In the late 19th Century, Psychologist Edward Thorndike proposed his ‘Law of Effect’ stating that behavior that produces a satisfying effect in a particular situation becomes more likely to occur again in that situation, i.e. be repeated. And any behaviour that results in an unpleasant effect is likely to be stopped.
Video describing Edward Thorndike's experiments with cats. Credit: Geert Stienisson, YouTube
In their latest paper published in Science. Prof. Rui Costa’s group from the Zuckerman Institute at Columbia University have uncovered the neural basis of the Law of Effect, using not cats, but mice. Using ingenious techniques, they show that the neural activity underlying rewarding behaviour is activated more frequently. In their experiments they show that mice learn to repeat patterns of brain activity that leads to the triggering of dopamine release in the reward areas of their brain. Moreover, they show that these patterns are progressively and continuously reinforced and shaped to make them more reliable in achieving the dopamine release reinforcement.
“It’s no secret that we derive pleasure from doing things we enjoy, such as playing our favorite video game,” said Prof. Costa, “These results reveal that the brain learns which activity patterns lead to feel-good sensations, and reshapes itself to more efficiently reproduce those patterns.”
Mid brain reward
Reinforcement learning is thought to be mediated via midbrain neurons that release dopamine, the ‘reward’ molecule in the brain. When animals receive a reward for an action, dopamine neurons in the ventral tegmental area (VTA) of the midbrain produce a burst of activity which leads to a release of dopamine.
Based on Thorndike’s behavioural principal and the evidence for dopamine’s release in the midbrain underlying reward, Costa’s group reasoned there must be a neural correlate for the ‘Law of Effect’.
And, taking this one step further, they asked the question, ‘could the brain be trained to learn the right pattern of neuronal activity normally involved in experiencing something enjoyable, and then replay that pattern at will to trigger a dopamine release?
Training the brain to reward itself
The team used a brain machine interface which transformed the electrical activity of groups of neurons in the motor cortex with the production of musical notes. If the mouse learned to produce the right notes in the correct order by coordinating is neuronal activity, its ventral tegmental area dopamine neurons would be activated via light stimulation. This was achieved by placing a light-guide into this area of the brain and by expressing channel rhodopsin, which activates neurons in response to blue light, in the dopamine-releasing neurons. In the control animals the light guide was still activated, however, the light-activatable protein that stimulated the dopamine-releasing neurons was not present. Meaning control animals could make the light work but did not get the dopamine release reinforcement.
Using this ‘closed-loop’ self-stimulation system the group could observe neural reinforcement taking place as the mice learned to coordinate their brain cell activity to play the correct sequence of notes. Whereas no learning was observed in the control mice that did not have the light-activatable proteins expressed in their ventral tegmental neurons.
“In essence, the mice learned to repeat the same pattern of brain activity that had been evoked previously by hearing those musical notes,” said Vivek Athalye, a doctoral candidate at UC Berkeley and the paper’s co-first author.
“In some ways, these results are entirely expected,” said Prof. Costa. “It makes sense that the brain would mimic the feeling of reward it gets from an enjoyable experience by producing the corresponding pattern of neural activity. But it had never been tested.”
Training reorganizes correlated brain cell activity
Delving deeper, the group explored how the patterns of neural activity changed with each daily repetition of the task. They found that the neuronal activity in the neurons became more aligned in the group that received the dopamine release in their VTA compared to the control mice.
Importantly, the authors note that VTA activation is unlikely to have directly caused changes in the neural activity of the neurons in the motor cortex, and the changes in this region are caused by separate mechanisms.
Understanding these mechanisms will provide therapeutic opportunities in the treatment of conditions like obsessive compulsive disorder (OCD) as co-author Prof. Jose Carmena explains, “This research also has important implications for the development of advanced neurotherapies, systems that would treat the underlying causes of brain disorders by modifying a patient’s neural-activity patterns.”
“For example, if a brain’s activity patterns are in overdrive, as is often the case for people with addiction or OCD, could we create a computer program that can help to retrain their brains and downshift this activity?” asked Prof. Costa. “This is something we’re actively exploring.”
Uncovering the neural basis for the Law of Effect is an important step in understanding the functional organisation and learning principles of the human brain, linking neuronal activity with behavior.
Athalye, VR., Santos, FJ., Carmena, JM., Costa, RM. 2018, "Evidence for a neural law of effect." Science