How does the brain 'think'? Pt. I

This is part one¹ of a series trying to explain my PhD, because people keep asking. I want to point out that my initial response to this is always “I promise that you won’t care”, but no one ever believes me. So, as a punishment, I’m going to make you read it. For three articles. You did this to yourselves.

–

You can make some rough distinctions in the sciences of mind. People who say they study psychology are usually interested in how and why people behave in certain ways. People who say they study cognitive science are usually interested in how people think about stuff. And people who say they study neuroscience are usually interested in how the actual architecture works—the brain and the nervous system more broadly.

I say I study cognitive neuroscience. What we usually care about is how people think, but specifically, how the brain does it. It’s all well and good to imagine the various possible different kinds of thinking, but what we really want to know is which of these the brain can actually implement.

Let me illustrate what I mean. First with a simple example, then with a couple of examples of thinking the brain doesn’t do. And I’ll wrap up with the most interesting (to me) question in cognitive science: what even is thinking?

¶Habits across the sciences of mind

To illustrate, let’s take habits. You probably know that most of the time, you go around on some kind of autopilot. Most of the stuff you do is routine, habitual. You learned it at some point, and learned it so well that you do it automatically. We could consider this a kind of thinking, in the sense that information moves through our brain.

Now psychologists will tell us things like, ’habits are formed through repeated behaviours reinforced by rewards.

Cognitive scientists will say things like, ‘we must have some kind of memory that allows us to learn new things, and probably some other kind that allows us to easily store and retrieve habitual things’.

Neuroscientists will notice things like how neural circuits adapt and change with repeated behaviour, and that the basal ganglia seems to play a role in this process.

Cognitive neuroscientists will say, hey, maybe the memory thingy that helps us learn new stuff uses the basal ganglia, and the memory thingy that happens automatically lies in the pathways of these neural circuits. They bridge the cognitive with the neuroscience.²

¶The value of cognitive neuroscience

So this is a silly example. But it teases the things apart. And the things should be teased apart, because on occasion, cognitive scientists will come up with ideas about how people think that don’t really seem like they make sense, in the context of the brain.

So, for example, back in the day, cognitive scientists were keen on the idea that human thinking was some sort of rule-based symbolic processing. They sort of imagined the brain manipulating logical statements—if x is true, then do y thing, else do z thing. I see the apple. I access my mental representation of an apple. Ah, yes, I know what this is. Connected to it is the eating concept. I am hungry. I activate that. If the apple is plastic, I have been tricked. I activate my humour concept. Something like this.³ We’d run these behavioural scripts over mental representations of things in the world—symbols that stood in for actual things, that we could manipulate in our minds. But the brain doesn’t really seem like it’s doing this.⁴ We can’t find anything in the brain that would allow it to store information literally, or operate on some set of symbols that represent things in the world. Instead, the brain has this kind of distributed set of neural networks, all running in parallel, that sort of map predictable inputs to predictably useful outputs.⁵

Or, similarly, we once thought that all our mental functions were highly modular—compartmentalised into precise, function-specific regions. The memory bit was over here, the other memory bit was over there. The bit that pays attention to things was somewhere else. And so on. Now, while it’s true that certain brain areas are specialised, research in more recent times indicates that cognition generally involves a web of integrative processes across multiple regions. Our brain functions are far more interconnected and dynamic then the lego model of mental modules would imply.

Basically, tying the cognitive science to the neuroscience keeps us honest. Gets us a little closer to understand how we really work, and stops us going down the garden paths of philosophical enthusiasm.

¶The interesting question in cognitive neuroscience: what is thinking?

Earlier, I used the example of habits to illustrate a kind of thinking. But it’s not exactly clear that this is thinking. It’s reflexive, isn’t it? Automatic. When we talk about thinking, we’re usually talking about something higher level than this. Habitual, associative processing in which we respond automatically is not really the kind of thinking people want to know more about.

Sometimes, people might assume that thinking is metacognition—our awareness of our thought processes, often accompanied by an inner voice. But, in fact, even this is often habitual. Many of your conversations, with yourself, or with others, is at some level just rehashing stuff you have said before, or thought before. It’s actually quite hard to think of forms of thinking that aren’t habitual. We really are animals first, and as behavioural science pioneer Edward Thorndike once said, it’s possible that all thinking is just the consequence “of an increase in the number, delicacy, and complexity of associations of the general animal sort”.⁶ But there are some quite striking, and puzzling, forms of thought that do seem to be truly higher-order.

So, for example, you shouldn’t text and drive. You shouldn’t text and drive because you can’t switch your attention between the two things fast enough to do either of them well. This is a problem, for driving.

Now, we can explain lots of driving in the context of the brain—how the brain does driving. It’s a form of mapping inputs to outputs, largely. See something in front, depress the brake. It becomes habitual, like anything else. That’s why you can drive somewhere familiar and not remember the entire journey.

We can also conceptually imagine how texting is done in the brain—mostly mechanical, inputs to outputs, automatic responding. See text, answer is yes, type yes. You don’t really need to think, alot of the time. Here, there are some more questions about how we do language, but language processing is also well-researched, and seems largely automatic.

But what no one has a good explanation for is what is responsible for switching your attention? What is deciding whether you should focus on your phone, or the road, at any given point in time?

Here, it seems like some higher level, more think-y process, is going on above our two lower level, more automatic processes. Something that arbitrates over stuff. This is probably closer to thinking, than either the automatic driving, or the automatic texting.

There are lots of examples of this kind of higher-level thing. Metacognition is probably, in part, an example of this.⁷ So is the ‘Stroop’ effect, or its relatives—any task where you have to inhibit some automatic process to do something less automatic instead. There are others too. Collectively, you might have heard them called ‘executive functions’ or, less commonly in pop-science, ‘cognitive control’. We also call it controlled processing, or in my lab ‘mechanisms of flexible attention/decision-making’. But, as the various names imply, no one really knows what these things are, or how the brain does them, with any kind of confidence. Only that, as these kinds of examples illustrate, they probably exist.

¶Outro

Often, executive functions seem pretty varied and disconnected from each other. In the literature, people carve them up in many different ways. So we might call the thing that switches attention between our phone and the road the ‘attention switcher’, and the thing that inhibits automatic responses to make less automatic ones ‘inhibitory control’, and so on. But naming these things doesn’t explain them. Indeed, we don’t even really know if these are different things, or aspects of the same thing—some more general control mechanism. And so, we end up with all these little people in the brain that are in there, pulling levers and whatnot—deferring the problem of solving these higher-order forms of thinking to the brains of these little people we have invented. It’s called the homoncular problem of cognitive science.

And that’s what my PhD is about. Trying to eliminate these little homonculi by actually explaining how the brain does this stuff. Trying to work out exactly what thinking really is, by understanding how the brain might achieve it.

You can check out part two and part three for examples of that.