Methodology Applicability — Review and Gaps

Status: Research document.. Third in the bounding-range series.

Inputs: methodology-bounding-range.md (twelve test cases + five failure modes); methodology-applicability-as-a-domain.md (meta-domain analysis with six primitives Cm/Cy/Eg/Jc/Gr/Fs and eight empirical attractors A–H).

Purpose: Stress-test the meta-domain analysis. Find what doesn't fit, where the six-primitive set has gaps, and what the methodology can and cannot do, stated sharply enough that the next iteration either fills the gap or accepts it as a structural limit.

Posture: the meta-domain analysis from yesterday is a first iteration. Its primitive set is provisional. The exercise here is to push against it: test candidate placings that resist the classification, look for kinds of "things" that we keep wanting to analyze but that don't fit, and sharpen the capability/incapability distinction. The output is a revised primitive-set candidate plus explicit gaps.

Part 1: Things that resist clean placement

The meta-domain has six primitives Cm/Cy/Eg/Jc/Gr/Fs. We test eleven candidate objects, asking for each: does this place cleanly in the meta-domain, or does it resist the classification in a way that surfaces a gap?

T1. Economics as a domain (re-examined under reflexive feedback)

Initial placement: (Cm=3, Cy=4, Eg=4, Jc=2, Gr=2, Fs=3) → attractor B (cyclic-domain zone). Partial value.

But this misses something. Economics has a structural feature the six primitives don't capture: publishing economic models changes economic behavior. The Efficient Market Hypothesis literature illustrates this directly — once an arbitrage opportunity is named, it disappears. Financial-market models that work in private use stop working when published. The analysis-domain coupling is itself load-bearing in economics.

Cy captures internal dependency cycles (markets constituted by participants constituted by markets). It does not capture the cycle between analysis and domain. These are different cycles.

Surfaced gap: the meta-domain needs a primitive for analytic reflexivity — the degree to which the domain changes in response to its own analysis. Call this Rx (reflexivity).

T2. The methodology itself

Initial placement: (Cm=4, Cy=1, Eg=3, Jc=3, Gr=4, Fs=3) → attractor A.

But the methodology applied to itself is the strongest possible case of analytic reflexivity. The methodology IS its own analysis; running the methodology on the methodology produces the four-layer self-analysis. The reflexive coupling between analysis and domain is total: there is no "domain separate from analysis" to analyze.

The six primitives place the methodology in attractor A but do not register that this is qualitatively different from the biology/entity/language placements in the same attractor. Rx is the missing axis.

T3. Self-help and behavioral-change literature

Initial placement: (Cm=2, Cy=3, Eg=4, Jc=1, Gr=2, Fs=2) → attractors B and G.

Also reflexive in a way the six primitives miss. Reading a self-help book changes the reader's behavior, which changes the domain the book describes. The instance base is partly constituted by readers of the literature.

This is the same gap as economics and methodology.

T4. AI safety and alignment research

Initial placement: hard to assign. Cm=3 (decomposable into specification / training / deployment / monitoring), Cy=2, Eg=2 (few mature AI systems), Jc=1 (deeply contested), Gr=2, Fs=3.

The harder problem: analysis of AI systems changes how AI systems are built. Safety research IS development. Reflexivity is near-total. Adding to the gap.

T5. A specific person's experience of grief

Try to place. Cm=0 or Cm=1 — the experience is atomic / weakly compositional. Cy=undefined. Eg=1 (one experience, one person). Jc=undefined (the person knows their experience; the analyst does not have access). Gr=0 or 1 — emotional intensity has gradation but the experience itself is largely holistic. Fs=0 — the function of grieving is non-structural; structural decomposition misses what grieving is.

Does it place? Technically yes: attractor C (function-mismatch) or E (no shared instance base). But the placement says "the methodology doesn't apply," which is correct.

Surfaced gap (subtle): the methodology's vocabulary doesn't have a clean way to talk about first-person experiential domains. The attractor labels capture the limit but the underlying structural reason is something like "the domain admits no third-person analytical decomposition." This may want its own primitive: Third- person observability (Tp). Or it may be subsumed under Cm and Eg together.

Hold this as a softer gap for now; revisit if more first-person candidates appear.

T6. The friendship graph of a social network

Initial placement: Cm=4 (decomposable into nodes / edges / clusters / centrality / community structure / temporal evolution), Cy=2 (networks have feedback effects), Eg=4 (many networks), Jc=3 (network science has matured), Gr=3, Fs=3 → attractor A.

This fits. Network domains have been analyzed extensively in network science, with primitives at the network-property level (degree, clustering, path length, etc.). The methodology would produce a primitive set similar to what network science already articulates.

But what about a specific network — the friendship graph of one specific Facebook population? This is more like a manifestation than a domain. Eg=1 if treated as its own instance. The methodology would position it in some arrangement, not analyze it as a domain.

Surfaced gap: the distinction between domain and manifestation within a domain may need to be sharper. The meta-domain analysis is about domains, but the methodology in practice analyzes both domains and instances within them. The meta-domain's primitives may not transfer cleanly to instance-level analysis.

T7. The French Revolution as a domain

Try to place. Cm=3 (decomposable into causes/events/actors/outcomes), Cy=2, Eg=1 (one revolution), Jc=2 (multiple historical traditions disagree on the causal structure), Gr=2, Fs=3.

Does it place? Numerically yes — somewhere between attractors A and G. The Eg=1 with Jc=2 is uncomfortable — for the genetic code, Eg=1 was compensated by Jc=4 (full agreement on the universal code); here, Eg=1 has only Jc=2 (multiple historical accounts), so the licensed-claim path is harder.

Surfaced gap: historical-event domains (single instance, contested interpretation) are a specific structural location, and the methodology's analytical instinct is to either generalize them (treat as one instance of "revolutions") or sub-decompose them (treat the causes as primitives at finer resolution). The meta-domain doesn't yet have a vocabulary for choosing between generalization and sub-decomposition.

T8. "The category of all mathematical structures"

Try to place. Cm=4 (every structure is itself compositional). Cy=0 (no cyclicity at the category level). Eg=? (every mathematical structure is an instance, so Eg=4 in some sense; but they're all stipulated, so the empirical-recurrence test is meaningless). Jc=4 (mathematicians agree on what mathematical structures are). Gr=4 (definitionally discrete). Fs=4 (function is structural).

Numerically high-value (attractor A). Actually attractor F (axiomatic-degeneracy). The Eg primitive doesn't distinguish "instances exist empirically" from "instances exist by stipulation." We already flagged this: split Eg=1 into Eg=1d and Eg=1s. The same split is needed at Eg=2, Eg=3, Eg=4: are these discovered or stipulated instances?

Confirming the earlier gap. The discovery-vs-stipulation distinction is not just a partial-level refinement at Eg=1; it applies at every Eg level. This may be a seventh primitive Origin of instances (Ds) rather than a within-Eg refinement.

T9. A still-developing domain (the entity system pre-release)

Try to place. The entity system is itself a domain being constructed; its primitives have been revised through six version cycles. Cm=4, Cy=0, Eg=1 (one developing instance — although Eg=4 if you count each version as an instance, but versions evolve into each other, not independent), Jc=3, Gr=4, Fs=4.

This places at attractor A with an Eg complication. The Eg level of an under-development domain is unstable: it changes as implementations and revisions accrue.

Surfaced gap: the meta-domain assumes the domain is a stable object. Domains in flux (active development, biological evolution in progress, social systems in transition) have time-varying primitive levels. The meta-domain analysis is a snapshot. Time isn't a primitive in the meta-domain.

T10. The classification of art styles

Try to place. This is itself an analytical artifact. Cm=3 (loose compositional structure — periods, movements, schools), Cy=1, Eg=4 (many art-style classifications proposed), Jc=1 (deeply contested), Gr=2 (boundaries fuzzy), Fs=2 (function is part analytical, part normative).

Places in attractor G (contested-judgment). But there's a structural feature: this domain is itself a methodology output of art-historians applying their own analytical procedures. The domain we're analyzing is downstream of another methodology.

Surfaced gap: the meta-domain doesn't distinguish primary domains (the analyzed object is the domain itself) from secondary domains (the analyzed object is the output of someone else's analytical procedure). The latter are meta-analyses of other analyses. Where the methodology applies to secondary domains, it is in some sense doing methodology-on-methodology one level removed.

This may be subsumable under Rx (reflexivity), but the structural character is different: in economics, the domain reacts to its analysis; in classifying art styles, the domain is an analysis.

T11. "Reality at finest resolution"

Try to place. This is the substrate-of-substrates. Cm=? (we don't know if reality is compositional at finest resolution — quantum gravity work suggests yes but the primitives are contested). Eg=1 (one reality). Jc=0 (deeply contested across physics programs).

Does it place? Only insofar as we have a candidate substrate to analyze. The Planck substrate is one such candidate; loop quantum gravity is another; causal set theory is another. Each places somewhere in the meta-domain, but "reality itself" is not a domain in the methodology's sense until a candidate substrate is fixed.

Surfaced gap: the methodology presupposes the analyst has fixed a candidate description of the domain. Where the description is itself in flux (open scientific frontier), the methodology can analyze each candidate but not the underlying object.

This is consistent with the project's earlier framing: the methodology is analyst-cartographic. What's interesting is that for some domains the cartography depends not just on the analyst's choices about decomposition but on which candidate description is being analyzed in the first place.

Part 2: Gaps surfaced, ranked

The eleven test cases surface real structural gaps in the six- primitive meta-domain. Sorted by severity:

Gap G1 (HIGH): Reflexivity (Rx)

The meta-domain does not capture analytic reflexivity — the degree to which a domain changes in response to its own analysis. Surfaced by T1 (economics), T2 (methodology), T3 (self-help), T4 (AI safety).

Proposed primitive: Rx with 5 partial levels:

Rx=0: no feedback. Analysis does not affect the domain. (Physics laws.)
Rx=1: weak feedback. Observation perturbs but does not transform. (Most natural-science measurement.)
Rx=2: moderate feedback. Analysis circulates and is incorporated into the domain. (Economics with public model release.)
Rx=3: strong feedback. The analysis becomes a normative resource for participants. (Psychology, self-help, organizational behavior.)
Rx=4: total reflexivity. Analysis and domain are inseparable. (Methodology, AI safety research, certain forms of philosophy.)

Heavy pairs: Rx-Fs (reflexive function-aligned domains converge; reflexive function-mismatched domains drift), Rx-Eg (reflexive domains tend to inflate Eg as analyses accumulate).

Effect on attractors: Rx changes how stable a methodology output is over time. High-Rx domains have outputs with short half- lives — the analysis ages because the domain has incorporated it. Attractor A high-Rx (methodology, AI safety) is different from attractor A low-Rx (chemistry, language at substrate).

Gap G2 (HIGH): Discovery vs Stipulation (Ds)

The meta-domain's Eg primitive does not distinguish empirically- discovered instances from axiomatically-stipulated ones. Surfaced by T8 (mathematical structures) and the genetic-code-vs-category- theory comparison from the prior document.

Proposed primitive: Ds with 3 partial levels:

Ds=0: stipulated. Instances exist by definition. (Pure mathematical objects.)
Ds=1: hybrid. Instances exist in mixed empirical-and-stipulated sense. (Information theory; the methodology's own primitive sets.)
Ds=2: discovered. Instances exist independently of the analyst. (Biology, physics-as-observed, languages-in-use.)

Heavy pairs: Ds-Eg (jointly determine licensed-claim path), Ds-Jc (stipulated domains have high Jc by definition).

Effect on attractors: attractor F (axiomatic-degeneracy) is the Ds=0 region. The genetic code at Ds=2 produces methodology output (it's discovered); category theory at Ds=0 degenerates (it's stipulated). Same Eg level, different Ds — different output.

Lighter alternative: rather than a new primitive, refine Eg's partial-level definition at every level to carry a discovered-vs- stipulated marker. We prefer the explicit Ds primitive because it appears at heavy-pair strength with Jc, suggesting it's a real structural axis.

Gap G3 (MEDIUM): Time / development state (Td)

The meta-domain is a snapshot. Domains in development (T9 — entity system pre-release), in transition (active scientific frontiers), or evolving (biology arrangements during major transitions) have time-varying primitive levels.

Proposed: either add Td (time-development state) as a primitive, or accept the meta-domain as a snapshot taken at analyst-chosen time with the time-variability captured at the Layer-4 Trajectory level.

We lean toward the second option: time-variability is what Layer 4 Trajectory already handles. A domain's meta-domain position changes over time; this is a trajectory in the meta-domain. No new primitive needed; we just need to remember the meta-domain placing is at a specific Sc=3 snapshot.

Effect on attractors: placements with low Td (still-developing domains) are unstable in the meta-domain — their attractor assignment may change as the domain matures.

Gap G4 (MEDIUM): Third-person observability (Tp)

The meta-domain's vocabulary doesn't cleanly distinguish third- person-observable domains from first-person-experiential ones. Surfaced by T5 (personal grief).

Provisional: Tp may be subsumed under (Cm, Eg) together. First- person domains tend to have low Cm and low Eg (single experiencer, no shared decomposition). Adding Tp as a separate primitive risks over-fitting; subsuming it under low-Cm-low-Eg gives reasonable behavior.

Hold as a softer gap unless more first-person candidates surface that force the distinction.

Gap G5 (MEDIUM): Domain vs manifestation distinction

The meta-domain analyzes domains. But the methodology in practice analyzes both domains and instances within domains. The meta- domain's primitives may not transfer cleanly to instance-level analysis. Surfaced by T6 (specific friendship graph).

Provisional resolution: the meta-domain is for Layer-1 domains (class-level, Sc=1 in the Layer-4 scope ladder). Layer-4 instances (Sc=3) are not in the meta-domain's natural target. A separate analysis would be needed for instance-level applicability.

Gap G6 (LOW): Primary vs secondary domains

The meta-domain doesn't distinguish primary domains from secondary (analyst-output) domains. Surfaced by T10 (classification of art styles).

Provisional: secondary domains place wherever the analyst's output sits. The fact that the domain is downstream of another analysis is captured by Rx (the domain incorporates the analyst's work into itself).

Gap G7 (LOW): Candidate-description dependency

For domains where the underlying object is contested (T11 — reality at finest resolution), the meta-domain analyzes a candidate description, not the underlying object. This is consistent with the analyst-cartographic discipline; the meta-domain doesn't need to handle it specially.

Resolution: no new primitive. The meta-domain analyzes whatever the analyst presents; if multiple candidate descriptions are available, each produces its own meta-domain placement.

Part 3: Refined primitive set (second iteration)

Combining the gaps surfaced:

The six-primitive meta-domain is revised to eight primitives:

Cm — Compositionality (unchanged)
Cy — Dependency-cycle density (unchanged)
Eg — Empirical groundedness (instance count; refined to drop the discovery/stipulation tracking, which moves to Ds)
Jc — Judgment convergence (unchanged)
Gr — Granularity (unchanged)
Fs — Function-structure alignment (unchanged)
Rx — Reflexivity / analysis-domain feedback (new)
Ds — Discovery vs stipulation status (new)

Two new primitives. We accept the cost in primitive-set growth because both gaps are structurally significant: Rx changes how stable methodology outputs are over time; Ds changes whether the empirical-recurrence test applies at all.

Eight primitives is above our typical band of six. Either we have undershot the methodology's natural scale on this meta-domain, or two of the eight primitives can be combined. We propose:

Keep all eight for now. The 3/3b iteration loop has only run twice (once in the prior document, once here). Further iteration may collapse two primitives into one or split an existing one.
Flag the potential combination Rx + Ds. Both relate to the epistemic-status of the domain (does the analysis become part of the domain; are the primitives discovered or stipulated). A unified primitive "Epistemic dependency on the analyst" might subsume both. We have not made this collapse because the two appear independent across our test cases.

Filter stringency estimate: with 8 primitives and 4–5 partial levels each plus the added dependencies (Rx depends on Cm; Ds depends on Eg), the coarse-level filter is probably 25–30% — still classifier-style, slightly tighter than the six-primitive version.

Core triads (revised):

{Cm, Cy, Gr}: Structural-decomposability (unchanged)
{Cm, Eg, Jc}: Empirical-grounding (unchanged)
{Cm, Fs, Jc}: Useful-output (unchanged)
{Eg, Jc, Ds}: Licensed-claim path (new triad) — joint determination of whether the methodology produces findings or cartography
{Rx, Eg, Fs}: Output stability over time (new triad) — how long the methodology's analysis remains accurate

Hub primitive: still Cm. Second-tier hub: Eg (4 heavy pairs in the refined set).

Part 4: Capability vs incapability — sharper statements

The user asked what the methodology can and can't do. Stated with the meta-domain in hand, by attractor:

What the methodology DOES in attractor A (Cm=4, Cy≤1, Eg≥3 or Jc=4, Gr≥3, Fs≥3, low Rx)

Produces a primitive set, dependency DAG, partial-level decompositions.
Constructs the coherent sub-lattice with filter stringency in the domain-type-typical range.
Identifies core triads, hubs, anchor pairs.
Predicts phase transitions, attractor states, and design opportunities at unpopulated coherent positions.
Detects cross-domain patterns when applied to many domains in this attractor (this is how SSA and Convergence Domain emerged).
Supports cross-instance comparison via unified manifestations.
Admits reflexive application to itself (the four-layer self- analysis is in this attractor).

What the methodology DOES NOT do in attractor A

Adjudicate between competing primitive sets when multiple sets pass the three-test extraction with similar quality.
Determine which partial-level meaning is correct when the same primitive admits multiple plausible interpretations (the cognition energy-semantics is the existing within-corpus example).
Replace domain-specific empirical work — the methodology produces structural maps, not measurements.
Predict specific empirical outcomes — it predicts structural consequences (a composite gate will not release without all prerequisites; an unpopulated coherent position is a candidate design point) but does not predict timing, magnitude, or specific realization.

In attractor B (cyclic-domain zone, Cy≥3)

Does: produces a primitive set and a cycle-broken dependency DAG; identifies cluster structure across instances; positions instances in a lattice with the cycle-breaking convention made explicit.

Does not: resolve which cycle-breaking convention is correct; predict equilibria of cyclically-constituted dynamics; replace cybernetic or dynamical-systems analysis.

In attractor C (function-mismatch zone, Fs≤2)

Does: produce a structural decomposition of the domain.

Does not: capture the domain's function or value when that function is non-structural; substitute for phenomenological, hermeneutic, or aesthetic analysis of the same domain.

In attractor D (granularity-bottleneck zone, Gr≤1 with Fs≥3)

Does: produce a regime classification with named phase regions.

Does not: capture the mechanism that the underlying continuous mathematics captures; replace differential-equation modeling.

In attractor E (empirical-bottleneck zone, Eg=0)

Does: nothing operational. Reasoning at most.

Does not: apply.

In attractor F (axiomatic-degeneracy zone, Ds=0)

Does: transcribe the axioms in primitive-set form.

Does not: generate new structural insight; produce licensed-claim output (the three-test extraction degenerates because the primitives are stipulated).

In attractor G (contested-judgment zone, Jc≤1)

Does: produce one analyst's primitive set, lattice, and patterns. Each analyst produces their own.

Does not: adjudicate between competing analyses; produce a domain-canonical primitive set.

In attractor H (mature-craft zone, all primitives moderate-high but practitioner literature already explicit)

Does: produce a structural map of the domain.

Does not: add marginal value over existing practice; structural- comparative readings here often duplicate practitioner knowledge.

Across attractors: methodology behaviors that hold

Does in every attractor:

Document its own discipline (cartography vs licensed claim).
Make its analyst-judgment dependence visible.
Mark retractions and downgrades cleanly when failure modes surface mid-analysis (the within-corpus retractions are the example).
Carry the empirical-recurrence test as a structural check on primitive-set stability.

Does not in any attractor:

Provide foundational answers about the nature of reality, the structure of mind, or the meaning of any specific domain.
Replace domain expertise.
Produce predictions whose confidence the methodology alone can set.
Guarantee that what it produces is what would be produced by another competent analyst applying the same procedure.

Part 5: What this means for Paper 11

The two iterations of the meta-domain analysis (yesterday's six- primitive draft and today's eight-primitive refinement) have produced:

A working primitive set for "analyzable domains" with two surfaced primitive-refinement signals from the 3/3b loop (Rx and Ds).
Eight empirical attractors that map the methodology's range.
Explicit empty regions (Cm=4 + Cy=4 most prominently) and forbidden regions.
A capability/incapability matrix by attractor.
Six structural gaps in the original six-primitive set, two ranked HIGH severity.

Paper 11's framing needs to absorb this. Specifically:

The previous "Range of Domains Analyzed" section, which used a catalog framing, should be replaced or augmented with a section that uses the meta-domain analysis. The honest framing is "the methodology applied to itself produces a structural map of where it applies, with eight attractors and two open primitive- refinement signals."
The "domain-general" claim needs to be qualified: domain-general within attractor A, graded value in attractors B/C/D/H, does not apply in E/F/G.
The methodological-discipline chapter should acknowledge that the cartography vs licensed-claim distinction depends on the attractor: licensed claims are only available in attractor A and partially in attractor B; the other attractors produce cartography only.
The conclusion's open-avenues paragraph should include the two primitive-refinement signals (Rx and Ds) as work in progress, not closed questions.

What this means for the methodology's general claims:

The methodology is a structural analytical procedure with identifiable structural assumptions (A1–A6 from yesterday's document, plus the Rx and Ds extensions today).
Its range is mapped by the eight-attractor meta-domain analysis.
Its outputs in each attractor are characterized.
Its failure modes are F1–F5 (yesterday's document) and have associated repair strategies (some implementable, some not).
Its claim of domain-generality holds within attractor A; outside it the claim is graded or false.

What the methodology still cannot resolve from inside itself:

Whether the candidate Layer-3 patterns (cyclic constitution, crystallization, substrate-vs-architecture, function-substrate- mismatch) are genuine abstractions or analytical artifacts. Only running them through the 12-step procedure will tell.
Whether the meta-domain primitive set will stabilize at eight or will collapse to seven (combining Rx and Ds under "epistemic- status").
Whether the empty regions (Cm=4 + Cy=4) are forbidden by deeper structure or merely unexplored.
Whether the methodology can be extended with native cycle handling (F1 repair) and a functional layer (F5 repair). Both require significant additional machinery.

Part 6: Next steps

The methodology has been applied to itself across three documents in this series. To carry this work into Paper 11, the natural next moves are:

Replace the existing "Range of Domains Analyzed" section in Paper 11 with material drawn from these three documents, structured around the meta-domain analysis. Keep the catalog of analyzed domains but reposition it as instances placed in the meta-domain landscape.
Add a capability/incapability subsection stating what the methodology does and does not do in each attractor.
Update the conclusion's open-avenues paragraph to include the Rx and Ds primitive-refinement signals plus the candidate Layer-3 patterns.
Update the methodological-discipline chapter to note that the cartography vs licensed-claim distinction is attractor- dependent.

These edits should be made carefully — the meta-domain analysis is still on its second 3/3b iteration, and any of the primitive labels, partial-level definitions, or attractor boundaries may be revised by further iteration. We should mark the meta-domain analysis as "current iteration" rather than as a fixed claim, and treat the eight primitives as a working primitive set rather than a definitive count.

The user has asked us to push deeper. The next iteration would push more candidate domains through, test whether Rx and Ds combine under a single epistemic-status primitive, and investigate at least one of the empty regions (likely Cm=4 + Cy=4) by trying to find a domain that inhabits it. These are exploratory research moves the methodology supports but does not direct.