How Your Brain Remembers Sequences: The Role of Synaptic Augmentation

From remembering a phone number just long enough to dial it, to keeping track of the steps in a recipe, we navigate sequences of items and events with apparent ease. We instinctively know what comes next, yet the brain’s ability to hold a few items in mind while simultaneously tracking their order and timing is a complex process. How does our “working memory” accomplish this feat?

The Limits of Traditional Working Memory Models

Traditional models of working memory have focused on persistent neural activity or learning through repetition. While these models explain what information is stored, they struggle to explain how we recall the order of items or the timing of events, particularly in novel sequences. Research demonstrates that people can immediately recall new sequences, including their order and timing, without practice, suggesting working memory stores temporal information in real-time. This raises the question: where is this information stored in the brain?

Synaptic Augmentation: A New Perspective

Now, researchers propose that this temporal information is encoded directly in the dynamics of synapses – the connections between neurons. Instead of storing time separately, short-lived changes in synaptic signaling may embed a sense of time into memory itself. Specifically, a process called synaptic augmentation is key.

How Synaptic Augmentation Works

Synaptic augmentation is a type of short-term plasticity that develops slowly and persists for tens of seconds. Most neural signaling relies on classical facilitation and depression, which occur much faster – between milliseconds and a few seconds. When a neuron fires repeatedly during a sequence, its outgoing synapses not only transmit the signal but also gradually increase in strength, retaining a record of how recently and how often they were active. This creates a time-dependent gradient across synapses, effectively encoding when each item in a sequence occurred relative to others.

Imagine a sequence of memory items represented by different neural populations. As each item is presented, synaptic strengths in the corresponding neurons gradually increase. Because synaptic augmentation builds and decays slowly, the resulting pattern of synaptic strength retains a temporal fingerprint of the sequence – earlier items show stronger augmentation than later ones. This gradient allows working memory to replay the sequence in the correct order and with approximate timing, at real speed or in a time-compressed form.

Dynamic Memory and Neural Activity

This approach aligns with observations that activity during working memory is dynamic rather than static. Memory representations fluctuate in strength over time, can be briefly reactivated, and may be replayed during sleep or rest, sometimes in a time-compressed form that supports learning and memory consolidation. Synaptic augmentation preserves a gradient that naturally encodes order and elapsed time, preventing the system from falling into a steady state where all past items are maintained equally, thus losing any temporal distinction.

Testable Predictions and Future Research

While currently conceptual, this model offers testable predictions. The slow buildup and decay of synaptic augmentation aligns with behavioral observations showing that working memory can maintain temporal intervals of several seconds without active rehearsal. The theory predicts that selectively disrupting augmentation mechanisms should impair memory for the order and timing of items, while preserving memory for the items themselves.

the model provides a framework for interpreting neural activity patterns observed in electrophysiological studies, accounting for ramping activity during memory delays and “activity-silent” states where latent information can be reactivated by cues. By linking these dynamic patterns to underlying synaptic changes, the model bridges the gap between observable neural activity and the synaptic processes that underpin working memory.

Time as an Emergent Property of Memory

This work reframes our understanding of working memory: time is not an add-on to memory but an emergent property of how synapses change in real-time. The brain doesn’t need a separate “clock” to timestamp experiences; instead, it uses its plasticity mechanisms to let time leave an imprint on memory. Models like this will continue to shape our understanding of cognition – not as static repositories of information, but as active, changing processes shaped by brain rhythms and slow changes in the connections between cells.