The human brain contains roughly 86 billion neurons, each connected to thousands of others through trillions of synapses, all coordinating to produce thought, perception, language, emotion, motor control — and somehow, also, memory. The fact that experiences from years ago can be reliably called back into the same brain that's currently processing this sentence is one of the deeper puzzles in biology.
Memory isn't one thing. It's at least four. And the way each kind works — what brain regions handle it, how it's encoded, how it gets stored and retrieved — has been one of the major projects of neuroscience over the last half-century.
Memory Isn't One Thing
The first useful distinction is between memory you can talk about (declarative or explicit memory) and memory you can't (procedural or implicit memory).
Declarative memory is what you usually mean by "memory" in everyday language — facts you know, events you remember, names of people. It splits further into:
- Episodic memory — memories of specific events, anchored in time and place. The first time you rode a bicycle. What you had for dinner last Tuesday.
- Semantic memory — general factual knowledge that's not anchored to a specific moment. The capital of France. That tomatoes are fruits. How many continents there are.
Procedural memory is everything you "know how to do" without being able to explain it. Riding a bike, typing on a keyboard, swimming, the way you reach for a doorknob without thinking. It is stored differently and in different brain regions from declarative memory, and it follows different rules.
You can lose one kind without losing the other. There are documented patients who have lost the ability to form new episodic memories but can still learn new procedural skills (like a new physical activity) — and crucially, they have no memory of having learned the skill, even as they demonstrate it.
The Three Stages
For any new piece of information to become a memory, it has to go through three stages:
- Encoding — the brain has to translate the experience into a form it can store. This involves attention (you can't remember what you didn't notice in the first place) and integration with what you already know.
- Storage — the encoded information has to be physically maintained somewhere in the brain. This appears to involve long-term changes in the strength of connections between specific neurons.
- Retrieval — when you need the memory later, the brain has to find and reactivate the relevant pattern. Retrieval is more reconstructive than reproductive — every time you remember something, you partially rebuild it, which is part of why memory is so often unreliable.
Memory failures can happen at any of these three stages. "I forgot to remember it" is often actually "I never encoded it in the first place because I wasn't paying attention."
The Hippocampus and Why It Matters
Most of what we know about memory localization comes from studying people whose brains have been damaged. The most famous patient in the history of neuroscience was a man known by his initials H.M., who had large portions of his hippocampus — a small, seahorse-shaped structure deep in each temporal lobe — removed in 1953 to treat severe epilepsy.
After the surgery, H.M. could no longer form new declarative memories. He could remember things from before the surgery. He could maintain a conversation by holding information in working memory. He could learn new motor skills (his procedural memory was intact). But anything he encoded as a new fact or event was gone within minutes. He famously had to be re-introduced to his doctors every day for decades.
H.M.'s case established that the hippocampus is essential for forming new declarative memories — but that once memories have been "consolidated" into long-term storage elsewhere in the cortex, they no longer need the hippocampus. That's why his older memories were preserved.
Modern neuroscience has filled in much more detail, but the basic story holds: the hippocampus is the brain's writing head, transferring experience into longer-term storage spread across the cortex.
How a Memory Is Physically Stored
At the cellular level, the leading model is long-term potentiation (LTP) — a long-lasting increase in the strength of the connection between two neurons that frequently fire together. Neurons that repeatedly activate each other strengthen the synapse between them. Over time, the pattern of strengthened connections becomes a kind of physical trace — an "engram" — of the original experience.
This is the cellular basis of the slogan "cells that fire together, wire together." Repeated co-activation strengthens the link. Eventually, activating one part of the pattern reactivates the others, which is what we experience as remembering.
The molecular details involve changes in receptors at synapses, modifications to the proteins that build them, and ultimately changes in gene expression in the neurons themselves. The neurons are physically restructuring themselves to record the memory. This is one of the most active areas of cellular neuroscience right now.
For the broader picture of how the brain fits into the rest of the body's coordinated systems, see How the Human Body Works: An Overview of Major Systems.
Why Memory Fails
Memory is unreliable in predictable ways:
- Information you never encoded. Most "forgotten" things were never properly encoded in the first place. Sustained attention is the price of admission for memory.
- Decay. Memory traces appear to weaken if not periodically reactivated. The classic forgetting curve, first measured by Ebbinghaus in the 1880s, shows that without rehearsal, most newly learned material is lost within days.
- Interference. Similar memories can crowd each other out. The Tuesday morning meeting and the Wednesday morning meeting blur into "a morning meeting."
- Reconstruction errors. Because retrieval is reconstructive, memories drift over time. Each retrieval is also a chance to alter the memory by encoding the current reconstruction back into storage. This is one reason eyewitness testimony, despite feeling vivid and certain, is notoriously unreliable.
The good news is that these failure modes are predictable enough that strategies for improving memory mostly work — spaced repetition, active recall, sleep (during which substantial memory consolidation appears to occur), and connecting new information to what you already know.
The Strange Status of Memory
Memory is biological — it has to be, because the brain is biological — but it is one of the points where biology runs hardest into the limits of what we currently understand. The cellular mechanisms of long-term potentiation are well-established. The neural circuits that handle different memory types are increasingly well-mapped. But the question of how a particular pattern of synaptic strengths actually contains the specific memory of the first time you saw the ocean — and not some other memory, or no memory at all — is not yet a question neuroscience can fully answer.
It is, in other words, one of the active frontiers of biology. And for once, the textbooks are honest that we don't quite know yet how it all works.
For how the broader endocrine and nervous systems coordinate, see How the Human Body Works. For the genetic story underneath every neuron's behavior, see Top 10 Fascinating Facts About Human DNA.