CLN Wiki · Memory & Cognition
Episodic Memory
= time · place · self @ Tulving 1972 · Zacks 2007 ~ 6 sections
01// Definition Foundation
// ##ep-def · Tulving's definition =Episodic memory — autobiographical record of personally experienced events, bound to time and place##ep-def .triple encoding: what happened + where + when .always includes a self-referential component — "I was there" .phenomenological quality — remembering feels different from knowing @Tulving, E. — "Episodic and Semantic Memory" (1972); "Episodic Memory: From Mind to Brain" (2002) // ##ep-vs · Episodic vs Semantic — the key distinction .semantic: "Paris is the capital of France" — decontextualized, timeless .episodic: "I learned that in school, sitting next to Marcus, raining outside" — contextualized →semantic memory is what you know — episodic is what you remember experiencing →episodic feeds semantic — repeated episodes abstract into context-free knowledge @see → Semantic Memory — Tulving's full taxonomy // ##ep-fragile · Why episodic is the most fragile !first to go in Alzheimer's — personal history lost before world knowledge !highly reconstructive — not a recording, a re-assembly from fragments @Loftus (1979) — eyewitness memory is episodic and highly susceptible to post-event information !temporal context decays fastest — "when" is the first coordinate lost →yet: unique events with high emotional salience can survive decades — §04 → Landmarks
02// Mental Time Travel Core Concept
// ##mtt-def · Tulving's "mental time travel" =Mental time travel — capacity to re-experience past events and pre-experience future ones##mtt .unique to humans — no convincing animal evidence of genuine episodic recall .same neural system: recalling the past and imagining the future use overlapping networks @Schacter & Addis (2007) — "constructive episodic simulation hypothesis" // ##mtt-implication · Why MTT matters for knowledge work →when you re-read a document, you're not just reading — you're re-experiencing writing it →"I remember being confused here" — episodic tag on a specific block →"I wrote this right after the incident with X" — episodic context aids interpretation .the episodic trace is not about the content — it's about the experience of engaging with it .richer experience at writing = richer trace at re-reading →CLN authorship is more episodically rich than pasting AI output — see §05 // ##mtt-ai · MTT and AI-generated content ~AI-generated text has no episodic trace — it was not experienced in the author's timeline ~copy-pasting AI output = zero episodic encoding — semantic only, maximally fragile →editing AI output creates episodic trace — the experience of deciding what to keep →choosing operators for AI content is an episodic act — "I decided this was a risk"
03// Event Segmentation Mechanism
// ##seg-theory · Event Segmentation Theory (EST) =the brain continuously parses experience into discrete events — bounded units with stable structure##est .event boundaries — moments of change: new actor, new location, new goal .at boundaries: working memory is updated, predictions are reset, encoding spikes @Zacks, J.M. et al. — "Event Structure in Perception and Conception" (2007)
Event A
stable context
low encoding
⚡ Boundary
encoding spike
new landmark
Event B
stable context
low encoding
⚡ Boundary
encoding spike
new landmark
Event C
stable context
low encoding
// ##seg-memory · Why boundaries improve memory .working memory updates at boundaries — previous event compressed into long-term store .boundary itself becomes a retrieval cue — "the thing after the shift to the new topic" .more boundaries in a reading session → more compression events → better recall structure !continuous prose has no boundaries — one long event, minimal compression landmarks →each CLN block transition is a manufactured event boundary — see §05 // ##seg-doorway · The doorway effect .walking through a doorway triggers an event boundary — context resets .content from before the doorway becomes harder to retrieve immediately after @Radvansky et al. (2011) — "Walking through doorways causes forgetting" →CLN section headers function as cognitive doorways — good for long-term encoding !but: too many rapid boundaries impair working memory continuity — anti-pattern
04// Landmarks & Retrieval Retrieval
// ##lm-def · Episodic landmarks =Landmark — distinctive event that anchors the temporal sequence, enabling forward/backward navigation##landmark .landmarks are recalled faster and more accurately than non-landmark events .retrieval uses landmarks as search anchors: "it was just after the DB crisis block" .landmark salience: unusual + distinctive + personally relevant = strongest anchor // ##lm-what-makes · What makes a CLN block a strong landmark .Operator rarity — a ! block surrounded by = blocks stands out more .Semantic distinctiveness — "// 04 · The Breach" > "// 04 · Issue" .Emotional charge — content that triggered a reaction encodes as landmark automatically .Position extremity — first and last blocks in a section have primacy / recency advantage →a document with 6 distinct blocks = 6 potential landmarks = 6 retrieval anchors →a 1000-word prose doc = ~1 retrieval anchor ("I read something about that")
01
=
Definition · what the system is
landmark ✓
02
.
Architecture · how it works
boundary
03
!
Critical failure · DB connection down
strong landmark !
04
→
Resolution · patched in sprint 3
boundary
05
?
Open · edge cases unresolved
landmark ✓
05// CLN as Episode Maker CLN
// ##cln-manufacture · How CLN manufactures episodic context →writing a CLN block is a cognitive act — choosing operator, title, content →each choice is experienced — "I decided this was a ! not a ~" = episodic moment →block title is authored, not generated — naming something is a memorable act .result: each block carries an episodic tag — "when I was thinking about the DB issue" .re-reading triggers MTT — the original thinking session is re-experienced .this re-experience re-encodes semantically — double benefit // ##cln-advantages · What CLN structure gives episodic memory +manufactured event boundaries — every block transition = encoding spike +ordinal sequence — blocks 01 → 05 provide temporal scaffolding for recall +operator-as-landmark — rare ! in a = document is immediately salient +title distinctiveness — semantically rich titles create stronger episodic anchors +authorship trace — every chosen word is an episodic memory, not borrowed text // ##cln-limits · What CLN cannot guarantee -passive reading still creates weak episodic trace — authoring is required for full benefit -blocks written in identical emotional state blur together — poor landmark differentiation -very long sessions without breaks dilute episodic tagging — context becomes uniform →write in sessions with natural breaks — each session becomes a distinct episodic context @→ CLN to Anki — converting episodic landmarks into spaced retrieval practice
06// The Memory Triad Synthesis
// ##triad-three · How all three systems work in CLN .Episodic — "I wrote block 03 right after the incident, I was frustrated" =retrieval cue: the authoring experience + emotional context .Semantic — "the ! block was about DB risk" =retrieval cue: operator meaning + content category .Spatial — "it was the red block in §03, middle-left of the page" =retrieval cue: position + color + depth →three independent paths to the same content — if one fails, two remain →prose activates semantic only — one path, least robust // ##triad-tension · The tension with AI authorship ~AI writes the content — no episodic trace from authoring ~human assigns operators + titles — partial episodic encoding from curation →the more actively you engage with AI output (edit, rewrite, restructure), the stronger the episodic trace →passive copy-paste = semantic encoding only = weakest possible retention // ##triad-open · Open questions ?does operator-choice act (deciding ! vs ~) create measurably stronger episodic trace? ?optimal session length before episodic context becomes too uniform? ?do CLN block titles create landmark effects equivalent to spatial landmarks in MTT? @→ Semantic Memory — the decontextualized partner system @→ Spatial Memory — the most durable of the three @→ Block Annotations — chunking and boundary effects in depth