Bouncing back from mistakes: how brain state improves decisions

Versão Portuguesa

The human brain is an intricate web of billions of neurons, firing electrical impulses in complex patterns every second of our waking day, an unending conversation from which our thoughts, memories, and perceptions arise. But what if the background noise of these neurons could unlock secrets about how our brains function and perceive the environment? A study by a team of researchers at the Champalimaud Foundation represents an important step forward for understanding how brain states might shape our ability to interact with and adapt to the world around us.

The Constant Chatter of Neurons

“The brain isn’t like a computer that turns off when it’s not doing a particular task”, explains Alfonso Renart, the senior author of the study published in eLife. “There’s always a kind of background hum, a baseline activity that can sometimes make it seem as if the brain is chattering to itself”. The team’s study lifts the lid on how that baseline activity, the continuous stream of electrical impulses sent by neurons, impacts behaviour and decision-making.

As Renart explains, “When you look at single neurons in the brain, you see that they are never quiet. Instead, there is a spectrum of activity. At one end of the spectrum, individual neurons change their activity in unison for brief periods of time, and at the other end, during other short intervals, they behave independently, with no real pattern between them”. This leads to periods of synchronised and desynchronised activity in the neurons.

As Davide Reato, one of the study’s first authors puts it, “The brain is like an orchestra. Sometimes the instruments play in unison, sometimes they each play their own tune”. Previous work had shown that these alternating brain states of synchrony and desynchrony correspond to different behavioural states. States of focussed attention are associated with lack of synchrony, whereas during idleness, neurons tend to be in sync. These studies therefore suggested that, in desynchronised states, brain areas might represent information about the outside world more accurately, and subjects would be better able to discriminate between similar sensory signals, such as images or sounds.

However, a peculiar paradox has since emerged: while brain areas in desynchronised states do indeed represent information more accurately, subjects often do not make better perceptual decisions during those states. 

Untangling the Paradox of Desynchronisation

To understand the solution to this puzzle and how the brain’s background hum affects perception, the team devised an auditory decision-making experiment with mice. The rodents were trained to distinguish between high-frequency and low-frequency sounds, with each sound linked to a different response.

Raphael Steinfeld, another one of the study’s first authors, explained the task: “Depending which sound we played, the animal had to report at the right or at the left response port. And we had the animal performing this task while we were recording the activity of neurons”. The team recorded neurons in the cortex, the brain’s outer layer and supposed seat of higher-order cognitive functions such as learning, memory and attention.

“We recorded in the auditory cortex, which is the part of the cortex that processes what we hear”, continues Steinfeld. “One of the challenges we faced was that we needed to record from many neurons at the same time. Having very few neurons at a time makes it very difficult to infer how synchronous the population of neurons is. It also took many months for each mouse to perform the task well enough for us to begin recording. However, once we had overcome these challenges, we were in a position to ask the crucial question: do mice make better decisions about the nature of a sound when the cortex is more desynchronised just before the sound is played?”.

Davide Reato picks up the story, “At first, we found that the state of synchronisation did not affect performance, consistent with the prevailing conundrum in the literature. But then we noticed that, after making a mistake, the animal’s next choice was more likely to be accurate, and this led us to investigate whether the effect of brain state on choice depended on the animal’s success in the previous attempt. When we looked into this, we saw a clear pattern: desynchronisation in the baseline activity of the cortex before the sound was presented did in fact lead to more accurate performance, but only if the mouse had made an error in the previous go. In other words, if cells in the auditory cortex each play their own tunes following an error, as opposed to playing in unison, the mouse’s next decision is more accurate”.

As Renart elaborates, “These findings provided a possible explanation for the puzzle. Clearly, desynchronised states are not always associated with better choices, but sometimes they are. The fact that the success of the previous choice determines whether or not they are made us think of the different ways in which we approach a challenge. If the problem is really hard, we focus all our attention on the relevant aspects in order to succeed. But certain tasks, typically repetitive ones in which we have acquired a certain level of skill, can be performed quite well without focussed attention. People refer to this as being ‘in the zone’, or ‘in flow’. During these flow states – such as in sports, video games or music – subjects are so engrossed in their task that they often report a heightened sense of detachment from their conscious awareness”.

“We think that what might be going on is that mice generally perform these laboratory tasks in a flow state, since the tasks are very repetitive, and the mice have become highly skilled at performing them. Mistakes, however, break the flow, and we think that this fundamentally changes how the brain controls behaviour. When an error is made, it’s as though the brain switches lanes on the information highway, suddenly recruiting the auditory cortex to help navigate the task. At that point, the state of the auditory cortex does influence the animal’s behaviour, so that if the auditory cortex is in a desynchronised state following an error, the animal performs better in the task. It’s as if the brain’s misstep acts as a wake-up call, and in responding to this, the desynchronised state of the auditory cortex sharpens the animal’s ability to discriminate between stimuli”.

Beyond the Study: Future Directions & Implications

The researchers’ findings suggest that in flow states of heightened performance and automaticity, higher-order brain areas like the auditory cortex may be less involved, until a mistake is made. Looking ahead, the team plans to further investigate their theory. “The nice thing about this theory”, says Steinfeld, “is that it’s very testable. If we interfere with the activity of the auditory cortex and see a bigger effect on the quality of the decisions the animal makes after error trials, this would support our theory”.

The team’s findings not only reframe our understanding of the interplay between the state of synchronisation and the accuracy of subsequent decisions, but could potentially open up new avenues of understanding about conditions where neural synchrony or error adaptation might be disrupted, such as in certain neurological or psychiatric disorders. Moreover, the results could inspire new approaches in AI and machine learning, in which the concept of learning from errors is a fundamental principle.

It seems that the constant hum of our neurons may not be the confused mutterings of a brain idly talking to itself, but rather a complex symphony that is continually shaping our perception of the environment, and our performance. As Steinfeld, an accomplished violinist, observes, “It’s something to think about the next time I play a wrong note”.

Original paper here.

Science Snapshot video of the study here.

Loading Likes...