Flexible working memory is enabled by “spatial computing”

Routine tasks require working memory to remember rules and content for each instance.


Follow us onFollow Tech Explorist on Google News

Working memory (WM) is a mental whiteboard for top-down management and short-term information storage. The central function of WM in cognition depends on this control. From WM, we can choose what to keep, read out, or delete and change the contents. However, what neural processes underlie such adaptable control is still being determined.

Routine tasks that require working memory, such as baking, require remembering general rules and specific content for each instance. A new study suggests a novel explanation for how the brain manages such cognitive demands’ general and specific components.

According to a new study conducted by scientists at MIT’s Picower Institute for Learning and Memory, the Karolinska Institute, and the KTH Royal Institute of Technology in Stockholm, Sweden, the brain creates distinct spaces in the cortex for each general rule and controls those patches with brain rhythms.

This system explains how the brain can consistently understand a process even when the details change. It also answers a few questions that neuroscientists have had about the physiological operations underpinning working memory. 

Earl K. Miller, Picower Professor in MIT’s Department of Brain and Cognitive Sciences, a member of the Picower Institute for Learning and Memory, and co-senior author of the study in Nature Communications, said, “Your brain can instantly generalize. If I teach you to follow some rules, like remembering C, A, and B and putting them into alphabetical order, and then I switch the contents to F, D, and E, you wouldn’t miss a beat. Your brain can do this because it represents the rules and the contents at different physical scales. One can just be plugged into the other.”

Working memory tasks, he explains, are governed by an interplay of brain rhythms at different frequencies. Slower beta waves carry information about task rules and selectively yield faster gamma waves when performing operations like storing information from the senses or reading it out when the recall is required. These waves operate on networks of millions of neurons, only a small number of which store the individual items of information relevant at any given time. 

Neurons that carry information about specific items can be found everywhere. Some become more electrically excited than others in response to different task rules.

How, then, can these somewhat imprecise rhythms selectively control the proper neurons to do the right things at the correct times? Why are redundant and dispersed neurons whose spiking is related to particular things? What causes “350 degrees” to specifically activate one neuron when it needs to be stored but not another with the same information when it needs to be recalled?

According to the theory of spatial computing, even though individual neurons that represent information items may be dispersed throughout the cortex, the rules that apply to them depend on the region of the network they are in. The pattern of beta and gamma waves determines those patches.

Lundqvist said, “By analyzing a lot of single neurons throughout the years, we had always wondered why so many of them appeared to behave similarly. Regardless of whether they preferred the same external stimulus, many neurons shared similar activity patterns during working memory. And these patterns switched from task to task. It also appeared that neurons closer together within the prefrontal cortex shared the same pattern more often. It started us thinking that memory representations might dynamically flow around in the prefrontal cortex to implement task rules.”

This allows the brain waves controlling the patches those neurons live in to selectively associate individual neurons encoding particular pieces of information with general rules. For example, when a friend calls you at the gym, you must turn the padlock dials to the combination numbers. 

According to spatial computing, the brain makes distinct patches for each step when you hear the combination asking you to retrieve a watch someone unintentionally left in their locker. Gamma waves applied when the rule is relevant cause the neurons within each patch that represent the combined number of that particular step to become especially excited.

This suggests that memory representations dynamically flow around the prefrontal cortex to implement task rules. The researchers made four experimental predictions about what they should see as animals played working memory games like remembering a sequence of images. 

Co-senior author Pawel Herman of KTH said, “This way, memory representations could be dynamically reshaped to fit current task demands independent of how individual neurons are connected or which stimulus they prefer. It may explain our impressive generalization capabilities in novel situations.”

This doesn’t mean that a patch is impermanently fixed. Wherever the brain decides to form them for the task at hand, the patches can appear and disappear for however long is necessary. The brain does not contain a permanent “remember oven temperature” patch. 

Miller said, “This gives the brain flexibility. Cognition is all about flexibility.” 

The researchers predicted three things about working memory: The first prediction was that there should be separate neural signals for rule and individual item information. The team discovered that bursts of waves carried rule information. On the other hand, individual neural spikes carried a mix of individual items and task rules, consistent with them representing individual items and being subject to specific regulations.

The second prediction was that rule information would be spatially organized, that rule-enforcing spatial patterns would be consistent, and that brain wave activity would cause neural spiking activity to represent the correct information at the correct times.

The final prediction was that brain wave activity would cause neural spiking activity to represent the correct information at the correct times. This was also reflected in the experimental results. The researchers observed distinct brain wave patterns when the brain needed to store images in memory and when it needed to recall the “correct” one. During recall, beta waves were reduced, and neurons spiked more and in a wider area than during storage.

They discovered that different rules had different locations for gamma bursts and that these remained stable even when the individual items changed during each game. Also, the researchers observed distinct brain wave patterns when the brain had to store images in memory and when it had to recall the “correct” one. 

The paper does not address every question about working memory, such as how neurons encoding specific information in one patch may be associated with their counterparts in another or how the brain controls the patches. More research is required to find answers to these questions.

The Picower Institute, the JPB Foundation, the Swedish Research Council, and the European Research Council all provided funding for the study.

Journal Reference:

  1. Lundqvist, M., Brincat, etal. Working memory control dynamics follow principles of spatial computing. Nature Communications. DOI: 10.1038/s41467-023-36555-4


See stories of the future in your inbox each morning.