Building a brain

20 April 2019

Interview with 

Helen Keyes, Anglia Ruskin University; Duncan Astle, Cambridge University

BRAIN

Brain schematic

Share

Can we build a brain? What does musical training do to your brain? We pick apart some of the latest neuroscience news with Duncan Astle from Cambridge University and Helen Keyes from Anglia Ruskin University. First up, Helen looked at a paper about whether musical training can improve your ability to filter out irrelevant information, and focus on the task at hand...

Helen - Well we know quite a lot about how music shapes the brain but we don't know anything yet about how music affects your attention. So we know that music can actually change your brain structure in ways that have to do with music. So they can change the auditory cortex if you're musically trained, it can make changes in your brain that lead to you being better at recognizing melody and tempo changes, things that you might expect. And slightly more interesting is that those skills can transfer into other auditory and skill improvements such as being able to recognize speech in a jumbled noisy environment. So there is a slight transfer already known and from these skills but this paper was being a bit more ambitious, looking at whether musical training can transfer to other cognitive skills and lead to improvements in your attentional skills.

Katie -  how did they go about trying to do that?

Helen - Well they looked at 18 professionally trained musicians who were pianists and compared them to 18 non musicians and these samples were matched for age and gender.They asked the participants to carry out an attentional network test to measure their executive control function. So in this test your job would be to say whether a centrally presented arrow was pointing to the left or right. A very easy task but that arrow would be surrounded by other arrows. Sometimes it'd be pointing in the same direction as the central arrow so to be congruent, and sometimes it'd be pointing in the opposite direction to the central arrow so incongruent.

Now this task is very easy if all the arrows are pointing the same way, but if the central arrow is pointing in a different way from the distractor and flanking arrows, it can take you a little bit longer to focus your control on the central arrow and put the irrelevant distracting information out of your mind. So that's what this test was measuring and they found that musicians were more efficient at ignoring the distractor arrows and focusing on the task at hand and getting the answer right more quickly, the direction that the central arrow was pointing.

Katie - So if being a musician is good for our ability to attend to things, does that mean we should all be teaching our kids to play the piano?

Helen - It depends on your perspective. So I would say yes! Teach for success. This is a great idea. The tiger parents are correct. The idea that if you teach your child or train your child to be and have a sustained attention on a musical task, this might transfer over to other cognitive skills and their ability to focus their attention better. However I'm sure other psychological colleagues would point out that this can also have negative impacts of putting too much pressure on your children. But from a purely and cognitive point of view yes this is a very good idea.

Duncan - I always think with these sorts of natural experiments, the chicken and the egg problem, as in how do we know for sure that it's the musical training that is causing these cognitive changes over time and not that these cognitive differences are there anyway, and they emerge at different points and in different ways across the lifespan and that's why they seem to co-occur with the musical playing.

Helen - I mean this is an excellent question and it's addressed by the paper in two ways. So firstly they ran additional analyses to show that the number of years of musical training that you have is strongly correlated with this improved cognitive function. So that suggests that it might be the training that's driving the effect, but I think you're right it's very difficult to say, you know, especially with classically trained musicians whether it's you know an educational effect or something else happening here to drive this.

The only study that I can think of that would support the idea that it's the training driving the improvement is a study that looked at 70 children who were between the ages of 5 and 7 and some of these children showed an interest in taking up musical training and some of them didn't. And at that particular stage, no differences, no cognitive differences or musical differences even were observed in that group at that stage. So it's not fantastically concrete evidence but it is suggesting that you can take people with the same baseline and musical training can perhaps have this positive effect.

Duncan Astle looked at a paper which built an artificial brain in order to better understand how the human brain works...

Duncan - A brain cell in this pretend brain is - imagine it just like a circle on a piece of paper. We could draw it out. But at a computational level it's just a simple equation. And usually it's a learning equation. So it takes information in and it adjusts itself depending upon what it's just received and then it sends on a message or an answer to the equation onto the next neuron and so on, and so we can then imagine that we have a whole sheet of those neurons, a whole sheet of those circles each with an equation. Then of course we can layer the sheets up to create multiple different layers of our pretend neural network and that in essence is how the simulation works.

Katie - You've got this pretend brain, this highly complex system. What did they do with it?

Duncan - Well they trained this pretend system to perform multiple different cognitive tasks by giving the brain input and then checking its output. And gradually over many thousands of iterations they managed to train it to perform 20 different tasks - memory tasks, some inhibition tasks, attention tasks, some target detection tasks. And when they had trained it to perform these different tasks they could then explore "what has this neural network learnt and how is it able to do this?". And they found some really important things so the first amazing thing is that it can do it. It can be so flexible that it can perform these 20 different tasks really well. That is a first and that's surprising.

The second thing they learnt is that naturally emerges different types of neurons or different types of brain cells within the network that offered different functions. So for instance some of them seem to be really important for the inhibition tasks. Others seem to be really important for the memory tasks, just like you might expect to find in the real brain.

The third thing they found is that some of the tasks will share neurons so there seems to be some general purpose neurons. We know that's true in the real brain.

Also the fourth thing they learnt was that the system learned to combine different clusters of networks to perform really complex tasks. So we learned that you can perform really complex tasks by combining much simpler tasks. Again that's something that the real brain does too. And finally they were able to show that you can damage this simulated brain in different ways and the patterns of impairment to performance that will produce are really similar to the pattern that you get when real people experience brain damage. So there was some really key surprising and kind of quite groundbreaking findings from that analysis.

Katie - First of all that's amazing. We've got a pretend brain going on in a computer. Looking ahead what kinds of applications could this be used for? Could you get to the stage where if it's good enough you can test out treatments on a pretend brain before testing on a real brain?

Duncan - I think artificial intelligence often gets a really bad rap. It's going to steal your credit card details and steal your job and so on. But actually it's an amazingly powerful tool that we can use to all sorts of brilliant ends. So from a scientific perspective it gives us a great way of exploring how the brain might work and provides us with a way of interrogating that system in a way that you can't do with a real human being. So for instance in our lab when we're designing new tasks or new experiments we not only run them with human beings, we also run them using these kinds of simulations so that we can compare and contrast how the two systems are doing it.

So that means that we make a lot more progress in our understanding of what's really going on under the bonnet. And you can start to simulate things, so for example maybe we could simulate different kinds of environments that children might grow up in by simulating the way in which we train the network. And that's the kind of thing that we would never be able to do as an experiment ourselves but because we can simulate a brain and we can experiment with it in all sorts of incredible ways to reveal new insights about how the brain works, how it develops, how it recovers from damage and potentially in theory what kinds of interventions might be beneficial to create a more robust or more resilient system in the future.

Katie - Is it fair to say that even though this pretend brain is incredibly sophisticated, there's always going to be room for a bit of human error, because I guess there's always going to be a coder behind the the equations that I've gone into in the first place.

Duncan - In essence even as complex and sophisticated as this particular simulation is, it's still not a patch on the real thing. So the human brain is the most complex computational piece of equipment known in the universe. And so simulating it in this simple way is really useful from a scientific perspective but it would always be a bit of a simplification relative to the real thing. That doesn't mean it's not useful, it still is massively useful but it's never going to be, you know, 20 different tasks. You and I can perform a lot more than 20 different tasks and I think we'll always struggle to produce a full simulation, but this could still be really useful.

Comments

Add a comment