Timedora

3 Storing and Consolidating Memories

Learn how working memory and distributed neural networks support remembering, and how consolidation stabilizes and reorganizes new memories over time.

Memory depends on neural networks

Memories do not reside in one single storage area of the brain. They depend on changing patterns of activity and connections across neural networks. keeps information available for immediate use, whereas supports retention over longer periods. Consolidation helps stabilize new learning and reorganize it over time.

keeps information available

temporarily maintains and uses information relevant to a current task. For example, holding a phone number in mind long enough to dial it uses . It also supports manipulating information, such as mentally adding two numbers or comparing two ideas.

is not simply a small, separate box in the brain. It depends on coordinated activity in networks that represent information, together with control processes that direct attention and protect relevant information from distraction. Depending on the task, sensory, language, memory, and frontal brain areas can all contribute.

Researchers continue to investigate how information is maintained. Possible mechanisms include sustained neural activity, temporary changes in synaptic connections, or a combination of mechanisms. is limited; silently rehearsing information, such as a name, can help keep it available briefly. Connecting information to prior knowledge or giving it meaning can also support learning beyond the immediate moment.

Long-term memories are distributed

Long-term memories are supported by changes in the strength and organization of connections among neurons. Learning a fact or event can involve : activity changes how effectively neurons communicate, helping establish or modify a memory representation.

Declarative memories include memories for facts, called , and personally experienced events, called . Memory representations are distributed across networks rather than kept as complete records in one location. Remembering a familiar song, for example, can draw on networks for sound, words, emotion, and knowledge about when or where it was heard.

The and related medial temporal lobe structures are especially important for forming many new memories and linking the elements of an experience. Other brain regions contribute according to the information being learned and remembered.

Consolidation stabilizes and reorganizes memories

is the set of processes through which newly formed, fragile memories become more stable. Two related timescales are useful to distinguish:

  • occurs over minutes to hours after learning. Cellular and molecular changes help stabilize altered connections.

  • unfolds over a longer period. Interactions between the and neocortex help reorganize the networks that support a memory.

is not a literal transfer of a finished memory from the to the cortex. Information is represented in cortical networks from the time of learning, while the helps link and reactivate relevant patterns. With time, cortical connections can become better able to support aspects of a memory. The degree of hippocampal involvement may vary with the memory and what is being recalled; detailed episodic memories can remain strongly dependent on hippocampal processes.

Rest, sleep, and the limits of consolidation

Periods of rest and sleep can support memory processing. Research has observed replay-like patterns of neural activity after learning, including during sleep; these patterns are thought to contribute to communication between the and cortex.

Consolidation does not guarantee that every detail will be preserved. Memories can be incomplete or altered, and later learning can influence them.

Example: remembering a route

After learning a new route, you may initially need to concentrate on each turn. Recalling or mentally rehearsing the route can keep it available in and strengthen learning. Over time, linked representations across brain networks can make the route easier to remember, while its detailed context may not be retained equally well.