How Example–Problem Pairs Work | See One, Solve One, Then Make the Method Yours

Quick Read: A Worked Example Is Not Finished Until the Learner Uses It

Worked examples are especially valuable for novices because they reduce unproductive search. Instead of spending limited working memory inventing a solution path, the learner can inspect how a valid path is organised.

But there is a danger. A learner can follow a finished solution fluently and mistake recognition for capability.

The paired problem exposes that illusion immediately.

The example lends the route. The problem asks the learner to carry it.

The Research Logic: Why Alternating Can Beat Problem Solving Alone

Research on worked examples has repeatedly shown that novice learners can benefit from studying complete solutions rather than solving every problem from scratch. Early work by Trafton and Reiser argued that examples are particularly useful when learners immediately apply what they have just studied to a similar problem.

Later research by Tamara van Gog, Liesbeth Kester and Fred Paas compared worked-example study, example–problem pairs and problem–example pairs. For novices, worked examples and example-first pairing produced stronger learning than problem solving alone in the studied conditions, and the ordering mattered. An example shown before the related problem gives the learner a usable schema before independent execution is demanded.

That does not mean every lesson must follow one rigid order. It means the order carries cognitive consequences.

Why “Try First” Is Not Always the Best Default for Novices

There is an appealing educational story: let students struggle first so discovery becomes meaningful.

Sometimes that works. Productive failure is a real research programme with carefully designed conditions. But unstructured failure and productive failure are not the same thing.

A novice who has no usable schema may spend ten minutes testing irrelevant moves, encoding accidental features and exhausting working memory before seeing the explanation.

If the instructional goal is initial acquisition of a new procedure, example-first can be a more efficient starting point.

For the contrasting design logic, see How Productive Failure Works.

The Pair Has Two Different Jobs

The example reduces search

It exposes a successful path, relevant decisions, intermediate states and verification moves.

The problem restores responsibility

It removes the finished path and asks the learner to retrieve, select and execute the structure.

If the learner only studies examples, responsibility never returns. If the learner only solves unsupported problems, search may dominate before useful schemas exist.

The pair creates a controlled handoff.

Distance Matters: The First Paired Problem Should Be Close, Not Identical

If the practice problem is identical except for one number, the learner can imitate surface moves without understanding the structure.

If it is too different, the bridge may disappear before the learner has stabilised the method.

A good first pair changes enough to require reconstruction while preserving the same governing principle.

• Change the numbers but keep the method.
• Change the wording but preserve the mathematical relationship.
• Change the data representation but preserve the scientific inference.
• Change the topic but preserve the paragraph function.
• Change the surface context while keeping the causal structure.

The Comparison Step Is Where Much of the Learning Lives

After the learner solves the paired problem, do not simply mark correct or wrong.

Compare the structures.

• Which cue was the same?
• Which step transferred directly?
• Which surface feature changed?
• Where did the learner depart from the model?
• Was the departure an error or a valid alternative method?
• What principle connects both solutions?

Without comparison, the learner may treat each problem as a separate episode. Comparison helps extract the invariant.

Example–Problem Pairs and Completion Problems

Completion problems create a smaller handoff. Part of the solution is provided and the learner completes the missing steps.

Example–problem pairs create a larger handoff. The entire example is available first; then the learner receives a fresh problem with the solution removed.

These designs can form a progression:

• complete worked example;
• worked example with self-explanation prompts;
• completion problem;
• closely paired independent problem;
• more varied independent problem;
• mixed problem set requiring method selection.

See How Completion Problems Work.

The Copying Trap

Students can use a worked example as a substitution template.

They scan for matching positions, replace numbers and reproduce symbols. The answer may be correct while the method remains brittle.

To break copying, ask one question before the learner starts the paired problem:

“What feature of this new problem tells you that the same method belongs here?”

Now the learner must identify structure before execution.

The Attention Problem: Students Often Study the Answer Instead of the Decisions

A worked solution contains visible algebra, sentences, diagrams or code. Learners naturally attend to what is printed.

But expertise often lives in invisible decisions:

• Why choose this representation?
• Why start here?
• Why ignore that detail?
• Why is this transformation legal?
• What would have made a different method preferable?
• How do we know the result is plausible?

A strong worked example therefore annotates thinking, not only steps.

See How Worked Examples Work for Performance.

The Timing Rule: Apply Before the Structure Goes Cold

The first paired problem should usually follow the example quickly enough that the learner can use the newly organised representation.

This is not because delayed practice is bad. Delay is valuable later for retention. The first immediate application serves a different function: it checks whether the learner can convert observed structure into action before the lesson moves on.

Later, return to the same method after spacing. Immediate transfer and delayed retention are different tests.

The Fading Rule: Pairs Should Evolve

If every practice problem forever follows a nearly identical example, the learner can become dependent on that adjacency.

Over time:

• increase the distance between example and problem;
• increase surface variation;
• mix neighbouring methods;
• remove topic labels;
• reduce explanatory prompts;
• require the learner to select which prior example is relevant.

The pair is scaffolding. It should eventually disappear.

From Near Transfer to Farther Transfer

Think of pairing in concentric rings.

• Ring 1: same method, nearly same surface.
• Ring 2: same method, changed wording or representation.
• Ring 3: same principle embedded in a different context.
• Ring 4: mixed set where the learner must decide whether the principle applies.
• Ring 5: novel problem requiring adaptation or combination with another method.

The learner should not be thrown directly into Ring 5 simply because transfer is the final goal. Transfer is built by progressively increasing what has to be recognised and reconstructed.

Example–Problem Pairs in Mathematics

Mathematics is the obvious home for the method.

Study one worked example of completing the square. Immediately solve a closely related quadratic. Compare the cue, transformation and final form. Then solve a version with a different leading coefficient. Later mix it with factorisation and the quadratic formula so method selection returns.

The pair teaches execution first. The mixed set teaches discrimination later.

Continue through the Mathematics Learning Hub.

Example–Problem Pairs in Science

A science example can model how to turn evidence into explanation.

Study one fully worked graph question: identify the pattern, connect it to a mechanism, state the limiting condition and explain the anomaly. Then receive a new graph from another biological or physical system and perform the same reasoning sequence.

The surface data change. The evidence-to-model operation transfers.

Continue through the Science Learning Hub.

Example–Problem Pairs in English

English examples should not become templates that students copy word for word.

Instead, pair by function.

Study a model paragraph that makes a claim, selects evidence and explains significance. Then give a different text or topic and ask the learner to construct a paragraph with the same functional architecture.

Or study how a sentence combines concession and argument, then construct a new sentence with different content but the same rhetorical relationship.

Continue through the English Learning Hub.

Example–Problem Pairs in Coding and Technical Learning

Copying a complete program can produce an illusion of fluency. Pairing forces reconstruction.

Study a small worked program that parses input, validates it and transforms the data. Then solve a new task with a changed data type or constraint. The learner must decide which architecture transfers and which implementation details must change.

As expertise grows, remove the adjacent example and require retrieval from a library of prior patterns.

When the Paired Problem Is Wrong, Diagnose the First Divergence

A wrong final answer can hide very different failures.

• The learner failed to recognise that the example applied.
• The learner selected the right method but recalled a step incorrectly.
• The learner copied a surface feature that did not transfer.
• The learner executed accurately but made an arithmetic or language error.
• The learner completed the method but did not verify the result.

Compare the attempted solution against the example at the first point of divergence. That point is often the real instructional target.

A Four-Pair Lesson Architecture

Pair 1 — Very close

Build the first successful transfer while the model is fresh.

Pair 2 — Changed representation

Force the learner to identify the same structure under different surface cues.

Pair 3 — Reduced guidance

Provide a shorter example or only a strategic hint.

Pair 4 — Mixed choice

Place the target method beside plausible alternatives so the learner must discriminate before executing.

The Retrieval Upgrade

Once the learner has succeeded with adjacent pairs, stop showing the example automatically.

Present the problem first and ask: “Which previous example family is this closest to, and why?”

Now the learner has to retrieve the relevant model. The instructional support has moved from visible example to memory-indexed example.

The Transfer-Appropriate Upgrade

If final performance requires uncued method selection, the mature version of example–problem learning must include uncued selection.

This is where pairing connects to How Transfer-Appropriate Processing Works.

The learner begins with a visible route, then practises the route, then retrieves the route, then chooses the route among alternatives.

The Variable-Practice Upgrade

Once the method is stable, change the surface until the invariant has to survive.

Series 0031 develops that logic in How Variable Practice Works.

The sequence is important. Variation without a stable starting schema can feel like unrelated chaos. Variation after a stable pair can teach generalisation.

A Student Protocol

• Study the example for decisions, not just symbols.
• Cover the example.
• State why the paired problem belongs to the same family.
• Solve without copying line by line.
• Compare both solutions.
• Find the first divergence if yours is wrong.
• Explain the invariant principle in one sentence.
• Solve a second problem with a changed surface.
• Return later without the example visible.

A Teacher Protocol

Design the pair, not just the example.

Before teaching, decide what the next problem will preserve and what it will change. Annotate the example with strategic decisions. Keep the first transfer close. Then deliberately widen the distance.

Do not leave twenty worked examples in front of students while they complete twenty near-identical questions. That can turn learning into visual matching. Remove or reduce support as soon as the learner can carry more of the route.

A Tutor Protocol

In small-group tuition, use pairs diagnostically. Let one learner study the model, then ask each learner independently to identify the cue and begin the paired problem. The first place their routes differ reveals whether the weak link lies in recognition, method memory, execution or checking.

Then choose the next pair to target that weak link rather than simply adding more questions.

A Parent Protocol

If a child says, “I understand the example,” place the example face down and give one similar question.

Do not immediately reopen the worked solution. Ask the child to identify what is structurally the same. If they cannot begin, the example was familiar but not yet usable.

The Deep Principle: Guidance Should Be Close Enough to Transfer and Far Enough to Require Thought

Instruction has a timing problem.

Give too little guidance too early and the novice wastes scarce capacity searching. Give too much guidance for too long and the learner never carries the method independently.

Example–problem pairs solve a small part of that problem elegantly.

Show the route. Remove the route. Ask the learner to rebuild it. Compare what survived. Change the surface. Repeat until the example is no longer needed.

Research and Further Reading

• Trafton & Reiser — The Contributions of Studying Examples and Solving Problems to Skill Acquisition
• van Gog, Kester & Paas — Effects of Worked Examples, Example–Problem, and Problem–Example Pairs on Novices’ Learning
• Leppink et al. — Effects of Pairs of Problems and Examples on Task Performance and Cognitive Load
• How Completion Problems Work
• Study & Learning Methods Hub

---

eduKateSG Learning Node Series · 0039 of the continuing series. Previous: 0038 — How the Imagination Effect Works. Continue through the Study & Learning Methods Hub and the wider eduKateSG Learning Hubs.