Using mathematical models of behavior, researchers see brain activity reliably preceding action
Patience may be a virtue, but apparently it can also be predicted. Apparently, all scientists have to do is hook up electrodes and study activity in the brain cells associated with it.
The study, published in Neuron, involved a collaboration between researchers from the University of Washington, Seattle; University of Oregon, Eugene; New York University; University of Yamanashi, Japan; and Champalimaud Centre for the Unknown, Portugal.
The researchers observed and analyzed patterns of brain cell activity in rats performing simple tasks. After studying these neural patterns, the scientists have created a dictionary of signals that link a particular neural pattern to the action later carried out by the rat. However, don’t dream of being able to predict football or basketball scores, because the predictions precede action by a few hundred milliseconds.
“This kind of research has not been conducted before, and we are the first to show how the behavior of an animal can be predicted,” Stefano Recanatesi, a lead author of the study, told Truly Curious. “Although some studies are there which analyzed the decision-making capability of animals, [their analysis is] not the way we did [it].”
The neural network
A brain consists of more than 80 billion neurons that work in a coordinated manner. These informational messengers use electric and chemical impulses to transmit a signal from one part of the brain to another, and to various parts of the body.
Depending on which body tissue these nerve cells innervate, a specific action is produced, no matter a person’s conscious or unconscious awareness of such action – e.g., the heart beats synchronously, legs and arms move on our command, and so on.
A classic neuron consists of three basic parts: a middle cell body, also known as the soma, and extensions on either side – a set of dendrites and an axon. Dendrites are thin filaments surrounding the soma and carry information from other neurons to the cell body; they are also considered the “input” part of the nerve cell. The cell body receives the information and provides the necessary energy for the message to pass through the other part of the neuron: the axon. An axon is a nerve cell long projection that carries the information from the cell body and sends it off to other neighboring cells (neurons or not). Simply put, the axon is the “output” part of the neuron.
Importantly, neurons only make up around 10% of the brain; the rest consists of other cell types that help support and nourish nerve cells.
Our brain produces specific impulses, which form patterns in response to any particular stimulus.
“There is already some literature on how the brain plans everyday functions and takes timely decisions,” said Recanatesi. “Thus, it was natural for us to seek a way to find out what the brain is up to while carrying out any special task.”
He further explained, “For this, we utilized one or more electrodes and installed them in different head areas [of rats] to record signals from different neural cells at a time.”
Telling the future
The researchers uncovered how brain activity can signal whether a rat will be patient or impatient while seeking water to drink.
They worked on 37 male adult Long-Evans hooded rats. The rats had free access to food, but water was restricted to the actual testing phase, to ensure they were thirsty enough to focus on the task.
The experiment had rats poking their noses into a particular port on the wall of their cage. The action resulted in a sound, a cue that there would be a reward (a small quantity of water) at a second port. However, if the rat lingered on at the first port for another randomly provided tone, it received a larger quantity of water at the reward port.
All the rats had 10-24 tetrodes (sets of four electrodes) to measure activity in their motor cortex, the area in the brain involved in producing movement.
Those rats willing to stand and wait to be served were termed “patient.” The patient rats usually waited 10 to 20 seconds to earn their greater reward. Conducting the experiments was not easy, said Recanatesi.
“We worked for around four to five years,” he said. “[It] was one of those kinds of projects where you start with an idea, but it doesn’t work at the end.”
What he meant was that their discovery was serendipitous.
“We started with a different idea and eventually realized that whatever we were looking for was not there in the data,” Recanatesi said. “We wanted to study the patience level of rats – whether or not rats waited for something or not.” In other words, the general patience thresholds of rats. The researchers simply learned that, like humans, some rats were more patient than others.
According to the summary of the study in Neuron, “The timing of self-initiated actions shows large variability even when they are executed in stable, well-learned sequences. Could this mix of reliability and stochasticity (randomness) arise within the same neural circuit?”
More interestingly, the researchers noted that neural activity could be used to predict the rats’ actions. Recanatesi said, “While studying the neural mechanism responsible for patience, we got data on how electric impulses in the brain change before taking action.”
The motor cortex weighs in
Researchers tracked the activity of neurons in the brain – the secondary motor cortex, to be specific, which has been shown earlier to be important for planning movement, working memory and self-initiated work.
They discovered that a rat’s future action was preceded by activity in a group of coordinating neurons in the cortex. A distinct neural signature was created in the cortex linked to a particular action. Thus, whenever a rat intended to undertake a particular action, a specific neural pattern emerged, which helped the scientists develop a dictionary with each signature predicting a particular action.
“If a person understands which pattern is related to which action, they can easily [predict] it,” Recanatesi said. Thus, the changes in the neurons in the brain make it possible to predict what the animal does next. On average, the behavior was predicted a few hundred milliseconds before the activity was undertaken. Within that time frame, the researchers could predict whether a rat would go for the greater or smaller reward.
Recanatesi said that animals show different types of behavior, both simple and complex. Thus, predicting activity much earlier would depend on the type of behavior; that is, whether it is conscious or subconscious.
The team found that changes in activity in a group of neurons in the M2 area “during the self-initiated waiting task unfolded through a sequence of patterns, with each pattern reliably predicting the onset of upcoming actions.” Based on the interactions between groups of neurons, they put together a remarkably predictive model that better reflects the neural elements in self-initiated action.
The discovery may be further applied to understand complex behavior in all animals, said Recanatesi. “Imagine predicting what a limb of an animal does next. Moreover, we may find similar correspondence in the human brain,” he added. “However, we know that human brains vary from others in several aspects.”
Researchers are now attempting to move on to other aspects of the research. “One of the many questions relevant to this project is observing interaction between multiple brain areas to see how these neural states [conduct their relays]” he said. “Moreover, we wish to find out whether any specific region in the cortex exists that is responsible for this [particular interaction].”
Parvaiz Yousuf is a writer who also doubles up as a researcher. He has publications on cancer biology and biochemistry and has an abiding interest in ornithology.
Edited by Catarina Nunes and Noel Figueiredo.
Click here for the original paper