at where we are now, at age eight, you can see there’s been some change. So now a proportion of the typically developing low risk group fulfil criteria for language impairment and 23 percent of the family risk group do.
And sorry, sorry, I’m reading up my notes from a different slide. Four percent of the language impaired group and 12 percent of the family risk group now have some kind of language problem and the children who resolve the difficulties in the language impaired groups. So really, time is quite important.
So let’s have a look at what’s happening over time to those children in our sample. And to do this, we extracted data from their language tasks, essentially two tasks of vocabulary, sorry, two tasks of grammar and expressive and receptive task and a test of vocabulary.
We formed a factor score and we looked at change in that factor score over time in the group and we then classify children according to the developmental trajectory that they showed.
So at the top of the slide, it’s actually blue, but you can see black, are children who had typical language throughout the study. So that group includes some children at family risk, as well as our typically developing control group.
These two lines are the group who in pre-school had a language impairment, and you can see that one of them, this one, has a persisting language impairment through time, but the other one has a resolving pattern of language impairment. So by the second point in time, I should have said this is three and a half, five and a half and eight, by the second point in time, the group with the resolving profile now is broadly within the average range. And the other very striking thing is that there’s another group of children who we weren’t expecting to be there. This is this group who have a declining language score relative to age. They have an emerging language difficulty. And even more surprising to me is the fact that 72 percent of those children are at family risk of dyslexia.
So here we’ve discovered something that we didn’t sort of realise, which is that heritable risk of dyslexia seems to carry with it some risk of language impairment.
So the other point just to make is that if you stick something like school entry, which is around about the middle on the slide, you have to start thinking again about who’s vulnerable to a reading problem.
Many years ago, looking at children with language disorder, Dorothy Bishop suggested that the crucial thing is kind of… It’s a critical stage hypothesis. Which is that whether or not a child will have a reading problem or dyslexia depends upon the status of their language system at the point at which reading instruction is introduced. And so this may have a bearing on who turns out to have a reading problem aged 8 in our study and what we need to do is recode the risk according to concurrent risk of a language impairment, not pre-school risk, and that’s what I’ve done in the next slide. That’s what I was mistakenly telling you the data from before.
So you can see if you now classify children as to whether or not they are dyslexic based on their language disorder at age 8, we get a slightly different pattern. Not much change in the low risk group, about eight point five percent now dyslexic, 23 percent of the family risk group. But we start to get more children with that diagnosis who have got a language disorder. So 33 percent of children with a persisting language disorder and/or an emerging one have dyslexia. If they’ve also had family risk of dyslexia, that’s as high as 52 percent and what this data shows is that here we can see that family risk and language disorder are additive risk factors for dyslexia.
So, just to summarise that, and it’s quite a complicated picture, the risk of a dyslexia outcome is if you’ve got pre-school children about equalling children who have got a language problem then and children who are at family risk.
In fact, we’ve done some analysis on the best ways of identifying later dyslexia in pre-school and what we find is that actually family risk is a better predictor than language, and that’s because language itself is going to vary. That paper published in JCPP and is open access. So if you want to look at it, you can look at it.
In-line with the meta analysis, we did find that dyslexia is more likely in children who have a language difficulty which is persistent at pre-school. In our data, that is children with the persistent profile and also some children who have got a growing language problem over time, and that we do see this pattern of additivity that being at family risk adds together with having a language problem in determining dyslexia. Now, how we interpret that, I’m not quite sure. One possibility is there really is some genetic impact over and above the effect of language. The other possibility is that the phonological deficit that you get because your at family risk is somewhat different from, or compounded by the one you get, because language isn’t able to mediate development of phonological awareness to the same extent.
OK, so next thing, then, let’s return to the question of what is the relationship between dyslexia and DLD?
So these are the hypotheses that I showed at the beginning. Dyslexia is the developmental outcome of language disorder. There is some evidence for that. But we need to think about the critical age here. Dyslexia is probable if language difficulties persist until the age of reading instruction. Dyslexia is a mild form of language disorder. Well, actually, it doesn’t look very simple. If you look at children resolving language problems, they don’t have mild dyslexia, actually, they’re pretty fine. They’re just like typically developing children. So I’m not quite sure that I think that hypothesis is supported.
The third hypothesis that dyslexia and language disorder co-occur, well, that’s always got to be true, that if you have two disorders, if they’re not the same disorder, they can, of course, co-occur. But maybe I mean, given that not everybody in this population is dyslexic in language disorder, population is dyslexic, there is some suggestion, but let’s have a look at the figures.
So, and there is a mistake on the slide which I have to say which I noticed at breakfast, this shows in our sample the comorbidity between dyslexia and DLD. This percentage is wrong, 60 something, but it doesn’t really matter because these are just percentages in our sample. The point is that there are children in what are essentially these map onto the four quadrants of the two dimensional model. So we’ve got some normal readers who I infer have neither phonological or language problems.
We’ve got thirteen point five percent of our sample were dyslexic, but they had typical language. But also we had ten point six percent of our sample were dyslexic with a developmental language disorder, and then we also have this group who are language distorted but are not dyslexic.
So, given that some children with DLD are not dyslexic, it means that when these two things come together, they can be comorbid. More recently we’ve been looking at the differences in developmental pathway between these three outcomes, but I’m not going to complicate the issue any further by talking about that today.
How do we think about the children with developmental language disorder without dyslexia? Well, I mentioned DLD and did I show you this? This is outside on the table just in case you don’t know about it.
This is the schematic diagram, which was in the paper, which is currently, I think, open access online in JCPP, which came out of this consensus group that we had to keep out the nature and characteristics of developmental language disorder.
And we have suggested that there’s a whole group of children who in the UK are called children with speech language and communication needs and there’s no reason to change that terminology. At least this is true in England, I’m not so sure in Wales, you’ll have to put me right on that one.
Within this group are lots of children with language disorders that are associated with different sorts of medical condition, including hearing impairment and also with various biomedical conditions, for instance, Down’s Syndrome. But there’s a group who don’t have those biomedical conditions and we call those children developmental language disorder. Now, of course, even within this group, there’s all sorts of different areas of impairment, including phonology. Now, our supposition is that the DLD children who are not dyslexic, are children who don’t have any involvement of the phonological system in their profile. When we actually look at that developmentally, they do early on, but it resolves.
OK, so dyslexia and language disorder are both heritable language learning impairments. They show similarities and differences in their oral and written language profiles, and they have overlapping comorbidities. I think dyslexia can occur as the developmental outcome of a language disorder and also I would want to note the language disorder may emerge later in some children at family risk of dyslexia and also can be thought of as comorbid disorders.
Now, let me skip that next conclusion slide and just talk about implications for intervention.
Oh, and how long do I have? Am I okay?
There’s now a very large body of research which shows that the most effective approaches for enhancing, improving decoding skills, is that the core of dyslexia are approaches that combine work on letter sound knowledge, phoneme awareness, but not in isolation, in the context of book reading. Interventions that just work on the phonological system simply don’t transfer to reading. And we and others have done a series of randomised control trials that show that these approaches are effective.
But one of the things that you notice when you do those trials is that there are a number of children who don’t respond very well. And when you look at the nature of those children, what you find often is that they have language difficulties and often they have attentional problems. And as we’ve seen, attention problems often go along with language difficulties. So it seemed to us about 12 years ago now that what we should be starting to do is more work on language intervention.
So I just tell you briefly about some of our work on language intervention. This was actually the second trial that we did. We did an initial proof of principle trial that showed that if you compare children in the early years to get a language intervention compared to those that get a phonology intervention with reading, they have very differential effects, but language can improve. We then said, OK, why wait until reception? That was done in reception, why not start in the nursery years.
So we developed an intervention which starts in nursery and goes for 30 weeks through the first two terms of reception class. There are three main components. Narrative… exercise and promoting narrative, vocabulary and listening skills. They go right through the programme and in the last 10 weeks we included also work on letter sound knowledge and phoneme awareness, though note that was also going on in the mainstream classroom as part of the standard reading instruction diet.
Now this intervention was delivered by trained teaching assistants and in nursery it was a small group, but after nursery it was a daily session alternating between a small group of four children and a one to one session on the alternate day. It was all done in the spoken modality and it was run as a randomised trial.
So we delivered the intervention to a group in each school. And then at the end of the intervention, actually after a year, a follow up, the waiting list control group then received intervention.
The next slide shows the effects of the intervention on various language measures. So any bar that’s above the line where the confidence interval doesn’t go below the line is significant.
So you just look eyeballing this, you can see it had a very significant impact. Essentially, the impact was very strong for measures of vocabulary. These are measures of vocabulary. This is a measure of the different words children use in their narrative. And this one is actually, again, in standard score on the KELF express vocabulary. They also showed improvements in grammar. These were the APT grammar scores and in listening comprehension, we found no significant difference in receptive grammar, or in the mean length of utterance. But overall a very positive impact.
There was also a significant impact on phoneme awareness and letter knowledge, because what they had was being in addition to what they were getting in the mainstream classroom, perhaps, but there was no significant impact on word reading. What hadn’t been part of the package? They had improved along with the waiting group. There was no difference between the two groups. Six months later, we looked at how well these effects were being sustained and this just shows the effect size of again, six months after the intervention finished.
You can see a very strong gain in language with an effect size of point eight three. Point three for narrative. Then we’ve got sort of moderate gains in phonological awareness and lesson knowledge. No gain, no specific gain for the intervention group in decoding, but very importantly for us and unexpectedly, we had a gain in reading comprehension, even though there had been no reading in this programme. So this gain in reading comprehension highlights the fact that promoting language can enhance children’s understanding of what they read.
We also went on to look at what mediated the gain in reading comprehension, and it was wholly accounted for by the gain in language six months earlier. So this was very promising. We were then able to get funding from the Education Endowment Foundation to run a larger field trial to replicate these effects, but also a trial which was run by an independent evaluator, and the training was delivered by ICAN, not by the research team. So this is where we stood back to see whether the intervention would still work.
And the other little thing we did here was we actually compared a 30 week intervention that I just described, with a 20 week intervention that only started in reception. The reason that we did that was that we found some difficulty always in recruiting schools that actually had nurseries attached. So we decided to see whether it would work as well if you started in reception compared to the 30 week version. Here’s the very complex, overly complicated analysis that tells you it did work.
And what we do here, essentially in this sort of model is we measure language at the start of the intervention, and then we measure it afterwards, and this is basically straight after the intervention, 20 or 30 weeks, depending on the group the children were in and delayed follow up six months later. We then look at the impact of the intervention on these outcomes, after we control for early language. So this language correlates well with later language. Just 60 percent of the variance in language at the second point in time is due to where children were at the first point in time. So to get a change, you have to have an effective intervention.
What we found was that the 20 week intervention had a significant impact. It didn’t dwindle over time. It was a factor of twenty one point twenty one. The 30 week intervention had a slightly stronger effect of point thirty, but the differences weren’t statistically significant.
But the take-home message is that this intervention works and it can be delivered successfully by trained and supported teaching assistants in nursery classes and also in early years settings. With the best implementation, it can have a significant impact on later reading comprehension. But what I want to say is there is no quick fix.
We’ve also done shorter interventions and they don’t work, they work in that you can train vocabulary, but it doesn’t generalise what we get here as a generalisation to other language skills and to reading comprehension.
With more and more cuts in services, there is a kind of pressure to have shorter interventions. But actually, it’s better to have a longer intervention for fewer children than to have short interventions, because I’m now of the opinion that they really don’t work. The other thing that’s really important is to ensure that implementation is good and that you support the people delivering it, deliver it well, and you can now buy it with the training from Oxford University Press. I’ve left some of these down on the table downstairs. It will be out in 2018.
I should say immediately, disclaimer, none of the profits comes to us. All of the profit sorry, the profit goes to Oxford University Press, but the royalties go to the Nuffield Foundation, who funded several of these trials and will be driven back into research on speech and language difficulties.
So just to sum up with a final plug for developmental language disorders, I think my take-home message to people in the dyslexia field is that oral language skills are critical to literacy development, and they’re a foundation for both decoding and reading comprehension.
We revamped our rally channel, so do share that link as well. And thank you very much for listening.
Do you want to take any questions?
Speaker 1 – Sorry, Maggie, this might be a silly question. You’ve said here that the foundation of decoding being so critical, can you explain, am I right in thinking the intervention didn’t make that much of a difference on decoding, but yet so… generally thoughts on how do we improve that as separate to the intervention or different interventions?
Professor Maggie Snowling – Yeah, that sort of sounds… does sound a bit confusing. In that trial decoding was, of course, being looked after in mainstream. So what we actually find, there is no benefit of the oral language intervention on the decoding, but the decoding is being taken care of in mainstream. So we were joining up, if you like, the language parts to improve reading comprehension. But what you say is important because what we’ve shown is that the language at age three is helping the foundation at age four and a half.
And that possibly implies that interventions at that stage are helpful in taking children with language difficulties and working on skills that will help them to be ready for learning to read. I think it’s not fail safe because these children who fall into the poor comprehender group who’ve got good decoding, they’ve got language problems and they have learnt to decode. So I think overall, I mean, what your point is and it is certainly not a silly question, what you’re pointing to is the disjunction between the longitudinal data and the data from the intervention.
It’s only really interventions that can prove causal connexions. And what we haven’t yet done is look to see how a sort of an intervention at three and a half year old can improve the reading foundation.
Having said that, in press in JCPP is a trial that we’ve recently finished called PACT, which is Parents and Children Together, and this trial was a version of the Nuffield language, early language intervention, but it was very much adapted to involve sharing books and dialogue around books, introducing vocabulary through the shared reading aspect. In that trial we found, we compared parents who were trained to deliver that language intervention with parents who were trained to deliver an intervention to promote fine motor and self care skills.
The parents who delivered the language intervention, their children’s language improved, but so did their early reading skills. So I think this issue about timing that we’ve still got to really understand more, because I think what you do depends upon what the language system looks like and also the age and stage of the child in reading instruction.
So it isn’t a simple prescription and hopefully we can work towards that. But I think you’re actually pointing to a very important thing there, which is you do need to do different things for different purposes and you think about what the outcome you want and you have to think what’s leading into that and where is the child in those different pathways? I think. Yeah.
Speaker 2 – That leads quite nicely into mine about timing actually, my question. Can you tell us a little bit about what the intervention looks like? And I think what I’m most interested in is how formal it is, because obviously at three, lots of people are choosing not to put their children into formal education at that age. So would it fit into the less formal settings like the schools and sort of pre-school settings and how’s it going down internationally and for example, Scandinavian, Scandinavian countries and Australia, New Zealand, where they’re looking at not putting children as early.
Professor Maggie Snowling – OK, so your question there is about implementation I think and it’s an important question. It is quite formal and it is better suited for children in school rather than in pre-school. And one of the reasons that we wanted to compare the 30 week and the 20 week version was that we wanted, first of all, to check it would just work if you started at reception and it does. You get a bit of added benefit from starting earlier, but it starts well then.
So we know it’s OK. It’s OK as a reception. So now you say, OK, what about the early years? And there we are, not nearly so happy with the version we’ve got precisely for the reasons you say, which is that it is a bit formal and it’s rather hard to implement. And also of course, with children coming in and out, not there every day, it’s sort of difficult.
So that’s one thing. The other thing is that we’ve run a ten week version of it and we’ve run a 15 week version just at the nursery part. And as I sort of implied, they weren’t very effective. They only improved vocabulary and there was no generalisation. So they’re not I mean, everybody should be working on vocabulary in nursery, clearly. But this intervention is quite a lot of stuff and you just getting vocabulary coming up so it’s not good.
So we’ve actually just applied for funding to do a whole year programme, but with this flexibility so that we will try to develop work for the whole setting, which will involve the teacher in understanding about language and how they can facilitate language in informal interactions and then actually identifying within the setting children who need a little bit extra and a little bit extra will be more along the lines of what we’ve got. So I think that’s the way we’re going.
It’d be great to collaborate if you’ve got ideas and if we get the funding, we’ll certainly be working with earliest practitioners to think about all those many, many different settings that that are necessary. Actually the published version, which OUPR are publishing and the training is for the reception part, we’ve ditched the the nursery part for the moment. Yeah.
Speaker 3 – You mentioned that ICAN ran part of the trial and I just wonder how… what you’re doing differs from their programme?
Professor Maggie Snowling – Well, it’s longer and it’s effective. I don’t really know the ins and outs of that. They I mean, it’s a competitor, so it’s a competitive programme. They claim, but they… if you look on the various what works sites in terms of security, of evidence, I think ours is higher and yeah. But they trained and they delivered it along our lines and alongside their programme, and they’ve, you know. Yeah.
Speaker 6 – And so the question comes from Jessica Selman and she says, you mentioned that the population mean reading standard score is one hundred and eight. Why is it not 100?
Professor Maggie Snowling – Well, it should be 100 and it would have been 100 when tests were standardised. You get this drift upwards called the Flynn effect I think, it also happens in IQ, which is why you have to re-standardise instruments. But one might say if it’s an English person at least, and I don’t know about what’s happening in Wales and Scotland, that it’s an effect of more systematic phonics instruction, which would have raised levels of reading at that stage in development.
I don’t think there’s been a comparable shift in reading comprehension levels at all. We are actually being pressed to re standardise our YARK test. I don’t know whether or not we’ll do it, but there is a pressure to kind of keep refreshing norms. For the whole thing, the whole shebang, I mean, they want a digital version and you can imagine.
Yes, I left lots of time for questions so I wouldn’t give you too much top down information. So come on, some more.
Speaker 7 – It’s not really a question, it’s more of a request. On a slide you have some references, but I have the feeling that possibly not all your references are there. Is it possible to have a reference list? Edited, I don’t want you complete words, but ones that…
Professor Maggie Snowling – Yes, I think most of them are on the slides, but I’ll definitely check that and I don’t know, presumably you can make that available? They are all open access, so you should be able to get them easily.
I was hoping you wouldn’t ask Joe. It was not Joe.
Speaker 8 – When you say there’s a group of children that language resolves, do you know anymore about that, about how it resolves and why?
Professor Maggie Snowling – Well, what we’ve done is we’ve compared some of the characteristics of children whose language results with children whose language persists and I suppose that’s the closest we can get to it. The children with the persistent problems tend to be lower in socioeconomic circumstances. They are lower in performance IQ, but I think that might just be a proxy for comorbidities. They seem to have more co occurring problems and there is other thing… No, I think they’re the two things that I would point to.
So I think they might be children with more, not just language, but other things as well and they also have lower SCS. Now, I don’t take that to mean necessarily that means their environment is poorer, but it may also mean they have more language difficulties in the family as well. So I think those children are persisting. They’re resolving ones obviously are better on all of those dimensions. As I mentioned, the resolving ones that they really have, they’re reading is fine, persisting ones have severe reading problems, but so do the children of these emerging profiles. They end up as poor as each other, and that’s striking. So a puzzle to me is why you get this emerging profile.
One of the things I think is that if you have a familial dyslexia with phonological problems that you might actually be able to keep your language going in, you know, when you’re still at home and in the home environment, but when you go to school, then maybe the demands of language get harder or, you know, peer group rather than adult interaction and you start to have difficulties. It would be good to follow those kids further.
Speaker 9 – Hi, yeah, we have some more questions online, if that’s okay, the first one is from Chardo* Vogiatsi, apologies if that’s not the correct pronunciation. Children with DLD are more at risk of developing dyslexia compared to the ones who don’t have DLD?
Professor Maggie Snowling – That’s correct, yeah, yeah. So if you’re… so we know that family risk of dyslexia is a risk factor because we know that reading skills are highly heritable, but if you look at DLD in school years, those children have much more dyslexic like difficulties than the FR children, the family risk children.
We also have a question from Deborah Flitcroft* who says, I’m interested in the family risk factors and were these identified? Did both parents need to have dyslexia to be a risk factor?
The family risk was defined in two ways. There was a self report questionnaire which had various sort of reading situations, literacy situations, and they had to rate how difficult they would find that. And then there was a kind of definition of people with dyslexia, have difficulty with these things. Do you think you’re dyslexic? So there was a self report. If they said yes or probable or possibly we counted that as a positive. And then we also assessed parents on a battery of tasks which included IQ vocabulary, reading from logical awareness, language.
They did a self rating scales of attention deficit hyperactivity and a self rated… Dorothy Bishop C.C.C. for adults, the communication checklist. So what we then did was we looked at… So there was missing data. So all the mums were assessed and about half of the dads and so we had to find a way of judging. We basically said if there is one, if there is a parent who’s got a definite reading problem and we define that as a score below 85, I think on non word reading and spelling and they self report, then we will call that family at risk of dyslexia.
We also have quite a lot of other data on the parents. As I indicated, we’ve got their language and we’ve got their attention. At some point, I hope to look to see how those factors play in as well to the picture. There is one thing that I would also say, which is that we’ve done some work looking at the impact of homelessness, the environment on reading in this sample, and as in all such studies, a rich home literacy environment in terms of books and frequency of book reading and knowledge of children’s authors and so on, is predictive of children’s phonological awareness and subsequent reading skills.
But we also in those models put in measures of parents. Actually, mothers, language and mothers literacy, and when you do that, the effect of home literacy environment drops out of your model. In other words, lots of things that are classed as the environment may themselves actually be about parental levels of language and literacy. That also implies if we’re trying to work with the literacy environment in families, which we should be doing, particularly in children with language impairment, to get the foundations in place, we do need to bear in mind the possibility that the parents themselves have great difficulty in these domains and that may be something for us to look at in more detail in the future.
Speaker 8 – There’s another question online from Angharad* Jones. It sounds as though there could be implications for children in Welsh schools where English reading is not introduced until age seven. Do you know of any research groups, or can you recommend any reading where I could learn more about the implications for children who are taught through a second language?
Professor Maggie Snowling – I’ve just been at a conference on this particular issue. The question of English as a second language is quite complicated. Some children who have, I mean there’s so many, even just in the U.K., there’s so many different languages. And I think also true in Wales. Right, there is more than 14. So, some of these children do fine and some of them have difficulties, and my impression coming away from the conference I’ve just attended, is that most of the difficulties for dual language learners, or children of English as an additional language, those difficulties are really linked more to socio economic and cultural factors rather than to the different languages.
So, yes, language proficiency is important. So in children learning in Welsh, children learning English, it’s seven. I think that’s good to wait because that presumably allows English proficiency to get to the point where that would be more developed than when Welsh is being introduced. There’s no, I mean, the two languages differ in their regularity. So learning Welsh should be actually a bit easier in the sense that you should develop more quickly than English will anyway.
So it sounds to me that that’s a very sensible position to take to start with Welsh. And then there should be transfer then of phonological awareness skills into English. Also, at a time when English language proficiency has come up. There is relevant research, certainly at Bangor University, but probably at other universities as well, looking specifically at the Welsh situation, I think.
Speaker 10 – We have one more question online, it’s from Susan Kite*, who says you mentioned the complexity of knowing what works in terms of decoding and comprehension skills and the importance of early intervention. Can you comment on the persistent decoding and spelling difficulties and appropriate interventions for key stage two?
Professor Maggie Snowling – There is a dearth of research looking at that question, particularly at spelling and writing. I think this should just really be a call for people to be doing that sort of research. I mean, we know a lot about good practise. Ever since, you know, the days of Samuel Orton and the Auton* Gillingham* method, trained specialist teachers have been using multisensory techniques for improving spelling. My impression is that sometimes teaching those sorts of rules, I mean, it does work. I know it does work, but it can take a long time. And then actually using those rules during writing, you know. It’s when you’re thinking about the writing, the composition, it takes a long time to put the whole thing together.
I think there’s a real dearth of work going on in intervention on writing and also research on it, although, of course, there are various government targets, at least in England, about that. So I think it really, really does need looking at and it’s a long time since I’ve been practising as a psychologist. But of course, what I even then was saying was, well, there’s these voice recognition software, various writing software applications that you could be teaching. And I think if a child has a very persistent problem with spelling and writing, it’s really quite important to try and get them to use technology to help them so that they don’t get too frustrated. But I’m not an expert on technology. There are people who are and I think that’s really a great way forward for kids who have those persisting problems. And I know they can be severe.
Speaker 11 – That’s lovely. Thank you very much. Thanks, Maggie. That was fantastic.
Professor Maggie Snowling – OK, thank you.