Working memory is a critical cognitive function that enables humans to manage and process multiple pieces of information simultaneously in short-term scenarios. For instance, it allows us to create mental lists while shopping or recall phone numbers for dialing. Despite the general agreement among scientists about the limitations of working memory capacity, there are differing theories regarding its mechanisms. However, recent research from Brown University’s Carney Institute for Brain Science has shed light on why these limits exist.
Michael Frank, a professor of cognitive and psychological sciences at Brown University associated with the Carney Institute, along with Aneri Soni, his graduate student, developed a new computer model focusing on the basal ganglia and thalamus—brain regions crucial for working memory. Their study published in eLife indicates that learning plays a key role in understanding these limitations.
The simulations run by Frank and Soni demonstrate that holding too many items simultaneously becomes increasingly difficult to learn, leading to confusion where stored information cannot be effectively utilized. Simultaneously, the research shows that when confronted with such constraints, the brain adapts by strategically employing mechanisms to conserve space. “The neurotransmitter dopamine is central in this learning process,” notes Soni. This finding offers new insights into disorders associated with altered dopamine function, including Parkinson’s disease, attention deficit-hyperactivity disorder (ADHD), and schizophrenia.
The team’s discovery was made through the development and testing of a novel computer model that mirrored human experiments conducted by Frank’s lab in collaboration with Matt Nassar. This 2018 study demonstrated how humans can “chunk” related pieces of information together to conserve space within working memory. Soni confirmed the accuracy of her brain-like model when she challenged it with tasks similar to those from the 2018 experiment, involving colored blocks oriented in various directions. Over multiple trials, the model learned to strategically compress and chunk related colors like blue and light blue.
Frank’s team’s simulations using this new model point toward learning as the primary driver of working memory capacity rather than mere storage limitations. When Soni ran tests with a version that had no chunking mechanism but ample storage space, it was found that without strategic compression, even abundant storage wasn’t effectively utilized.
A crucial element in the model’s learning process is its simulation of the human brain’s dopamine delivery system. This component activates when information is recalled more efficiently due to effective chunking, reinforcing this strategy for subsequent trials. Soni also manipulated the dopamine levels in her model to reflect known conditions found in patients with Parkinson’s disease, ADHD, and schizophrenia. The results indicated that without a healthy dopamine delivery system, these models struggled to utilize their storage space as effectively.
“These findings highlight how computational brain science can advance our understanding of psychiatric disorders,” Frank states. “For example, while many consider Parkinson’s a movement disorder due to its visible symptoms, it also affects working memory. Current treatments primarily target the prefrontal cortex. Our research suggests exploring drugs that could influence the basal ganglia and thalamus for symptom improvement.”
This study was funded by various institutions including the Department of Defense (ONR MURI Award N00014-23-1-2792) and the National Institute of Mental Health, among others. The computing hardware utilized in these experiments received support from the National Institutes of Health.
