What Is Long-Term Memory? Definition, Types & Capacity
Nobody has ever run out of room. So when something goes missing, the trouble is almost always somewhere other than the shelf.
In the 1920s a Moscow newspaper sent a young reporter to a neuropsychologist, for an unusual reason: he never took notes, and his editor found it unnerving. He didn't need them. He could hear a briefing once and give it back whole.
Alexander Luria studied that man, Solomon Shereshevsky, on and off for roughly thirty years, and by his account never established a clear limit on what he could retain — long strings of digits, nonsense syllables, material he was given once and asked about years later without warning.
The obvious conclusion is that Shereshevsky had more storage than the rest of us. Luria's account points somewhere less convenient. He described a rare and extreme form of synesthesia — inputs arriving wrapped in colour, texture and taste — alongside elaborate imagery and deliberate mnemonic strategies that Shereshevsky built for himself. Later historical work has argued his memory was neither effortless nor flawless, so the case can't settle exactly how much came from encoding and how much from technique. What it does illustrate is that extraordinary retention travelled with extraordinary encoding, not with evidence of a larger tank — and it came at a price, since he struggled badly with abstraction and metaphor and reportedly tried, without much success, to forget things on purpose.
His case is an extreme version of a limit everyone lives inside. The room was never the constraint.
The Bottleneck Isn't Space
Long-term memory is the system that holds information from minutes to a lifetime, and it has no known practical capacity ceiling. Nobody has learned so much that they became unable to learn more. A lifetime of faces, routes, language, facts and skills accumulates without anyone hitting a wall.
Which relocates the question. When something goes missing, the bottleneck is usually encoding, consolidation, interference or retrieval — not a tank that filled up. Traces can fail to form richly enough, fail to stabilise, get crowded out by similar material, or sit intact but unfindable.
Shereshevsky was exceptional at the first of those. Ordinary forgetting can begin at any point: weak encoding, failed consolidation, interference or retrieval trouble — and as we'll see, a startling amount of it isn't loss at all.
One House, Four Rooms
Before the pipeline, the layout — because "long-term memory" isn't one archive but several systems that fail independently of each other.
Two rooms belong to explicit memory, the knowledge you can consciously bring to mind. The photo album is episodic memory: events you personally lived, each tagged with a where and a when. The encyclopedia is semantic memory: facts and meanings with the learning occasion stripped away. You know a year has 365 days; you have no idea when you learned it.
The other two belong to implicit memory, which shows up in performance and altered responses rather than in deliberate recall. The workshop is procedural memory — skills and habits, running automatically. A smaller reflex room holds priming and conditioning: responses shaped by past exposure you may have no memory of at all.
These rooms are genuinely separable, not just conveniently labelled. Brain injury can lock one door while the others stay open — which is why "how's your memory?" has never had a single answer, and why the Memory & Recall section tests them separately rather than producing one score. The wider brain tests collection covers the abilities that draw on them.
How Something Gets In and Stays
Three stages, and the first does most of the work.
Encoding converts an experience into a storable trace, and the depth of that conversion matters enormously. Material processed for meaning — what does this connect to, what does it resemble, why would it be true — is retained far better than material processed only at the surface. That difference is the entire gap between Shereshevsky's involuntary multisensory encoding and everybody else's effortful version of it, and the encoding guide covers the mechanics.
Consolidation stabilises what was just encoded. New declarative memories stay relatively fragile while hippocampal–cortical consolidation develops over hours and days, and sleep can support that process. The other rooms don't work the same way — procedural and other nondeclarative memories consolidate through partly different mechanisms, which is one more reason the house metaphor holds up.
Retrieval brings it back. Not as a readout, though. Each retrieval reassembles the memory from fragments, which is why remembering can quietly rewrite what's remembered — and why testing yourself strengthens material more than rereading it does.
The Yearbook Study
If encoding and retrieval are the real limits, then a memory could be perfectly intact and still fail — depending entirely on how you're asked for it. That prediction has been tested about as thoroughly as anything in the field.
In 1975, Harry Bahrick, Phyllis Bahrick and Roy Wittlinger tracked 392 people's memory for high school classmates, across intervals stretching from two weeks to more than fifty years.
Recognition barely moved. Shown a face or a name, people identified classmates at around 90% accuracy for roughly fifteen years after graduation, and remained well above chance for decades after that.
Free recall — generating the names cold, unprompted — fell away far more steeply over the same span.
Same people. Same memories. Two completely different answers, depending on whether the question came with a prompt attached. That gap is the practical face of everything above: durable storage, unreliable search. It's why "I can't remember his name" so often resolves the instant somebody else says it. The Verbal Memory Test samples the recognition side of that contrast, asking whether each word has appeared before.
When a Room Goes Dark
The separateness of the rooms becomes vivid, and grim, under injury.
Severe hippocampal damage can profoundly impair the ability to add anything new to the photo album, while the workshop keeps running and some skill learning carries on — though how completely one is lost and how well the other is spared varies considerably from case to case.
The rooms don't merely look separate on a diagram. They fail separately, on their own schedules.
What This Actually Means for You
Every reliable study technique is one of the three stages, done on purpose. Space your practice out instead of massing it. Test yourself rather than rereading, because retrieval strengthens what rereading only makes familiar. Sleep on anything that matters. Connect new material to something you already know, since meaning is what encoding runs on.
None of that is a trick, and none of it will make you Shereshevsky — whose gift, it's worth remembering, came bundled with difficulty thinking abstractly and a mind he couldn't quiet. Cognitive Train's free brain training exercises work the systems that feed this archive, starting with the Short Term Memory Test for the brief holding systems that support much deliberate learning.
But the reframe is worth keeping, because it changes what you do about forgetting. The space was never the problem. What you put into it, and how you go looking for it later, is where all the leverage lives.
From here, follow any room into its own guide — explicit vs. implicit memory for the top-level split, or the full architecture in the types of memory guide.