NSM (Natural Semantic Metalanguage) — Domain Analysis

Threshold check: is it a domain?

The methodology's domain criteria, applied to NSM:

• Decomposable structure? Yes. NSM defines concepts via paraphrase using only semantic primes. Decomposition is the central operation.
• Irreducible primitives? Yes (claimed). Wierzbicka and Goddard maintain ~65 primes empirically derived through cross-linguistic testing; the discipline is that no prime should be definable in terms of other primes without circularity.
• Instances? Yes. Every human language is an instance. Every concept paraphrased in NSM is a unit of analysis.
• Dependencies between primitives? Lightly developed in NSM literature but present. Some primes presuppose others structurally (e.g., quantifiers presuppose substantives; mental predicates presuppose someone to think/feel).
• Empirical recurrence? Yes — that's the central claim of NSM. Primes have lexical equivalents in every studied language.

Verdict: NSM is a domain. It admits the methodology's analytical apparatus. Proceeding with the 12-step.

Step 1 — Information gathering

NSM is a research program in linguistic semantics initiated by Anna Wierzbicka in 1972. Three claims:

1. The semantic-prime claim: there exists a small set of cross-linguistically universal semantic primes — concepts like SOMEONE, GOOD, BEFORE — that have lexical equivalents in every studied human language.
2. The paraphrase claim: any concept in any language can be paraphrased (defined) using only the primes plus a small grammar (the "natural semantic metalanguage").
3. The irreducibility claim: the primes are mutually irreducible — no prime can be defined in terms of others without circularity.

Methodologically, NSM is empirical: primes are added or removed based on cross-linguistic testing. The inventory grew from ~14 primes (1972) to ~65 today through decades of iterative work, with the Goddard-Wierzbicka team being the primary curators.

The literature is substantial. Semantics: Primes and Universals (Wierzbicka 1996) is the canonical statement. Words and Meanings (Goddard and Wierzbicka 2014) is the most recent comprehensive treatment.

Step 1b — Domain type declaration

NSM is most naturally classified as a surface domain sitting on top of the cognitive substrate. Its primitives describe what cognitive substrate produces at the symbolization (Sy) interface — the discrete semantic units the substrate emits for use in language and thought.

Alternative classification considered: NSM as a substrate-style domain (cross-linguistic invariants suggest substrate-like universality). Rejected because NSM's outputs are semantic content that downstream cognition consumes, which is a surface-level function, not a substrate-level structural commitment. NSM is what the cognitive substrate emits, not the substrate itself.

R11 prediction for surface domain: filter stringency ~25-40%, functional-integration core triad, ~9-12 primitives. The 16 NSM categories sit slightly above this range, which is a hypothesis to test.

Step 1c — Level of description declaration

Level: semantic primes for human meaning, treated as the universal lexical-conceptual atoms of natural language.

Floor (accepted as primitive at this level): the prime concepts. We do not attempt to decompose primes further into cognitive operations or neural substrates; that lives below NSM's level.

Background (unmodeled context): the cognitive substrate (Rp/Ct/Sy/...) that produces semantic representations; the linguistic competence that maps representations to surface forms; the cultural transmission mechanisms that propagate meanings.

Adjacent levels:

• Above: complex concepts (paraphrasable from primes), discourse, narrative structure, technical vocabulary. NSM's claim is that these are recoverable from the primes by paraphrase.
• Below: the cognitive substrate operations that compute semantic representation. NSM treats primes as outputs of these operations without modeling the operations themselves.

This declaration matters because primitive-extraction debates often turn out to be level-of-description disagreements. NSM-skeptics frequently object on cognitive-architecture grounds ("primes can't be primitive, they're computed by neural circuits"). NSM's answer is structural: at NSM's level of description, primes are what survives the iterative empirical reduction. Whether they're computationally primitive at the cognitive substrate level is a different question for a different analysis.

Step 2 — Landscape analysis

Instances: every human language. NSM has been tested against English, Russian, Polish, Japanese, Lao, Yankunytjatjara, Acehnese, Arabic, Mandarin, Spanish, French, German, and dozens of others. The cross-linguistic empirical surface is broader than almost any other primitive-decomposition project in the humanities.

Recurring patterns:

• Every studied language has a lexical equivalent for each NSM prime (the universality claim, repeatedly tested).
• Some primes have fused lexicalizations in some languages (e.g., a single word covering multiple primes), but the prime concepts are distinguishable through usage.
• Concept-paraphrase definitions in NSM English translate naturally into other languages when the paraphrase uses only primes; non-prime words cause translation breakage.

Landscape position of NSM itself: NSM is the most empirically developed primitive-decomposition program in linguistics, comparable in maturity to formal-semantics traditions (Montague grammar, situation semantics) but methodologically distinct (NSM is decomposition-into-primes; formal semantics is composition-via-functions).

Step 3 — Primitive extraction (with 3/3b iteration)

This is the central question. Three candidate primitive resolutions to consider:

Resolution A — The ~65 primes as primitives

The natural reading of NSM as a methodology output: each of the ~65 primes is a primitive. Each has been tested for irreducibility against the others.

Problem: the methodology's substrate-clustering pattern predicts ~6 primitives for substrate domains, ~9-12 for surfaces. 65 primes is far higher than either band. Either NSM is at a different domain type (e.g., compound/vocabulary-style), or the ~65 primes are not at the primitive level but at the partial-level resolution.

Resolution B — The 16 categories as primitives

Wierzbicka organizes the ~65 primes into ~16 categories (substantives, determiners, quantifiers, mental predicates, speech, actions, existence/possession, life/death, time, space, logical, intensifier, taxonomic, evaluators, descriptors, similarity). At this level, NSM might be a surface domain with ~16 primitives (slightly above the 9-12 surface band, but in the right range).

Test: are the categories irreducible? Can any be reduced to combinations of others? Let's apply the three-test criterion:

• Substantives (I, YOU, SOMEONE, SOMETHING, PEOPLE, BODY): irreducible — without substantives, nothing to refer to.
• Determiners (THIS, THE SAME, OTHER): partially reducible? "The same" and "other" presuppose a comparison; THIS is deictic. Could potentially merge with substantives or quantifiers.
• Quantifiers (ONE, TWO, SOME, ALL, MUCH/MANY): irreducible.
• Mental predicates (THINK, KNOW, WANT, FEEL, SEE, HEAR): irreducible at NSM's level.
• Speech (SAY, WORDS, TRUE): partly reducible to mental predicates? SAY ≈ "do words to communicate something thought." Possibly mergeable with mental predicates or actions.
• Actions/events/movement (DO, HAPPEN, MOVE, TOUCH): irreducible.
• Existence/possession (BE, THERE IS, HAVE): irreducible.
• Life/death (LIVE, DIE): possibly reducible to existence + temporal change. LIVE ≈ "to be alive over time"; DIE ≈ "to stop being alive." But "alive" is a state, which requires the life/death primitives to define. Probably irreducible at NSM level.
• Time (WHEN, NOW, BEFORE, AFTER, A LONG TIME, A SHORT TIME, MOMENT): irreducible.
• Space (WHERE, HERE, ABOVE, BELOW, FAR, NEAR, SIDE, INSIDE): irreducible.
• Logical (NOT, MAYBE, CAN, BECAUSE, IF): irreducible.
• Intensifier (VERY, MORE): possibly reducible to comparison + degree. MORE = "comparison with quantity"; VERY = "intensified degree." Could merge with quantifiers or descriptors.
• Taxonomic/partonomic (KIND OF, PART OF): irreducible at NSM level.
• Evaluators (GOOD, BAD): irreducible.
• Descriptors (BIG, SMALL): possibly merge with evaluators (both are graded properties).
• Similarity (LIKE): irreducible.

Iteration outcome: 16 categories reduce to maybe 11-13 after collapsing some pairs. Still in the surface-domain band.

Resolution C — Higher-order semantic primitives

A bolder reading: the categories themselves cluster into a smaller set of semantic primitives at the substrate level. Candidate set:

1. Reference (Rf) — picking out objects of discourse. Subsumes substantives + determiners + quantifiers. Levels: 0 = no reference; 1 = deictic; 2 = personal; 3 = quantified; 4 = compositional reference (sets, classes).
2. Mental activity (Mn) — cognitive engagement with referents. Subsumes mental predicates + speech. Levels: 0 = none; 1 = perception (SEE, HEAR); 2 = belief (THINK, KNOW); 3 = desire (WANT, FEEL); 4 = communication (SAY).
3. Action (Ac) — change agents and events. Subsumes actions/movement. Levels: 0 = static; 1 = happen; 2 = do; 3 = move; 4 = touch (physical interaction).
4. State (St) — being, possession, life. Subsumes existence/possession + life/death. Levels: 0 = no state; 1 = exist; 2 = possess; 3 = live; 4 = die (the transitions).
5. Spatiotemporal (Sp) — when and where. Subsumes time + space. Levels: 0 = no positioning; 1 = present moment / here; 2 = before/after / above/below; 3 = duration / distance; 4 = full topology.
6. Modal (Md) — logical and modal relations. Subsumes logical + intensifier. Levels: 0 = absolute; 1 = negation; 2 = possibility; 3 = causation; 4 = conditional.
7. Evaluative (Ev) — qualitative judgment. Subsumes taxonomic + evaluators + descriptors + similarity. Levels: 0 = no judgment; 1 = good/bad; 2 = size/degree; 3 = kind/part; 4 = comparison/similarity.

This gives 7 primitives at the substrate level — exactly in the methodology's substrate-clustering band.

3/3b iteration verdict

After iteration: Resolution C is the most promising candidate at the substrate-style level. The 16 categories are partial levels within these 7 primitives. The 65 primes are sub-level positions within the 16 categories.

Three resolutions, three different views of the same domain:

This matches the methodology's scale-invariance: the same domain can be analyzed at multiple levels, each producing a primitive set appropriate to that level. NSM's literature presents the ~65 primes as primitive; the methodology suggests there's a substrate-level layer at ~7 primitives that the NSM tradition has not surfaced.

Working hypothesis for this analysis: proceed with Resolution C (7 substrate-level primitives) as the primary analysis, with the ~16 categories as partial levels and the ~65 primes as sub-level positions.

Step 3c — Evaluator identification

NSM does have an evaluator in the methodology's sense: the paraphrase operation. Given a concept, the paraphrase operation unfolds it into a structure of primes. The operation is Kd1-Kd2 (variable, context-dependent) — different analysts produce slightly different paraphrases of the same concept, and the discipline is empirical iteration rather than mechanical derivation.

This is interesting: NSM's evaluator is soft (Kd1-2), distinguishing it from biology's ribosome (Kd4) or the entity system's dispatch (Kd4). Like cognition's symbolization (Sy), NSM has a split evaluator: formal-mode paraphrases (definitions of technical terms) can be precise and reproducible (toward Kd3-4); natural-language paraphrases (definitions of cultural concepts) remain Kd1-2.

NSM is therefore a soft substrate at the semantic-content level — analogous to cognition's symbolization layer but operating on meaning units rather than perceptual/representational primitives.

Steps 4–6 — Dependencies, pairs, load classification

4 — Dependencies

Using Resolution C's 7 primitives:

• Rf → (none): substantives are the analytical root. To refer to things, no prior structure required.
• Mn → Rf: mental activity requires something to be mentally engaged with. THINK presupposes SOMEONE thinking about SOMETHING.
• Ac → Rf: actions require actors and patients (or at least entities being acted upon).
• St → Rf: states are states of something.
• Sp → Rf: spatiotemporal positioning presupposes things to be positioned.
• Md → {Mn, Ac, St}: modal relations modify mental, action, and state predicates. Without something to be possible/necessary/negated, no modality.
• Ev → Rf: evaluation is evaluation of a referent.

Dependency root: Rf (Reference). All other primitives presuppose Rf.

5–6 — Pairs and load classification

C(7,2) = 21 pairs. Classification (qualitative, first-pass):

Heavy pair count: ~11 of 21 ≈ 52%. Filter stringency for substrate-style domain expected at ~12-20%; if this analysis stabilizes, the coherent sub-lattice should be tight.

Steps 7–8 — Coherent sub-lattice and Hasse walks

Coherent sub-lattice

The dependency structure forbids any subset containing Mn/Ac/St/Sp/Md/Ev but not Rf. The dependency-coherent count of subsets:

• Total 2^7 = 128 subsets.
• Subsets containing Rf: 2^6 = 64.
• Subsets NOT containing Rf: only the empty set is coherent (no prims that require Rf).
• Md additionally requires Mn/Ac/St; subsets containing Md but none of Mn/Ac/St are incoherent.

Rough count: ~50-55 coherent subsets out of 128, i.e., ~40%. This is above the substrate-typical 12-20% band — closer to surface (25-40%). The relatively loose filter is consistent with Resolution C being substrate-flavored but with some surface character.

Better fit may be that NSM is genuinely a bridge domain between cognitive substrate and cultural ecosystem. Bridges typically have filter stringency in the 12-25% range; the 40% here may indicate additional dependencies not yet captured. The 3/3b iteration would tighten this.

Build-up narrative (Hasse walk)

A coherent build-up sequence:

1. Rf alone: pure reference. No mental activity, no action, no states. Just the ability to point at things. Trivial degenerate case.
2. Rf + Mn: reference + mental engagement. Can think about things. Pre-linguistic infant cognition?
3. Rf + Mn + Ac: reference + thinking + action. Embodied cognition with motor activity.
4. Rf + Mn + Ac + St: + state-tracking. Allows persistence of mental contents and action consequences.
5. Rf + Mn + Ac + St + Sp: + spatiotemporal positioning. Full event-grounded cognition.
6. Rf + Mn + Ac + St + Sp + Md: + logical/modal operators. Counterfactual reasoning, possibility, negation. Adult cognition.
7. Full set + Ev: + evaluation. Judgment, preference, normative reasoning. Fully meaning-capable.

This sequence has structural shape similar to the entity system's build-up (E → EI → EIT → EITM → EITMX → EITMXP) — informational primitives first, then mental/temporal, then evaluative/modal.

Step 9 — Load-bearing compositions

Triangles

Core triad

{Rf, Mn, Ev} is the strongest candidate: reference + mental activity + evaluation. This is the operational core of semantic meaning — picking out referents, having mental contents about them, judging them. Without any one of the three, semantic meaning collapses:

• Without Rf: nothing to mean about.
• Without Mn: no mental engagement with meaning.
• Without Ev: no judgment, no normative content, no distinguishing important from unimportant.

This matches the methodology's "functional-integration core triad" prediction for surface domains, though NSM behaves more like a substrate-style domain in the cross-linguistic universality of its primes.

Step 10 — Emergent property map (with derivations per M3)

All ten derive cleanly from the seven proposed primitives + their dependencies + composition rules. M3 passes for NSM.

Step 11 — Cross-domain pattern observation

Primitive count. 7 primitives at the substrate-style level. In the substrate-band (~6) or just above. Consistent with the methodology's clustering prediction.

Filter stringency. ~40% (high for substrate, mid for surface, possibly high for bridge). Needs tighter dependency analysis to reduce.

Core triad. {Rf, Mn, Ev} — operational core of semantic meaning. Has the same "anchor primitive + activity + evaluator" shape as biology's {G, T, R} (encoding + reader + evaluator) or the entity system's {E, I, T} (typed data + identity + namespace).

Hub primitive. Rf participates in 6 of 11 heavy pairs. Substantive hub primitive, similar to the entity system's E or biology's G.

Dialectical pattern. The 3/3b iteration revealed a structural finding NSM literature does not name: that the ~65 primes appear to be partial-level positions within ~7 substrate-level primitives. If this analysis holds, the discovery is that NSM's primitive inventory is actually a finely-resolved partial-level scoring system across a smaller substrate primitive set. This is exactly the relationship the methodology predicts between coarse and fine resolutions.

Step 12 — Literature alignment

Where NSM agrees with this analysis:

• The 16 categories Wierzbicka uses to organize primes match approximately to the 7 substrate-level primitives' partial-level categories. The organizational intuition was there; this analysis names it structurally.
• The hub status of substantives in NSM literature (every paraphrase requires at least one substantive) matches Rf being the dependency root and pair-relationship hub in this analysis.
• The empirical iterability of NSM (primes added/removed through cross-linguistic testing) matches the methodology's 3/3b loop.

Where this analysis differs from NSM:

• NSM treats the ~65 primes as the primitive level; this analysis suggests they are partial-level positions within ~7 substrate-level primitives.
• NSM does not formalize dependencies; this analysis surfaces Rf as the root and identifies Md's dependencies on Mn/Ac/St.
• NSM does not develop multi-arity compositions; this analysis identifies the core triad {Rf, Mn, Ev}.

These are not contradictions — they are different levels of analysis applied to the same domain. NSM's prime inventory remains valid as a fine-grained semantic vocabulary; this analysis re-positions it within a substrate-style framework.

Connections to existing arrangements

The user explicitly asked about connections to cognitive architecture, cognitive substrate, culture, language, meaning, artifacts. The candidate connections:

NSM ↔ cognitive substrate. NSM's 7 primitives describe what the cognitive substrate produces at its semantic output. The cognitive substrate's Sy (symbolization) primitive is what computes the semantic content; NSM is the content set that Sy produces. This is the same relationship as biology's genome (G) producing proteins (P) — biology substrate computes; the protein set is the substrate's output.

NSM ↔ cultural ecosystem. NSM primes are shared across all human languages. The cultural ecosystem transmits and refines meanings using NSM primes as the substrate-level vocabulary. The cultural ecosystem operates on NSM content the same way digital ecosystems operate on application-architecture content.

NSM as a bridge domain. NSM may be the cognitive-substrate ↔ cultural-ecosystem bridge — the substrate-level vocabulary that cognition produces and that culture transmits. This would make NSM structurally analogous to:

• Biology's biology-to-organism bridge (substrate → surface)
• Cognition's neural-to-cognitive bridge (substrate → architecture)
• The entity system's substrate-bridge extensions (substrate → application architecture)

This is the W2 workstream from the holistic-review note: cognitive-substrate ↔ NSM bridge analysis. The work above suggests the bridge exists and is structurally meaningful. A dedicated bridge analysis would extract the bridge primitives (analogous to biology's ~12 developmental mechanisms or the entity system's 11 substrate-bridge extensions).

Findings worth recording

1. NSM passes as a domain. The 12-step analysis produces a coherent primitive set, dependency structure, core triad, and emergent-property map.
2. Substrate-level primitives = 7. Resolution C produces {Rf, Mn, Ac, St, Sp, Md, Ev}. The ~65 NSM primes are partial-level positions within these 7. The ~16 NSM categories are intermediate-resolution groupings.
3. The substrate-clustering pattern holds. 7 substrate-level primitives is in the ~6 substrate band (slightly above), consistent with the methodology's cross-domain observations.
4. M3 passes for NSM. All 10 claimed emergent properties derive from the 7 primitives + dependencies + composition rules.
5. NSM is structurally a bridge between cognitive substrate and cultural ecosystem. The candidate position in the realization chain is between cognition's symbolization layer and the cultural transmission machinery.
6. The dialectical move. NSM literature treats ~65 primes as primitive; the methodology surfaces a more reduced ~7 substrate at a coarser level. The two are not in conflict — they are different levels of the same domain, exactly as the methodology's scale-invariance predicts.

Open questions

1. The bridge analysis (W2) — extract the cognitive-substrate ↔ NSM bridge primitives. What's between the cognitive substrate's Sy and NSM's Rf+Mn+Ev core?
2. The 40% filter stringency — high for substrate, low for surface; suggests additional dependencies the first-pass analysis hasn't captured. Tighter iteration would reduce.
3. The cultural-ecosystem connection — NSM primes are transmitted culturally; the cultural ecosystem domain (Cd, Tr, ...) operates on NSM content. The full ecosystem-bridge story is unfinished.
4. Wierzbicka and Goddard's responses — would the NSM team accept the substrate-level reduction to 7 primitives, or would they argue the 65 are genuinely primitive? This is a literature-alignment question worth pursuing if the analysis is ever published.

What this analysis adds

NSM was the closest literature analogue to the methodology (see primitives-history-nature-and-literature.md Part V). Running the methodology on NSM-as-domain has produced:

• A structural reframing of NSM: the ~65 primes resolve to ~7 substrate-level primitives at a coarser resolution.
• A bridge candidate: NSM as the cognitive-substrate ↔ cultural-ecosystem bridge.
• A second non-trivial M3 test: NSM's emergent properties all derive cleanly from the 7 substrate-level primitives.
• A literature touchpoint: the methodology and NSM converge on the same domain via different routes — the methodology arrives at ~7 substrate primitives; NSM arrives at ~65 fine-grained primes. Each finds something the other doesn't surface (NSM has empirical cross-linguistic validation the methodology lacks; the methodology has dependency structure + level-relativity + bilateral fixed-point criterion that NSM doesn't formalize).