There is a specific and slightly maddening experience that most readers know. You can picture the book. Right-hand page, about two thirds down, blue cover, read on a train. What you cannot produce is the actual point.
That is not a failure of one memory. It is two different memory systems, one of which worked.
Tulving's distinction
In 1972 Endel Tulving proposed that long-term memory is not a single store but at least two, distinguished by what they hold and how it feels to use them.
Episodic memory holds events, located in time and place. It is autobiographical and it has a first-person quality: using it feels like re-experiencing. Where you were, what the weather was, who said the thing.
Semantic memory holds facts, concepts and meanings, stripped of the occasion on which you learned them. You know that Lisbon is in Portugal without any sense of when that entered your head. Using it feels like knowing, not remembering.
Tulving's shorthand was that episodic memory involves remembering and semantic memory involves knowing. The dissociation is real enough to appear in brain injury: patients exist who cannot form new episodic memories at all while still acquiring new facts, which is difficult to explain if the two are one system.
Why this bites note-takers specifically
The reading experience above is the two systems separating in real time. The context, which is episodic, encoded effortlessly and without your involvement. The content, which needs to become semantic, did not, because becoming semantic requires work you did not do.
Facts start life episodically. You encounter something once, in a place, at a time, and the trace is bound up with that occasion. Repeated encounters across different occasions gradually strip the context away and leave the fact standing on its own, at which point it is semantic and durable. A fact met once, in one context, stays episodic and decays on the forgetting curve with everything else that happened that afternoon.
This produces two familiar and opposite failures.
Context without content. You know exactly where you read it. That is an episodic trace doing its job on material that never made the transition.
Content without context. You are certain of a claim and cannot say where it came from. That is semantic memory working normally, and it is also how confident misattribution happens, which is a good argument for the citation discipline in literature notes.
The practical consequences
One pass through a book cannot produce semantic knowledge. It can only produce an episode. Re-encountering the same idea across different sittings is the mechanism that converts it, which is the deeper reason spaced repetition works and why massed rereading does not.
Notes written in the language of the source stay episodic. Copied phrasing is bound to the occasion of copying. Rewriting an idea in your own words forces you to extract the meaning from the packaging, which is the transition to semantic happening under your hands. This is the whole substance of why writing helps memory, and it is why progressive summarization puts the value in the summarising rather than the highlighting.
Retrieval cues are episodic hooks. The reason a smell or a song can restore a forgotten afternoon is that episodic traces are richly bound to their circumstances. You can exploit this deliberately: a note that records where the idea came from and what you were trying to solve at the time gives you two routes back to it instead of one. See context-dependent memory.
A note system is mostly a semantic prosthesis. It stores meanings, searchably, without the occasion. That is exactly what external memory is good at, and it explains a common disappointment: people expect their archive to restore the feeling of having understood something, and it never does, because the part they wanted was episodic and was never written down.
How to write for the transition
The useful test on any note is whether it would still make sense to a stranger who was not in the room.
- Write claims, not gestures. "Interesting point about attention" is episodic shorthand and will be worthless in a month. The claim itself, stated fully, is semantic and survives.
- Keep the source, and keep it separate. The idea in one place, the citation attached. See how to take notes from a book.
- Record the question you were holding. One line about why you cared. This costs nothing and gives future-you the episodic thread back into your own reasoning, which is the part that never survives on its own.
- Meet it again, elsewhere. Link the note to a different note in a different context. Every fresh context is another pass at stripping the original one away. This is what linking your thinking is doing underneath the vocabulary.
The honest limit
The two-system model is a simplification and has been argued about for fifty years. The boundary is porous, semantic knowledge probably scaffolds episodic recall rather than sitting beside it, and Tulving himself revised the account substantially in later work.
None of that disturbs the practical point, which is robust: knowing where you learned something is not knowing it, those states feel more similar than they are, and only one of them survives a year.
More in how memory works, active recall, and the wider memory and focus section.