How Analogical Encoding Works | Compare Cases to See Deep Structure

eduKateSG Learning Node Series · 0018

One example can teach a solution. Two carefully compared examples can teach the structure that survives when the surface changes.

That is the promise of analogical encoding.

A student sees two stories, two equations, two negotiations, two scientific systems or two historical episodes. The details differ. The names differ. The numbers differ. The visible world changes.

But something underneath is the same.

When learners are asked to compare cases explicitly, they can begin to align relations rather than merely remember details. The comparison helps extract a schema: a portable structure that can later be recognised in a new situation.

Quick Read: What Analogical Encoding Does

Dedre Gentner, Jeffrey Loewenstein and Leigh Thompson studied analogical encoding as a way to improve learning and transfer from cases. In a 2003 Journal of Educational Psychology study, learners who explicitly compared two cases were better able to abstract the underlying problem-solving schema and transfer it than learners who studied the same cases separately. Earlier work in negotiation likewise found substantial transfer advantages when learners drew analogies across examples.

Recent work continues to show the educational importance of structured comparison. A 2025 study of comparison-based learning in higher education found that explicit comparisons can support cognitive and metacognitive learning, while a 2025 study of analogous video cases reported stronger application of learned strategies when learners were prompted to notice shared structure.

Do not ask only, “What happened in this case?” Ask, “What relation in this case is also operating in that one?”

The Surface Trap

Novices often classify problems by what they look like.

A mathematics problem about paint seems different from one about fuel. A story about two friends sharing a prize seems different from a business problem about profit allocation. A biology system involving blood pressure seems different from a plumbing system involving water pressure.

Experts are more likely to notice relational structure: proportionality, conservation, feedback, constraint, accumulation, trade-off, equilibrium, symmetry.

Analogical encoding helps learners cross that gap by making similarity itself an object of study.

Comparison Is Not Merely Looking at Two Things

Two examples can sit side by side without producing useful comparison.

A learner may read Case A, then Case B, and remember both independently. The shared principle remains invisible.

Analogical encoding adds an explicit task: align the cases. Which role in A corresponds to which role in B? What relationship is preserved? What differs only on the surface? Which causal chain appears in both?

The comparison is the learning event.

Objects Versus Relations

Consider two systems.

In one, water flows through a pipe because of a pressure difference. In another, electric current flows through a circuit because of a potential difference.

The objects are different: water, pipes, electrons, wires.

The relation can be similar: a difference across a system drives flow while resistance constrains the amount of flow.

An analogy becomes educational when the learner maps relations rather than merely notices that “both things flow.”

This distinction protects against shallow analogies.

Schema Abstraction

When two cases share a relational pattern, comparison can support abstraction of a schema.

A schema is not a memorised sentence floating above examples. It is an organised pattern that can be instantiated in different situations.

For example:

Two parties value several resources differently. Instead of splitting each resource evenly, trade across differences in priorities so both sides gain more.

That principle can appear in business negotiation, household planning, international agreements or classroom resource allocation. The examples differ; the integrative-trade structure remains.

Why Separate Study Can Fail to Transfer

A learner can understand Case A perfectly and still fail to recognise its relevance to Case C.

Understanding a case is not the same as abstracting the principle.

When examples are studied separately, memorable details can dominate encoding. The learner remembers the story, numbers, names and local solution. Comparison suppresses some of that surface noise by asking what survives across both cases.

Transfer improves when the learner has a representation that is less tied to one surface.

Near Analogies and Far Analogies

Not all comparisons are equally difficult.

A near analogy shares both structure and visible features. Two ratio problems about recipes are easy to compare. A far analogy may preserve the same structure across very different domains, such as comparing ecological carrying capacity with server capacity in computing.

Novices often benefit from near analogies first because surface similarity helps align the cases. Once the relational pattern is visible, farther analogies can test whether the learner truly owns the structure.

Similarity can therefore function as scaffolding.

Difference Detection

Comparison does more than reveal similarity. It makes differences informative.

When two cases share many features, the one changed feature becomes salient. This is why “spot the difference” works perceptually and why carefully matched examples can work conceptually.

Two arguments may have the same claim structure but differ in evidence quality. Two graphs may have the same intercept but different gradient. Two scientific experiments may be identical except for one controlled variable.

Matched cases turn difference into signal.

The Alignment Prompt

Useful analogical prompts force correspondence.

What plays the same role in both cases?
Which relationship is preserved?
What changes only on the surface?
What causal sequence appears in both?
Which feature breaks the analogy?
What general principle could generate both examples?

Without alignment, learners may simply list similarities: “Both involve money.” “Both involve triangles.” “Both have conflict.” Those observations may be true but structurally weak.

The Mapping Table

A simple two-column table can externalise analogy.

Case A element on the left. Case B counterpart on the right. A third column explains the shared role.

This is especially useful when domains use different vocabulary. “Pressure difference” may align with “potential difference.” “Resistance” may align with “friction.” “Input constraint” may align with “budget limit.”

The third column prevents superficial pairings because the learner must justify why the elements correspond.

Analogical Encoding Versus Ordinary Analogy

Teachers often use analogy by presenting one familiar source and one unfamiliar target: “An atom is like a solar system.”

Analogical encoding often works differently. Learners compare two or more cases to abstract a common relation before applying it elsewhere.

The direction is less “use A to explain B” and more “compare A and B to discover C, the schema they share.”

That distinction makes analogical encoding particularly valuable for transfer.

The Analogy-Breaking Question

Every educational analogy should include one question: where does the comparison stop working?

Electric current is like water flow in useful ways, but electrons do not behave exactly like parcels of water. Memory is like a library in some respects, but retrieval is reconstructive and context-sensitive in ways a shelf is not.

Asking where the analogy breaks protects the learner from literalising the metaphor.

It also deepens understanding by forcing comparison of differences as well as similarities.

Analogical Encoding in Mathematics

Mathematics is a natural home for structural comparison.

Compare two equations with different numbers but the same transformation. Compare a direct-proportion problem about recipes with one about currency. Compare a quadratic graph with a projectile path. Compare a geometric proof and an algebraic proof that preserve the same invariant.

The learner should explain what stays the same when the surface changes.

This helps prevent method selection from becoming a keyword game. If “speed” always signals one formula and “percentage” another, students may learn lexical cues rather than mathematical structure.

Continue through the Mathematics Learning Hub.

Analogical Encoding in Science

Science uses models constantly. Analogy can connect unfamiliar mechanisms with better-known systems.

Compare heat flow with diffusion. Compare electrical circuits with fluid systems. Compare feedback in body temperature with control systems in engineering. Compare natural selection across antibiotic resistance and pesticide resistance.

The purpose is not to prove systems identical. It is to reveal a shared causal architecture.

Then ask where each analogy fails.

Continue through the Science Learning Hub.

Analogical Encoding in English

English learners can compare texts, paragraphs and arguments at the level of function.

Compare two introductions that establish stakes differently. Compare two passages that create tension using different vocabulary. Compare two arguments with the same evidence but different reasoning. Compare two narratives where a small early detail becomes important later.

The learner abstracts craft rather than copying wording.

That is a safer route from model text to independent writing.

Continue through the English Learning Hub.

Analogical Encoding in Vocabulary

Vocabulary learning benefits from comparison when near-synonyms are aligned by function and constraint.

Compare reluctant and hesitant across parallel sentences. Compare claim, assert and allege in matched contexts. Ask what semantic feature changes.

The learner stops treating words as interchangeable dictionary entries and begins seeing the relational geometry among them.

Vocabulary-specific owners remain in the Vocabulary Learning Hub.

Analogical Encoding in History

Historical comparison is powerful and dangerous.

Compare two revolutions, two monetary crises or two postwar recoveries and relational patterns can emerge: legitimacy, resource constraint, institutional breakdown, coalition formation, feedback between policy and public response.

But historical analogy can mislead when surface similarities are used to erase differences in institutions, technology, culture or scale.

Good historical analogical encoding therefore requires both mapping and disanalogy. What structure is genuinely shared? Which contextual difference changes the outcome?

Analogical Encoding in Economics and Finance

Economic concepts often become clearer when learners compare structurally similar systems.

Compare household budgets and government budgets to identify both shared constraints and critical differences. Compare insurance pools and diversification. Compare bank runs with coordination failures in other systems.

The danger is transferring a metaphor past its valid boundary. A government is not simply a household; a central bank is not simply a commercial bank.

Analogical encoding should reveal structure without flattening institutional differences.

Analogical Encoding in Coding

Programming learners can compare algorithms that solve different-looking tasks with the same control structure.

A loop counting votes, a loop summing prices and a loop scanning sensor readings all instantiate iteration over a collection.

Comparing them helps the learner see the loop schema rather than memorise a domain-specific script.

Later, a new task can be classified by structure before syntax is written.

Analogical Encoding in Professional Training

Case-based professions depend heavily on recognising when a new situation resembles a known one in the right way.

Managers compare negotiations. Engineers compare failure modes. Lawyers compare fact patterns. Clinicians compare presentations while also attending to differences that change the diagnosis.

Case comparison can build judgement when the learner is required to articulate the relation rather than merely say, “This reminds me of that.”

Why Novices Miss Deep Structure

Experts possess categories that novices do not yet have.

An expert sees “feedback instability.” A novice sees “a machine that keeps overshooting.” An expert sees “conservation.” A novice sees “water moving around.”

Analogical encoding helps construct these categories by giving learners multiple instances and asking for the invariant relation.

The learner is effectively learning a new lens.

Why One Example Is Often Too Concrete

A single example contains both the principle and accidental details.

Suppose a teacher explains proportionality using a recipe. The student may encode cups, flour and servings along with the ratio structure. When a currency-conversion problem appears, the recipe cues are absent.

A second carefully chosen example strips away some accidental detail. What remains shared becomes more likely to represent the principle.

The Contrast Pair

Sometimes the best comparison is not two correct instances but one correct instance and one near miss.

Two arguments may differ only in whether evidence is explained. Two geometric figures may differ by one property. Two scientific experiments may differ in whether a confound is controlled.

Near contrasts make the decisive feature visible.

This connects analogical encoding with Series 0019 on concept boundaries, where examples and near nonexamples teach where a category stops.

The Sequence of Cases Matters

Case order can either support or obstruct abstraction.

Beginning with two extremely different cases may make correspondence difficult. Beginning with nearly identical cases can make alignment easy but produce a narrow schema.

A useful progression is often near comparison → explicit schema → farther comparison → novel transfer.

The learner first sees the relation, then learns to recognise it under increasing surface variation.

Teacher-Supplied Schema Versus Learner-Generated Schema

Should the teacher simply state the principle?

Sometimes yes. Clear explanation can save time. But learner comparison can produce a richer representation because the principle is extracted from concrete evidence rather than floating above it.

A strong sequence can combine both: learners compare first, propose a common structure, then the teacher names and refines it.

This mirrors the larger educational pattern in Productive Failure and generative learning: generation creates material for consolidation.

Analogical Encoding and Productive Failure

Productive Failure can generate multiple solution methods. Analogical encoding can then compare them.

One student uses a table. Another uses an equation. A third draws a diagram. The teacher asks which relations all three are trying to preserve and where each representation succeeds or fails.

Now failed and successful attempts become a comparative dataset.

See How Productive Failure Works.

Analogical Encoding and Worked Examples

Worked examples reduce search by showing an expert route. Comparison can make worked examples even more useful.

Instead of studying one worked example and solving ten near-identical questions, compare two worked examples with different surfaces. Ask which step performs the same function in both.

This helps move the learner from procedure copying toward structural recognition.

The existing worked-example owner remains How Worked Examples Work for Performance.

Analogical Encoding and Transfer

Transfer is the reason analogy matters.

If the learner only understands the two training cases, comparison has produced local insight. A new case must be introduced.

Crucially, the new case should not announce the relevant analogy. Real life rarely says, “Use the proportionality schema from Tuesday.”

The learner must retrieve and map the principle independently.

See Why Transfer Is the Real Proof of Learning.

The Retrieval Problem in Analogy

Even a well-learned analogy is useless if the learner does not retrieve it when needed.

This is the classic transfer problem: the source knowledge exists, but the target case does not trigger it.

One solution is to practise classification across varied cases. “Which earlier structure does this new problem resemble?”

Another is to retrieve the schema independently of any one example. “What are the defining relations of a feedback loop?”

Then use successive relearning to keep that schema accessible over time.

The False Analogy Failure

Humans are good at seeing resemblance—even when the resemblance is irrelevant.

Two countries may both be small and wealthy but have radically different institutional histories. Two students may both score 60% but have entirely different weak links. Two graphs may look similar while representing different scales.

Analogy must therefore answer: which relation is shared, and is that relation causally or functionally important?

The Surface-Match Failure

Learners often choose an analogy because words match.

A problem mentions “growth,” so they retrieve exponential growth even though the relationship is linear. A text mentions “pressure,” so they import a physical-pressure model into a metaphorical context.

Train students to justify mappings by relations, not keywords.

The Overextension Failure

A useful analogy can become harmful when extended too far.

The solar-system model of the atom may support an early spatial intuition but fails badly as a literal model of electron behaviour. Memory as a computer can highlight storage and retrieval while hiding reconstruction and context.

Every analogy needs an expiry boundary.

The Memorised-Analogy Failure

Students can memorise the teacher’s analogy without learning to reason analogically.

“Electricity is like water” becomes another fact.

To train analogical reasoning, require learners to generate a new comparison, justify mappings and identify where it breaks.

The skill is not remembering metaphors. It is aligning structure.

The Expertise Boundary

Experts can compare very distant cases because they already possess abstract schemas. Novices may need more concrete alignment.

This means teachers should not interpret failure to see a far analogy as lack of intelligence. The learner may not yet have the relational vocabulary required to map the cases.

Use nearer cases, explicit correspondence and a named schema first. Then increase distance.

Analogical Encoding as a Learning Ladder

Level 1: notice visible similarity.
Level 2: align corresponding elements.
Level 3: align relationships.
Level 4: state the common schema.
Level 5: explain where the analogy breaks.
Level 6: recognise the schema in a farther case.
Level 7: generate a new analogy independently.
Level 8: use the schema to solve a novel problem.

This progression moves from resemblance to transfer.

A Five-Minute Classroom Comparison

Show two cases with a common principle.
Ask students to identify three correspondences.
Require one relational sentence: “In both cases, X causes Y because…”
Ask for one important difference.
Name or refine the abstract principle.
Give a third case and ask whether the principle applies.

The sequence is short because comparison need not become a separate unit. It can be embedded wherever examples already exist.

A Student Study Protocol

When studying examples, stop collecting them one by one.

Choose two examples of the same concept.
Write the visible differences.
Map corresponding roles.
Write the shared relationship.
State the rule in abstract form.
Find a third example from a different context.
Test whether the rule still works.
Write one case where it would fail.

A Parent Protocol

Parents can use comparison without turning home into a lecture.

Ask, “What is the same about these two questions even though they look different?”

If the child answers with surface features, follow with, “What relationship is doing the same job?”

This single prompt often reveals whether the learner sees the deep structure or only the topic vocabulary.

A Tutor Protocol

When a learner succeeds on one example, do not immediately give ten more of the same.

Give a second example with a changed surface. Ask for comparison before solution. Then introduce one near contrast that does not use the same method.

This three-case sequence tests recognition, abstraction and boundary control.

Analogical Encoding and AI

AI can generate cases rapidly, but speed creates a quality-control problem.

A useful system can ask for two examples that share one target relation while differing in surface domain, then request a near nonexample. The learner—not the AI—should perform the comparison.

If the tool generates the mapping and the abstract rule immediately, the learner may receive the answer without doing analogical encoding.

Use the machine to supply the case library. Keep the structural comparison as the learner’s job.

Comparison as Metacognition

Comparison can also reveal how a learner evaluates quality.

Place the learner’s answer beside a model answer and ask for three differences that matter. Compare two essays and identify why one argument is stronger. Compare two solutions and decide which is more robust.

Recent higher-education research on comparison-based self-assessment suggests that explicit comparison can support metacognitive as well as cognitive learning.

The learner is learning both the domain and the criteria by which work in that domain should be judged.

The First Weak Link in Analogical Reasoning

When transfer fails, locate the broken stage.

Did the learner understand each source case?
Could the learner align corresponding elements?
Could the learner identify the shared relation?
Could the learner state the schema abstractly?
Could the learner retrieve the schema later?
Could the learner recognise it under surface change?
Could the learner resist a tempting false analogy?

“Cannot transfer” is not one problem. It is a chain of possible weak links.

Use the Diagnostics & Recovery Hub when the failure extends beyond one comparison task.

Analogical Encoding and Memory

Comparison can compress several cases into one schema, but details may still matter.

The learner should not erase all differences in pursuit of abstraction. Strong knowledge stores both general structure and important case-specific conditions.

In medicine, the same general mechanism can produce different decisions when age, dose or comorbidity changes. In law, similar doctrines can produce different outcomes under different facts. In mathematics, the same algebraic structure can have different domain restrictions.

A mature schema knows its parameters.

The World Is Full of Repeated Structure

Feedback appears in thermostats, ecosystems, organisations and human physiology. Networks appear in transport, friendship, computing and language. Scarcity appears in time, money, attention and energy. Thresholds appear in materials, markets, exams and biological systems.

Education becomes more powerful when learners can see these repeated structures without pretending the domains are identical.

Analogical encoding is one route from isolated knowledge to a connected world.

The Deep Principle: Learn the Relation That Can Travel

A case is local. A schema can travel.

The learner who memorises one solution waits for a familiar surface. The learner who understands a relation can recognise the same structure in a new costume.

That is why comparison matters.

Two cases create an opportunity to separate what is accidental from what is structural. A third case tests whether the separation worked. A distant case tests whether the schema is genuinely portable.

Learning becomes transfer when the learner can say, “This looks different, but underneath it is the same kind of problem.”

Use This Tomorrow

Choose two examples from the same topic that look different. Put them side by side. Map corresponding parts. Write one sentence describing the relation they share. Then find a third example from another context and test whether the relation still holds.

If the schema travels, the learning has moved beyond the examples.

Research and Further Reading

eduKateSG Learning Node Series · 0018 of the continuing series. Previous: 0017 — How Pretesting Works. Continue through the Study & Learning Methods Hub.

How Analogical Encoding Works | Compare Two Cases Until the Shared Structure Appears