01// Definition Foundation

// ##def-tulving · Tulving's taxonomy (1972) =Semantic memory — long-term store of general knowledge: facts, concepts, meanings, relationships##sm-def .context-free — "Paris is the capital of France" has no personal episode attached .content-addressable — retrieved by meaning, not location or time .acquired by repetition and meaningful encoding — not single exposures @Tulving, E. — "Episodic and Semantic Memory" (1972)

Episodic =personal events =time + place anchored ="I was there"

Semantic ← here =world knowledge =meaning-addressable ="it is true that"

Spatial =cognitive maps =location-addressable ="it was over there"

// ##def-distinct · What makes semantic distinct .survives amnesia that destroys episodic memory — H.M. case .degrades before spatial in Alzheimer's — spatial is most durable .not tied to a body in space — purely conceptual, abstract →the "weakest" of the three long-term systems — but the most flexible

---

02// Architecture Structure

// ##arch-network · Associative network model =semantic memory is a network — nodes (concepts) connected by weighted edges (relationships)##network @Collins & Quillian (1969) — hierarchical network model @Collins & Loftus (1975) — spreading activation model (non-hierarchical) .edge weight = semantic distance — "car" closer to "truck" than to "justice" .retrieval = traversal — finding a concept means activating a path through the network // ##arch-schema · Schemas =Schema — structured cluster of related concepts with default values and slot fillers##schema ="restaurant" schema: menu · waiter · table · bill · tipping =new information is assimilated into existing schemas — or triggers schema update @Bartlett (1932) — "Remembering" — schemas distort recall toward expectations →schemas explain why CLN block titles matter — they activate the right schema on first read →"Risk Analysis" activates risk schema; "Section 3" activates nothing

---

03// Spreading Activation Core Mechanism

// ##act-how · The mechanism .activating one node spreads excitation to connected nodes — weighted by edge strength .activation decays with distance — far nodes receive subcritical activation .priming — activating "nurse" makes "doctor" faster to retrieve (Meyer & Schvaneveldt 1971)

risk

→

danger

→

warning

risk

↘

mitigation

→

insurance

// ##act-cln · How CLN exploits spreading activation →! operator label activates "risk" node — primes danger, mitigation, consequence →= operator label activates "definition" node — primes precision, boundary, scope →? operator label activates "unknown" node — primes inquiry, curiosity, gap .pre-activation happens before content is read — semantic context is set in advance .content lands in an already-primed semantic neighborhood — encoding is stronger @Craik & Lockhart (1972) — levels of processing: deeper semantic encoding → better retention

---

04// Interference & Decay Fragility

// ##int-types · Two interference types =Proactive interference — old knowledge overwrites new##pi =learned "risk = financial" → new "risk = technical" is harder to encode =Retroactive interference — new knowledge corrupts old##ri =learning a new phone number makes old one harder to recall !interference is proportional to semantic similarity — close neighbors compete hardest →this is why CLN operator semantics must be consistent — inconsistent use creates interference // ##int-prose · Why prose is maximally susceptible !prose encodes by meaning alone — no positional or color anchor !two similar paragraphs about risks compete in the same semantic neighborhood !no discriminating cue — retrieval must rely on content alone →CLN reduces interference by adding spatial and color anchors to semantic encoding →"the red block in §03 about DB risk" — three independent retrieval cues, not one // ##int-decay · Decay without retrieval ~semantic memory decays when not retrieved — paths weaken without traversal ~retrieval itself strengthens the path — the act of finding = the act of reinforcing →CLN ? markers flag content for review — explicit retrieval practice built into the document @see → CLN to Anki — operator-to-card-type mapping for spaced retrieval

---

05// CLN & Semantic Memory CLN

// ##cln-helps · What CLN does for semantic encoding +Schema activation — meaningful block titles activate correct schema before reading ="// 03 · Risk Analysis" vs "// 03" — the former pre-primes the network +Operator priming — each operator activates a semantic category before content arrives =reader is in "definition mode" when seeing = — content encoded more precisely +Interference reduction — position + color anchors differentiate similar content =two risk items distinguished by §02 vs §05 position — separate retrieval paths +Retrieval cue richness — semantic + spatial + visual = three independent paths to same content // ##cln-hurts · Where CLN can hurt semantic memory -operator inconsistency — = used for actions breaks the priming effect →inconsistent operator = noise in the semantic network, not signal -weak block titles — "// 04 · Notes" activates no useful schema →generic titles waste the activation slot — schema priming fails -over-structuring — too many blocks fragment content below schema granularity →schemas need a critical mass of content to activate meaningfully

---

06// vs Spatial Memory Comparison

// ##vs-core-tension · The fundamental trade-off ~spatial: robust, fast, ancient — but requires stable layout; breaks when structure changes ~semantic: flexible, reorganisable — but fragile, interference-prone, degrades faster →neither is "better" — optimal documents encode via both systems simultaneously // ##vs-dual · Dual-coding in CLN .same CLN line encodes two ways simultaneously: =spatially — "red operator, column 0, §03, middle of page" =semantically — "! means risk · 'DB connection' is the content" →retrieval can start from either path — spatial cue leads to semantic, semantic leads to spatial →Paivio (1971) — dual-coding theory: two encodings are more retrievable than one @see → Spatial Memory §04 — full comparison table // ##vs-open · Open questions ?which retrieval path dominates for CLN — spatial or semantic — in practice? ?does operator training (learning CLN vocabulary) create a new semantic schema for operators? ?can CLN operator consistency measurably reduce proactive interference in recall experiments? @→ Spatial Memory — the complementary system @→ Pre-Attentive Processing — how operators reach semantic memory @→ Block Annotations — episodic boundary effects