'Functional organization for verb generation in children with developmental language disorder' with Saloni Krishnan

WEBVTT
00:00:05.969 --> 00:00:38.310
Welcome to Episode Seven of the Language Neuroscience Podcast. I'm Stephen Wilson and I'm a neuroscientist at Vanderbilt University Medical Center. First of all, I'd like to say a heartfelt thank you to everyone for listening and for all the kind emails and comments that many of you have sent me. I really appreciate knowing that people are enjoying listening, learning new things and staying entertained while commuting, working out and doing dishes. If you have a colleague who you think might be interested in the podcast, especially maybe one who isn't on Twitter all the time, it would be great if you could tell them about it. My guest today is Saloni Krishnan.
00:00:34.679 --> 00:01:16.290
Saloni is a lecturer at Royal Holloway, which is part of the University of London, where she directs the N-CoDe lab, which stands for neuroscience of communication development. She's been doing outstanding work on developmental language disorders, word learning, dyslexia, and many other topics related to the neural basis of language development. In particular, she recently won the Neil O'Connor award from the British Psychological Society for her paper that we're going to be discussing today, which is called "Functional organisation for verb generation in children with developmental language disorder". It just came out in NeuroImage. In this study Saloni and her colleagues compare the neural basis of language processing in children with DLD, developmental language disorder, to typically developing kids.
00:01:16.680 --> 00:01:24.750
They recruited the largest sample ever of children for a study like this, and it's also a registered report, which means that they had to commit in advance to their analysis plan.
00:01:24.959 --> 00:01:30.090
I think this makes it a very authoritative study. Okay, let's get to it. Hey, Saloni. How are you?
00:01:30.870 --> 00:01:32.040
Hello, I'm good.
00:01:32.730 --> 00:01:50.640
30 in the morning in Nashville. But it kind of feels a lot later, because I had to get up at four in the morning this morning. I actually woke up in Milwaukee, and then had to fly back here in time for our chat. So I might be a little bit loopy because of that. How about you? Where are you at?
00:01:51.049 --> 00:01:56.209
Gosh that sounds really stressful. I don't think I've flown in a long time now.
00:01:57.379 --> 00:02:01.189
It was the first time that I've flown since the pandemic. Yeah, it was pretty weird.
00:02:01.400 --> 00:02:06.319
Yeah, I don't envy people coming into like London's airport, which is Heathrow at the moment.
00:02:06.680 --> 00:02:08.870
Currently, there are six hour long queues.
00:02:09.120 --> 00:02:09.629
Oh, wow.
00:02:10.830 --> 00:02:13.199
I am, as you might have guessed, in London.
00:02:14.159 --> 00:02:22.650
I'm using my daughter's bedroom as a home office for the time being. It is nice and sunny outside, which is lovely.
00:02:22.870 --> 00:02:33.789
Okay, yeah, no, it's very nice and sunny here, too. So our main topic today, as we talked about, is going to be your recent prize-winning paper.
00:02:29.680 --> 00:02:45.099
But before we get to that, I wanted to kind of learn more about your background and how you got interested in this field. So I was wondering if you could start by talking about, when you're a kid, did you have any interests that kind of pointed to your later career in language in the brain?
00:02:45.719 --> 00:02:57.419
That's a really interesting question. So I don't usually talk about this. But I have a brother and he has very severely refractory epilepsy.
00:02:53.430 --> 00:03:50.939
And what that basically means is kind of epilepsy that doesn't respond to medication. He's had this for a long time. But one thing that kind of happened because of his epilepsy is that he never really learned to speak. So he was developing normally till about one, one and a half. And then he started getting these seizures. And suddenly, he couldn't learn language anymore. So a lot of my childhood was actually spent, you know, interacting with speech and language therapists, who were working with him or neurologists who were studying his brain. And so even though I never intended for that to eventually be something I'd be interested in, it was clearly something that was going on in the background quite a lot. And I think when I finished high school, and I had this kind of existential moment of, what am I going to do next. Because all through high school, I thought I wanted to be a dentist. And no offense to dentists, but I'm kind of glad that I'm not one today.
00:03:51.868 --> 00:03:57.479
I'm glad that you're not one too, because there's nothing I hate more than like interacting with a dentist in any way, shape, or form.
00:03:58.800 --> 00:04:20.189
But I think that at the time that I realized I didn't want to be a dentist was literally, you know, when everyone's done the exams, and you're kind of waiting for offers from university. I had to have a quick change of plan. And so I thought, you know, what else might I like to do? And I thought that the job that speech and language therapists did with my brother was incredibly cool.
00:04:17.100 --> 00:04:30.779
It seemed to me this really creative, but also really cool way to kind of keep some of the science I'd been really interested in alive. So I ended up doing an undergrad in Speech and Language Pathology in India.
00:04:31.829 --> 00:04:44.970
And I really loved it. And then I guess this kind of idea of working with brains stuck around and so when it kind of came to picking a master's degree I ended up ended up choosing cognitive neuroscience.
00:04:45.259 --> 00:04:54.079
Right. And so in India, did you have did you have to choose that career path like right out of high school? You kind of had to choose an undergrad degree that was going to define your career?
00:04:54.110 --> 00:05:16.639
Yeah, it's pretty specialized quite quickly. Most people, you know, in India, especially if you do the sciences, you're kind of told like, well, are you going to do medicine or you're going to do engineering. And I was going to be very different by being a dentist. But as I said, that didn't quite pan out. So went down a very, very different route to most people.
00:05:17.180 --> 00:05:18.290
Yeah, for sure.
00:05:18.589 --> 00:05:22.579
But yeah, we do specialize early in undergraduate. Yes.
00:05:23.610 --> 00:05:28.379
And was your brother's epilepsy ever able to be controlled or managed?
00:05:28.839 --> 00:05:43.660
Um, yeah. So I think when I was doing my undergraduate degree, we were really lucky because we had really good neurologists. And I kind of built up a relationship with some of them. And I managed to get my brother seen. And of course, like 13 years down the road, treatments were better.
00:05:43.839 --> 00:05:55.660
And we finally found medication that kind of controlled his epilepsy. So that's been brilliant. But obviously, like we all know, language learning has a kind of window. And I don't think we quite caught that window.
00:05:56.860 --> 00:06:01.240
Right. So how is how is your brother's language at this point?
00:06:02.949 --> 00:06:34.990
He is about five years younger than me. So that kind of puts him in his late 20s. I'd probably say his language is closer to a two year old's language ability. He doesn't really, in fact, actually, I have a two-year-old, not even sure that's quite where I would peg him. He does some things really well. And he does kind of comprehend things. And interestingly, he can carry tunes from songs he's heard quite a bit. But actually expressive language has definitely always been an area that he's struggled.
00:06:35.519 --> 00:06:45.658
Right. That's a really fascinating and unique way into our field, I think. And I'm glad that he has epilepsy has better control than it was in the past.
00:06:46.079 --> 00:07:02.370
Yeah, I think sometimes it can sound a little bit like a sob story. I think it is. And it is what it is. And it ended up taking me down a path that, you know, showed me some really interesting career options that I wouldn't have known had I not been hanging out in these clinics with him as a little child.
00:07:02.550 --> 00:07:10.680
Yeah. And so when you finished your speech pathology degree in India, you practiced there for a while, is that right?
00:07:11.649 --> 00:08:04.600
Actually, so when you finish an undergrad in India, you're kind of contracted. As part of your degree, you kind of work for a year. And that's to kind of get the relevant range of experiences. And also because like, we have a fairly severe shortage of speech and language therapists, so it's quite good to kind of be on the ground and be working with people. So I did that for a year. But then I came to London to do a master's in cognitive neuroscience. I went back after that, for a little while. I hadn't quite figured out what I was doing. So I was sort of following my then husband around. Well, sorry, I should say that again, because he's still my husband. So I did this year in India, where I was doing all of my clinical training as part of my undergraduate degree. And then I moved to London for my master's.
00:08:05.290 --> 00:09:22.809
But I was pretty sure I wanted to be a clinician. So the master's was kind of a, you know, interesting way to study more about things I liked. And then the point was always to go away and become a clinician. So a year later, we moved to Dubai, and I was a clinician that for about a year. But by then the research bug had bitten me a little bit, because that's the first time in London that I encountered research. That changed things to me a little bit. I think when I went back to clinics after my master's, I kind of felt a little bit like, you know, we don't always have evidence for exactly what we're doing. We didn't quite have the right categories and the right labels. And yeah, just ignited lots of questions. And I was really interested in you know, why we didn't have evidence for certain programs. There were lots of things that will be marketed to people. So I used to work with people with autism and dyslexia, and consequently a lot of people with DLD as well. And they don't these kind of programs that, you know, companies would come in and market. Then you'd be like, actually, I don't think this works. Why am I being asked to do this? Or why am I even being asked to recommend this? And I just felt like we had very little evidence in terms of what was actually working in the clinics and good labels to communicate with people in terms of what their problem is might actually be.
00:09:23.460 --> 00:09:33.269
So you decided to go to the UK for your PhD and postdoc. Was that difficult to get back there? Or did you find an easy path back into research?
00:09:34.049 --> 00:10:00.000
I was really lucky because my MSc supervisor basically helped me get funding for my PhD, which is incredibly tricky to do as an international student coming to the UK. So I was really lucky that he kind of helped push some of that through and made sure that I could come back to the UK. Once I was in the UK became a little bit easier because we've always loved it as a country and so it kind of became easier to stay.
00:10:02.129 --> 00:10:05.610
And what did you work on during your PhD and postdoc?
00:10:07.340 --> 00:11:01.700
So I guess in my PhD, as I said, I've been really interested in expressive language for a long time. And I was thinking a little bit about what makes up language production. So what does the neural organization for language production look like. And while I was doing my PhD, I kind of had this idea. So I worked with Fred Dick, who used to work with Liz Bates as you know. There was this big idea that, you know, perhaps there were these building blocks to language, like auditory skills and motor skills. And if we look for individual differences in the skills we'd kind of find that, you know, if you were particularly bad at motor scales, maybe that would show up in your functional activation for language production. And you might see that, you know, you get less neural activity in core motor regions. So you'd have more activity because you were maybe having to put in more effort. And so that was kind of where I started my PhD going, like, that's what I'm gonna investigate. That's really cool.
00:11:01.700 --> 00:11:20.629
And that's partially what I had done my master's on as well. And during my master's, I had this beautiful, lovely correlation that showed exactly that. I had scanned about 25 kids or so. And it was great. So I was like alright, I'm going to come back to my PhD, scan some more children, and then this correlation is going to be there. And it went away.
00:11:23.809 --> 00:11:32.809
So you learned an early lesson about sample size? And yeah, reproducibility.
00:11:33.289 --> 00:11:34.309
Exactly.
00:11:36.470 --> 00:11:54.860
You recently landed a faculty job, which is a very remarkable achievement, especially in recent years. And I read this really interesting article that you write about setting up your lab. So can you kind of talk about what it was like to set up your lab? I guess that's about two years ago now, like, what were those first steps like for you?
00:11:55.320 --> 00:12:08.580
Yeah. So I recently got a faculty position at Royal Holloway, which is a member of the University of London. But unlike most of the other colleges, which are part of the University of London, it's not actually in London.
00:12:08.580 --> 00:12:34.440
It's slightly outside of London in Surrey. It's this really beautiful campus. It used to be an old women's college and like everything now has become, I don't know what the word to say here is, but admits everybody now my guess. Coed. But yeah, it's a really beautiful college.
00:12:30.929 --> 00:13:19.470
And I was very lucky, because right at the time that I was applying, a few things kind of fell into place. Someone who's very good friend, Carolyn McGettigan, used to be there for quite a few years. And she had recently landed a faculty position at UCL. And so she was moving. And one of the things about most UK departments will be that they'll tend to have one kind of person that does similar things, they won't have a whole suite of people that do very similar things. Actually, it was a good time to apply. Because, you know, they didn't have someone who was doing MRI and language anymore. So it was kind of a nice little niche that opened up. And I was always very lucky, because like, Carolyn helped me quite a bit in terms of thinking about what to say as part of that application and things like that.
00:13:19.929 --> 00:13:26.740
All right, that's really good. And so she knew what they'd be looking for in terms of filling that gap that she was creating by leaving.
00:13:27.220 --> 00:13:39.039
Yeah, I think so. I mean, she gave me a mock interview. That mock interview was just almost the same as a real interview. So very, very helpful to have networks, I guess.
00:13:39.100 --> 00:13:41.889
Yeah. And then what was it like setting up the lab?
00:13:42.590 --> 00:14:09.230
So I was gonna get into Royal Holloway, because it's kind of out in Surrey. And a lot of people commute. It was a very different culture to where I previously been in Oxford where, actually I was the only one commuting in. So I kind of felt like I fit a bit more with the other faculty that were there. But equally, when you start up a lab, I think it's just really lonely. And no one had quite prepared me for that.
00:14:09.230 --> 00:14:25.580
So, you know, there's this excitement of like, oh, my goodness, I have like my own office. And after a couple of years in Oxford, where we'd had like, you know, a building closure, temporary offices, and all that stuff. I was just like, what, I've got a room to myself.
00:14:21.769 --> 00:14:59.210
This is really bizarre. But then, you know, you're kind of sitting in this room and you're like, but I don't have anyone to chat with or like, you know, I don't have anyone to go to lunch with. And so there's a lot of kind of, how do I even start to build up a lab? How do I build a community around me? And I was very lucky because everyone at Royal Holloway was very welcoming. And at least it was the days before the pandemic. So you could meet people like, you know, in the local break room or stuff. But yeah, which newer staff starting obviously haven't had.
00:15:00.980 --> 00:15:26.600
Yeah, I really feel for anybody that starting any kind of new position in the last year. That must be even harder than normal. But yeah, so how do you kind of get a get up and running? You know, you need grants to do research, and you need data to get grants, and you need students to get data and you can't have students because you don't have grants like, how do you how do you break into that cycle?
00:15:29.480 --> 00:15:41.809
It was tricky, actually. I also wanted to try and do something a little bit different to what I had been doing in my postdoc because I wanted it to be really mine and show that I could set stuff up.
00:15:38.840 --> 00:16:04.460
Not to prove to anyone else, but just to myself to be like, yes, I can lead a new line of work and think about this. And Royal Holloway has a nice model where they do protect faculty early on from teaching. So you're only given a half load over the first few years, which helps tremendously. Otherwise, being thrown into teaching on top of everything else is like, hard.
00:15:59.899 --> 00:16:18.830
And you also get a little bit of a startup. I got some really useful advice, actually, from someone who is at Oxford, who wasn't in our field. But just you know, someone who was generally chatting to. And he was just saying, like, you know, spend your money. Don't hold on to it, like just start spending.
00:16:16.159 --> 00:16:55.879
You've got to do something with it. And also, like, don't bet it all on one idea. Try to do a couple of different things because you will find that one doesn't work and then in that case, you've got a couple of things going. So I felt like I was in a relatively lucky place because I did have links with my old postdoc project. Kate and I had negotiated, the papers that I would still continue to work on and that I was still writing up and had access to. But equally, I had some new stuff to start and some new ideas. And I ended up hiring a part time undergraduate research assistant. And the two undergrads who worked with me have just been so brilliant.
00:16:51.649 --> 00:17:12.710
They really helped me, you know, just meeting them forced me to think about research so I didn't slip with all of those other deadlines and things that were happening. But equally, they were really good at just, you know, making sure stuff was happening and taking on some of that early legwork of you just have to do this. You just have to collect this data.
00:17:13.859 --> 00:17:32.549
Yeah, that's really cool. I had a similar experience when I started my first faculty job too. Some really brilliant and hardworking and just great undergrads really got things off the ground before I had any significant amount of funding to like hire full time personnel. It just never could have started without them.
00:17:33.509 --> 00:17:44.759
Yeah, I think that was the best thing that like really worked. They were great. Yeah, and then I wrote a couple of little grants and started to get these little grants. And yeah, it's kind of going okay.
00:17:45.480 --> 00:18:01.079
That's really good. Should we talk about the paper? So, first of all, congratulations on getting the Neil O'Connor award, which I think is from the developmental section of the British Psychological Society. Can you tell me about what that award is about?
00:18:01.930 --> 00:18:58.450
Yeah, I was incredibly honored and surprised to get it. Because if you look through that list of previous winners, it's got so many people that I just really admire in terms of like developmental work. It's for a piece of work on a neurodevelopmental disorder, actually. And so it's quite a specific prize in some respect, but it's tended to go to people who have really made an impact in terms of, you know, thinking about clinical signs and symptoms, overlap between different neurodevelopmental disorders. And traditionally, you know, people who tended to work on autism. So I think when I submitted stuff, I was like, well, you know, I'm language disorder, is that really going to float people's boat? It's also neuroimaging. It's not telling us anything that is specifically clinical about this population. So I was very, very surprised and honored to get it.
00:18:59.890 --> 00:19:16.450
Well, I'm really glad that you did because that was kind of where I heard about the paper. And then I looked at it and I was like, oh, this has heaps of things in it that are really interesting to me. Before we talk more about it, do you want to talk about like, who your co authors are on the paper, and you know, what their roles were?
00:19:16.569 --> 00:19:26.170
So the Oxford BOLD study is actually a grant that Kate Watkins is the PI on.
00:19:20.769 --> 00:20:31.420
We wrote it together, I think back in 2016. So it was a grant that we submitted with Dorothy Bishop and me. The idea was really we had been doing a lot of kind of reading around the DLD literature. And it was something I was really, really interested in from my clinical experience, but also kind of thinking about how to put that together with stuff I had worked on in the past. And Kate and Dorothy have previously done work in this area already, but like a much smaller study. And so we were really keen to try and get some funding. We'd applied previously and actually that MRC grant got rejected. And we were told that, I don't remember what we were told, but what we were given was an invitation to resubmit, which for the UK is incredibly unusual. You don't usually see invitations to resubmit a grant once it's rejected. But so I spent a year kind of, you know, holing away and trying to help Kate with this grant. But she finally submitted it, and we got the money to start this project in I think 2017. And it's been a really, really exciting project.
00:20:31.420 --> 00:21:09.759
Because the idea was that actually, despite 20 years of research or more using fMRI, we surprisingly still know very little about the neural basis of DLD. I think what's particularly galling is, if you're really a researcher interested in this area, and you do like a quick Google Scholar for like, MRI and DLD, you'll find that about, like 20 papers come up. And if you then do that same search for something, say, like autism or ADHD, then of course, you see 1000s of paper come up. And I mean, DLD is a really common disorder. So it's a little bit surprising that we don't know more about the neural bases.
00:21:11.859 --> 00:21:35.470
This grant was like an opportunity to really try and unpack that a little bit. It was just one study at the heart of it. And it was just to say that we're going to collect as much data from kids with DLD as possible. And then we're going to have them do a sort of series of scans in the scanner, and also get a very detailed characterization of their language and cognitive skills.
00:21:36.109 --> 00:21:56.329
Yeah. Can you talk a bit about the these DLD kids? So I understand that the field has kind of converged on this new terminology, is it replacing SLI? Or is it just a different concept than SLI? Are you guys thinking about it differently now, in terms of how it is distinguished from other things?
00:21:57.500 --> 00:23:03.710
So when I was in clinics, even the term SLI in India wasn't used. I think SLI has tended to be a term that was quite popular in research and I think didn't quite make it to clinics in quite the same degree. And even today, a lot of clinical practices will actually use very different labels. So they might have things like language delays, sometimes language disorder. In research, you'll find terms like developmental dysphasia, and things like that. And so back in 2016, Dorothy led this kind of consortium exercise called CATALISE. And the purpose of CATALISE was to try to agree on two things. One is specifically like, what are the criteria which we think constitute the language disorder. And then the second part of this was, what is the label we should use? And I think we realized that it was important, not just, you know, it was obviously important what that label was, but actually what that label was, was somewhat less important than people converging on using it.
00:22:56.089 --> 00:23:56.390
Because in order for a diagnosis to gain popularity, actually, you have to have people know what it is. And with autism, or ADHD, we seem to do this very well. Or dyslexia, for example, we just don't have that same recognition for terms like SLI and DLD. And of course, with SLI, the idea was that you had to show this discrepancy between language skills and IQ. And I think, in practice, it was recognized that that doesn't happen, or that happens very, very rarely. And so we were kind of focusing not just research but also services on a very small subset of people who had language problems. And so kind of DLD as a label is a little bit more inclusive. So I think with a nonverbal IQ of greater than 70, you can get a kind of diagnosis of DLD. You don't have to show that discrepancy between nonverbal and verbal skills as you would have had to do for SLI.
00:23:57.500 --> 00:24:17.900
That's really interesting. I think that the term SLI was promoted a lot by people that had a lot of ideological commitment to the modularity of language. And, yeah, it's pretty clear from your paper that that's the exception rather than the rule.
00:24:12.740 --> 00:24:19.849
So how do you define DLD in your study and in general?
00:24:20.289 --> 00:24:22.900
So yeah, people disagree on this a little bit.
00:24:22.900 --> 00:25:04.029
And I think the idea is that, you know, you can imagine that language ability is some kind of trait, right, and it's a trade in a similar way to how we might calculate like, BMI for instance, which is body mass index. And the idea there is it lies on a continuum, but at some point, you're gonna say like, this is a bit excessive. I think these people need support, that these people need help. And with BMI, we have established categories to do that along the continuum. And then we argue that you would do a similar kind of thing for DLD as well. So I mean, different researchers differ a little bit in where they draw about specific cut.
00:24:59.619 --> 00:25:51.700
For example, in the kind of larger studies that have tried to estimate the prevalence of DLD, what people tend to do is use a criterion of one and a half standard deviations below the mean, on two or more language tests. We've gone slightly gentler, because recruitment was obviously a slightly more challenging and in research settings, I think it tends to be accepted that you use about one standard deviation below the mean, on two or more tests of language. And that kind of two or more tests of language is actually really important, because that's where you're saying it's not just some kind of normal variation. But say, if you're testing different aspects of language, you shouldn't be following kind of in that category, by chance on two or more of those tasks. Does that make sense?
00:25:51.789 --> 00:25:54.339
Yeah, it does.
00:25:51.789 --> 00:25:58.150
Don't the kids in your study also have to have a clinical diagnosis, or am I mis-remembering that?
00:25:58.180 --> 00:26:10.359
No. So actually, you know, because DLD is such a contentious term. It's not used comprehensively by clinical practices, as I said.
00:26:07.240 --> 00:26:43.420
There was quite a lot of debate around the time we started, because it's a bit more accepting and adopted now. But you'll still find on Twitter, the occasional SLT will pop up and say, like, you know, for example, our clinical practice doesn't do diagnoses. We think that, you know, we should respond functionally to the child's profile. It's also quite a new diagnosis. So you might expect that older children haven't really received it. For a variety of reasons, and sometimes people don't want the diagnosis. So we didn't mandate that people have to have a diagnosis. But we tested them on language tests.
00:26:44.109 --> 00:26:48.430
Didn't they have to have a history of treatment for speech language issues?
00:26:48.640 --> 00:26:53.109
They had to have a history of speech and language problems, but I think we just asked parents.
00:26:53.000 --> 00:27:18.920
Right. So it seems like the rationale for your study is essentially to see if there's any functional abnormalities in language processing that might explain the language deficits that these kids have. Can you tell us about prior studies of functional and structural differences in this population that might explain that language disorder? And evidently, you guys weren't satisfied with the state of knowledge. So maybe talk about that a bit?
00:27:20.069 --> 00:28:12.210
Yeah, sure, I'll try and pull this up to remind myself about what different people had done. But as I said, especially functionally, there had just been a really, really small number of studies. So I think I'm looking at a literature review that we kind of carried out. On my lovely excel sheet that Harriet Smith, who is one of the co authors on this might remember, I think we have seven rows of data. So there were about seven functional studies when we started doing BOLD. And basically one of the issues with all of these studies were that the numbers of people they had used have been quite small. And then each of them have used a completely different task. To give you an example, one of the studies did a kind of task switching executive function task, another data kind of implicit language learning task.
00:28:12.750 --> 00:28:51.809
And then the study that Kate and Dorothy did, which was first authored by Nick Babcock, had a kind of covert auditory response naming tasks. You'd kind of hear like, a little definition, and you had to kind of come up with the word for it in your own head. So really, very, very few studies. And every study did something slightly different, and then had a slightly different definition for what they consider DLD to be. So perhaps it was kind of unsurprising that we didn't have this clear picture of like, okay, well, kids with DLD show less activity over the left IFG.
00:28:46.410 --> 00:29:09.089
There were kind of converging themes, perhaps, that people tended to focus on kind of superior temporal regions, or inferior frontal regions, particularly in the functional tasks. But it wasn't cohesive, and you couldn't really extract a clear pattern for some of the reasons that I've mentioned.
00:29:10.589 --> 00:29:11.130
Okay.
00:29:11.000 --> 00:29:25.519
I mean, also, it was fairly typical of the times, right. The studies I'm talking about are sort of, like, early 2000s. And, you know, looking back at studies that it was fairly typical to just have a few participants in those designs.
00:29:25.569 --> 00:29:43.539
Right. So let's talk about some of the decisions that you made to kind of go beyond these past studies. So first of all, you all wanted to recruit a large number of kids, right? So it's obvious that more kids is good, the more kids the better. But how did you decide how many kids you are going to need?
00:29:47.680 --> 00:31:32.829
This is kind of a long but pretty interesting story, I think. So at some point, we had started planning this project and in one of our project meetings, Dorothy suggested this idea of like, why don't we do a pre-mortem of the project. I don't know if you've ever heard of this before, I certainly hadn't. But it's this idea that you sit around, and you kind of say, the project has completely failed. Now, here are the reasons I think it's completely failed. And so we kind of sat around for a couple of hours coming up with a whole bunch of different reasons that we thought BOLD could fail. And some of them are really obvious, like, we just wouldn't be able to recruit people. Some of them were kind of less obvious. So for example, I was really concerned that we were going to say that we were going to fully characterize language and cognitive abilities. And I'd be like, well, I'm sure we'll go to a reviewer and they're going to point out that we missed this really important question or, you know, this really key thing and we're just not going to think about it till it's too late. And obviously, at the time, the open science movement was kind of kicking out and kind of just very close to this meeting, we saw an announcement saying that NeuroImage was going to accept registered reports and kind of neuroimaging data. And we thought this could be kind of cool. Maybe we can submit one task as a kind of protocol or a basis. This would kind of help us decide how many participants we're going to kind of collect, or at least minimums that we need to collect. I'm also doing Kate a disservice here, because actually, in the grind, she does have a really detailed section on power analyses. But I think, which obviously led to kind of the numbers we propose for the grant. But I think we wanted to formalize it a bit more and formalize some of the decisions we were taking in this paper.
00:31:34.119 --> 00:31:53.950
But yeah, the overall grant, I think, when we wrote it, the objective was to scan 160 children. We said that half of those would be kids with DLD or what we hoped would be kids with DLD, because as I said, we didn't necessarily know before they came to us. Then the other half would be kind of matched controls.
00:31:54.849 --> 00:32:24.279
Right. Okay. So that makes sense. So there's a power analysis in the grant. But you're kind of addressing another very interesting question, which is that this is a registered report. So I've never done one of those. I don't think many people have. Like you said, I think open science has been growing in prominence as a philosophy. What was that like for you to have to define your plans in such detail, prior to getting any data?
00:32:25.240 --> 00:32:55.089
Really, really useful I would say. Especially in a project that was this complicated. You kind of asked me about what we consider the standard for DLD. Of course, there's stuff we define, and we define it on the basis of like, Kate and Dorothy's previous papers, but we weren't sure whether it would be accepted by others as being sufficient or not. There's also stuff like, you know, just with MRI data, what are you going to consider motion? What are you going to throw out? What will you keep?
00:32:51.430 --> 00:33:48.339
Where are you going to draw these lines? And it's really easy to blur those lines once you've already collected the data, because you see it, and then you're just like, well, that makes complete sense. And I had had experience of fooling myself before where, you know, I collected this set of behavioral data, which I thought was going to lead on to this beautiful MRI experiment. And then I analyzed the behavioral data, and it was this pattern. And I was like, Oh, yeah, that makes complete sense. And then I went back to my pre-registration, I was like, that's not what I said would happened. Interesting. So, you know, I think it just sort of helps in terms of really, really defining a protocol. And I think once we did this particular paper, it became like a real reference for the rest of the project as well, because we were like, well, that's what we said in the paper. We know we can kind of rely on some of these criteria, like, you know, what is the definition of DLD? What are the tasks we're including?
00:33:48.490 --> 00:33:56.980
What tasks feed into the diagnosis versus what tasks don't feed into the diagnosis?
00:33:51.430 --> 00:34:03.700
Yeah, so there were many decisions like that along the way. And the registered report really helped clarify our thinking on many of those issues.
00:34:03.789 --> 00:34:30.730
Yeah. But I mean, you also have very specific plans for your analysis of the functional imaging data, which I find terrifying to imagine, you know. I don't want to admit to p-hacking here on my podcast. But, you know, sometimes you don't really know how you want to analyze your data until you see it, like you said. You guys committed, you committed to the very specific analysis plan. You define your ROIs, you said exactly what tests you're gonna do. That must've been daunting.
00:34:30.730 --> 00:35:06.099
Yeah, I think that was probably the hardest part. Actually, one of the interesting things here was, so as I said, BOLD has a series of scans. So you have things that are, you know, possibly a little bit less like, tricky to collect. You have stuff like the DTI and the resting state data and so on, but you don't have to make as many decisions. But when you do task MRI, you have to make this decision about, what task am I going to use. And we wanted one task that was language processing and another task which was a little bit more like language learning. And in fact, the language learning task for a variety of reasons was a little bit easier to decide on.
00:35:06.280 --> 00:35:16.300
But the language processing task, I think we went back and forth over multiple meetings about what we would use. And at some point, Kate was like, why don't we have a menu of tasks?
00:35:16.449 --> 00:35:35.949
And we'll choose from this menu of tasks. And I think you said, we'll talk about this a little bit later in terms of performance. But actually, it was incredibly hard to pick which task we were going to use to basically look at a very simple idea of, what is language processing look like in a population that struggles with language?
00:35:37.840 --> 00:35:42.429
Yeah, so let's not keep the listeners in suspense, which task did you choose?
00:35:43.599 --> 00:36:00.789
We chose verb generation in the end. And again, it's a very simple task, the idea would be, you'd see, say, a picture of a ball. Then you would have to say something you could do with that. So you might say, kick the ball, throw the ball, any of those things.
00:35:54.309 --> 00:36:17.349
And so it's a really like common task. It's been used in developmental work, I think it's been used quite a bit in adult MRI work, and also the aphasia and literature. So yeah, quite a simple task. But we had evidence that kids with DLD could do it.
00:36:12.610 --> 00:36:19.480
And that was probably the most important part of choosing the task.
00:36:19.869 --> 00:36:22.210
Right. So the feasibility for your population.
00:36:22.719 --> 00:36:28.090
And you're right, I think it is one of the most commonly used tasks like in language mapping.
00:36:24.789 --> 00:36:37.059
It's certainly got a long history. And so when you submitted this initial registered report, what was the review process like? Was it similar to a paper being reviewed?
00:36:38.079 --> 00:37:21.130
Yeah, so we wrote the Introduction and Methods just like you would for any paper, but as you've seen, it's a little bit longer and a little bit more specified, I think in terms of the analyses we propose to do. And one of the reasons I didn't mention, but the reason we ended up picking verb generation is that in the KE family, which is this kind of, it's not DLD, but they do have a profile that is quite similar to DLD. There was a task that had used overt verb generation, and then kind of seen differences in the frontostriatal network that we were interested in. And so that at least gave us some basis for kind of ROIs in the process. But anyway, we wrote this all out.
00:37:17.110 --> 00:37:21.130
And we sent it to NeuroImage.
00:37:21.130 --> 00:38:13.539
And we wrote down the dates, because I thought that was kind of interesting. But we submitted it on the 24th of July 2018. And I remember that very well, because that's exactly the day I went on maternity leave. It was a kind of scramble to get that in. But we submitted that and I think we received reviews, towards the end of August. So at that point, my baby daughter was about two weeks old. Yeah. So I guess Kate really was amazing, because she didn't ask me to do stuff during my maternity leave, and really kind of pushed on with responding to the reviewers. But actually, most of the questions that reviewers asked were things that we thought we knew about. We were anticipating a much harder review process. Around some of those things I said in the pre mortem, like, why are you defining DLD like this? Why are you doing this particular task?
00:38:13.809 --> 00:38:16.480
You know, were really gung ho about defending these decisions.
00:38:16.480 --> 00:38:35.230
But actually, what the reviewers asked about was, are you sure you can do a speech production task with children in the scanner? And things I guess, you know, we were like, yes, I think we can do that. But we I guess we just had to provide more evidence that we could do that.
00:38:29.889 --> 00:38:54.280
So yeah, we submitted revisions by the middle of November. And we had an in principle acceptance by the 5th of December 2018. So you know, as papers go, given the fact that I was in the maternity leave, I think that was pretty good, like timing, to make sure that our key protocol for this project was accepted.
00:38:55.860 --> 00:39:21.300
Yeah, about six months. It's not nothing. But it does give you a very firm basis, right, to work from. So you mentioned the KE family. Was that the main source for your hypotheses in your preregistered analyses, where you're going to basically look for differences in the inferior frontal gyrus and in the basal ganglia?
00:39:21.389 --> 00:40:20.460
It certainly influenced us because Kate was obviously the person who did some of that work on the KE family. It wasn't the only thing though, because I think as I said, I'd done this review of the literature in DLD. And quite a few studies had pointed to differences in the striatum. It wasn't always consistent. Most studies suggested there was kind of a reduction in the volume of the caudate nucleus. And some studies kind of suggested that there was an increase or kind of a change that might be modulated by like intracranial volume or age in some way. So again, not a completely consistent picture, but at least more than one study converging on this idea that there might be a difference in this area that's not traditionally thought of as a language area. And given that the kind of similarities and behavioral profile perhaps to the KE family, we thought that was a good place to kind of start in terms of building a hypothesis.
00:40:20.579 --> 00:40:31.949
Okay. Can you talk about how you characterized language and non-language functions in your DLD group and in your control group?
00:40:32.530 --> 00:41:59.079
Yeah. So as I said, we wanted to have this really very detailed picture of people's language ability. And the way we do this is quite similar to some of the work that say Bruce Tomblin and Courtney Norbury have been doing in the behavioral domain. But we said that for each kind of language ability that we might be interested in, so like vocabulary, grammar, or narrative, we were actually going to look at a receptive measure and an expressive measure. So we ended up having six measures of language ability. So either receptive vocabulary, expressive vocabulary, receptive grammar, expressive grammar, receptive, expressive narrative. So that was kind of our key language set of tasks. And then on top of that, we collected a few extra things. So we've always been really interested in nonword repetition. So that was one of the extra measures that we think is language-y, but is not core language and wouldn't feed into a diagnosis of DLD. We also collected measures of reading fluency, because obviously, we were working with 10 to 15 year olds, and that's something that's starting to be quite important. UmWhat else did we get. So in addition to that, we got a few verbal working memory measures, things like the digit span. And also some declarative measures. So things like lists learning tasks, where you kind of hear a series of items, and then you have to remember it, and then you kind of do that a few times. That's a lot of the tasks that were sort of verbal.
00:41:59.079 --> 00:42:24.610
We then had some some of the classic kind of nonverbal tasks like colored progressive matrices, and block design, but also a task that was kind of called coding, which is where you kind of translate these symbols into numbers. I think those are the main tests. Oh, and we had a pegboard task to kind of just look at fine motor ability.
00:42:24.670 --> 00:42:31.659
And then you end up doing a factor analysis of some sort and coming up with a language measure and a memory measure.
00:42:32.260 --> 00:43:23.380
Yeah, so again, this is a really collaborative project. So Dorothy, Bishop, and one of the statisticians in her lab, Paul Thompson, came up with that code because again, it had to be pre-specified before we had any kind of sight of the data. And one of the things we were really worried about is that if you have all of these different neuropsychological tests, I mean, you could imagine running a correlation with each one to the brain. Then saying, well, this is clearly the one that is the most important because it shows me a difference. So early on, we decided that actually, the thing we were really interested in was language and memory factors. And Dorothy wrote this bit of code, which actually said that either all of our tasks are going to load onto one language factor, or they might load onto two. And so she kind of built this model, which would allow us to test which of those two ideas for kind of better fits to the data.
00:43:18.550 --> 00:43:28.900
And so when we did it, we found that the kind of memory and language model was a slightly better fit to the data. So that's what we ended up using.
00:43:28.929 --> 00:44:10.480
Yeah. So you found that you do need to have two factors. But I guess one of the first striking things I noticed in your paper is you plotted a correlation between the language factor and the memory factor in all your participants. And they're really like, shockingly, highly correlated, right? That's like the end of SLI right there and the beginning of DLD. It's not just in the factors, but you kind of show like every single one of your measures that you report, the DLD group always does worse than the other than the controls. And you can have a third group in there too, that we haven't really talked about.
00:44:07.750 --> 00:44:16.269
But they're always they're always performing less well than the other groups, whether it's language or not language, right.
00:44:17.260 --> 00:44:32.980
Yeah. And actually, that was one of our concerns. So one of the debates we've always had is like, ideally, what you want is also a task that the kids with DLD do well on. And this has kind of led and traditionally they should do okay on IQ, right?
00:44:28.900 --> 00:45:05.650
And, in fact, what the problem is that our controls are almost super controls, right? So you like end up recruiting from like, you know, parents of children around the university, and that probably ends up giving you quite a high IQ control group. So it's not the DLD group that's the problem, but it's really the controls. And we tried really, really hard to, you know, recruit controllers that wouldn't be these kind of super normal controls for want of better term. But yeah, I don't think we fully succeeded.
00:45:08.050 --> 00:45:20.199
Yeah, that's always difficult. So let's talk about how the kids did on the verb generation task. Most of them could perform it, did the behavioral data come out how you were expecting?
00:45:20.420 --> 00:45:59.809
Yeah, I think so. And I think it was actually really encouraging because, so we defined performing it as kind of getting more than 75% of the trials on our task being accurate. In other words, they had to agree then 75% accuracy, which we thought was pretty good. Because you know, you can sometimes like, it's a very quick task. Anyone could be stumped for a moment and not recognize a word that comes up in a very strange scanning environment. And yeah, most of our kids with DLD, so I mean, we've got this long thing in the paper about how we ended up excluding children, but we only have to exclude four on the basis of accuracy. So I think that's pretty good retention for this task.
00:45:59.889 --> 00:46:13.929
Right. I know you end up including those for in some of your analyses, but were you worried that by excluding the worst performing kids, you might be kind of like, you know, making it harder for your to find a difference between the groups?
00:46:14.710 --> 00:47:04.989
Yeah, potentially. But I think this issue of performance was something that has been quite a feature in the development literature. So for example, when I was doing my PhD, there was this paper by Brad Schlaggar in Science in 2002. And that was a really important paper in the sense that it suggested that you could have differences, that might look like age-related differences between, say, children and adults, but actually completely attributable to like, performance, or how children perform the tasks that they were doing. And so we were really concerned that we would put kids in the scanner, they already had language problems, we'd give them a language task, and they would just not be able to do it. And that's all we were picking up the fact that they can't do the task. And that's not a true brain difference, or at least not a true brain difference that we were kind of hoping to see or understand.
00:47:05.199 --> 00:47:24.639
Yeah, this is just the bane of all clinical studies of language disorders, right? It's literally impossible to, you know, present people with a task that's gonna kind of implicate their language difficulty and yet have them be matched to a control group in any meaningful way.
00:47:24.929 --> 00:47:45.510
Yeah, so they're not matched perfectly because they are still a little bit worse than their control group. But at least we have tried really hard to kind of ensure that we do know about their performance, and we do know what they're doing. And, yeah, I think we did possibly as good a job as I could have hoped for at the start of the study.
00:47:46.320 --> 00:48:07.980
Right. So should we talk about what you found in your imaging data? In your primary analysis, you have a number of pre registered analyses, the main one is to compare BOLD signal between groups in four regions of interest. Can you tell our listeners, what those regions of interest were and what you found?
00:48:07.000 --> 00:48:14.889
As I said, we kind of based our regions of interest on this work that had been done with the KE family.
00:48:15.280 --> 00:49:41.079
Specifically Frdrique Ligeois had done this work where he showed that when affected members did this overt verb generation tasks, they showed reduce activity in the inferior frontal gyrus, pars triangularis, and in the putamen bilaterally. We thought these were kind of good candidates, because we wanted to look at the frontostriatal system. Here was a task that would be done with someone with a very similar profile. And they had shown these frontostriatal differences. They kind of became our ROIs that we chose. And we also wanted to show that actually, it wasn't just that kids with DLD would just have reduced activation all over the brain. So we thought we would pick an area that would be activated by the task because, you know, you're seeing pictures, so you would expect to see activation in right lateral occipital cortex. But you shouldn't necessarily see it affected by group. So we expected quite beautifully, in our naive way that we'd see this lovely, very tight pattern. And of course, in the graphs, what you can see is that completely didn't work out. There's a lot of variability. But the two groups pretty much look identical in those two ROIs. So they do kind of activate the left IFG very strongly, but you know, to the same extent. They don't activate the putamen very much. Both groups. They only region where we might potentially kind of looking like we might find an effect was that control region or the right lateral occipital cortex?
00:49:41.380 --> 00:49:43.119
Oh, you don't want to even talk about that.
00:49:45.489 --> 00:49:55.119
Yeah, so a null result on that first analysis. Surprising, but maybe. I don't know. I guess you were probably surprised.
00:49:55.360 --> 00:50:04.630
Yeah. I guess the first time I did, I looked at the data. I was a little bit disappointed. And then I went well, you know, that's the science. They can do this task.
00:50:04.869 --> 00:50:11.800
We don't see differences, maybe that's a good thing. Maybe it's an interesting story actually.
00:50:07.389 --> 00:50:24.940
Because maybe now we know that when it's simple, and they can do the tasks, these regions can function in a way that looks like controls. And, and maybe the differences we need to look at are maybe earlier in development or, you know, with tests that are more complex.
00:50:25.809 --> 00:50:28.599
Yeah, a whole variety of things.
00:50:25.809 --> 00:50:29.619
But you know, it's a good starting point.
00:50:29.739 --> 00:50:44.050
Yeah, I mean, I guess we should talk through some of the other results before we sort of get into like, the big interpretation. Then you looked at kind of whole brain correlations between your language and your memory measures with your BOLD signal, did you find anything there?
00:50:44.030 --> 00:50:45.230
Oh, we skipped laterality.
00:50:45.900 --> 00:50:48.150
No, we're gonna get to that.
00:50:49.079 --> 00:51:32.219
Okay. One of the things about doing a registered report, and I think that sometimes this can be a little bit hard to appreciate is that obviously, it doesn't tie your hands. All those analyses you wanted to do with your MRI data can still be done. It's just that, you know, you need to explicitly label them as exploratory because that's what they are. You don't have a stronger hypothesis about those regions. Otherwise, it would have been a kind of pre-registered hypothesis, maybe. One of the kind of planned exploratory analyses was that these are the ROIs we're betting on. But it might be that even if we did see differences there, we might see differences in the kind of broader language network in regions that we hadn't specifically anticipated.
00:51:32.219 --> 00:51:42.750
And so that's why we did what would have been the most standard analysis if, you know, I had just kind of started with this data, which is just to do a GLM comparing the two groups.
00:51:44.940 --> 00:51:59.250
When you do that, what's really really surprising, actually, possibly even more so than the ROI, was that we just didn't see any differences. So any statistically-significant differences. And, yeah, that's probably the thing that threw me the most.
00:51:59.730 --> 00:52:32.489
Yeah. It's not that different than what we see in aphasia, honestly. I mean, I'm not saying that there's no differences between people with aphasia and, you know, matched controls. There are obviously there are differences in the parts of the brain that are missing. But it is remarkable how the language network just always kind of looks the same, even in these groups that have very significant language deficits. And I thought it was kind of neat that you saw that.
00:52:33.449 --> 00:52:47.219
Yeah, it was really, yeah, it wasn't what we were expecting. And so I guess, as I said, it opens up all of these new interesting questions.
00:52:41.610 --> 00:52:51.090
But certainly, I wouldn't have put in my money on not finding any brain regions that are different.
00:52:51.570 --> 00:52:59.159
I mean, if that had been talked about in the pre-mortem, would that have been considered a failure? Or is that a success, and just an unexpected result?
00:53:00.000 --> 00:53:21.659
No, I think the data is the data, right? Someone put it nicely, you can't expect nature to give up her secrets too easily. And sometimes I think studies just open up more questions. So obviously, it would be lovely if every study confirmed your hypothesis, but most of the time it won't.
00:53:22.349 --> 00:53:42.150
Yeah, I think what your study does, is it kind of says, okay, all those preliminary findings that we had from the smallest studies, as well intentioned as they were, and is reasonably designed as they were, they were probably just mostly, you know, false positives.
00:53:47.190 --> 00:54:00.900
The thing is, though, that, you know, perhaps another study will come along in a few years which has 500 kids with DLD, and they'll find something and they'll be like, well, this study was a false positive. I think science is always kind of moving on, right?
00:53:57.719 --> 00:54:19.349
So I feel like I don't want to characterize other people's work, perhaps as quite that strongly. Because we didn't use the same task. We haven't used the same populations. And it could be that if you did the task, exactly, it would replicate. But certainly, I think we're reasonably confident that this was well powered to detect differences. And we didn't.
00:54:20.159 --> 00:54:34.019
Maybe I don't want to characterize it that strongly either. I guess performance confounds, too, right. I mean, another thing that you've talked about, like some of some of these other studies might have had significant performance differences between groups that could lead to their findings.
00:54:34.559 --> 00:56:04.949
Yeah, and actually one of those studies was the one I mentioned by Nick Badcock, that Kate was the senior author on. When we were talking about this data, she was actually really interested in this idea that, why is it that that data look different to this. And could we conceivably drill down and try and find reasons for that. So one of the analysis that we did in this paper was to kind of look at almost the flipside of that Brad Schlaggar work where they had tried to match performance. We tried to completely not match performance. So we said, let's take the poorest performing, that's not probably the best term. Maybe let's take the kids with DLD, who aren't performing as accurately, or kind of the lowest quartile in terms of accuracy, and then compare them to typically developing kids who are, you know, typical. Really at that kind of nearly 100% range, because that's where most of our typical data are. That's where most of our kind of control cohort was. And so we did possibly what a standard study would do. We took that kind of low performing group and we compare them to a control group who are kind of age and gender matched. And actually, interestingly, there we are able to replicate some of those frontostriatal differences. So we see differences in the left IFG and we see differences in the caudate nuclei. And we think actually, those differences might be due to performance, suggesting that in previous studies maybe that's what was being picked up.
00:56:05.610 --> 00:56:10.110
Right, yeah.
00:56:05.610 --> 00:56:19.409
Even those differences were pretty small, though, in your study, right? I mean, they didn't come out, like in a sort of corrected whole brand analysis, but they're kind of there. Maybe if you're looking for them, you might be able to find them.
00:56:19.409 --> 00:56:32.969
Our performance subgroup analysis, yeah, it wasn't quite a z of like 3.1, which was probably what I use as standard, but they did come out at a whole brain corrected threshold of 2.3. So at least with some level of correction.
00:56:33.510 --> 00:56:36.840
Right. I didn't mean to skip over laterality.
00:56:38.219 --> 00:56:51.150
I'm very interested in laterality. You had this hypothesis that the DLD kids might be less lateralized than the typically developing kids.
00:56:45.659 --> 00:56:55.650
Can you tell us, you know, first of all, what distribution of laterality did you observe? And was it different between the groups?
00:56:55.949 --> 00:58:15.030
Yeah, so I mean, one of the things we all learn quite early on is this idea that language is left lateralized, right. In a variety of ways people have looked at this, like how does that lateralization change over development. And one of the things that Dorothy is particularly well known for is, so she's really interested in lateralization and how it patterns with language impairment. She has this really beautiful paper in Science kind of discussing the multiple ways lateralization might be kind of associated with language impairment. But I think, again, because there aren't very large studies of this we wanted to basically look at, do we find evidence for kind of either, right lateralization, which would be kind of atypical relative to kind of the normal pattern of left lateralization or even more weak lateralization. So where you have more bilateral representation rather than strong left lateralization. One of our kind of pre registered hypotheses was to look at that basically. We specifically wanted to look at the frontal lobe because of some previous research in this area. When we did that, what we found is that, well, so in our typically developing controls, we did find that overall, we saw this pattern of left lateralization.
00:58:15.030 --> 00:58:30.449
Although surprisingly, it wasn't quite as strong as we might have expected. So our mean is just about the kind of 0.2 range which would just traditionally be considered the kind of starting point of where you'd expect left lateralization on that LI index.
00:58:30.510 --> 00:58:45.000
So just to fill in the listeners that don't live and breathe lateralization. It's usually measured in terms of Laterality Index, which is abbreviated LI. 0 means you completely bilateral, 1 means you're completely left lateralized, -1 means you completely right lateralized.
00:58:45.690 --> 00:58:51.090
You're saying you saw a mean of about 0.2, which is only slightly left lateralized.
00:58:51.570 --> 00:59:23.519
Yeah, and that was for the typically developing kids. Overall, you can kind of see that that distribution tends towards like left lateralization, but there's a substantial number of children who are kind of below that 0.2 number. So who either kind of more bilateral or even right lateralized. But I guess the interesting thing is to think about whether you'd see more children with DLD who are right lateralized or atypically lateralized. So you'd kind of see that distribution almost shift downwards. And you don't.
00:59:23.550 --> 00:59:29.940
Like there are there are plenty of children who have this kind of weaker or right lateralization in the kids DLD.
00:59:30.389 --> 00:59:32.099
But it's not different to the controls.
00:59:32.130 --> 01:00:19.650
Yeah, I think your data definitely speak against there being a lateralization difference between the groups. I was like you, in your data, I was surprised by how not very lateralized they were. It kind of left me wondering, and I don't know what you think about this. I think there's two possibilities. One is that kids are less lateralized than adults. And that's what Elissa Newport thinks and, you know, her collaborators like William Gaillard and Madison Berl. That might be part of the story. But it also could be kind of task-related and analysis method related because we would not see anything like this with adults, you know. I mean, we just don't see very many bilateral or right lateralized people at all. Do you think it's more methodological? Or do you think it's a difference?
01:00:20.480 --> 01:00:55.639
So we didn't, in our studies, screen for handedness. I don't think we do see strong relationships with handedness. But we didn't say we only have to have right handed people do the study. That might have contributed in some way. In children, there is evidence that they tend to kind of increase lateralization, or the LI indices, with age. There's some studies that have shown that. I can't remember whether we looked at this with BOLD yet. But yeah, I guess I don't want to speak if we haven't looked at it. It might be that there is a change over development. But these children are already 10.
01:00:55.940 --> 01:00:59.059
Yeah, they're not that young. Yeah, they're really not that young.
01:00:59.659 --> 01:01:30.110
Yeah. And I think task probably has quite a lot to do with it. So we did think that this task has been shown to kind of, yeah, pattern with left lateralization. But perhaps it's not as good as other tasks that do this. I don't know, perhaps word fluency or something like that might be kind of a stronger or better measure. I think that might be part of the explanation. But we were certainly surprised to kind of see that they weren't that lateralized.
01:01:30.420 --> 01:01:56.250
Yeah, I don't want to get too down in the weeds. But you know, I think that if the task has a fairly high cognitive demand as this might for children, you might kind of start to get like bilateral anterior insula, parts of the multiple demand network, washing out the left lateralization in the language network. Maybe if you had ROIs that were more, just focal on the IFG. Like you might see more convincing lateralization.
01:01:55.320 --> 01:02:03.929
And actually, the thing that I didn't talk about, is that our baseline here to rest, right. And so again, if we had had a sort of higher level baseline you might have expected to see
01:02:09.139 --> 01:02:18.949
I guess like for the main purpose of what you're trying to do with it, you really did, I think, convincingly show that these kids are not less lateralized.
01:02:19.559 --> 01:02:27.719
Yeah, and I mean, Dorothy has kind of shown this with some transcranial doppler work as well. So that's kind of nice in that it paints a kind of more consistent picture.
01:02:28.079 --> 01:02:39.090
There are some ideas emerging out of her lab now that perhaps suggests that maybe it might not be lateralization on one task, but in a kind of suite of tasks.
01:02:35.190 --> 01:02:59.699
So you know, you might, as someone who has language impairment, you might have a greater risk of having language impairment if you have this more diffuse pattern of lateralization. Across one task, you're kind of left lateralized and the next you're right lateralized. The next you're kind of like bilaterally lateralized, relative to someone who is left lateralized.
01:02:59.000 --> 01:03:21.949
I don't know I find that a little hard to buy, actually, on second thought. But anyway, that's probably a topic for another day. So should we conclude from your study that there are no frontal or striatal differences between these populations? Or is that premature? Do you think there's other lines that need to be investigated?
01:03:25.139 --> 01:03:42.150
This is a hypothesis I'm quite invested in. So I feel like a little bit biased. But I think in general, one study is not the end of the story. In the laterlization work, where I'm not as biased I don't think, I still think that this study isn't the end of it.
01:03:38.219 --> 01:03:56.369
I think that as people look at this idea of inconsistent lateralization, for one. And I think similarly, with the frontostriatal differences, we might expect that actually, maybe we'd see them for like tasks that a little bit harder.
01:03:51.750 --> 01:04:05.849
One thing that I'm looking at at the moment, is to look at some of the structural data from the BOLD study. And that is kind of suggesting that there might be differences in the DLD group.
01:04:03.030 --> 01:04:08.460
Although obviously, that's kind of work in progress. So I don't want to make too much of it.
01:04:09.329 --> 01:04:14.579
Which would kind of leave us with this interesting picture where maybe structure and function don't correspond fully.
01:04:14.699 --> 01:04:37.679
And we need to start thinking a bit more about the task we're using. We haven't looked at things like connectivity. I think all you can conclude is, at least for this kind of simple language task that they can perform, we don't see a hint of a difference. And so it can't be a kind of difference that characterizes the language system, whatever the language system is doing.
01:04:38.230 --> 01:04:48.250
Right. And, you know, you're talking about functional and structural differences. Whenever we don't see a difference, it really just means that we don't see a difference. It doesn't mean that there isn't a difference, right.
01:04:48.250 --> 01:04:54.909
So I mean, there's probably there's got to be a difference because these kids have language deficits. There's something different about the way their brain is processing language.
01:04:56.559 --> 01:05:01.809
Like you said, there's just many other avenues you could look into. Diifferent tasks, connectivity, all kinds of things.
01:05:02.320 --> 01:05:04.869
Yeah, I think it's a project that's gonna keep giving for a little while.
01:05:05.739 --> 01:05:09.789
Oh, yeah. I bet.
01:05:05.739 --> 01:05:14.800
So is the major focus of your work in your lab gonna be to follow up on this, or have you got other projects that you're working on too?
01:05:14.989 --> 01:05:17.000
A little bit,.
01:05:14.989 --> 01:05:19.909
I mean, this is a project that's very, very close to my heart.
01:05:17.000 --> 01:05:45.590
And I'm really lucky that Kate is the kind of PI who's happy to collaborative even though I've left. But I think something that I've become really interested in, in chatting with people with DLD, is when they come in, they tend to talk, at least at this sort of older age range, that they really avoid language environments. So they don't tend to kind of you know, so for instance, they don't read very much, which you might expect.
01:05:42.619 --> 01:06:17.900
But they also say that, you know, they might avoid audiobooks, they might avoid television, and the kinds of things they do tend to be not very language heavy. And, and I've been thinking a little bit in this idea from the sense of like, you know, what motivates us? Because we know from the dyslexia literature, that there's this idea that, you know, the rich get richer. So you, as you read, you expose yourself to more complex grammar, more complex vocabulary, and therefore, your language gets better, and you feel more able to tackle more challenging books. And so I've kind of been thinking a little bit about that kind of cycle.
01:06:13.639 --> 01:06:52.039
And also some cool work that Pablo Ripolls at NYU has been doing, where he's been showing these links between sort of reward systems and language systems, which I think we don't really think about at all. And they're not core language systems. But I do think they're interesting in terms of what we want to do, what motivates us, and what we kind of end up learning as a result. So some of the new work in my lab is kind of focusing on some of those ideas. So how can we measure intrinsic reward? How can we measure reward that's related to language? Is that going to be different in people with say, dyslexia or reading disorders?
01:06:53.300 --> 01:06:59.119
And does that kind of affect the kind of like, reward memory link that you might expect?
01:06:59.570 --> 01:07:07.579
Oh, that's really interesting. Do you think that that would have clinical applications, that you could develop maybe therapeutic approaches around?
01:07:07.760 --> 01:07:41.090
So one of the ways I'm trying to think about this is not kind of in a person specific way. So like, you know, this person is motivated, this person isn't motivated. But more in the kind of state kind of thing. So when you're motivated in this particular moment in time, or when you find something rewarding in this particular moment of time, what does that do to your kind of learning? And how can we get kind of more moments like that, perhaps. And I think that might be a cool approach to take into clinics, because then you're not kind of saying like, well, this person really can't benefit from this, because they're not this particular way. I think that that's kind of really exciting.
01:07:39.199 --> 01:08:12.889
I guess, one thing might be that we find that motivation isn't different at all. And those kind of motivation-memory links are preserved, in which case, that might be a really cool system to target for intervention. Because actually, you could do more of this, and then you get better benefits because you're enjoying this more. Or we might find that actually, those links are affected. And then we might want to think about ways that, you know, we can boost that kind of reward system. I'm not saying that, you know, that's the cause of language problems. I would think that it's very unlikely.
01:08:12.889 --> 01:08:20.270
But I think it's something that kind of possibly develops alongside having a language problem that you perhaps don't enjoy language as much.
01:08:20.270 --> 01:08:33.800
Yeah, I mean, it could kind of create a negative feedback loop, right, like you said. So you mentioned your daughter, has watching her acquire language changed any of your views on the science of language development?
01:08:34.550 --> 01:09:16.489
Oh, yeah, absolutely. I think theoretically, I was a bit more of a constructivist. I kind of think of things not from a very modular perspective. Certainly, seeing her I feel like she does more with input than I give her, which is perhaps a kind of naughty thing to say, given my PhD upbringing. So it shaped things a little bit. It's just fascinating to watch her read, to watch the words she learns, but also to kind of see how much kind of emotion and perhaps reward affects things. She loves her dad much more than she loves me. And whatever he says, is learned and received. And things I say aren't picked up quite as much.
01:09:16.489 --> 01:09:25.489
Oh, I'm sure that's not true. Having kids took my thinking in a similar direction. I have to admit.
01:09:26.779 --> 01:09:32.180
There's something about watching it unfold that sort of makes the biological basis more salient.
01:09:32.779 --> 01:09:33.590
Yeah.
01:09:34.039 --> 01:09:35.689
Well, thanks very much for talking with me.
01:09:36.170 --> 01:09:47.029
It's been a pleasure. It's been really interesting. I kind of went back to think a little bit about this issue about performance and reread some papers that I had read as a PhD student, which was fascinating, actually.
01:09:47.720 --> 01:09:49.760
Oh, yeah. Which ones did you go back to?
01:09:50.960 --> 01:10:00.470
I went back to the Tim Brown Cerebral Cortex paper, which I'd read loads of times, but just hadn't read for a while. I was like, wow, this is a really impressive piece of work.
01:10:00.710 --> 01:10:16.609
And you have more of a perspective on it now right having done this. Well, it's a really cool study. And I enjoyed reading it and talking to you about it. I think it really sets the record straight.
01:10:12.409 --> 01:10:23.060
I mean, this is much more well powered than any of the previous literature and kind of like, just sets down the baseline of what we know and what we don't know.
01:10:23.569 --> 01:10:31.550
Well, thank you. If I haven't done it before, I should really acknowledge that, you know, I was in maternity leave for a significant chunk of the paper.
01:10:32.510 --> 01:10:43.520
There are loads of coauthors on this who did so much work in terms of the heavy lifting of like, testing and lots and lots of things. So, yeah, a big thank you to all of them, and particularly Kate.
01:10:43.609 --> 01:10:45.470
Yeah. Cool.
01:10:43.609 --> 01:10:47.090
Thanks a lot. I hope we can catch up in real life sometime soon.
01:10:47.210 --> 01:10:50.600
Yeah, that would be lovely. And thanks so much for having me. It's been a real pleasure.
01:10:50.630 --> 01:10:52.520
Great. Take care. You too.
01:10:52.689 --> 01:10:56.770
Bye. Okay, that's it for Episode Seven. If
01:10:52.699 --> 01:10:52.909
Bye.
01:10:56.770 --> 01:11:04.810
you'd like to learn more about Saloni's work, I've linked her lab website and the paper we discussed on the podcast website, which is www.langneurosci.org/podcast.
01:11:06.130 --> 01:11:11.859
I'd like to thank Latane Bullock for editing the transcript of this episode. Thank you all for listening. Bye for now.