Analogy¶
Core Idea¶
Analogy is structural mapping between two domains — the source domain (familiar, understood, reference system) and the target domain (unfamiliar, to-be-understood, recipient of inference) — in which the structural mapping (relational role-alignment) projects the relations and roles among source elements onto target elements, so that inferences available in the source domain become candidate inferences in the target domain. The abstraction has six core structural specifications: (1) there are two distinct domains with their own elements, relations, and causal structure; (2) the relational alignment operates not by surface similarity but by the roles elements play in the underlying structure (the Sun maps to the atomic nucleus because both occupy the "center with smaller bodies orbiting" role, not because they are spatially or materially similar); (3) the systematic mapping preserves higher-order relations—causal, functional, dependency—so that valid inferences in the source become conjectures in the target (if gravitational attraction holds bodies in orbit, then perhaps an analogous attractive force holds electrons in orbit); (4) the systematicity principle (Gentner 1983) states that analogies with richer, deeper, and more interconnected relational structures are stronger and more cognitively useful than those based on isolated feature matches; (5) the candidate inference emerges from the projection: given the mapping, what can we infer about the target that we could not infer from the target alone; (6) the analogy is evaluable—not all mappings are equally good, and analogies succeed or fail based on the depth and systematicity of the relational correspondence. [1] The structure-mapping theory of Dedre Gentner (1983) gave the modern formal account, distinguishing analogy (relational mapping) from mere-appearance matching (surface similarity) and literal similarity (relational plus surface). [1] Analogical reasoning is a central topic in cognitive science, AI, philosophy of science, and creativity research, with formal models including constraint-satisfaction approaches and neural-symbolic architectures.
How would you explain it like I'm…
Like-This-Like-That
Same-shape match
Role-to-role mapping
Structural Signature¶
A directed relational mapping from source to target that preserves the relational structure (especially higher-order causal and functional relations) while discarding surface attributes, generating transferable inferences that would not be available from either domain considered alone. The signature is the combination of (a) relational role-alignment—elements of the source are matched to elements of the target based on their functional roles in the respective domains, not their superficial properties; (b) selective preservation of higher-order structure—causal chains, dependencies, and functional relationships transfer, while concrete attributes (color, material, scale) do not; © inferential transfer to the target—inferences drawn in the source about relational dependencies become conjectures about the target; (d) the surface-vs-structural-similarity distinction—a principled rejection of shallow feature-matching in favor of deep relational correspondence; (e) the systematicity principle—analogies are stronger when the source's relational structure is interconnected and forms a coherent system, and when multiple independent relations all support the same target mapping (convergence is cognitively powerful); (f) the analogical retrieval process—the cognitive or computational mechanism that selects the source domain in the first place, which depends on structural similarity cues and prior experience. [2] The signature distinguishes analogy from similarity judgment (which asks "how alike are A and B?") and from metaphorical assertion (which makes a semantic claim of identity).
What It Is Not¶
Analogy is not simile (#220): a simile is a compact, explicit single-attribute comparison marked by "like" or "as" that foregrounds one shared feature for rhetorical effect; analogy is a deeper relational mapping that preserves a system of correspondences, not just one feature. Simile is structurally simpler and verbally marked; analogy is structurally richer and often implicit. It is not metaphor: metaphor is a figurative assertion of identity ("the mind is a computer") that can have analogical depth but is primarily a semantic and rhetorical device operating in language; analogy is the cognitive operation of relational mapping, which underlies many metaphors but also operates in non-linguistic reasoning (case-based reasoning, scientific modeling, problem-solving). Metaphors without systematic relational mapping are not analogies; analogies without metaphorical expression (e.g., transferring a mathematical solution from one problem to a structurally similar one) are still analogies. It is not surface similarity or appearance matching: the critical distinction is that analogy privileges relational correspondence. The solar-system model of the atom is an analogy; saying "atoms are round" because small marbles are round is mere appearance matching with no inferential power. It is not identity or equivalence: an analogy is a partial mapping with acknowledged disanalogies; treating analogies as identities produces category errors (mistaking "the brain is like a computer" for "the brain is a computer"). It is not induction or deduction as such: analogy generates conjectures by relational projection, which is a distinct inferential form, though analogical inferences can support or be tested by inductive and deductive reasoning.
Broad Use¶
Science and engineering use analogy pervasively for hypothesis generation and explanation: the hydraulic analogy for electrical circuits (voltage = pressure, current = flow, resistance = narrow pipe) teaches circuit behavior through a familiar mechanical system; Rutherford's atom-as-solar-system, Maxwell's vortex models for electromagnetic fields, and Darwin's selection-as-breeding analogy each generated productive research programs. [3] [4] Biomimicry explicitly maps biological designs onto engineered artifacts (gecko adhesion → dry adhesives, lotus leaf surface → self-cleaning coatings, bird wing → airfoil). Mathematics uses analogy to transfer proof techniques across structurally similar problem classes, and category theory formalizes structural correspondence as a universal construct. [5] [6] Law relies on analogy as its primary inferential form (precedent reasoning projects the structure of decided cases onto new cases by role-alignment of facts and parties). [7] Case-based reasoning in AI stores past solutions and retrieves them by relational similarity to solve new problems. Teaching and pedagogy explain unfamiliar concepts through analogies with familiar ones (computer memory as a filing cabinet, DNA as a recipe, evolution as tinkering). [8] Creative problem-solving exploits distant analogies to import solution patterns from one domain into another (the Velcro invention from burrs, synectics methodologies in design). [9] Policy analysis, history, and comparative social science use analogies between historical and current situations to project outcomes and risks, with varying degrees of rigor. [3] [10]
Clarity¶
The abstraction clarifies that effective analogies operate at the level of relational structure, not surface appearance, which resolves a persistent confusion about what makes an analogy strong or weak. It separates the deep-structural operation (role-alignment and relation-preservation) from the rhetorical surface (whether the analogy is explicit, metaphorical, or buried). It also clarifies that analogies are partial mappings with disanalogies — they are tools for inference, not statements of equivalence — which prevents both over-reliance (treating the analogy as a proof) and over-dismissal (rejecting useful inferences because the analogy is "imperfect"). The concept of higher-order relation preservation, introduced by Gentner and colleagues, gives a principled basis for preferring deep causal/functional analogies over shallow featural comparisons.
Manages Complexity¶
A novel target domain has no pre-organized inference structure, so reasoning about it from first principles requires building models, generating hypotheses, and testing them in isolation — a computationally and cognitively expensive process. Analogy compresses this by importing the entire relational scaffold of the source domain in a single mapping, giving the reasoner an immediate hypothesis space, conceptual vocabulary, and set of candidate inferences to test. The cost is the disanalogy-filtering burden (identifying which source inferences do and do not transfer), but this is generally far less than starting from nothing. The same compression operates at the individual-learner level: a student who sees the river-flow analogy for electrical current imports years of intuitive fluid dynamics into a new domain in seconds.
Abstract Reasoning¶
Analogy surfaces a general pattern — transferring relational structure from a known system to an unknown one via role-alignment, preserving higher-order relations — that is itself a meta-abstraction unifying several reasoning modes. The same structural pattern recurs in: category theory and functor preservation (mapping structure between mathematical categories); type systems in programming languages (generic types reused across differently-typed instances by structural correspondence); transfer learning in machine learning (pretrained models transferred to new tasks by fine-tuning, where the learned representational structure is the analogical source); precedent and stare decisis in common law; modeling and simulation in science (models are analogs of target systems with preserved relational structure). The reasoning unit is relational-structure transfer — the cognitive and formal operation that makes reasoning across domains possible at all.
Knowledge Transfer¶
Role-mapping table:
| Role in analogical reasoning | Counterpart in transfer learning |
|---|---|
| Source domain | Pretraining domain (e.g., ImageNet, large text corpus) |
| Target domain | Downstream task (e.g., medical imaging, domain-specific NLP) |
| Source elements / relations | Pretrained model weights encoding learned features / relations |
| Role-mapping | Architectural reuse + fine-tuning alignment |
| Higher-order structural correspondence | Shared feature hierarchies (edge detectors, syntactic structure) |
| Inferential transfer | Fine-tuned performance on target task |
| Disanalogies | Domain gaps requiring adaptation or retraining |
| Systematicity bias | Preference for transferring deeper layers over surface features |
| Analogical failure | Negative transfer — source structure hurts target performance |
Transfer paragraph: the practical transfer is that deep learning's transfer-learning paradigm is analogical reasoning formalized at scale. A pretrained model carries relational structure (learned features and their compositions) from a source domain; the fine-tuning process performs role-alignment (mapping the model's learned representations to the target task's structure); the success of transfer depends on systematicity — how well the source's higher-order relational structure aligns with the target's. When alignment is strong (e.g., natural-image features transferred to medical images), transfer is powerful and data-efficient. When alignment fails (e.g., natural-language features applied to tabular financial data), negative transfer can occur, where the source structure actively harms target performance. The same analogy-strength dynamics that cognitive scientists identified in human reasoning (depth of relational correspondence predicts utility of transfer) are precisely what determines which pretraining regimes generalize to which downstream tasks. Understanding transfer learning as machine analogy makes the conditions of its success and failure predictable.
Example¶
Formal scientific example (Rutherford's solar-system model of the atom, 1911): Ernest Rutherford's analogy from the solar system to atomic structure is a textbook case of relational mapping. The mapping was: Sun → nucleus; planets → electrons; gravitational force → electrostatic attraction; orbital motion → electron motion; central-mass concentration → nuclear-charge concentration. The surface attributes did not match (atoms are not hot, luminous, or planetary-scale; nuclei are not made of plasma) — the relational structure did. Rutherford's analogy made predictions available in the target domain that were otherwise inaccessible: that atoms have a compact positively-charged center with near-zero volume relative to the atom, that most of the atom's mass is concentrated there, and that most of the atom's volume is empty space. The gold-foil scattering experiment confirmed the relational structure. Importantly, the analogy also exhibited disanalogies that later physics had to address: electrons do not follow classical orbital trajectories (quantum-mechanical corrections were needed), and the analogy had to be replaced with quantum-mechanical models for finer-grained prediction. But the initial analogical import of the solar-system's relational scaffold was the conceptual breakthrough that organized early atomic physics. Structurally, this is prototypical analogy: relational role-alignment from source to target, higher-order relation preservation (centrally-concentrated-attractive-force generates bounded-orbits), inferential productivity, and partial disanalogy.
Structurally faithful non-formal example (software-engineering design-pattern transfer): A developer designing a configuration system for a new application recognizes that the domain is structurally similar to a previously-solved problem: the Observer pattern in UI event handling. The role-mapping is: subject → configuration store; observers → application components using the configuration; notify → configuration-change broadcast; state-change → configuration-update. The relational structure is preserved: loose coupling via subject/observer separation, single-source-of-truth with propagation to dependents, extensibility through observer addition without subject modification. The developer imports the entire solution scaffold (pattern structure, known pitfalls, typical implementation) from the familiar source to the unfamiliar target, saving the cost of re-deriving the solution. Disanalogies are noted and handled: configuration changes often require atomic updates, unlike single-event UI notifications, so a transactional wrapper is added; configuration dependencies can be cyclic, unlike typical UI hierarchies, so cycle-detection logic is added. The structural pattern is identical to Rutherford: relational role-alignment, higher-order relation preservation, inferential transfer, and explicit handling of disanalogies. The emergent benefit is a working design in a fraction of the time first-principles derivation would require.
Structural Tensions and Failure Modes¶
T1 — Surface-similarity trap. Analogies that rely on surface feature matches rather than relational structure produce shallow and often misleading inferences (e.g., "both have wings" to compare insect and bird flight without attention to lift mechanisms). The failure mode is analogies that feel compelling because of surface resonance but lack inferential productivity or import incorrect conjectures. The corrective is explicit attention to whether the correspondence is at the level of surface features or relational structure, with preference for the latter.
T2 — Disanalogy neglect and analogical overreach. Analogies are partial mappings, and every analogy has points where the correspondence breaks. The failure mode is treating the analogy as total equivalence and importing source inferences that do not hold in the target (e.g., running "the brain is a computer" to infer that the brain must have a separate memory and CPU, which it does not). The corrective is explicit disanalogy identification — naming the ways in which source and target differ — before acting on analogically-derived inferences.
T3 — Analogy as rhetoric vs. analogy as inference. Analogies serve two functions: persuasion/communication (rhetoric) and conjecture-generation (inference), and these have different quality criteria. Rhetorical analogies can be evocative without being inferentially sound; inferential analogies must be structurally faithful even at the cost of rhetorical elegance. The failure mode is using rhetorical analogies as if they were inferentially warranted, which is a common source of policy errors (historical analogies for political decisions) and scientific misdirection (metaphors that organize thinking poorly). The corrective is explicit role-labeling: are we using this analogy to communicate or to infer?
T4 — Entrenched analogy lock-in. A successful analogy, once adopted, often dominates thinking in a field and suppresses alternative framings that might be more productive (the computer analogy for mind dominated cognitive science for decades, obscuring embodied and dynamical alternatives). The failure mode is monoculture of analogy — a single mapping becomes load-bearing in a way that prevents the field from considering structurally different framings. The corrective is cultivating a repertoire of alternative analogies and periodically testing whether the dominant framing is still the most productive.
T5 — Surface similarity as seduction. Analogies that rely on salient surface-feature matches rather than deep relational structure produce superficially compelling but inferentially weak mappings. [11] For example, "gravity attracts bodies just as love attracts people" shares the surface feature "attraction" but lacks systematic relational correspondence—gravitational force follows an inverse-square law, is instantaneous at classical speeds, is proportional to mass, and has no learning or volition, while romantic attraction has none of these properties. [11] The failure mode is the false-positive analogy: it feels right because of surface resonance but imports incorrect or useless conjectures into the target. The corrective is explicit structural scrutiny: does the mapping preserve relations beyond the initial surface match, and do those relations form a coherent system?
T6 — Generative versus convergent analogy. Some analogies open up new exploration spaces and invite multiple alternative framings of a problem (suggestive, productive analogies); others impose structure so forcefully that they lock reasoning into a single trajectory and suppress competing interpretations. [3] Hofstadter and Sander (2013) emphasize analogy as generative—the source domain provides not a unique solution but a space of possible interpretations and extensions. However, once a mapping is adopted, it can become convergent—a single dominant reading that constrains perception. The failure mode is mistaking a productive initial analogy for a closing of the question; the corrective is maintaining flexibility and testing whether alternative analogies might be equally or more productive, especially in research and design contexts where the goal is novelty rather than closure.
Structural–Framed Character¶
Analogy sits at the structural end of the structural–framed spectrum: it is a pure relational pattern, the same in any domain where it appears, and nothing about its meaning depends on a particular field's vocabulary or assumptions.
The prime names a directed mapping from a source domain to a target domain that preserves relational and causal structure while discarding surface features, so that inferences available in the source become candidate inferences in the target. That mapping operates identically whether the domains are physical systems, mathematical structures, or social situations. It carries no evaluative weight, and its definition is formal — a structure-preserving alignment of roles and relations — with no appeal to human institutions. Applying it feels like recognizing a shared structure already there. On every diagnostic, it reads structural.
Substrate Independence¶
Analogy is about as substrate-independent as a prime can be — composite 5 / 5 on the substrate-independence scale. Its signature is fully substrate-agnostic — a relational-role alignment that maps source to target, preserving higher-order relations while discarding surface attributes. As a core reasoning mechanism it spans cognitive science, philosophy, rhetoric, mathematics, and effectively every knowledge domain, since it is the very engine of cross-domain transfer and learning. The examples are sparse in the input, but the concept is foundational enough — abstraction is what makes substrate-independence possible at all — that it sits comfortably among the canonical 5s.
- Composite substrate independence — 5 / 5
- Domain breadth — 5 / 5
- Structural abstraction — 5 / 5
- Transfer evidence — 4 / 5
Relationships to Other Primes¶
Parents (2) — more general patterns this builds on
-
Analogy is a kind of Comparison
Analogy is a kind of comparison specialized by its alignment mechanism: the items compared are entire domains, and the alignment rule projects structural roles and relations from a source domain onto a target domain rather than matching surface features. It inherits comparison's commitment to placing items under a shared frame with an alignment rule producing a relational output, and supplies the specific case where commensurability is achieved through relational role-correspondence, licensing inferences in the target by transporting the source's relational structure across the mapping.
-
Analogy presupposes Abstraction
Analogy presupposes abstraction because mapping between two domains by relational role rather than surface similarity requires those roles to have been abstracted from their concrete embedding in each domain. Without abstraction's purpose-relative retention of structure — selecting load-bearing features while discarding the rest — the source domain's relational skeleton could not be lifted from its specific contents to project onto the target. Abstraction supplies the structure-extracting operation that makes the source's roles transportable; analogy supplies the cross-domain mapping that uses those abstracted roles to license inferences in the target.
Children (4) — more specific cases that build on this
-
Metaphor is a kind of Analogy
Analogy is the general pattern of structural mapping between two domains in which relational roles align so that inferences in the source become candidate inferences in the target. Metaphor is the specific case in which the source domain is typically concrete and embodied, the target is typically abstract, and the mapping organizes how the target can be thought about and talked about at all. It inherits analogy's relational-alignment machinery but commits additionally to constitutive cognitive-linguistic projection — a specialization of analogy applied to thought and discourse.
-
Simile is a kind of Analogy
Simile is a kind of analogy specialized to surface-marked, single-attribute mapping: an explicit linguistic marker — like, as, resembles — flags the mapping between tenor and vehicle, and the projection typically foregrounds one shared attribute rather than a full relational system. It inherits analogy's commitment to structural mapping between two domains where source features support inferences in the target, and supplies the specific case where the mapping is overtly marked in language and bounded in scope, contrasting with the unmarked, deeper relational mapping of metaphor and full analogy.
-
Conceptual Blending presupposes Analogy
Conceptual blending presupposes analogy because constructing a blended space by selectively projecting elements from two or more input spaces requires a cross-space mapping that identifies correspondences between elements — exactly the structural role-alignment that analogy supplies. Without analogy's relational mapping between source and target domains, there would be no correspondence structure to feed selective projection, and the blend could not preserve a coherent generic space capturing what the inputs share. Analogy supplies the mapping operation; blending adds the further move of constructing a new integrated space with emergent structure not present in either input alone.
- Precedent (Stare Decisis) presupposes Analogy
Stare decisis treats a prior case's outcome as presumptively binding on the present case when the cases are materially similar, and the work of identifying material similarity is analogical reasoning. Analogy provides the structural-mapping machinery — relational role-alignment between source and target — that lets the court project the rule from the precedent case onto the new fact pattern, and that lets the craft of distinguishing operate. Without analogical mapping, there is no way to determine which prior cases govern, so precedent cannot function without analogy as its inferential engine.
Path to root: Analogy → Comparison
Neighborhood in Abstraction Space¶
Analogy sits among the more crowded primes in the catalog (16th percentile for distinctiveness): several abstractions describe nearly the same structure, so a description that fits it will tend to fit its neighbors too — transporting it usually means disambiguating within this family rather than landing on it exactly.
Family — Language, Symbol & Cultural Form (32 primes)
Nearest neighbors
- Juxtaposition — 0.84
- Comparison — 0.83
- Paradigmatic vs. Syntagmatic Relations — 0.83
- Simile — 0.82
- Icon–Index–Symbol Distinction — 0.81
Computed from structural-signature embeddings · 2026-05-29
Not to Be Confused With¶
Analogy must be distinguished from Metaphor, though they are closely related and often co-occur. Metaphor is a linguistic or cognitive assertion of identity through implicit comparison—"the mind is a computer" is a metaphor that claims identity (not just similarity) between mind and computer, importing the computer's conceptual structure to shape how we think about the mind. Metaphor is primarily a semantic device operating in language, rhetoric, and conceptual blending. Analogy is the structural correspondence that underlies many metaphors but is logically independent of metaphorical expression. A metaphor without analogical depth (superficial feature matching with no relational correspondence) is not a strong analogy; an analogy without metaphorical wrapper (structural mapping used in mathematical problem-solving or case-based reasoning) is still a powerful analogy. The distinction is that metaphor is a utterance form (saying "A is B") that may carry analogical depth; analogy is the structural operation of role-alignment that may be expressed metaphorically. Metaphors can be rhetorically powerful without being informationally valuable, and analogies can be informationally sound without being rhetorically compelling. A psychologist transferring a mathematical solution from one domain to another is using analogy without metaphor; a poet using a metaphor without structural correspondence is using metaphor without analogy.
Nor is analogy equivalent to Paradigmatic vs. Syntagmatic Relations, which describe structural relations within a domain or language system. Paradigmatic relations are contrasts—the units that can substitute for each other in a given position (in English, "cat," "dog," "bird" are paradigmatically related as animals that can fill the subject-noun role). Syntagmatic relations are sequential or combinatorial—how units combine in sequence or structure (how nouns, verbs, and adjectives combine to form sentences). Both are within-domain relations: they describe how elements of a single system relate to each other. Analogy is an across-domain correspondence that maps elements from one domain to another based on relational roles. Paradigmatic and syntagmatic analysis would describe a language system's internal structure; analogical reasoning uses that structure as a template to map onto a different domain. The distinction is that paradigmatic-syntagmatic describes the organization of a single system; analogy bridges systems.
Analogy is distinct from Inversion, the reversal or flipping of relationships or hierarchies. Inversion asks "what happens if I reverse this relationship?"—if A dominates B, what if B dominates A instead? Analogy asks "what role does A play in system X, and what plays that role in system Y?"—role-correspondence rather than reversal. Inversion is a unidirectional transformation; analogy is bidirectional mapping preserving structure. An inverted hierarchy is a different hierarchy; an analogical mapping is a correspondence between hierarchies that preserves relational structure. A leader and team member could invert their roles (team member becomes leader); in an analogy between business management and molecular biology, a manager might map to a regulatory protein and an employee to regulated proteins—this is role-mapping, not inversion. The confusion arises because both involve transformation, but inversion flips direction while analogy preserves direction through relational role-alignment.
Analogy is also not Representation, the use of a symbol or proxy to stand for something else. Representation is about substitution—using X to stand in for Y because X is more convenient, more manipulable, or more accessible. A map represents a territory; a model represents a system. Representation focuses on the proxy's fidelity to what it represents. Analogy is about structural correspondence—mapping roles and relations between two domains to enable inference transfer. A map of a territory is a representation (fidelity matters); mapping territories in order to transfer knowledge about cartography from one region to another is analogical reasoning (role-correspondence matters). A mathematical model that represents physical behavior is a representation (the model stands in for the physical system); using the model's structure to analogize to a different physical system is analogy. The relationship is that representations can enable analogies (because the representation has tractable structure), but they serve different purposes: representation is about substitution for analysis or communication; analogy is about structural correspondence for reasoning transfer.
Finally, analogy is not Transfer of Learning, though analogy is a primary mechanism enabling transfer. Transfer of learning is the application of knowledge acquired in one context to a new context—a skill learned in school applied at work, a principle learned in one course applied in another. Transfer is the learning process or application outcome. Analogy is the relational structure that makes transfer possible. Not all transfer is analogical (rote copying of a procedure to a new context is transfer without analogy); not all analogies are deployed in learning contexts (mathematical analogy for pure reasoning is not about transfer of learning but about problem-solving). The relationship is that analogical reasoning is a primary mechanism of successful transfer, but the two are not equivalent. A training program that enables positive transfer has enabled learners to recognize structural correspondences and apply them; that success is grounded in analogical reasoning, but the program is labeled "transfer" not "analogy." Conversely, a perfectly valid analogy in a specialist domain might never be deployed for learning transfer because the reasoner uses it only for conjecture generation in research.
Solution Archetypes¶
Solution archetypes in the catalog that build on this prime — directly (this prime is a source ingredient) or as a related prime.
Built directly on this prime (3)
Also a related prime in 7 archetypes
- Ambiguity-Exploitation in Visual Metaphor
- Archetype Overmatching Guardrail
- Archetype Pattern Indexing
- Contrastive Differentiation
- Design-Principle Extraction and Reapplication
- Literal-vs-Figurative Boundary Preservation
- Symbol-System Coherence in Visual Art
Notes¶
Fifth draft of batch 11. Tight pair with #220 simile (simile is the minimal marked single-attribute form of analogical comparison; analogy is the full relational-structure transfer). Thematic link to #221 metaphor (pilot draft; metaphor and analogy overlap substantially in cognitive function, with metaphor being primarily linguistic-rhetorical and analogy being primarily relational-structural, though major theories including conceptual metaphor theory treat them as deeply continuous). Also linked to #222 metaphor_visual_artistic (visual-art metaphors draw on the same relational-mapping operation). The transfer-learning mapping is structurally faithful because deep learning's transfer paradigm implements analogical structure transfer at scale. Gentner's structure-mapping theory (1983) provides the formal backbone for the definition; Hofstadter and Sander's Surfaces and Essences (2013) treats analogy as the foundation of cognition, a position aligned with but stronger than the claims made here. Origin noted as cognitive_science with substantial alternate-domain contributions from philosophy (Aristotle's Analogia, Hume, Mill on inference by analogy), rhetoric (Aristotle's Rhetoric), and mathematics (formal analogy in category theory and model theory).
References¶
[1] Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. Cognitive Science, 7(2), 155–170. Structure-mapping theory of analogy: the alignment rule is to preserve higher-order relational structure while dropping surface features, distinguishing analogy from literal similarity and other comparison readings. ↩
[2] Forbus, Kenneth D., Dedre Gentner, and Kathy Law. "MAC/FAC: A Model of Similarity-Based Retrieval." Proceedings of the 17th Annual Conference of the Cognitive Science Society, 1995, pp. 504–509. Retrieval model distinguishing surface-matching (MAC) from deep similarity (FAC) in analogical access. ↩
[3] Hofstadter, D., & Sander, E. (2013). Surfaces and Essences: Analogy as the Fuel and Fire of Thinking. Basic Books. Argues that the same mapping-across-domains structure underlies word use, scientific conceptualization, and everyday categorization; treats translation-style analogical mapping as the core engine of cognition. ↩
[4] Vosniadou, Stella, and Andrew Ortony, eds. Similarity and Analogical Reasoning. Cambridge University Press, 1989. Edited collection synthesizing empirical and theoretical work on analogy in learning, development, and reasoning. ↩
[5] Polya, G. (1945). How to Solve It: A New Aspect of Mathematical Method. Princeton University Press. Foundational pedagogical pattern catalog: codifies recurring problem-solving heuristics (understand the problem, devise a plan, carry out the plan, look back) into teachable vocabulary; prototype of pattern-language thinking in pedagogy. ↩
[6] Falkenhainer, Brian, Kenneth D. Forbus, and Dedre Gentner. "The Structure-Mapping Engine: Algorithm and Applications." Artificial Intelligence, vol. 41, no. 1, 1989, pp. 1–63. Structure-Mapping Engine (SME) algorithm, foundational computational implementation of structure-mapping theory. ↩
[7] Kolodner, Janet L. Case-Based Reasoning. Morgan Kaufmann, 1993. Comprehensive treatment of case-based reasoning as computational implementation of analogical retrieval and transfer. ↩
[8] Vygotsky, L. S. (1978). Mind in Society: The Development of Higher Psychological Processes (M. Cole, V. John-Steiner, S. Scribner, & E. Souberman, Eds.). Harvard University Press. Develops internalization as the reconstruction of an initially external, interpersonal operation into an internal, intrapersonal one — externally scaffolded regulatory speech becoming private inner speech for self-regulation — supports the developmental-learning exemplar. ↩
[9] Hofstadter, Douglas R. Fluid Concepts and Creative Analogies: Computer Models of the Fundamental Mechanisms of Thought. Basic Books, 1995. Philosophical and computational treatment of analogy as fluid concept-blending, introducing Copycat architecture for analogical reasoning. ↩
[10] Holyoak, K. J., & Thagard, P. (1989). Analogical mapping by constraint satisfaction. Cognitive Science, 13(3), 295–355. Develops a theory of comparison-as-analogical-mapping in which the operation is run by jointly satisfying structural, semantic, and pragmatic constraints; supplies a formal model of the five-role machinery (comparands, alignment-rule, output relation) the prime names. ↩
[11] Gentner, Dedre, and Arthur B. Markman. "Structure Mapping in Analogy and Similarity." American Psychologist, vol. 52, no. 1, 1997, pp. 45–56. Synthesis of similarity judgment and analogy processes, establishing systematicity principle and relational prominence. ↩
[12] Hofstadter, Douglas R., and Melanie Mitchell. "The Copycat Project: A Model of Mental Fluidity and Analogy-Making." Advances in Consciousness Research, vol. 13, 1995, pp. 205–267. Letter-string analogy task and Copycat implementation, canonical example of context-sensitive analogical reasoning.
[13] Hummel, John E., and Keith J. Holyoak. "A Symbolic-Connectionist Model of Relational Inference and Analogical Mapping." Psychological Review, vol. 110, no. 3, 2003, pp. 220–264. LISA (Learning and Inference with Schemas and Analogies) model integrating neural and symbolic approaches to analogy.
[14] Lakoff, George & Johnson, Mark. (1980). Metaphors We Live By. University of Chicago Press. Cognitive theory of metaphor as central to semantic change and conceptual structure; metaphorical extensions as motivated by embodied cognition; foundational for cognitive semantics. CROSS-DP-22.
[15] Lakoff, G. "The Contemporary Theory of Metaphor." In A. Ortony (Ed.), Metaphor and Thought (2nd ed.). Cambridge University Press, 1993. Systematizes contemporary metaphor theory and addresses the relationship between metaphorical mapping and blending; clarifies how metaphor constrains but does not fully determine blended-space construction. [^lakoff-1993]