Morphological Analysis

Quick facts

Mechanism: Produces a morphological field (a Zwicky box of independent parameters by their possible values) plus a cross-consistency-pruned set of internally consistent configurations.

Produces: morphological-field

Evidence: P · Domain: Divergent Ideation

Use when a solution is genuinely a configuration - a choice on each of several semi-independent dimensions (product architecture, service bundle, policy package) - and the risk is defaulting to one familiar combination and never seeing the rest of the space. Not a top-down MECE tree and not a provoked option list.

Most design and option work jumps straight to a few familiar combinations and never sees the rest of the space. Morphological analysis (the “Zwicky box”) refuses that shortcut. It treats a solution as a CONFIGURATION - a single value chosen on each of several semi-independent dimensions - and lays the whole space out so the unobvious corners come into view. The durable move has two halves, and the second is what separates it from a plain options list: (1) decompose the problem into a small set of independent PARAMETERS with their discrete VALUES, giving the full combinatorial cross-product; then (2) run a cross-consistency assessment - pair off every value with every other value across parameters, strike the internally incompatible pairs, and keep only the configurations that survive. The output is a morphological field: a parameter-by-value box plus a consistency-pruned set of internally consistent configurations. It is not a tree, not a ranked list, and not the chosen answer - it generates and prunes-for-consistency; a downstream step scores and chooses.

When to Use

• The solution genuinely is a configuration - a choice on each of several semi-independent dimensions (a product architecture, a service or pricing bundle, a feature set, a policy package, a research design, a go-to-market shape).
• The real risk is tunnel vision: the team keeps defaulting to one familiar combination and would not otherwise see the rest of the space, including the corners no forward search would visit.
• The parameters are real and discoverable, and some value pairs are clearly incompatible - so the cross-consistency pass does real pruning rather than rubber-stamping everything.
• The goal is breadth-with-structure (cover the space, then narrow to the viable region), with scoring and selection handled as a separate downstream step.

When NOT to Use

• Do not use it to provoke fresh options by negating the premises. Negating a problem’s foundational assumptions to jolt out new ideas is think-assumption-reversal; its product is a provoked option list. Morphological analysis negates nothing and provokes nothing - it builds a structured combinatorial space and reduces it by consistency. They share only the abstract goal “expand the option set,” which is well under the overlap line.
• Do not use it to transform one seed idea. Applying fixed transformation verbs (substitute, combine, adapt, modify, and so on) to a single existing concept is think-scamper. There is no decomposition into orthogonal axes and no cross-product there; that is a different machine.
• Do not use it to import structure from a distant domain. Borrowing the deep structure of a far-off field to spark an idea is think-far-analogy-ideation. Morphological analysis uses no analogy; its lift comes from recombining the problem’s own parameters.
• Do not use it to invert toward bad ideas and flip them. Deliberately generating the worst possible ideas and then reversing them is a provocation technique, not an enumeration. Morphological analysis does not invert; it enumerates a space and prunes it for internal consistency. (This library ships no skill for that provocation, so reach for it only outside this catalog.)
• Do not confuse it with a top-down MECE tree. Decomposing one question into a mutually-exclusive, collectively-exhaustive hierarchy you read as a tree is think-issue-tree. A tree never multiplies its branches together and never runs a cross-consistency pass; morphological analysis decomposes into PARALLEL orthogonal axes and then RECOMBINES and prunes them. Decomposition is the only shared part; the recombination-and-pruning is exactly the part a tree lacks.
• Do not treat the box as the evaluator. The field generates configurations and prunes them for internal consistency; it does NOT rank them by value. Scoring surviving configurations against weighted criteria to pick one is think-decision-option-review, which runs AFTER. Reading a surviving configuration as “the chosen answer” skips the trade-off comparison the box was never designed to make.
• Do not use it on a non-configurational problem. If the answer is a single insight, a reframe, or one number, forcing it into orthogonal axes is overhead. The method assumes the solution factorizes; many do not.

Instructions

When asked to map a solution space or systematically generate configurations, follow these steps:

1. State the focal problem in one line. Name the thing being configured and what “a complete solution” must specify. The field exists to serve this; if the answer is a single insight or one number rather than a configuration, stop (that is the non-configurational anti-pattern).
2. Choose the parameters (the independent dimensions). Identify a small set (typically 3-6) of orthogonal parameters - the dimensions any solution must take a position on. Keep them genuinely independent and few; adding parameters to feel thorough multiplies the space toward the unmanageable. The decomposition is the whole ballgame, and nothing in the method forces a good one - justify each parameter.
3. Enumerate the values for each parameter. For each parameter, list its discrete possible values (the cells in that column). Aim for values that are meaningfully distinct and reasonably exhaustive for that dimension. This is the raw morphological field - the full cross-product is the product of the value counts.
4. Run the cross-consistency assessment (the reduction step). Pair every value with every other value ACROSS parameters and judge each pair as compatible or incompatible. Mark the pairs that cannot coexist in a sensible solution (technical conflict, contradiction, mutual exclusion). This is “inference by exclusion”: it is what collapses a huge raw space to a small viable residual.
5. Derive the internally consistent configurations. Keep only the configurations (one value per parameter) that contain no incompatible pair. These are the surviving, internally consistent candidates - the residual set worth examining. Note the rough before/after size to show the pruning did real work.
6. Sanity-check coverage and the axes. Confirm the parameters are still independent and that the surviving set includes the unobvious corners, not just the familiar combination you started from. If the field merely reproduces the preferred solution, the axes are probably arbitrary - say so and reselect.
7. Hand off, do not rank. Present the consistent configurations as the candidate set for a separate evaluation step (for example think-decision-option-review). Do not score them by value here; the box generates and prunes for consistency, it does not choose.
8. Emit the morphological field artifact per references/TEMPLATE.md: the parameters and their values (the box), the cross-consistency judgments, and the pruned set of internally consistent configurations.

Output Format

Use the template in references/TEMPLATE.md. The deliverable is the filled morphological field - the parameter-by-value box, the cross-consistency assessment (which value pairs are incompatible), and the pruned set of internally consistent configurations - not a prose essay. Do not rank the configurations by value; that is a separate downstream step.

Quality Checklist

Before finalizing, verify:

• The focal problem is stated in one line, and the answer genuinely is a configuration (not a single insight, reframe, or number forced onto axes).
• The parameters are few (typically 3-6) and genuinely independent - not a sprawling list added to feel thorough, and not arbitrary axes that merely justify a preferred solution.
• Each parameter’s values are discrete, meaningfully distinct, and reasonably exhaustive for that dimension.
• A real cross-consistency assessment was run: value pairs across parameters are judged compatible or incompatible, with the incompatible pairs named.
• The pruned set contains only internally consistent configurations, and the rough before/after size shows the pruning did real work.
• The surviving set surfaces unobvious corners, not just the familiar starting combination.
• The configurations are handed off as candidates, NOT ranked by value here (scoring is think-decision-option-review, downstream).
• The output is the morphological field artifact, not prose.
• No overclaiming: the evidence is practitioner-grade and transferred; claim a coverage-and-consistency aid, not a measured gain in solution quality, and remember the coverage promise is often only partly realized in practice (see evidence/dossier.md).

Evidence

Tier P (governing). Morphological analysis is a real, named, long-lived method with a clear lineage (Zwicky 1969; Ritchey from 1998) and a large descriptive base - Alvarez and Ritchey (2015) catalogue roughly 80 published applications across engineering design, technology forecasting, and policy. Unusually for this library, controlled experiments DO exist, yet they hold the grade at P rather than lift it, for three reasons. First, they measure a lighter sibling, the morphological CHART, without Ritchey’s cross-consistency step. Second, the headline results are mixed-to-unflattering: the strongest comparison (Daly, Seifert, Yilmaz and Gonzalez, 2016, n = 102 students) found the method raised concept ELABORATION but was beaten on practicality by design heuristics and on quantity by plain brainstorming, and Smith, Troy and Summers (2012) found adding chart functions did not improve concept quality and that designers explored only about a quarter to a seventh of the space - so the “complete coverage” promise is largely unrealized in use. Third, every study is on human design students or practitioners; none studies an AI-produced field. The strong “90-99% reduction” descriptions come from Ritchey’s own methodological writing, not from controlled validation, and Zwicky’s “100x brain efficiency” / “philosopher’s stone” claims are excluded as self-promotion. The transfer caveat is double - human-to-agent AND chart-to-full-GMA - which is exactly why the honest grade is held at P. The skill ships as a coverage-and-consistency aid with a hard “the parameterization is only as good as the analyst, and the box is not the evaluator” wall, never as a guaranteed exhaustive search or a decision-quality booster. Full grading, sources, and caveats: evidence/dossier.md.

Examples

See references/EXAMPLE.md for a completed morphological field on a real decision.

Deep dive: worked example

A full worked run (the shared Northwind scenario)

Morphological Field (Zwicky Box) - Worked Example

A completed run of the morphological-analysis skill on a real, consequential decision. This is the quality bar a generated morphological field should meet.

Uses the shared recurring scenario (Northwind, a B2B SaaS weighing a self-serve free-tier launch). Northwind fits this method well because the free-tier offering itself is genuinely a CONFIGURATION - the team must take a position on several semi-independent dimensions (what to gate, how conversion is triggered, how free users are supported, where they sign up) and tends to default to one familiar combination. Where think-scenario-planning models the uncontrollable external worlds the bet lands in, this skill maps the internal design space of the offer and prunes it to the internally consistent shapes. See docs/internal/AUTHORING.md.

The configurations below are GENERATED and pruned for consistency, not ranked by value. Choosing among the survivors is a separate downstream step (think-decision-option-review).

---

Focal problem

• What is being configured: the shape of Northwind’s self-serve free-tier offering - the concrete design of the free product and its path to paid, not whether to launch it.
• What a complete solution must specify: what the free tier gives away (the gate), what triggers the upgrade to paid (the conversion lever), how free users are supported (the support model), and where/how users first land (the acquisition surface).
• Why a field and not a single answer: the offer genuinely factorizes into these semi-independent dimensions, and Northwind keeps defaulting to one familiar combination (“free-forever seat-limited, upgrade when you add seats, community-only support, website signup”). Laying out the space forces the unobvious corners into view.

Parameters (the independent dimensions)

Parameter	Why it is a real, independent dimension
P1: Gate (what free gives away)	The core “free vs paid” boundary; the most consequential design choice and orthogonal to how users arrive or are supported.
P2: Conversion lever (what triggers paid)	How a free user becomes a paying one; independent of the gate (the same gate can convert on seats, usage, or features).
P3: Support model (how free users are helped)	The cost-and-experience choice for the free population; independent of gate and conversion.
P4: Acquisition surface (where users first land)	The top-of-funnel entry; independent of the product’s internal shape.

The morphological field (the box)

P1: Gate	P2: Conversion lever	P3: Support model	P4: Acquisition surface
Free forever, seat-limited	Add seats (per-seat upgrade)	Community / docs only	Website self-signup
Free forever, feature-limited	Hit a usage cap (metered)	In-app self-serve + email	Product-led (in-product invite/share)
Time-limited full trial	Unlock a premium feature	Assisted (human onboarding)	Marketplace / integration listing

• Raw configuration count: 3 x 3 x 3 x 3 = 81 full configurations before pruning.

Cross-consistency assessment (the reduction step)

Incompatible value pairs (cannot coexist in a sensible offer) and why:

Value	Incompatible with	Why they cannot coexist
P1: Time-limited full trial	P2: Add seats (per-seat upgrade)	A trial converts on the clock running out, not on seat growth; pairing them sends contradictory upgrade signals.
P1: Time-limited full trial	P2: Hit a usage cap (metered)	A full trial intentionally removes caps; a metered cap contradicts “full.”
P1: Time-limited full trial	P3: Community / docs only	A short trial that must convert fast cannot rely on slow community support to get users to value in time.
P1: Free forever, feature-limited	P2: Unlock a premium feature	Redundant/circular: the gate already withholds premium features, so “unlock a premium feature” IS the gate, not a separable lever.
P3: Assisted (human onboarding)	P1: Free forever, seat-limited	Human onboarding for an unbounded free-forever population is economically incoherent at self-serve scale (cost grows with free users who may never pay).
P4: Marketplace / integration listing	P2: Add seats (per-seat upgrade)	Marketplace acquisition lands single users/integrations, not teams; a seat-growth conversion lever has nothing to act on at entry.

(Note how the cross-consistency pass does real work here: the “time-limited full trial” gate is incompatible with three of the other column’s values, which collapses a large share of the 81.)

Internally consistent configurations (the pruned set)

Surviving configurations (one value per parameter, no incompatible pair). Showing the representative residual after exclusion:

#	P1 Gate	P2 Conversion lever	P3 Support model	P4 Acquisition surface	Note
C1	Free forever, seat-limited	Add seats	Community / docs only	Website self-signup	The familiar default - team-expansion PLG.
C2	Free forever, seat-limited	Add seats	In-app self-serve + email	Product-led (in-product invite)	Default, but viral entry + lighter-touch support.
C3	Free forever, feature-limited	Hit a usage cap	Community / docs only	Website self-signup	Usage-metered freemium; converts on consumption.
C4	Free forever, feature-limited	Hit a usage cap	In-app self-serve + email	Product-led	Metered freemium with assisted self-serve and viral entry.
C5	Free forever, feature-limited	Hit a usage cap	In-app self-serve + email	Marketplace / integration listing	The unobvious corner - acquire via an integration, convert on usage.
C6	Time-limited full trial	Unlock a premium feature	In-app self-serve + email	Website self-signup	A trial-led shape (kept consistent: no seat/metered lever, no community-only support).
C7	Free forever, feature-limited	Unlock a premium feature - EXCLUDED	-	-	(Struck by CCA: circular with the feature-limited gate.)

• Pruned count vs raw: the cross-consistency pass removes the large majority of the 81 raw configurations; roughly a dozen survive as internally coherent, of which the six above (C1-C6) are the meaningfully distinct families.
• Unobvious corners surfaced: C5 (marketplace acquisition + usage-metered conversion) is a coherent shape a forward search starting from the seat-limited default would almost certainly have missed - it lets Northwind ride an integration partner’s distribution while still converting on consumption.

Hand-off (not a ranking)

These six consistent configurations (C1-C6) are the candidate set, not a recommendation. The field has done its job: it took the 81-cell space, struck the internally incompatible combinations, and surfaced both the familiar default (C1) and an unobvious-but-coherent corner (C5). Scoring them against Northwind’s criteria - expected conversion, support cost, time-to-value, fit with the buyer structure - is the separate downstream step (think-decision-option-review). Morphological analysis chooses none of them.

---

Note how this differs from its neighbors on the same Northwind decision. think-scenario-planning builds the uncontrollable EXTERNAL worlds the free-tier bet lands in and asks which moves survive all of them. think-issue-tree would decompose one question (for example “why is self-serve conversion low?”) top-down into a MECE diagnostic tree. think-assumption-reversal would negate a premise (“what if the free tier gave away the whole product?”) to provoke ideas. This skill does none of those: it decomposes the OFFER into parallel parameters, enumerates their cross-product, and prunes it to the internally consistent configurations. The deliverable is a consistency-pruned field, not a set of external futures, a diagnostic tree, or a provoked list.

Grounding: the full evidence dossier

What the research does and does not show, with graded sources

Evidence Dossier: Morphological Analysis

The single source of truth for the morphological-analysis skill. The SKILL.md, the sidecar (skill.meta.yml), and the eval cases all derive from this file. If a claim is not here, it does not belong in the skill. Promoted from frameworks/_proposed/morphological-analysis/dossier.md and admitted as a Build at tier P.

Skill	thinking-framework-skills.morphological-analysis (installable name think-morphological-analysis)
Family	divergent-ideation
Evidence tier	P governing (controlled studies exist but measure a sibling variant on humans and do not favor the method, so P is a ceiling as well as a floor - see “What the evidence shows”)
Confidence	Moderate that decompose-enumerate-and-prune-for-consistency surfaces unobvious configurations and breaks default-combination tunnel vision; low that any specific solution-quality effect transfers to agents, and the coverage promise is often only partly realized
Status	draft (admitted as a Build at tier P; honors the catalog’s prior cand / build / P tag, which the research run corroborated rather than overturned)

---

1. The mechanism (what actually does the work)

Morphological analysis (the “Zwicky box” or morphological box) lays out the whole solution space of a multi-dimensional problem and then searches it systematically. You decompose the problem into a small set of independent PARAMETERS (the dimensions any solution must specify), enumerate the discrete VALUES each parameter can take, and arrange them as the columns and cells of a field. A candidate solution is one value chosen from each parameter - a single path through the box - and the set of all such paths is the configuration space. The promise is coverage: instead of jumping to a few familiar designs, you force the full Cartesian product of options into view, including the unobvious corners no forward search would have visited.

The durable move has two halves, and the second is what separates the method from a plain options list:

1. Parameterize and enumerate. Reduce the problem to N orthogonal parameters with their value sets, giving the full combinatorial space (5 parameters with a handful of values each is already tens of thousands of configurations).
2. Cross-consistency assessment (CCA), the reduction step. Because most of those configurations are nonsense, you pair off every value with every other value across parameters and strike out the internally incompatible pairs. Tom Ritchey named this “inference by exclusion”: a field of 100,000 raw configurations routinely collapses by 90-99% to a small, internally consistent residual set worth examining.

The artifact is therefore not a tree and not a ranked list - it is a parameter-by-value field plus a consistency-pruned set of viable configurations. This is the feature that distinguishes morphological analysis from its neighbors: it is the only method in the family whose product is a combinatorial space that has been DECOMPOSED into independent axes and then RECOMBINED and pruned, rather than a set of options generated by provocation, negation, transformation, or analogy.

2. Lineage

The method is the work of Fritz Zwicky (1898-1974), the Swiss-American astrophysicist at Caltech (better known for dark matter and supernovae), who developed the “morphological approach” from the 1940s-1960s to structure problems in astrophysics and jet/rocket propulsion, and set it out in Discovery, Invention, Research Through the Morphological Approach (1969). Read Zwicky for the origin and the totalizing ambition - but note that his “100x” and “philosopher’s stone” claims are rhetoric, not evidence.

The modern methodological authority is Tom Ritchey of the Swedish Morphological Society / Swedish Defence Research Agency, who computerized the method, formalized the cross-consistency assessment as “inference by exclusion,” and extended General Morphological Analysis (GMA) toward “wicked” / policy problems.

“Morphological analysis” is a generic descriptive term in common use - Zwicky’s method is unbranded and untrademarked - so this entry is documented descriptively with attribution to Zwicky (and to Ritchey for the modern CCA formalization), not flagged as branded.

3. What the evidence shows, and what it does NOT show

The honest grade for the move - enumerate the solution space by parameter combination and prune to the consistent set - is P (practitioner). Morphological analysis is an unusual case for this library because controlled experiments DO exist, yet they push the grade toward P rather than up, for three reasons: they measure a lighter sibling (the morphological CHART, without Ritchey’s cross-consistency step), the headline results are mixed-to-unflattering, and they are run on human design students, never on AI agents.

What the record supports. Morphological analysis is a real, named, long-lived method with a clear lineage (Zwicky 1969; Ritchey from 1998) and a large DESCRIPTIVE base of application: Alvarez and Ritchey (2015) catalogue roughly 80 published GMA applications across engineering design, technology forecasting, and policy. As a structuring discipline it is genuinely used and teachable. The morphological-chart variant has been put through real controlled studies in engineering-design research, which is more than most catalog entries can claim.

What the controlled studies actually found, and why it does NOT lift the grade. The strongest experiment, Daly, Seifert, Yilmaz and Gonzalez (2016, n = 102 first-year engineering students), compared morphological analysis against individual brainstorming and design heuristics. Morphological analysis raised the ELABORATION of concepts - but it was NOT best on any headline outcome: plain brainstorming produced the most concepts, and design heuristics (not morphological analysis) produced the most PRACTICAL ones. So the best direct comparative evidence shows the method is respectable but not dominant, and beaten on practicality by a sibling technique. Smith, Troy and Summers (2012) ran chart-design experiments and found that ADDING functions to the chart did not improve concept quality and that designers explored only about a quarter to a seventh of the available configuration space - empirical confirmation that the combinatorial-coverage promise is largely unrealized in use. Richardson and Summers (2011) found the variety and novelty of means did not depend on how functions were represented. Heller, Feldhusen and colleagues (2016) document the “dilemma”: arbitrary parameter selection and combinatorial explosion can make the method either useless or unfeasibly effortful. None of these measures the distinctive cross-consistency-assessment reduction step on its own; the CCA’s strong “90-99% reduction” descriptions come from Ritchey’s own methodological writing, not from controlled validation.

What the record does NOT support. No nameable primary source establishes a quantitative effect of morphological analysis on decision or solution QUALITY. The one strong comparative number is the unflattering one - it lost on practicality to design heuristics. The critical literature also names a structural risk directly: a morphological box “can easily be used to justify the designers’ preferred solution without covering the intended complete solution space,” which makes a poorly-decomposed exercise theatrical rather than exhaustive.

4. Transferred-evidence flag (required honesty for this library)

Every study above is on human subjects - engineering students or design practitioners - in design-ideation settings. None studies morphological analysis performed by or with an AI agent, and the most-studied object is the morphological CHART, not Zwicky/Ritchey’s full GMA with cross-consistency assessment. The evidence is transferred from human design contexts and from a sibling variant, and is not validated for AI-augmented use. That double transfer (human-to-agent, and chart-to-full-GMA) is exactly why the honest grade is held at P and not raised to M.

The AI value is mechanical and modest: an agent makes the method cheap to run, forces the discipline (a genuine parameter decomposition, an actual cross-consistency pass, an inspectable residual set), and produces a durable artifact - benefits that do not depend on any contested outcome claim. The skill ships honestly as a P-tier coverage-and-consistency aid with a hard “the parameterization is only as good as the analyst, and the box is not the evaluator” wall, never as a guaranteed exhaustive search or a decision-quality booster.

5. Excluded figures (required)

• Zwicky’s own claim that the morphological method could “increase the efficiency of our brains by 100 times,” and his “philosopher’s stone” framing, are author self-promotion with no measurement behind them and are excluded - they may not move the grade.
• The frequently-quoted “a field reduces by 90-99%” is a methodological description of the procedure from Ritchey, not a measured outcome of decision quality, and is reported as a property of the procedure, not as evidence of effectiveness.
• No nameable primary source establishes a positive quantitative effect of morphological analysis on solution or decision QUALITY; the one strong comparative result (Daly et al. 2016) has it losing on practicality to design heuristics.

6. When it works / when it fails (drives the eval negative cases and “When NOT to Use”)

Works best when:

• The solution genuinely is a CONFIGURATION - a choice on each of several semi-independent dimensions (a product architecture, a service bundle, a policy package, a scenario set) - and the risk is tunnel vision: defaulting to one familiar combination and never seeing the rest of the space.
• The parameters are real and discoverable, and some value pairs are clearly incompatible, so the CCA does real pruning.
• The goal is breadth-with-structure rather than a single quick idea. It pairs naturally with a downstream evaluation step: generate the consistent configurations here, then score and choose them with a weighted decision review.

Fails or misleads when (poor-fit / anti-patterns):

• The parameters are arbitrary. The method’s validity depends entirely on the decomposition, and nothing in it forces a good one. A box “can easily be used to justify the designers’ preferred solution without covering the intended complete solution space.” Garbage axes give a garbage space that merely looks systematic.
• The combinatorial explosion is left unmanaged. Without serious pruning the space is too large to inspect - the experimental record shows designers actually examine only a fraction of it (roughly a quarter to a seventh in controlled studies), so the “complete coverage” claim quietly fails in practice. Adding more parameters to feel thorough makes this worse, not better.
• The problem is not configurational. If the answer is a single insight, a reframe, or one number, forcing it into orthogonal axes is overhead. The method assumes the solution factorizes; many do not.
• It is treated as the evaluator. The box generates and prunes-for-consistency; it does NOT rank by value. Reading a surviving configuration as “the chosen answer” skips the trade-off comparison it was never designed to make.

7. Distinctness (why it is a Build, and against which shipped neighbors)

Verdict: Build, at the conservative governing tier P. The Build burden is to name one distinct, durable cognitive move that no shipped skill produces, name the closest shipped skill, and show that neither a mode nor a short chain of existing skills already produces it above the overlap ceiling. Morphological analysis clears that burden because its move - decompose a solution into independent parameters, enumerate the full cross-product of their values, then prune to the internally consistent configurations - and its artifact - a parameter-by-value morphological field plus a consistency-pruned configuration set - are produced by nothing in the catalog:

• The closest shipped skill in its own family is think-assumption-reversal, and the wall is clean. Assumption-reversal negates a problem’s foundational premises to provoke fresh options; its product is a provoked option list. Morphological analysis negates nothing and provokes nothing - it builds a structured combinatorial space and reduces it by consistency. They share only the abstract goal “expand the option set,” which is well under the overlap ceiling. The same disposes of think-scamper (seven fixed transformation verbs on one seed - no decomposition, no cross-product) and think-far-analogy-ideation (imports deep structure from a distant domain - no analogy here). The “invert to bad ideas then flip” provocation is likewise a provocation, not an enumeration; this library ships no skill for it, so it is not a routing target here.
• The closest STRUCTURAL cousin anywhere in the catalog is think-issue-tree (synthesis), and it is a different machine. An issue-tree decomposes one question top-down into a MECE hierarchy you read as a tree; morphological analysis decomposes into PARALLEL, orthogonal axes and then RECOMBINES them into a cross-product, pruning inconsistent pairs. Decomposition is the shared part; the recombination-and-consistency-pruning - the entire generative heart of the method - has no analogue in a tree. A tree never multiplies its branches together, and never runs a cross-consistency pass. So issue-tree does not subsume it, and the gap is exactly the distinctive part.
• think-decision-option-review is downstream, not overlapping. A weighted decision matrix SCORES given options against criteria to pick one; the morphological literature is explicit that the box runs FIRST to generate the candidate configurations and a decision/Pugh matrix runs AFTER to evaluate them. Generating-by-combination and scoring-to-choose are different operations on different artifacts (a configuration field versus an options-by-criteria scorecard). No overlap that threatens the ceiling.
• It is not a recipe. The cross-product enumeration plus the cross-consistency assessment is one integrated apparatus, not a fixed “skill A then skill B” chain of shipped moves. The nearest chain one could propose - issue-tree to decompose, then decision-option-review to evaluate - still never produces the parameterized cross-product or the consistency pruning, so the move is not an emergent property of any sequence the catalog already owns. Build, not Recipe; and since no shipped mode emits the field, Build, not Fold.

The learning value of the YES: where inversion turned out to be a famous STANCE that every concrete instantiation already owned, morphological analysis is a less-famous METHOD with its own irreducible artifact - a consistency-pruned combinatorial field - that the catalog genuinely lacks. It ships at P with eyes open: the controlled evidence is real but does not crown it, the coverage promise is often unrealized, and the parameterization is only as good as the analyst, so the when-NOT-to-use wall (non-configurational problems, arbitrary axes, treating the box as the evaluator) is load-bearing.

8. Output artifact

The skill must emit a morphological field, not prose: the focal problem in one line; the parameters (the independent dimensions) and the discrete values of each (the box); the cross-consistency assessment (which value pairs across parameters are incompatible); and the pruned set of internally consistent configurations (the surviving candidates), with a rough before/after size to show the pruning did real work. The configurations are handed off as candidates for a separate evaluation step and are never ranked by value within the field.

9. Sources

Named sources

• Fritz Zwicky, Discovery, Invention, Research Through the Morphological Approach (Macmillan, 1969). The foundational statement of the method and the morphological box. Foundational / practitioner; the “100x brain efficiency” and “philosopher’s stone” claims are self-promotion and excluded as evidence. (P)
• Tom Ritchey, “General Morphological Analysis: A general method for non-quantified modelling” (Swedish Morphological Society; presented 1998, revised editions on swemorph.com). Defines modern GMA and the cross-consistency assessment / inference-by-exclusion reduction step. Practitioner / methodological. (P)
• Tom Ritchey, “Principles of Cross-Consistency Assessment in Morphological Modelling,” Acta Morphologica Generalis (2015). The detailed account of the CCA pruning move and the 90-99% reduction description - a methodological description, not a measured effectiveness outcome. Practitioner / methodological. (P)
• Aleksander Alvarez and Tom Ritchey, “Applications of General Morphological Analysis: From Engineering Design to Policy Analysis,” Acta Morphologica Generalis 4(1) (2015). Catalogues ~80 published GMA applications - a descriptive, selection-biased case base, not controlled evidence. Practitioner. (P)
• Shanna R. Daly, Colleen M. Seifert, Seda Yilmaz and Richard Gonzalez, “Comparing Ideation Techniques for Beginning Designers,” Journal of Mechanical Design 138(10): 101108 (2016). Controlled study, n = 102 students: morphological analysis vs brainstorming vs design heuristics. MA raised concept elaboration but was NOT best on practicality (design heuristics won) or quantity (brainstorming won). The strongest direct comparative evidence; measures the chart variant on human students, not AI agents. (M for the experiment’s internal rigor; counts toward this entry only as P-level support because it measures a sibling variant on humans and does not favor the method.)
• Gregory Smith, Timothy J. Troy and Joshua D. Summers, “Concept Exploration Through Morphological Charts: An Experimental Study,” Journal of Mechanical Design 134(5): 051004 (2012). Two design experiments on chart size/shape: adding functions did not improve concept quality, and designers explored only ~1/4 to ~1/7 of the configuration space - the coverage promise is largely unrealized in practice. Experimental, human students; on the chart variant. (M for rigor; P-level support here, same transfer caveat.)
• Bart R. Richardson and Joshua D. Summers (2011), experimental work finding that the variety and novelty of means did not depend on how functions were represented in the chart. Experimental, human students; on the chart variant. (P-level support, same transfer caveat.)
• Patrick Heller, Jorg Feldhusen et al., “Rethinking Morphological Analysis Application for Concept Synthesis in Engineering Design,” Athens Journal of Technology and Engineering 3(2) (2016). Critical analysis: arbitrary parameter selection lets the box rationalize a preferred solution, and combinatorial explosion forces a useless-or-unfeasible trade-off. Critical literature. (P)

Excluded under the evidence rule: Zwicky’s “increases brain efficiency 100x” / “philosopher’s stone” claims (self-promotion, no measurement) and the “90-99% space reduction” property (a procedural description from Ritchey, not a measured decision-quality outcome) are not counted toward the grade. No nameable primary source establishes a positive quantitative effect of morphological analysis on solution or decision QUALITY; the one strong comparative result (Daly et al. 2016) has it losing on practicality to design heuristics.