Soha Farboud

Why do we do what we do, even when it isn't actually smart?

An email intended for your boss about that delayed project has been sitting in your drafts for a week – still unsent. It's not actually that complicated, yet you keep pushing it to tomorrow every day. Or you're at a birthday party and before you know it, you're grabbing a third handful of chips, even though you really didn't plan to anymore.

These types of moments feel almost automatic. That is also exactly what they are. Our brain is not a perfect rational machine that carefully weighs every decision. It has built-in shortcuts that help us act quickly, without having to think about them consciously. Usually, those shortcuts are useful, but sometimes they work against us.

As soon as those shortcuts really get out of hand, we recognize them in conditions such as addiction, compulsive behavior, or anxiety disorders. People with an addiction know very well that the substance is bad for them, but the automatic urge to use it anyway prevails. The same applies to people with an anxiety disorder: they know their fear is exaggerated, yet avoid new situations. In all these cases, the brain does something the person actually doesn't want.

To better understand this, it is important to know which brain areas are involved in this type of choice behavior and what those areas concretely do. That is what this thesis is about.

What brain scans can and cannot tell us

Neuroscientists have been using brain scans such as MRI for decades: a technique that allows you to make both detailed photos and movies of the brain. With those movies – which show in real time which brain areas are active – we can see which areas light up at the moment someone makes a decision. But there is a fundamental problem with that approach: it only shows a relationship. You can see that a certain area lights up when someone does something, but that doesn't mean that same area actually plays a role in that behavior. This is the concept of correlation versus causality (Figure 1).

To prove that a brain area is actually involved in a specific behavior, you have to temporarily disrupt the activity in that area to see what changes then. You want to do that in a safe and reversible way, without surgery. But until recently, that appeared impossible for the brain areas that are most interesting for choice behavior and motivation, such as the nucleus accumbens and the amygdala (Figure 2). These areas lie deep in the brain and far beyond the reach of existing non-invasive techniques such as magnetic or electrical stimulation. The non-invasive techniques can only reach the superficial cerebral cortex. Medications can reach those deep areas, but spread throughout the entire brain, making it impossible to determine which specific area causes the effect.

So we needed a method that could stimulate deeply, precisely, and safely. That method now exists, namely: transcranial ultrasound stimulation, abbreviated as TUS.

Ultrasound: from echo to brain stimulation

Ultrasound is focused sound that passes through tissue without damaging it and is perhaps best known from the ultrasound used to view a baby in the womb. The same technology can also be directed at the brain. By placing a small device on the head and sending directed ultrasound vibrations, we can safely influence brain activity from the outside, deep into the brain, with a precision of just a few millimeters (Figure 3).

Animal research already showed nice results, but at the start of my PhD research, there was still little evidence if it also worked in humans. That was a problem, because without that evidence, TUS could not be deployed to answer major scientific questions about the human brain.

Therefore, this thesis had two goals. First: show that TUS can influence behavior in humans. Only then: use TUS to investigate which brain areas are involved in how we make choices.

A nudge in the right direction

To test whether TUS can truly change behavior in humans, it is smart to choose a situation of which we already know exactly how it works in the brain. Because if you test two unknowns at once, you never know for sure what you find. Imagine testing a new medicine for a disease you don't understand yet: if you then see something, you don't know if that's due to the medicine or just the usual course of that disease. So you need a well-known system as a starting point.

For my research in Chapter 2, that was the control of eye movements. We know very well which brain areas are involved in the choice to move your eyes left or right: the frontal eye fields in the frontal cortex (Figure 4, top). Earlier animal research with great apes had also demonstrated that TUS of the eye fields could actually influence the choice of where the ape looks. The question now was: does this also work in humans?

Thirty-five participants played a computer game with which we could study the influences of TUS on their choice behavior (Figure 4, bottom). The results showed that – also in humans – TUS is capable of influencing choice behavior. Stimulation of the left eye fields caused more eye movements to the right, and stimulation of the right eye fields caused more eye movements to the left. The effect was specific and not driven by placebo: if we applied the same technique to the hand motor cortex as a control measurement, nothing changed in the eye movements. Moreover, the effect occurred within a fraction of a second after the start of stimulation.

Every human is different and that was clearly evident here as well: both the brain chemistry and behavior varied strongly between participants. In each of them, we measured the concentration of GABA, an inhibitory neurotransmitter, a substance with which brain cells communicate with each other, in the eye fields. People with higher GABA concentrations naturally had a stronger preference to look to the right. After TUS, this difference disappeared: everyone moved to the same level. TUS brought the brain back into balance, as it were, regardless of how your brain was naturally inclined.

These findings were one of the first convincing indications that TUS can directly and specifically steer behavior in humans. This cleared the way for the main question of this thesis: which brain areas play a role in choice behavior and motivation?

Two systems that steer choices

Before I discuss the results, it is good to understand which two mechanisms are central to this thesis. The first mechanism is learning from experiences: we learn by doing and discovering what works. If you try something and it has a good outcome, you remember that and do it again next time. This is the rational, flexible system that adapts based on experience. The second mechanism is the autopilot of our brain: if something good is expected, we are automatically inclined to take action. If we expect something bad, we are instead inclined to stay still. This is fast, automatic, and evolutionarily old. You also see it in other animals: a prey animal that smells a predator freezes spontaneously.

Usually, these two systems work together and reinforce each other. Think of someone seeing a mango for the first time at the market: the fresh smell and bright yellow immediately evoke a positive feeling, which prompts buying. At home, the mango turns out to be delicious, and the next time at the market, the movement to the fruit stall is almost natural. The autopilot took the first step, and learning from experience (‘this was tasty, I'll repeat that’) anchored the behavior.

But sometimes they work against each other. Think back to those chips at the birthday party. You've learned that you shouldn't eat too many of them, but the sight of the bowl automatically triggers a reaching movement. Or someone in recovery from addiction walking past a pub: rationally he knows he shouldn't go inside, but the automatic urge is there anyway. Especially in those moments, when the two systems do not want the same thing, it becomes clear how strong the autopilot can be (Figure 5, left). To measure this, I developed a computer-based behavioral task with which I could measure both the influence of the autopilot and the learning from the consequences of choices (Figure 5, right).

The brain in action: what happens during choices?

In Chapter 3, I first focused on the question of which brain areas light up during these types of choices. In the MRI scanner, 29 participants played the space adventure (Figure 5, right), while I measured their brain activity in the meantime. Which areas light up when a participant sees a planet? What happens in the brain at the moment of choice? What if they see positive or negative outcomes?

What we saw was that the nucleus accumbens (Figure 2), a small nucleus deep in the brain that has long been associated with reward and motivation, reacted more strongly to blue planets than to red planets. This fits with a role in our autopilot: the area seems to pick up reward signals and put us in action mode. The amygdala, another deep-seated brain area classically known as the ‘fear center’ but in reality involved much more broadly in processing emotionally relevant information (Figure 2), showed a different pattern. When seeing planets, it reacted particularly strongly to planets that participants saw for the very first time. Already by the second presentation, that strong reaction was gone. It seems that the amygdala functions as a kind of alarm system for the unknown: ‘new symbol, no idea what to do with it yet, watch out!’.

Both areas also reacted strongly at the moment the outcome appeared: the moment participants saw whether their choice had been good. So they are both involved in processing outcomes.

These were, however, correlational findings. The ice cream and the shark attacks. Time to take the ice cream off the shelves.

Cause or effect? The nucleus accumbens and amygdala under the microscope

In the final empirical chapter, Chapter 4, the same 29 participants played the space game again, but now three times, and each time they received TUS in one of three conditions beforehand: stimulation of the nucleus accumbens, stimulation of the amygdala, or sham stimulation where everything was the same but the ultrasound did not actually reach the brain. The order was random so we could compare fairly. By comparing behavior after the three conditions, we could find out what changed in their choices and learning capacity when we temporarily influenced these brain areas.

After stimulation of the nucleus accumbens, we saw a twofold effect: both the influence of the autopilot and the speed of learning increased. Participants started with a stronger inclination to act upon reward signals and to do nothing upon punishment signals, but subsequently also learned faster. The result was that the influence of the autopilot was initially greater, but also decreased faster as participants learned more.

That might sound contradictory, but actually, these two processes are less the opposites of each other than is often thought. They have the same goal: optimal behavior. When you don't know something yet, your autopilot provides a handy starting point: do something if there's a reward to be had, do nothing if punishment threatens. Fast, automatic, and often good enough for a beginner situation. But as soon as you have more experience, what you've learned starts to carry more weight and you can react more nuancedly. The nucleus accumbens seems to play a central role in both of those processes.

For the amygdala, we saw a different specific contribution: after amygdala stimulation, participants learned faster from planets they saw for the first time. So the amygdala indeed seems involved in the rapid first processing of new stimuli, exactly what we had seen in the scanner.

Together, these two areas form an important circuit. The amygdala detects new stimuli and quickly assigns a value to them, and then sends that information through to the nucleus accumbens. That functions as the actual hub: it is there where motivation signals – from both the amygdala and other brain areas – are brought together and converted into concrete action. The amygdala gives the starting signal, as it were, and the nucleus accumbens determines what you do with it. These findings are unique: for the first time we have causally demonstrated that the nucleus accumbens and amygdala in humans actually play a determining role in how motivation is converted into action (Figure 6).

What does this mean for practice?

These findings are not only theoretically interesting, but they also help us understand why choice behavior derails in certain conditions, and perhaps how we can restore that.

Think, for example, of addiction. One of the core symptoms is an overwhelming, automatic urge to use, even if you know it's bad for you. The autopilot can then become so strong that what you've learned offers insufficient counterweight to that urge. Both processes, the too strong influence of the autopilot and the reduced capacity to learn from negative outcomes, can be traced back to the nucleus accumbens. My findings support that view and show that TUS of the nucleus accumbens can influence that pattern.

Anxiety disorders show a different pattern. People with anxiety disorders avoid new situations and consider the unknown as threatening. That is associated with an overactive amygdala. Interestingly enough, we saw that amygdala stimulation actually accelerated the learning of new stimuli, suggesting that targeted influence of the amygdala can help people give the unknown a place more quickly.

Just as in Chapter 2, where TUS brought the brain back to a common level regardless of how everyone was naturally composed, TUS also offers the possibility in the context of choice behavior to work in a customized way. Not everyone has the same brain, and not everyone will react to a treatment in the same way. TUS offers the precision to be able to address precisely those individual differences.

Exactly there lies the clinical promise of TUS: not as a brain controller, but as a tool. The brain still has to do the work itself. But a targeted ultrasound treatment could offer the brain a helping hand in relearning, in breaking old automatic patterns. A nudge in the right direction.

Conclusion

This thesis shows that ultrasound is an effective technology to influence human behavior, that the nucleus accumbens and amygdala each contribute in their own way to how we make choices and learn, and that those contributions really have a causal and not merely a correlational link. Together, these insights bring us a step closer to a better understanding of the human brain and to new possibilities for the treatment of conditions where choice behavior has derailed.