Skip to content

First Direct Evidence Detected of Learning-related Memory Replay in Human Brains

Key findings

  • In this study, two men with quadriplegia who were having a brain–computer interface implanted were asked to repeatedly move a computer cursor through a specific sequence of targets, using their thoughts
  • Patterns of neural activity used for controlling the cursor were found to be replayed during rest
  • Replay was stronger for learned than unlearned cursor movement sequences
  • Replay of learned sequences occurred both while participants were awake and while they were napping

Decades of interdisciplinary research suggests that after new memories are encoded, offline mechanisms are involved in their long-term consolidation. A leading hypothesis is that consolidation occurs because the brain ''replays'' the neural firing sequences that occurred during learning.

Non-invasive neuroimaging and intracranial electrodes have provided indirect evidence of replay. However, these methods can't test for replay at the level of individual neurons.

Now, researchers have published direct evidence that firing rate patterns corresponding to previously learned motor sequences are replayed in the human cortex during rest. Leigh Hochberg, MD, PhD, and Sydney S. Cash, MD, PhD, co-directors of the Center for Neurotechnology and Neurorecovery at Massachusetts General Hospital, and colleagues describe the findings in Cell Reports.

Study Design

Mass General is part of BrainGate, an ongoing effort to develop brain–computer interfaces that will allow people with quadriplegia to control computer cursors, robotic arms and other assistive devices with their thoughts. For the research reported here, two men with quadriplegia each had two microelectrode arrays implanted in their motor cortex.

Intermixed with other research sessions, the patients participated in five sessions of a sequence copying game. A specific sequence of four colored targets was presented on a computer screen. After watching, the participant was asked to move a cursor under neural control from the center of the screen to the same sequence of targets as quickly and accurately as possible.

In a given session, the same sequence was repeated 66 times, along with control sequences that did not include any of the same target transitions. As intended, the participants learned the repeated sequences more successfully than the control sequences.

For 20 to 30 minutes before and after the sequence game, neural signals were recorded as the participant was invited to close his eyes, relax and nap if desired.

Testing for Replay

Firing rates of the 40 neural features used to control the cursor were examined during the rest periods. Neural activity was identified as either matching the order of targets in the repeated sequences (hits) or not (misses).

Across sessions, the proportion of hits for repeated sequences was significantly higher in rest period 2, after the sequence game, than in rest period 1. The proportion of control sequence hits did not differ between the two rest periods.

Waking vs. Sleep

By visually examining neural activity patterns, the researchers subdivided the rest breaks into periods of waking and periods of non–rapid eye movement sleep. During both states, replay of the repeated sequences was stronger than replay of control sequences. This is consistent with the proposal that replay during waking stabilizes memories until sleep can occur and establish long-term memories.

Future Directions

The next step in this research will be to test whether replay has a causal role in memory consolidation. One approach would be to test for a relationship between the strength of the replay and the strength of post-nap memory recall.

Learn more about research in the Department of Neurology

Refer a patient to the Department of Neurology

Related topics


For the first time, researchers have direct evidence of learning-related replay in the human brain during sleep. A recent study using implanted multi-electrode arrays in two participants found identical neural firing patterns during game-play awake periods and sleep periods.


In cognitively intact older adults, impaired metamemory—the ability to monitor one's memory accurately and engage in compensatory strategies when necessary—was linked to greater deposition of tau in brain regions that are vulnerable to pathologic changes in early Alzheimer's disease.