Welcome to Episode 20 of the Language Neuroscience Podcast. I'm sorry for the intermittent podcasting, but I do have a bunch of new episodes in the pipeline and I'm really happy to share today's episode. My guest is Fred Dick, Professor of Auditory Cognitive Neuroscience in the Department of Psychological Sciences at Birkbeck, University of London. Fred is an old friend of mine from grad school around the turn of the millennium and we've collaborated on several papers
over the years. He does beautiful cognitive neuroscience research around the themes of audition, learning, attention, language, and music. His papers are richly detailed, thoughtful
and always a joy to read. Today, we're going to focus in particular on his paper entitled 'Extensive tonotopic mapping across auditory cortex is recapitulated by spectrally directed attention and systematically related to cortical myeloarchitecture', with co-authors, Lehet, Callaghan, Keller, Sereno and Holt, came out in the Journal of Neuroscience in 2017. I'd also like to mention that Fred is the person who actually taught me how to do stats in real life. He taught me in JMP, which is what
he was using at the time. And I don't know if he's still using it, but I have to admit that I still am. It's my native language for ANOVAs. Okay, let's get to it. Hi, Fred. How are you?
Hey, Stephen. Good to see you again.
Good to see you, too.
I'm very pleased to, to actually be back in touch after a few years.
Yeah, it's been a while I was thinking about like, when was the last time we caught up? It was when I still lived in Tucson, right?
It was. Yeah, we've been in email contact but not actually in, even in virtual space.
Yeah. So you, I kind of got to see a bit of this while we were like, you know, futzing with Zoom for the last 10 minutes. But like, can you kind of share with me like where you are today? And so
I'm sitting in the in BUCNI, the Birkbeck-UCL Centre for Neuroimaging, which has now two scanners. So we just put in a new Prisma now actually a couple of years ago after a remarkable almost decade long saga, that in, which it definitely aged me and but now it's everything is happy. Of course the the scanner went in a few months before the pandemic hit.
All right.
So, so it has not been used as much as we'd like, but, but it's starting up again now. And even, even the director is starting to use it again.
So are you the director?
I am indeed the director. (laughter) for my sins.
Even the director. (laughter)
Yeah, which honestly, it sounds so highfalutin. But honestly, it really involves doing things like getting the scanner cleaned, and mopping up and wet vaccing when there's a flood around the RF cage.
Oh, yeah. No, it's funny, like, when I was in Arizona, I got quite involved in the day to day setup and running of our scanner, and I remember that, the head of Radiology at the time, like coming to me one day and going like, Stephen, this is what, this is what leadership is. And I was like, really? (laughter) Like leadership is like stepping into the gap that nobody else will fill. You know, it's like, when you see something that's not working, you're like that, but
that just has to work. I'm gonna have to make that work because nobody else was making it work. Apparently that's leadership.
Evidently. I guess leadership is, is, you know, trying to to get reasonable passwords for the Internet routers and failing multiple times.
Well, yeah. So you're blaming yourself for our zoom connectivity problems, is that what you are saying?
Ah, ah, only in part. Only in part.
Okay. (laughter)
We have lots of them, so I can just keep going around rooms. So currently, I'm actually sitting in our very child friendly MRI suite with a, MRI compatible monkey that can actually go in with the kids.
And that's lovely. I should take a screenshot of that and use that for the podcast, the MRI compatible monkey. (laughter)
Exactly. And we have a MRI compatible Leo on the other side.
All right. (Laughter) I want those. Yeah.
They're really, they're cute.
I would think that most plushies are MRI compatible, but you probably have to pay like $300 to get one that's guaranteed to be, right?
I'm sure that's true. But, we can scan them with abandon after a little bit of metal detection. But um, yeah, but they make the kids happier. And they're really nice to have on a Friday afternoon.
Yeah, sometimes. I mean, yeah, I think we should take more lovies into our offices, things would be better. Okay.
Next to me, I'll just I'll point out also a picture that you may have remembered so this is our former director, Marty Sereno's drawing of a neuron with all its boutons in the nucleus isthmi of the turtle.
Wow.
Yeah. So...
Yeah, I know that. In the turtle?
In the turtle. Yes.
Yeah, I know that Marty was always studying the brains of creatures that you didn't expect to be studied, like, you know, catching squirrels on the UC San Diego campus and putting electrodes in their brains. (Laughter)
Yeah, it's either it was either Marty or the coyotes.
Yeah. Yeah. So I've known you for more than 20 years. And like me, like your interests have evolved a lot over that time. So the first thing I wanted to ask you is, like, what are your core interests as a scientist right now?
As a, as a cognitive neuroscientist, my interests, strangely have actually remained the same, but I think that, that, like you, the way that that we've approached the problems has perhaps changed over the years. And if nothing else, in my case, actually gone
back full circle. So, so, my interests started out really being driven by by my previous experience, both as a, as a somewhat successful but not always successful learner of second, third, fourth and fifth languages, as well as being a musician and, and how those processes kind of inter intersected, whether they did individual differences in, in kind of basic things like attention to sound sensitivity, due to different sound features, as well as as personal characteristics like confidence
and, and kind of overall arousal levels and so on, contributed to success in, in learning both of these, these expert skills. And then as I went on, I became more and more interested in the process that actually allows you to do that. So what are the mechanisms that, that occur in, in brains more generally, that, that allow us to perform these expert skills? What are the constraints that need to be kind
of built into the brain? But also how does the brain change, change through that how many different ways are there to actually solve that problem? So, when I was at UCSD, I kind of explored a number of different reasons, reasons, excuse me, explored a number of of different routes towards trying to figure out that problem. So, initially, again, like you looking at at people who have strokes, and what happens to them after, after their stroke, so initially adults, but then
also kids who had strokes. And first what happens to their behavior, how, why do they show sometimes very, very dramatic, and what looks like selective deficits in their in their ability to understand different sentence types or words and so on. But then also what, how, what is, is kind of preserved, what, what interactions with the environment, actually drive that, that profile of deficits or springs? And how does that
change with the development. So this is really inspired in large part by Liz Bates's research.
Yeah.
As well as the rest of my amazing set of of advisors at UCSD. And so I kind of think of the of actually the work that I'm doing now, which is, which is, again, looking at at the way that that brain changes itself, when faced with novel learning problems to be quite similar to that initial set of questions that I was asking,
Okay, so you've got like a real theme running through all the various things that you've worked on?
Yeah, I may be the only person who sees it.
I think that I realized that, that I didn't (Laughter) have a theme in my in my scientific pursuits when I had to write my first job applications. And you know, you have to write the research statement. I was like, well, language in the brain. But apart from that, it's like, whatever, like caught my fancy on any given day. I've become more
focused since then. But yeah, I remember back when I realized that I'd basically spent grad school just, you know, doing whatever caught, you know, caught my eye.
You had good taste, though.
Well, I did you know, we did some cool stuff. I mean, and yeah, we worked together. So yeah. Um, so you know, just kind of stepping back even before that, like, you know, did you become interested, was it as a child that you started the, the second and third language acquisition and the violin or like, is that how old were you when these interests first developed?
Well, I was old by, by musician standards when I started. So I was always interested in music and, and in my family, we listened to a lot of music, just in house, mostly kind of, you know, European art music to classical music. But my dad was also a real jazz aficianado, which unfortunately made me really hate jazz up until I was an adult, I always regret. But, but so I started violin when I was almost 10 and piano a few years before that.
And then, you know, being American in the, in the 80s, I didn't really start learning languages until I was in like, seventh grade. So like 12, and 13. But I was really lucky because I had some really good Spanish teachers.
Ah, yeah.
But I never really thought of, of either of those from a scientific angle,
When did the scientific interest in them develop?
Much later. So a part of it was actually to in trying to figure out, especially as a as a musician, how to get myself to learn better. Because that is essentially what as an instrumentalist, you're always trying to do as you spend six hours a day, practicing, and trying to essentially make yourself a better, a better machine in a lot of ways, and a much more automated and efficient and accurate machine.
And so the one thing that I was interested in, was, you know, how, why are these, you know, what drives these huge differences in, in, you know, how good people are at that. And in particular, how does attention play a role, both in, in the both kind of moment to moment attention, but particularly sustained attention to maybe I'll come back later. And, and your ability to control that sustained, sustained attention to particular
dimension. So it's something that musicians think about all the time, like, what are you, as you're, as you're performing, but particularly as you're practicing, what are you actually focusing on moment to moment? Is this, this amazing amount of mind control over minutes, which you really don't get in almost any other pursuit.
I wonder if it's kind of similar to meditation, right? I mean, like, I've never been a meditator, but it sort of sounds like the way that people describe learning to meditate. It really is. Do you, did you ever like try that? Or is it similar?
I have, except that I always have problems not falling asleep when I'm meditating. Which, would be more catastrophic when you're laying a violin at your hands.
Yeah. (Laughter) Yeah, I know how much those violins cost. I know that you need to get like a separate insurance policy and like a separate airline seat for your violin and so on.
Yeah.
Huh. Okay, so, yeah, when I first met you, you were working on this really cool Psych Review paper on why, it was basically like, you know, why do people with aphasia seem to show these particular patterns of grammatical deficits? And you kind of like, explained that in a very, I don't know, from, from an outside perspective, right. It was not a grammatical explanation. It was a processing explanation.
Yeah.
Yeah.
Which I should actually say, as with I think maybe the thing that is characterized my a lot of my work, sadly, is that it's a bit of being a scientific magpie, in that I've picked up ideas along the way and then kind of put them together in, in something that hopefully has a, it has a,
you know, functional role. So, in this case, I actually came to UCSD from Morton Gernsbacher's lab at the University of Wisconsin, Madison, where she had just done a collaboration with Mark St. John, who was actually at UCSD in the cognitive science department and they had had done a early connectionist stimulation of, of what happens when you just look at the relative frequency and consistency of grammatical structures when in networks that are learning to, to understand
who's doing what to whom, and what Mark and and Morton Ann did was to show that, that the relative resistance to break down when you lesioned these, these pretty simple artificial neural and networks, was really strongly predicted by a frequency by regularity
interaction. And so when I came to UCSD, Liz Bates actually put me on a project that was really related to that, because it was something that I'd worked on as a as a special student in Morton Gernsbacher's lab, after stepping off wet eared from the plane from from Germany as a violinist and, and so the work that, that Liz and, and I did on, with the, on the psych review paper was very much based on, on kind of that general
framework. So what, how can you apply different levels of an stress or a noise to a system externally, to kind of probe it's appropriate secrets, and what's, the what's the most parsimonious explanatory framework for the pattern of, of
behavior that you see. So, using, the sort of simplest toy paradigm ever, with four sentence types, but it's this differed along along these frequency and regularity dimensions, we could actually show that that the pattern of results that you see in a phasic patients with his particular paradigm was really well predicted by, by how frequently you you produced or, or encountered that particular structure, and also, how often that structure mapped onto a sort of more, more general word
order frequency.
So you could explain what appeared to be this rather selective, like syntactic deficit that had been explained by sort of linguistically inspired people in those terms, but you could explain it based on more superficial properties of the system.
Exactly. So how does, how does it sort of general associative system, break down under, under different conditions, when it has been placed in a in a given learning environment?
Yeah.
And, and so this was, you know, it's an idea that's been around for a long time. But, but it was quite a nice demonstration that, that you can actually show this, not only in people with brain damage, but also in, in college students before they acquire brain damage, just by giving them (Laughter) a little bit of, a lttle bit of stress either acoustically or, or making the task a little bit harder.
Yeah, simulating aphasia in college students.
Exactly.
So, yeah, that that was cool stuff. But then like, you know, you kind of like, disappeared and moved to London and how did that come about?
So that was that was another one of these, these strokes if fate where, where I was just finishing my PhD, and, and very, very close friend of mine from UCSD, who had gone to Britain, and ended up at Birkbeck College, University of London called me one day and said, Hey, there's a job you should apply for at Birkbeck. And I'd never heard of this place before. Actually, I had because of Annette Karmiloff-Smith and Mark Johnson. Mark was actually there
at that point. It just sort of floated on the horizon and, and I thought, Oh, this is ridiculous, I hadn't even had my PhD, but, but you know, the application was easy and I thought it would be like, good experience. And, and then I applied and, and to my great surprise, they said, hey, you should come out and do this interview, in three weeks, in
London. And so, I sort of scrambled together a talk and, and got, I think you were there at that point, I had, you were one of the people who was subjected my, my practice talk.
It's possible.
Yeah. (Laughter) And, and then went over and and in Britain, things happen really fast with jobs. So I did a couple of I did a talk and an interview a couple of days later, and then at 4.30 that afternoon, they're like, hey, do you want the job? If so, you have two weeks to decide.
Did you take the two weeks or did you know right
I did take the two weeks, to figure out what I
I forgot about that. Yeah. away? could do? Because I'd actually also at that point, taken a postdoc position with Steve Small.
Yeah, and, and so Steve very, very kindly allowed me to do the postdoc, but also allowed me to leave early at quite a tumultuous period, I have to say as well, because Liz was sick at that point.
Yeah.
And, and, and so I did kind of flyover postdoc in Steve's lab and then, in the, in the spring, went over to London and started it and I've been been here ever since.
Yeah, you never came back. I mean, has it ever come up that you might come back? Or are you just kind of like, totally settled into being like a, you know, at this, transplanted American expat now?
No, it has. I have definitely thought about it. More recently, and less recently, so but, but London is an amazing place to, to do work in language and, and in Audition. It's really concentrated environment, not only at Birkbeck and UCL where I am now but just sort of around the, around the entire city as well as Oxford and Cambridge, which, you know, are really close by.
Yeah.
And, and so it's, it's, it's a fantastic environment, and really a very collaborative one. So, and London is obviously a great, a great city, although I know that you have your reservations about it.
About London? I don't know. I mean, like, No, I think it's great city. (Laughter) I don't remember what my reservations were.
I think it might have been with some nasty weather and, and, and kind of smog. But
Yeah, I mean, it's no, it's no Sydney, but I mean, I'd live there. (Laughter) Cool. So, about a decade ago, I guess it must have been maybe a little bit more, you got into what you call myeloarchitectonic mapping, which is just a super cool term, just as a word.
Yeah.
And so, you know, we're going to kind of talk a bit about this paper, recent paper, but um, you kind of have some myeloarchitectonic papers that precede our sort of
That is a good question. So, hopefully, first, focus paper for today. Can you tell us like, you know, first of all, like, what is myeloarchitectonic mapping? not a misnomer, when you, when you talk about it using MRI. So, so in, from the good old days, indeed, over over a century ago, one of the first the first types of stains that people were able to do in the brain was, was to look at the degree of myelin in tissue, but post-mortem tissue.
And, and, for instance, Fleschig in about 1910, all the way up into the 50s, did these remarkable studies, looking at the development of myelin patterns, both subcortical and cortically and showed that, that they proceed at quite a systematic, in a systematic way, where, where primary cortices are tend to be early myelinated, and more heavily myelinated, which had been shown by a number of other anatomists as well. And, and this proceeds in a
relative stereotyped manner. But it really gives you an idea of basically where the, where the, the linchpins are, of, of brain connectivity are, and also where, where different modalities may lay. And, and in the intervening period, of course, a lot of interest has been generated by these maps, both in terms of understanding what the, what the potential functional importance of myelin is, both for, for kind of neural processing, but also
particularly for learning. My interests in using that in MRI was really driven by the fact that the auditory system in particular is really organized or seems to be organized by the degree to which myeloarchitectonic patterns, or the degree of myelination, segregates different auditory areas.
So are you talking, just just before we get too far down, like are you talking myelination, like in the gray matter ribbon or in the only in the white matter?
So in generally in the, in the, in the cortex itself.
So, we're talking like the start of like the axon when it's just come out of the neuron, but it hasn't like kind of gone into those big white matter bundles that sort of make it white matter already. It's myelinated right close to the cell body.
Exactly. So you're looking at the actual strip of, you know, three, four millimeter strip of cortex and you're looking at myelin content there. So, yes, so, so, for instance, if you do a post-mortem stain, so, Gallyas stain for instance of, of the, the Heschl's gyrus. What you'll see, is that the kind of mid layers so for instance, layer four, distinguished by a particular pattern of, of cytoarchitectonics, the kind of cell lines that are in there, tends to really stain very
deeply for, for myelin. That's characteristic of primary areas like A1 which is actually quite difficult to identify in human cortex. But it is, it is really a characteristic of particular cortical layers and this, the heaviness of that stain is one of the characteristics that people use in, in a number of non and non-human species to identify what's called auditory core, so they kind of input regions from the medial geniculate into cortex, but also secondary and tertiary auditory
areas. So, So what has, what has been frustrating with in I think human auditory and language neuroscience, is that, that you know, up until recently, we would say okay, well you know, we're looking at at these early auditory areas, but we didn't actually know how to identify them we sort of had an idea okay there on heschl's gyrus mostly based on on post-mortem work.
But, it was quite unclear where they kind of began and ended, how many of them are, were how much individual variation there was, but, but one of the ways to, to find them, was to combine looking at frequency for reference, so tonotopy, so the gradiation of of frequency preference across, across auditory cortex, with patterns
of myelination. And MRI is of course, highly sensitive to the degree to which brain matter is myelinated which is why you can see in a typical T1 weighted image, that, that you know, that that white matter is is bright in T1 weighted imaging, gray matter is gray. And, and, and it
distinguishes between those. But what's been shown in over the last 15 years or so, is that that we can actually look within cortex itself, in more subtle patterns of of myelination using with quantitative techniques, where you can actually estimate how long it takes for protons to
relax. And there is a more direct method of measuring the myelination or if you actually just look at at the relative signal brightness in both T1 and T2 weighted images and take any kind of ratio measure and this also gives you a kind of estimate of, of relative
myelination. So by combining a couple of different measures we can we can really triangulate on where in a particular person, primary auditory cortex is and use that as, as a interpretive base for, for understanding a particular person's brain.
Right. Okay, so you say you were driven to this interest by wanting to identify where primary areas are in an individual on a structural basis, which isn't obvious any other way?
Exactly.
But you're still going to combine that with functional measures as well such as tonotopic gradients to identify primary auditory cortex in this case. And it's not, I mean, any scanner can do these myeloarchitectonic mapping, right, it's not like you know, you don't need like, particularly specialized sequences. Is that right?
So, it depends. So there are a number of different ways of doing the scan. So, thanks to, to my great physicist friend, so Nick Weiskopf, Martina Callaghan, Antoine Lutti, as well as Marty Sereno, who has been involved with with all this. There, they have developed a number of quantitative sequences that, that actually are special insofar as they're not something
that comes on the scanner. And, and so they allow you to estimate with really, like, quite good precision what the true relaxation rates are for, for different tissues. And therefore that, that is a more direct index of the degree of myelination. When you compare that with with post-mortem work. You can also use what I was referring to before these, these ratio measures. So when you take T2 or T1 ratios from just clinical scans, and these gives you very similar patternings in
many cases. So, this is a, this image that you now see in the Human Connectome Project. And that, that Matt Glasser and David Van Essen in particular have been really very much at the forefront of developing as a technique. And so you can use those as well as kind of a, as a marker of these of these
different areas. And they've done really comprehensive work had looking at the, at the correspondence of, of gradients of these, of these, of image brightness in these in these ratio measures, how these correspond, for instance, to functional connectivity differences, or activation and so on. Or the basis of, for instance, their their parcellation schemes.
Yeah, this is so fascinating. I mean, I remember when we were in grad school, like, there was just no way to, like, make any inferences about psycho architectonics. In except, you know, in post-mortem. And, and it seemed like a real limitation, but it's just exciting that that, you know, that these new imaging approaches are kind of overcoming that roadblock.
Yeah.
Yeah, sorry.
So one thing, I think that is really exciting and as, as sequences get better, and we get better at at analyzing them, that, that the amount of data that we have really looking across individuals and a particular species, so in humans usually, gives us a really different view of, of not only consistency, but also the variability in normal
myeloarchitectonic mapping. So, this is something that, that we know, up until now has not really been amenable to quantification, because in, in post mortem tissue, it's, you know, there are only a few people in the world who can actually flatten it, there is massive distortions. And obviously, you can only do a few, a few subjects at, you know, at a time, and they're very difficult to kind of put
together. But, but with these new imaging techniques, you can really get, you know, hundreds of people and and ask, how, you know, how similar are people? How different are people, but how, what was the kind of level of consistency in what we think of is really a kind of fundamental organizing principle in the brain.
Yeah.
And so this is a fun thing to look at, in terms of cortical evolution, in terms of making cross species comparisons, and but also thinking very kind of more broadly about, for instance, cortical development,
Right. Yeah, I know. And I noticed in the paper that we're about to talk about in more detail, like, you know, you always have all the individual subject functional maps and structural maps presented in your papers. And I think that's something that you probably like, you know, Marty Sereno was always doing that in his work, and I can at least see that influence there. So that's kind of like talk more about this paper that we agreed to
talk about. It's called 'Extensive tonotopic mapping across auditory cortex is recapitulated by spectrally directed attention and systematically related to cortical myeloarchitecture.' Love that title. Kind of like explains, just explains everything about the paper like. So, you know, we've talked a bit about how like kind of this myeloarchitecture is like a really important part of this, because it allows you to identify regions structurally.
And then we're going to, and then let's talk a little about tonotopic mapping. So we talked about that already. It's just this just the idea that, you know, there are these gradients across the cortex, where neurons differ in which frequency they prefer to respond to. And there's many different fMRI paradigms that can be used to map that out as we have actually
worked on together. But can you tell us how you do it in this particular paper that we're talking about like how do you do the tonotopic mapping?
The way that we do just the tonotopic mapping is pretty simple here. And so, usually, what, what we do and this is really based on on Marty's techniques from the 90s, is to present slow sweeps of, of either bandpass filtered sounds or white noise or tones that, that ascend or slowly descend
at, at a periodic rate. And, and then you look to see whether or not there are, there, there are signals, for instance, at a voxelwise level, where you see locking, greater responses to one phase of that, of that cycle as you sort of slowly go up, or
go down. So, so, we can see, for instance, if a if a voxel prefers low frequencies, and we start a sweep with low frequencies, going up to high frequencies that, that, that voxel will show in enhanced response at the beginning of that cycle, or all and then if there's another voxel that prefers high frequencies, then we see it, see that, that increased response at the end of the cycle. So it's a kind of simple, but really robust
technique. And so what we did here is actually make a kind of stepwise version of that, using five frequency bands, for reasons that I'll explain a little bit, a little bit later, but we use the little what I call mini sequences, which always irritates Adam Tierney. They're just sequences. (Laughter)
There's nothing mini about them.
They are mini, there are four tones. And it's a kind of cute little, cute little paradigm, it sounds like (Vocalizes). But actually it goes up or down and in a staircase and, and people's task is simply to detect whenever you hear repetition in one of those, those little mini sequences. And sorry, Adam, and, and so this, this does mean that you have to really pay attention. And, and, as Marty is has shown in, in many of his papers, attention to the the mapping stimulus is
really important. People to like focus in on those on the details of that. So then in this case, we simply ask at each of the five levels of frequency. What what frequency evokes the strongest, the strongest BOLD response? And we call that the winner response. And that's a really typical way of doing
mapping. But we can also ask what what frequency is actually the least the evokes the least response, and we call that the loser takes all which is my favorite, my favorite thing to sneak into a paper.
I mean, I think this paper is just so full of Fredisms. (laughter) I really enjoyed reading it. It's just, you know, you, you write like you speak. And yeah, there's, there's just like a lot of turns of phrase that I enjoyed.
Thank you.
Another thing that I know, nothing I noticed early on in the methods is like, you're like, Okay, we acquired more than 7000 functional volumes per participant. And I was like, well, that's a lot of data. But they must have had a really, really short TR, like, I'm sure you doing multiband you've got like a 300 millisecond TR, or something.
And then I get further down the methods like no, you don't it's like, your TR is a second, you've acquired more than almost two hours of functional data per subject.
It's a lot, we really wanted to make sure it worked. Yeah, so the, you know, as you know, auditory areas, they're small. And, and attention effects can be small. And, and the just to get have enough power, it was, but also just to give people enough time to actually do that, do the task, and you kind of have been trained to it. It was important to get a lot of a lot of
scanning done. Part of this was also, it's that I mean, a lot of the, this paper was very much explicitly inspired by Ayse Saygin's work with Marty much earlier, showing that that there's this really widespread attentional mapping of retinotopic features. And, and so this in a lot of ways, this paper is sort of the auditory analog of of that.
Yeah. Yeah, it's clear. So, with the tonotopic organization, what do you find here like which where do you see tonotopic organization, I mean, not getting into the attention part yet, but just with your basic tonotopic mapping stimulus.
We see basically, that the entire temporal plane
is tonotopically organized. And which really corresponds to what it In a lot of other groups kind of most prominently I think, Michelle Moerel and Federico de Martino have shown in the really beautiful work using often natural sounds, which are honestly better better than tones for, for establishing tonotopy, but, but really see very robust tonotopic maps across the temporal plane and then down into the, onto the, the crown of the of the superior temporal gyrus and little bit of
the the superior temporal sulcus as well, depending on where you look.
Yeah, maybe the dorsal bank. You know, the paper is full of these very beautiful figures, so I definitely recommend to our listeners to check it out. But, you know, without, you know, if people are maybe just doing the dishes or driving and they don't want to do that right now, can you kind of relate that to like, you know, how does, how do the areas that are tonotopic relate to the myelo identified primary auditory cortex?
It's a complicated map. So, when we look at, at the correspondence between how tonotopic a voxel is and how heavily myelinated it is, what we see is that, that when you look across, across cortex or across the entire temporal lobe, that, that the kind of local dips and troughs of, of myelination that, are that can be quite severe, can be explained or at least correspond with depths and and rises in the degree of tonotopicness, both at, within the circular sulcus, so you see a shared drop in, in
tonotopicness and degree of myelination as you go from the medial part of the superior temporal gyrus down into the circular sulcus abutting the insula. And then equally, you see this shared drop in, in how tonotopic an area is and how myelinated is along the lateral superior temporal gyrus. And that's really conserved across
people. What's interesting, is that you also see a disconnect between those and more sort of in the middle of the temporal, superior temporal plane, where there's a high degree of tonotopicity, you actually see a drop in in relative myelination. So it's not a one to one correspondence.
Yeah. And in general, I mean, you did see tonotopy that extended well beyond primary auditory cortex, right, like well beyond.
Way beyond. But, what you also do see is that, there so it's not just that, you see, for instance, that, that our primary auditory cortex, is highly myelinated and then everything else is just sort of less myelinated. What, what we've seen and of course, what, what other people have seen as well, and the most prominently in the in the Human Connectome Project, is that they're multiple gradients of, of, of increases and decreases in
myelination. And more laterally, there's this, that this increase in myelination and then a kind of sharp decrease as you go around the, around the crown of the superior temporal gyrus, that corresponds with the degree, the change in the degree of tonopacity. And so that was really interesting thing to see. And what we've seen thus far unpublished data, even though I analyzed it ages ago, is that, that actually gives you a clue about where kind of more speech selective regions seem to start
firing up. So it's quite a sharp boundary. And and you can actually, when you look kind of person by person, at where you start seeing kind of speechy looking responses, that's where, where this gradient is actually sharpest.
So tell me where that is again, like.
So if you look along the the central part of the superior temporal gyrus going down in the sulcus, there is this kind of bifurcation where, where you see stuff that, that is, you know, very reactive to speech, kind of coming online. And then, the degree of tonopicity and the degree of myelination fall off in tandem,
As, I guess I'm just trying to understand exactly where that gradient is maximal, is it like as the gyrus curls round and becomes the sulcus, like the superior temporal sulcus?
Exactly. Yes.
Okay. That corresponds really well to like what we published in this 2018 NeuroImage paper, using veins corrected fMRI to kind of try and look more in more detail at like the, I mean, we have like auditory stimuli and written stimuli and kind of, it's very interesting because the lateral surface of the STG is totally auditory and the written stimuli
want nothing to do with it. And then as soon as you turn the corner and get into the STS, all of a sudden, it's like totally fine with written or auditory, you know, it's like become a language area. So that's really exciting for me to like, hear that you guys are seeing the same kind of boundary there.
So I will publish this at some point in the near future. And it's something that I've talked to a number of people about, and I think, one one thing that that is really interesting about these, the kind of upper back of the of the superior sulcus, is that at least when you look at what presumably is the analog in macaques, that there are these interleaved, auditory and visual areas, they're very, very thin. And so oops, someone is coming
in. Who is it? (voice of someone coming in) I'm just in the middle of an interview. Later, yes.
I feel so important.
I just said 'interview', which was very funny.
I know! That is why I feel important.
So, so what do you see, I'll just sort of back up a little bit to when you look at the superior temporal sulcus, would you see in these these beautiful papers from, from Pandya and, and...
Schmahmann?
No, it's, it's terrible.
Makris?
No, so it's earlier than that.
Petrides?
Yes. Thank you!
Okay.
So they did all these beautiful tracer studies looking at at these connections of these abutting areas. So it makes sense that you would see shared representation, or rather overlapping representations that at least using fMRI would look, you know, kind of blurred together and surely they talk to
each other. A lot. And, and, and one thing that I always put it in a plug for is, is that when we talk about, you know, auditory cortex, even in A1 itself, in ferret for instance, a substantial minority of of single neurons, in primary auditory cortex are actually driven mostly by visual input.
Wow.
Yeah, Jenny Bizley, my colleague here at UCL, has some really beautiful papers, as you know, showing how many neurons there are that are really excited about visual input in ferret in primary auditory cortex. And there's no particular reason not to think that that that is true and in, in primates as well.
Yeah. I mean, you know, humans, ferrets were are essentially the same. Neither of us are lettuces.
Yeah, and we can all wear cute outfits.
Yeah. (Laughter)
For some reason obsessed with with ferrets, and cute outfits, I don't know why.
Okay, well, we'll have to, I'll have to google that later. (Laughter) So, okay, we get a little sidetracked from the main actual point of view paper, although, you know, I'm pretty obsessed with the STS so as soon as you like, start talking about that I'm gonna like it down that road. But you know, this paper like, so we've been talking about sort of the, you know, the myeloarchitecture, we turn it up, but the paper is actually really about attention, auditory, like spectrally
directed attention. And we haven't even really talked about that yet. But that's actually the main point of the paper. So, can you now talk about like, you know, what you were trying to investigate with spectrally directed attention and what your hypothesis was?
So it is a, in some ways, a really obvious question to ask in Audition. So just as in envision, were the kind of primary representational axis is the patterning, and rapatterning of the retina across the cortex, which you see in the auditory system is the pattern and repatterning of the primary cochlear representation or axis
frequency. And, and we know that in vision that, that a huge amount of the kind of oomph of what you see in in, in cortical activity is due to to changes in attention, and particularly in attention to particular parts of visual scene as mapped by retinotopy. So one question was, do you actually see the same thing in audition? Of course, it's, it's one thing to actually ask people, okay, pay attention to this part of a visual, of
visual scene. So if you get someone to keep looking at a central fixation point, or just an object, but have them pay attention to something that that's a little bit further away from the fovea, they can do that really easily. It's much harder to figure out how to do that in audition. Because people are not used to, you know, saying, okay, well, let's pay attention to like 1200 Hertz, now. There's no, there's no natural framework
of doing that. So I think that's that's made it difficult to figure out a, how to do that. That was one of the things that we had to figure out practically, but, but in terms of why we would even ask this question, what is relevant for to begin with? It's a good question. (laughter) And, but there's a lot of different information in different
frequency bands. So one thing that Lori Holt and Adam Tierney and I had been particularly interested in over a number of years is how we kind of allocate attention to different dimensions of the acoustic signal, when they are more or less informative about a behavioral goal that we want to accomplish. And one of this, the, the kind of most striking dimension is indeed the spectrum. So either particular frequencies, but more often
frequency bands. And so if you take the example of speech, as we've worked on, there are certain bands of spectrum that contain different information about for instance, vowel identity or consonant identity and so on. And particularly when you have multiple talkers, or for instance, a crowded acoustic environment, it may be really helpful to kind of place more attentional gain on that particular part of the spectrum.
And it allows you to ferret out behaviorally relevant information more easily, but perhaps also,
There are those ferrets again!
Exactly, the ferrets, we can give you that later and have a little ferret next to me. But we may also be able to suppress irrelevant parts of the spectrum, which is something that kind of intuitively do we do all the time, for instance, in a bar, so you have, you know, all this kind of like noise, or that might be broadband, or a kind of low level, low rumble, or a high hiss. And, and the information that we really want to eke out might be in a
somewhat more narrow band. So by attending to that, that particular part of the spectrum, and filtering out other parts, if you kind of take an engineering analogue, then we can really add to kind of increase the, the efficacy of our listening processes. So, we wanted to ask a, can we actually see that? Can we actually get people to do that in an, in a,
an experimental task? And also, if we ask people to selectively attend to one frequency in a, in a more kind of more fully populated acoustic space, can we actually see something that looks like, like a added activation in, in some parts of the tonotopic map, when you're attending to differences, the preferred frequency of again voxel can attending to that frequency up activation, and conversely, if we ignore that, the best frequency are the preferred frequency of that
voxel, can we actually make activation go down?
Okay, so just to kind of summarize, your hypothesis is that like, you can get a, that you'll be able to get a tonotopic map, without even physically changing the frequency of the stimulus, but just by having people attend to different frequency bands, you want to see if you can recapitulate those same tonotopic maps but but driving it by attention alone. So attention to a given a frequency band will be enough to activate the neurons corresponding to that band.
Exactly. So we should be able to just map out, most of of tonotopic cortex by actually just asking people to pay attention to different frequencies over time.
So can you talk about like what the stimulus? How do you set up the stimulus and task for that?
That was really hard. So what we ended up doing was to have always two bands of these many sequences. So you can take a bit of in musical terms is kind of like the, the, you know, soprano voice and the bass voice. And, and both of these voices are or, or parts are, are defined by these mini sequences. So you have these two melodies going in different parts of the spectrum. And, and we just ask people to pay attention to the
high one or the low one. But over time, what we do is we change which band, those, those two minute sequence streams are in, such that over the course of a run, we actually sample the entire space of the spectrum that we're able to stimulate
with the headphones. And each band is either is attended to in the presence of each other band, such that, that we can ask, okay, when when, for instance, the highest frequency is in the presence of any of the other frequency bands, and we actually asked you to attend high or attend low that you can see which voxel prefers it when you're attending to the high band versus the low band, and then work out on the basis of sampling across that entire
frequency range. What the voxels preferred attentional frequency is. So really hard task. Really, really hard.
Yeah, and I'm sure that people just love doing that for two hours. (laughter)
You know, oddly, it's quite, it feels quite improving. People didn't mind doing it. It's difficult. But you felt it was the first test that I've ever done where actually felt more awake after an hour and a half of doing it. Actually feels a lot like doing your training.
Okay.
Yeah,
I'll take your word for it.
I'll subject you to it.
I don't think so. (Laughter) So what did you find?
So what we found was, indeed, we could actually basically map out most of the tonotopic, tonotopic cortex, just by having people attend to these, these frequency bands. And what was quite cool is that we could also do that in primary auditory cortex, because it's not really similar to what you see in vision. So that was one of our kind of primary questions, can you actually modulate, all across auditory cortex by attention, including, including primary auditory
cortex. So that that did distinguished some degree with the auditory system from from the visual one. And that and based in some studies, for instance, in Ayse Saygin's paper, V1 was really not attentionally mappable.
Do you think that A1 is just a more high level area than V1? Like, maybe like the V1 equivalent is like in the thalamus or something for the auditory system?
Yeah, so let's just say so unlike the, the poor, impoverished visual system, which only has, you know, it has the very complex retina, and all of its its machinery, and then the LGN, but then you just get shunted immediately to V1.
Right, because the auditory system has already gone through like four or five nuclei before it gets to cortex.
And they're all talking to each other. And, and, and, and, and so by the time you get to cortex is sort of what, what David McAlpine, a hilarious and fantastic auditory researcher, refers to as 'the cooling blanket for the auditory brain'. (Laughter)
That's great. And, you know, you mentioned in your paper that, you, in future work, you might want to see if there are any tonotopic gradients in frontal areas, like Saygin and Sereno, saw for the, for the visual analogue of this experiment. Did you, have you gotten around to doing that yet?
I feel, that is I have to say, it's a bit of a Captain Ahab moment for me. I keep trying, and we see like tantalizing, tantalizing suggestions of it, and then, and then it kind of goes away. Um, there definitely are, what look like, spectrally segregated fiber pathways, potentially, across, across a few frontal regions. But in terms of really finding, for instance, attentionally driven maps, we're not like, we don't have the evidence for that yet.
Interesting.
So that's not to say that they aren't there. But it wouldn't be surprising to me if they're really, really small compared to the visual ones, and more and more kind of individual varied compared against the visual ones. But certainly they're not coming up in the same way that, that, that the the ones in, in frontal eye field and
Yeah. Because those are really quite striking.
Yeah. And, and, and really quite consistent across individuals. It's worth also saying that, that auditory, auditory areas are just way smaller than visual ones. Way, way smaller, so they might just be really hard to see.
Right. I guess you probably just haven't kept your subjects in the scanner for long enough. If you had them in there for six hours, you might be able to answer these questions.
Exactly.
Cool. So, are you still kind of actively working on this, you know, follow up to this project, or?
Yeah, very much. So we, we have a number of projects now we're trying to kind of more naturally direct people's attention to, to different parts of the spectrum, using the kind of these dimensional selective approaches where we get people tacitly to pay attention to a part of the spectrum and other in the service of learning some kind of auditory category, for instance, or in the service of not dying in a game. (Laughter)
There's a plus. And, and with Adam Tierney, we've also been exploring, not really mapping approaches, but more kind of dimensionally, dimension based approaches, for instance, of what happens when you, when you have to attend to, to pitch cues versus more amplitude or duration based cues in order to do a task. How does, how does your brain react to, to that? How does how does your functional connectivity between different regions, change as a function of, of this kind of
perceptual reweighting? And, and we've actually, so Kyle Jasmin, who is a postdoc in with Adam and me, showed really nicely how music participants seem to actually have quite different patterns of functional activity in these tasks, or to, to match non music, music controls. So, so we're doing quite a lot, looking at, at this kind of perceptual, reweighting attentional reweighting approach currently, as well as doing just more mapping.
Mapping. (Laughter) There's always more mapping to be done if you've been trained by Marty Sereno.
Exactly.
Cool. Well, I should I think I've taken up all of our time. So, you probably need to get back to your directing duties.
Yes, probably. Yes. I think I have to do some consulting on, on some new new earbuds, so.
Okay. Yeah. And yeah, some cleaning and checking that needs to happen.
Yes.
Thanks so much for you know, catching up with me today.
Absolutely. A pleasure. And, and we should really do this in the flesh soon.
Yeah, I mean, do you have any plans to come to the US anytime soon?
Yeah, I'm gonna be actually I'm going to Pittsburgh in a couple of weeks. So, which I was really hoping was closer to Nashville than it is.
Yeah. It's a good solid day's drive.
Yeah,
But if you feel like doing a road trip, you know, ready and waiting.
Okay. Yeah. (Laughter)Yeah, I was hoping that my personal helicopter, which would take me down but.
Yeah, not to, not to be.
No.
All right. Well, have a good trip.
Thank you.
And, yeah, I'll catch up with you soon.
Okay, great. Thanks a lot Stephen.
Yep, sure. Take care.
Okay, you too.
Bye.
Bye.
Okay, well, that's it for Episode 20. Thanks to Fred for joining me on the podcast. And thank you all for listening. I've linked the papers we talked about in the show notes and on the podcast website at langneurosci.org/podcast. I'd like to thank the journal Neurobiology of Language for supporting transcription or today's episode, and I'd like to thank Marcia Petyt for transcribing the episode. See you next time.