While it might not be time to toss the paper-and-pencil model of testing, researchers in Illinois have discovered that an MRI can predict an individual's level of success in a complex task. By studying a specific brain region of subjects who were learning a strategic video game, experts developed insights into the ways we learn.
I spoke recently with Art Kramer, director of the Beckman Institute at for Advanced Science and Technology at the University of Illinois, about his work.
You found that an MRI can predict how much people will learn over a few weeks?
Traditionally, learning is predicted by various paper and pencil or computer-based tests. We asked whether we could gain some accuracy in predicting how much people could learn by using non-traditional measures such as MRI and functional magnetic resonance imaging. We had people in an MRI when they were inexperienced with the task -- before they had any practice -- and used those data to predict learning over about a month.
Did you use a traditional MRI machine?
It's not what you do in a hospital. This is a research MRI. The machine is very similar. It's an MRI machine with some additional software and hardware. We record the activity that's generated -- both the structure of the brain, as well as the function -- as people were playing this video game.
What did you see in the MRI that allowed you to predict the learning curve?
We looked at activated voxels. A voxel is a three-dimensional pixel. It represents different locations in the three-dimensional structure of the brain. We were interested in the basal ganglia, which sits toward the middle of the brain. We know that [certain] basal ganglia regions play a role in new learning. We were interested in exploring the signal we could get from the MRI and whether we could look at the signal in regions that have been implicated in the learning of new skills and whether the signal would be predictive of learning in this complex simulation.
We had different conditions: when people played the game, when people were just lying in the MRI relatively passively. The signal we looked doesn't seem to depend upon doing anything in particular. This is a signal that didn't depend upon playing the game per se because we got the same predictability whether individuals were playing the game or not. But this signal that we recorded probably has a lot to do with the connections among regions in the particular area of the brain. That's still being explored.
The interesting news is that this particular signal is highly predictive of learning, much more so than most performance-based measures.
Talk more about the video game that study participants were tasked with learning. I read that it's 30 years old.
Thirty years plus. It was developed when I was a graduate student. We developed this game to come up with a complex task. Most of the tasks used in cognitive neuroscience or cognitive psychology laboratories are very limited, usually because the researchers want to look at specific aspects of memory or decision making or verbal processing. Most of these tasks are very focused and don't represent the complexity and richness of anything we do outside the lab. We were interested in going in the other direction when we developed this.
We developed this video game as a platform to study the learning of complex skills. We built in aspects of shifting your attention around to different objects in the display, remembering different pieces of information, using different rules depending upon the context and the changes in the game. There's also a very complex psychomotor control in which you use a joystick to control a spaceship. We designed a somewhat entertaining task that was complex to learn, but built in various aspects of memory and decision making and control and attention. It was used very productively in studying learning and studying strategies that might be applied to enhance ideas about how we learn and how much and how quickly we learn.
Why is this work important and what are the potential applications?
These data give us knowledge about how the brain works [without using invasive measures]. There are no tubes or electrodes. Putting people in a big MRI can be a valuable way to look at neuro-circuits in the brain that play an important role in learning. From this study and other studies, we can learn about how the brain helps us learn new skills and remember those skills.
This was actually done for the Navy. The biggest trainer in the United States and maybe the world is the U.S. military. When you think about it, it makes sense. You have all these young people, tens of thousands or hundreds of thousands, come in each year and they have to be trained. There's a lot of high-tech equipment, so there's a lot of training and retraining. They're very interested in optimizing training and learning and retention of skills and transfer of skills. That's why we got the money to look at models of training and learning strategies that might enhance how much or how quickly you learn. But [we're also looking at] various biomarkers that might help us predict who is going to be a better learner than others.
What's the next step for this work?
There are a number of next steps:
- One next step is to understand the mechanism that underlies this remarkable prediction of learning. There are a lot of possibilities. One of them is that it has to do with blood flow. Another has to do with the white matter, the different connections among brain regions. The measure we looked at here also provides some index of iron content in the brain. That might be important too, but we don't know.
- Another way to go is to expand our zoom lens and not just focus on the basal ganglia, but open it up more. We need to look at other areas of the brain and ask whether we can predict different kinds of learning or predict learning more effectively than we have.
- The third way to go is to ask whether these brain regions that we've shown these individual differences in whether these brain regions are amenable to intervention -- whether we can change them to enhance learning. It's also interesting in learning whether these kinds of predictions work uniformly across the lifespan. All we looked at was the 20 to 30-something group. Are there different predictors in kids? What about older people?
Photo: Art Kramer