A machine for making predictions

Researchers in Brazil and the United States are investigating how the human brain perceives time and uses this information to plan actions.

By Karina Toledo, in Rio de Janeiro

Agência FAPESP – Chatting, playing music and videogames, driving, practicing sports and any other activity that requires foresight and planning would be impossible without the human brain’s ability to measure time.

Understanding how the brain processes time is the goal of a research project supported by FAPESP and led by André Mascioli Cravo, a professor at the Federal University of the ABC’s Center for Mathematics, Computation & Cognition (CMCC-UFABC) and one of the coordinators of the Timing & Cognition Laboratory (São Paulo State, Brazil).

Partial results were presented during the Ninth World Congress of the International Brain Research Organization (IBRO 2015), held in Rio de Janeiro in July.

“Our brains use temporal information in a highly automatic manner to plan activities,” Cravo told Agência FAPESP. “It’s like a machine for making predictions. For example, when you press the button to switch on a computer, you know how long it takes to boot up. If takes a bit longer, you immediately conclude something must be wrong. Either it’s broken down or there’s no power. When you buy a new computer with a different start-up delay, you adapt very quickly. We’re interested in understanding how the brain learns this temporal relation and uses it in future actions.”

To discover the neural mechanisms involved in this process, Cravo and his collaborators at CMCC-UFABC use electroencephalography (EEG) to record electrical activity in the cortex of healthy volunteers while they are playing a first-person shooting game on a computer.

The target appears and disappears on the screen at different intervals. The volunteer is asked to press a button to shoot at the target. Once the task is performed with excellence, the pause between the action of pressing the button and firing a shot is increased.

“The first time this happens, the volunteer obviously misses the target. However, to our surprise, a single miss is sufficient for the brain to learn the new temporal relationship and correct the next motor action,” Cravo said.

According to Cravo, the mechanism used to encode the error and adapt behavior appears to be based on the different phases of the brain’s oscillations. “These oscillations reflect the excitability of neural populations,” he said. “This is an indirect measure of the extent to which a given group of neurons is prepared to process new stimuli.”

In the scientific literature, brain oscillations have been related to various cognitive functions, such as attention, memory and decision making. According to one theory, the brain uses these oscillations to predict when something will happen.

“Studies have shown that if you know when something is going to happen, these oscillations appear to be a good mechanism for preparing a region of the brain to process the incoming information. Apparently, this same mechanism is involved in the case of temporal information,” Cravo said.

According to him, the signal recorded by the EEG device represents the sum of the cortical brainwaves’ different amplitudes, phases and rhythms. In an experiment, the group focused on so-called delta oscillations, which have frequencies ranging from 1 Hz to 4 Hz.

“Previous studies showed that delta oscillations as well as theta oscillations (4-8 Hz) appear to play a crucial role in controlling cortical excitability. Moreover, our task involved temporal dynamics that produced an inherent rhythm of 1 Hz so that these endogenous delta oscillations could synchronize with the rhythm of the task to improve the participant’s performance,” Cravo explained.

The results of the experiment showed that knowing the phase of delta oscillations at the instant that the target was presented to the volunteer enabled the researchers to predict what the volunteer would do the next time the target appeared on the screen.

“It’s clear from our studies that the greater the timing error, the more volunteers adjusted their action the next time they fired,” Cravo said. “However, the only way we could measure this error was by observing the delay with which the target was presented. We therefore decided to see whether neural information – in this case, the phase of delta oscillations at the moment the target appeared – would help us predict what the volunteer would do the next time it appeared. It was as if we were measuring not just the error itself but also the volunteer’s expectation of when the target would be presented.”

This result suggests that in making a prediction, we are preparing not just for what will happen but also for when it will happen, he added.

“These oscillations appear to be fundamental to the encoding of this foresight,” Cravo said. “They enable timely preparation of regions linked to the task to be performed.”

Brain clocks

Does the brain use electrical oscillations to tell time, similar to a man-made clock with a pendulum or a quartz watch with a crystal oscillator?

For Dean Buonomano, a researcher affiliated with the Departments of Neurobiology and Psychology at the University of California Los Angeles (UCLA) in the US, the answer is no.

“Oscillations are a powerful way to tell time, but that requires counting the number of oscillations, something the human brain isn’t good at,” said Buonomano, who came to Brazil to take part in IBRO 2015 with FAPESP’s support.

In Buonomano’s view, the mechanism used by the brain to tell time is based on the dynamics of neurons. Excitatory and inhibitory neurons influence neighboring neurons to create patterns of activity that evolve over time in a complex, dynamic process.

“One possible analogy is an auditorium full of standing people. If you shove the people in the front row, each of them bumps into the next person and this creates a pattern of motion that evolves over time,” he said. “Analyzing this motion like a film, you could mark the passage of time on the basis of each person’s position. You could take several photos of the process, shuffle the images, and then put them all back in the right order because your brain understands the dynamics of this system.”

Previously, it was believed that there was a sort of central timekeeper in the brain, in the form of a specific neural circuit responsible for most temporal functions, analogous to a computer’s clock.

“We now know that isn’t how it works. Time is so fundamental to our interaction with the world that practically all neural circuits are involved in this processing at some level,” Buonomano told Agência FAPESP.

Each circuit specializes in a different way of processing time according to the problem in question.

“For example, the auditory cortex must detect sounds and determine their duration, as well as the intervals between them and their rhythms, in order to process speech or music. The motor cortex needs temporal functions to control the actions of our muscles, and so on. Each one processes time differently,” Buonomano said. We need to know how this processing works, he added, to understand the brain and also to find out what causes the neurological disorders that impair learning, memory and cognition.

Subjective time

Cravo and Buonomano are beginning a collaboration with the aim of understanding subjective impressions of time.

While the brain continues to keep time objectively, Buonomano explained, we feel time passing more or less quickly, depending on the activity in which we are engaged.

“We’re designing experiments to show why this subjective impression of time passing is so often an illusion. There appear to be several different ‘clocks’ in our brain, and our research aims to find out whether they’re all subject to the same temporal illusions,” he said.