The model, described in a paper published in the journal *Psychological Review*, draws inspiration from theoretical neuroscience, psychology, and artificial intelligence. It offers a comprehensive framework that not only simulates how the brain processes sequential memory but also explains how memory errors can occur.
"It's the first model that brings together memory distortions and the core processes that support sequential memory," says Mehrdad Jazayeri, an associate professor in MIT's Department of Brain and Cognitive Sciences and the Picower Institute for Learning and Memory. "It may provide new insights into understanding cognitive deficits in psychiatric disorders in which sequential memory is known to be impaired."
Sequential memory is critical for everyday activities such as remembering a list of items in the grocery store or the steps in a recipe. It is closely tied to our sense of time and our ability to navigate through the world.
The new model captures two core processes thought to support sequential memory: the ability to predict upcoming items in a sequence, and the ability to recall the order of items in a sequence.
To achieve these processes, the model's "brain cells" — mathematical units representing neurons — have several layers of interconnectedness, reflecting the hierarchical architecture of brain regions involved in memory. The model also includes mechanisms for representing time, learning from past experiences, and making errors.
The researchers tested the model by comparing its performance to human memory data on different sequential memory tasks. The results showed that the model accurately captured human performance, not only in terms of correct responses but also in the types of errors that people made.
"It has long been thought that sequential memory involves prediction, but our model is one of the first quantitative frameworks that explicitly shows how prediction contributes to memory distortions and errors," says lead author Mohammad Amin-Nouri, a research scientist in the Department of Brain and Cognitive Sciences.
For instance, the model predicts that people are more likely to mistake an item for an adjacent item in the sequence due to strong predictive signals between nearby items. It also explains how our expectations about the order of items can influence our memory for the sequence.
The researchers believe the new model could potentially be used to develop tools for assessing memory deficits in psychiatric disorders and for guiding the development of cognitive rehabilitation interventions.
"We could use this model to pinpoint specific memory processes that are impaired in different psychiatric disorders," says Amin-Nouri. "This could help us understand the cognitive roots of these disorders and design targeted treatments to improve memory functioning."
The researchers plan to further refine the model by incorporating additional cognitive processes, such as attention, and by comparing it to brain imaging data to gain a deeper understanding of the neural mechanisms of sequential memory.
