One example can teach.

A family of examples can teach the shape of a concept.

This distinction matters because learners rarely need only one successful encounter.

They need to know:

• what clearly belongs;
• what still belongs when the surface changes;
• what almost belongs;
• what definitely does not;
• which variation matters;
• which variation does not;
• what happens when competing cases appear together;
• whether the idea survives transfer.

A training case family is a deliberately designed set of related examples that collectively reveal the structure, boundaries, variation and transfer conditions of a target capability.

The important word is collectively.

Each case has a job.

The learning does not live in any one example.

It emerges from the relationship among them.

---

Quick Read: The Case-Family Stack

Base Case → Aligned Variation → Contrast Case → Near-Miss → Counterexample → Mixed Case → Transfer Case → Stress Case

Not every training target needs all eight.

But this stack gives tutors a way to think beyond the worksheet.

A strong case family often contains:

• one case that makes the rule easy to see;
• several cases that preserve the rule while changing the surface;
• one or two cases that expose the decision boundary;
• a case that represents a common misconception;
• a mixed set where category labels disappear;
• a transfer case from another context or representation.

Why a Family Is Better Than a Random Collection

Random variety can look sophisticated.

It can also be instructionally noisy.

One question uses difficult arithmetic.

One changes the representation.

One introduces unfamiliar vocabulary.

One changes the underlying rule.

The learner experiences variety but may not know which difference matters.

A case family controls variation deliberately.

That makes it closely related to Training Example Selection, Training Contrast and Training Perturbation.

Case Family Member 1: The Base Case

The base case is the cleanest useful instance.

It should make the target relationship visible with minimal irrelevant difficulty.

For direct proportion:

y = 3x

For evidence-bounded inference:

a short passage where the clues clearly support one reasonable conclusion.

For experimental reasoning:

a simple controlled setup with one changed factor and one measured outcome.

The base case is the learner’s first stable reference point.

Case Family Member 2: Aligned Variation

The next cases should change surface features while preserving the target structure.

For Mira:

y = 5x

then a map-scale problem.

then a currency-conversion problem.

The numbers and stories change.

The proportional structure remains.

This helps the learner find the invariant rather than memorise one surface.

Case Family Member 3: Contrast Case

Now place a related but different category beside the target.

Direct proportion beside inverse proportion.

Literal retrieval beside inference.

Observation beside explanation.

The contrast asks:

Which feature changes the correct decision?

This is how a category gains a boundary.

Case Family Member 4: Near-Miss

A near-miss resembles the target closely but violates one defining condition.

For direct proportion:

y = 3x + 2

For English inference:

a plausible answer whose adjective is one level stronger than the evidence.

For Science:

a seemingly controlled experiment where a second relevant variable quietly changes.

Near-misses train the learner to notice conditions rather than broad resemblance.

This is the role of Training Nonexamples.

Case Family Member 5: The Misconception Case

Some cases should represent the wrong rule learners actually use.

This is different from inventing an arbitrary error.

If many students believe every straight line means direct proportion, make that misconception visible.

If students repeatedly infer personality from weak behavioural clues, build a case around that overclaim.

If Science learners confuse observation with explanation, build a model answer that restates the result without giving a mechanism.

A 2025 systematic review of erroneous and contrasting erroneous examples found that such examples can support learning when prompts, feedback, prior knowledge and cognitive load are appropriately managed.

The misconception case turns common error into instructional material.

Case Family Member 6: The Mixed Case

Once the learner understands the family, remove the labels.

Mix related categories.

Mira receives direct, inverse and non-proportional relationships together.

Jonas receives literal, inferential and language-effect questions together.

Nadia receives observation, explanation and experimental-design questions together.

Now the learner must classify before acting.

This is where explicit case-family study turns into independent decision-making.

Case Family Member 7: Transfer Case

Move the relationship into a different context or representation.

Change topic.

Change genre.

Change apparatus.

Change from table to graph.

Change from worked example to fresh problem.

The learner should recognise the same structure despite the new surface.

This connects to Training Invariants and the broader transfer owners elsewhere in the eduKate estate.

Case Family Member 8: Stress Case

A stress case tests whether the capability survives increased demand.

Add time pressure.

Add irrelevant information.

Add a plausible competing method.

Add a longer passage.

Add unfamiliar apparatus.

The point is not to make the question cruel.

The point is to see where the performance envelope currently ends.

A Case Family Is a Small Curriculum

When case design is good, a sequence of six or eight examples can function like a miniature curriculum.

The sequence introduces the structure.

Shows what remains invariant.

Shows what breaks the rule.

Shows a common misconception.

Then removes the labels and tests transfer.

This is much more deliberate than selecting six questions because they happen to be numbered 1–6.

Case Families and Variation Theory

Variation Theory provides a useful lens for case families because it treats learning as discernment of critical aspects through patterned variation and invariance.

A 2025 intervention study on what examples to use in instruction compared induction, contrast and random pairing in a kindergarten Chinese-character task. Contrast produced the strongest outcomes in that setting.

The result should not be universalised mechanically.

But the design lesson is useful:

The relationship among cases can teach something that isolated positive examples cannot.

Case Families and Contrasting Cases

Contrasting cases have a long history in learning research.

A 2025 Instructional Science article describes how carefully chosen contrasting classroom-discussion cases can help learners notice important features that a solution or analysis needs to account for.

The wider principle transfers well to school tuition.

Do not merely present cases.

Choose cases whose differences expose what the learner should notice.

Case Families and Example Ambiguity

A case family can fail if its early examples are ambiguous.

The 2025 Contemporary Educational Psychology research on worked-example selection found that examples built from more ambiguous problems could produce poorer learning and more misconceptions.

This suggests a practical sequence:

clarity first, then controlled complexity.

Do not make the base case ambiguous merely because later authentic tasks will be.

Teach the structure before stress-testing it.

Mathematics Case Family: Direct Proportion

Mira’s family might look like this:

• Base: y = 3x.
• Aligned variation: y = 5x.
• Representation change: table with constant y/x.
• Near-miss: y = 3x + 2.
• Contrast: inverse relationship.
• Context transfer: map scale.
• Mixed case: classify six different relationships.
• Stress case: word problem with irrelevant numbers.

Now Mira is not merely “doing proportion questions.”

She is learning the topology of the concept.

Mathematics Case Family: Quadratic Method Selection

A different family trains decisions rather than definitions.

• easy difference of squares;
• clean trinomial factorisation;
• case where completing the square exposes the vertex;
• case where the quadratic formula is the robust route;
• near-miss that looks factorisable but is not;
• mixed set without method labels.

Each case exists to sharpen method selection.

English Case Family: Inference

Jonas’s family might contain:

• short obvious inference with nearby evidence;
• same reasoning with evidence spread across sentences;
• literal question for contrast;
• plausible but overclaimed answer;
• change from narrative to article;
• change from attitude to intention;
• mixed comprehension set;
• timed passage.

The evidence rule stays stable while the surface changes.

English Case Family: Paragraph Development

• strong claim + explanation;
• claim + example;
• claim + repeated restatement as nonexample;
• strong paragraph with consequence;
• same architecture on a different topic;
• two connected paragraphs;
• whole composition.

The case family helps Jonas see development as a functional relationship, not a model paragraph to copy.

Science Case Family: Experimental Reasoning

Nadia’s family might include:

• simple controlled plant experiment;
• same logic with heat;
• same logic with electricity;
• near-miss with two variables changed;
• case with correct design but misleading expectation;
• case with reversed outcome;
• mixed questions where experimental reasoning is not announced.

The apparatus changes.

The experimental spine remains.

Vocabulary Case Families

Vocabulary also benefits from families.

Take “reluctant.”

• clear positive example in context;
• different context preserving hesitation;
• near-synonym “unwilling”;
• contrast with “eager”;
• near-miss where “reluctant” is too weak or too strong;
• sentence generation;
• retrieval inside composition.

The word becomes part of a semantic neighbourhood rather than an isolated flashcard.

Case Families Can Be Built From Student Errors

A good tutor can turn recurring errors into future case families.

Mira’s sign mistake becomes one near-miss in an algebra family.

Jonas’s overclaim becomes one English nonexample.

Nadia’s uncontrolled experiment becomes one Science boundary case.

Over time, a tuition programme develops a library not merely of questions but of diagnosed teaching cases.

That library can become increasingly precise because each case is associated with the misconception or decision it reveals.

Case Families and Training Repetition

A case family makes repetition intelligent.

The same underlying relationship repeats.

The surface does not have to.

See Training Repetition.

This creates repeated encounters with the target while progressively widening the conditions under which it must be recognised.

Case Families and Training Recombination

Early family members may isolate the target.

Later members should recombine it with realistic demands.

Mixed sets restore method competition.

Transfer cases restore unfamiliarity.

Stress cases restore timing and attention competition.

This connects to Training Recombination.

Case Families and Training Compression

After enough cases, ask the learner to compress the family.

What rule explains the positive cases?

What condition separates the near-miss?

What invariant survives variation?

Which mistake is tempting but wrong?

This is Training Compression.

The learner should not carry eight separate examples forever.

The family should compress into a portable model.

Learner-Generated Case Families

Advanced learners can build their own case families.

Ask Mira to create:

• one clear example;
• one varied example;
• one near-miss;
• one counterexample;
• one harder transfer case.

Ask Jonas to create a family around evidence-bounded inference.

Ask Nadia to create a family around controlled experiments.

This combines case design with Training Generation.

If the learner can build the family, the concept boundary has become increasingly generative.

Failure Mode: Too Many Cases Before the Base Case Is Stable

The learner sees eight variations before understanding the core relationship.

Everything becomes noise.

Reduce the family temporarily.

Stabilise the base structure.

Then widen again.

This respects Training Readiness.

Failure Mode: Family Members Are Too Similar

The learner can solve every case from surface memory.

Add variation.

Change context.

Change representation.

Add a neighbouring category.

The family should force abstraction eventually.

Failure Mode: Family Members Are Too Different

The learner cannot align them.

Bring the cases closer.

Preserve more surface similarity.

Highlight corresponding relationships.

Then widen after the invariant is visible.

Failure Mode: The Family Has No Negative Cases

Every example belongs.

The learner never learns where the category ends.

Add near-misses and counterexamples carefully.

Boundaries turn general familiarity into precise knowledge.

Failure Mode: The Family Never Reaches Transfer

The learner performs beautifully on the designed family.

A school question arrives in a new form.

The learner fails.

A family is incomplete until the relationship survives beyond the family itself.

The Parent Case-Family Audit

• Does the child see more than one positive example?
• Are there near-misses and nonexamples?
• Does the set vary context while preserving the target structure?
• Does the learner practise mixed classification?
• Does the family include a transfer case?
• Can the child explain what connects the family?
• Can the child create a new family member independently?

The Tutor Case-Family Audit

• What is the base case?
• What should remain invariant?
• What feature should vary next?
• Which contrast exposes the decision boundary?
• Which misconception deserves a near-miss?
• When should labels disappear?
• What transfer case will prove the family has generalised?

The Deeper Idea: A Concept Is a Landscape, Not a Single Example

One example gives the learner a point.

A case family gives the learner a landscape.

There are central cases.

Edges.

Neighbouring categories.

False friends.

Variations.

Transfer routes.

When learners understand that landscape, an unfamiliar question is less likely to feel like an entirely new country.

A well-designed case family teaches not just an answer, but the neighbourhood of decisions around the answer.

Research Foundations

Useful current sources include the 2025 research on worked-example problem selection and ambiguity, the 2025 Variation Theory intervention on example selection, the 2025 systematic review of erroneous examples, and the 2025 Instructional Science work on contrasting cases. Together they support the design principle behind case families: learning depends partly on how examples relate to one another, what distinctions they make visible and whether learners can carry the resulting structure into new cases.

Continue Through How Training Works

Read this with Training Example Selection, Training Contrast, Training Nonexamples, Training Invariants, Training Perturbation and Training Compression.

Next: How Training Works | Training Distractors — Use Plausible Wrong Alternatives to Train Discrimination.