A student can copy every step of a worked example and still not understand why the method works.

The page looks complete.

The learner’s model may not be.

This is why one of the most useful training questions is so simple:

Why is this step allowed?

Not “what comes next?”

Not “what formula did the teacher use?”

Why.

Training self-explanation is the learner’s active attempt to explain the reasoning, relationship or principle behind a step, answer, example or decision.

It makes hidden reasoning visible.

That helps the learner connect new material to existing knowledge.

It helps the tutor see where understanding is incomplete.

And it creates a bridge from studying an example to generating the next step independently.

---

Quick Read: The Self-Explanation Loop

See or Attempt → Explain Why → Connect to a Rule → Check the Explanation → Reattempt Without the Original Support → Transfer

Useful prompts include:

• Why does this step follow?
• What rule is being used?
• What changed from the previous line?
• What stayed invariant?
• Why is this method better here?
• What assumption is being made?
• Why is this answer supported by the evidence?
• What would make this explanation invalid?
• How is this example similar to the previous one?
• Can you explain it without copying the teacher’s words?

Why Self-Explanation Is Different From Repetition

Repetition can make an action smoother.

Self-explanation asks the learner to model the action.

Mira can repeat:

3x + 5 = 17

3x = 12

x = 4

She may remember the sequence.

Now ask:

Why did subtracting 5 preserve the equation?

If she says, “because the 5 moves to the other side,” the procedure exists but the relational model may still be shallow.

If she says, “because subtracting the same quantity from both sides keeps the equality true,” the explanation reveals a more general structure.

That general structure can transfer.

Why Self-Explanation Is Different From Teacher Explanation

A teacher explanation can be excellent.

But listening to an explanation is not the same cognitive act as constructing one.

The teacher may connect the steps perfectly while the learner follows only the surface.

Self-explanation requires the learner to generate the connection.

A 2026 replication study in the British Journal of Educational Psychology describes self-explaining as an engagement activity that supports active use of instructional scaffolds such as worked examples. A 2025 systematic review of student-generated explanation in undergraduate mathematics and statistics synthesised 45 studies and found that self-explanation can support conceptual and procedural learning, while also identifying important boundary conditions such as material complexity, prior knowledge and the quality of prompting.

The implication is not that students should narrate every obvious step forever.

It is that strategically chosen self-explanation can make important reasoning visible when passive following would hide it.

Self-Explanation and Worked Examples

Worked examples reduce the need for beginners to solve everything from scratch.

That can free attention for understanding.

But examples can also become passive if learners simply read the steps.

Add self-explanation prompts:

• What was the goal of this step?
• Why was this transformation valid?
• How does this line depend on the previous one?
• Could another method work?
• What feature of the problem made this method appropriate?

The 2025 Educational Psychology Review systematic review of erroneous examples also highlights explanation prompts as one of the common ways learners are asked to process examples more deeply, while warning that prompts can become cognitively expensive when the material is already complex.

Prompting should therefore be selective.

Explain the critical step.

Do not force a paragraph of explanation after every trivial arithmetic operation.

The Self-Explanation Target Should Be Chosen Carefully

Some steps deserve explanation because they contain a general relationship.

Others are routine execution.

If Mira is learning differentiation, explaining why the Chain Rule requires multiplication by the derivative of the inner function can be valuable.

Explaining why 2 × 3 = 6 every time may not be.

If Jonas is learning inference, explaining how the evidence supports the conclusion is valuable.

Explaining why he wrote every full stop may not be.

Self-explanation should concentrate on the structural decisions the learner needs to carry into new tasks.

Self-Explanation as Diagnosis

A correct answer can hide weak understanding.

Ask for the explanation.

Mira gets the answer right because she copied the method pattern from the previous question.

When asked why the method applies, she cannot answer.

The score says success.

The explanation reveals fragility.

Jonas writes a defensible inference but explains, “It just sounds right.”

The answer is correct.

The process is not yet reliable.

Nadia identifies the correct variable but cannot explain why it is the independent variable.

Again, self-explanation exposes whether the relationship has been learned.

Self-Explanation as Error Correction

After feedback, ask the learner to explain the correction.

Not:

Do you understand?

Ask:

What was wrong with your first decision, and what will you do differently next time?

This converts feedback into a learner-generated rule.

Then use a fresh reattempt.

If the explanation is good but performance remains wrong, the learner may understand conceptually but lack procedural stability.

If performance improves, the explanation may have become operational.

Self-Explanation and Training Contrast

Put two cases side by side.

Then ask the learner to explain why the method changes.

This combines Training Contrast with self-explanation.

Mira sees two quadratic equations.

One factorises cleanly.

One does not.

She explains the structural feature that changes her first method choice.

Jonas compares an evidence-bounded inference with an overclaim and explains where the second answer crosses the boundary.

Nadia compares a controlled and uncontrolled experiment and explains why the same numerical result supports different levels of causal confidence.

The explanation makes the contrast explicit.

Self-Explanation and Training Nonexamples

Nonexamples become especially useful when the learner explains why they fail.

This is stronger than merely circling “wrong.”

• Which rule did this solution violate?
• Where did the inference exceed the evidence?
• Which uncontrolled variable weakens the conclusion?
• What would need to change to make this nonexample correct?

This connects to Training Nonexamples.

The learner develops negative knowledge: not only what works, but why tempting alternatives fail.

Self-Explanation and Training Analogies

An analogy becomes much stronger when the learner explains the mapping.

Do not ask only:

Does this equation work like a balance?

Ask:

What corresponds to what, which relationship is preserved, and where does the analogy stop?

The explanation converts a memorable image into relational knowledge.

Mathematics Self-Explanation: Why This Method?

Mira solves a quadratic equation by factorisation.

The tutor asks:

Why did you choose factorisation instead of the quadratic formula?

A weak explanation:

Because this chapter is factorisation.

A stronger explanation:

Because the quadratic can be written as two simple linear factors immediately, so factorisation is efficient and exposes the roots directly.

The second explanation is portable.

It can guide choice when the chapter title disappears.

Mathematics Self-Explanation: Explain the Transition

Instead of explaining an entire solution, focus on one consequential transition.

For example:

y = (3x + 1)⁵

Ask Mira why:

dy/dx = 15(3x + 1)⁴

rather than:

5(3x + 1)⁴

She needs to identify the outer function and inner function, then explain why the derivative of the inner function contributes the extra factor of 3.

The self-explanation trains nested structure, not only formula recall.

English Self-Explanation: Why Is This an Inference?

Jonas reads a passage and answers correctly.

The tutor asks:

What evidence did you combine to reach this conclusion?

Jonas must now reveal the chain.

The passage never says he is reluctant directly. I inferred it from the delay, the short response and the fact that he avoids the decision.

This is stronger than an answer that happens to be right.

The learner knows what evidence paid for the conclusion.

English Self-Explanation: Why Does This Paragraph Work?

Give Jonas two paragraphs.

Ask him not merely which is better but why.

• Which sentence states the claim?
• Which develops it?
• Which example changes the reader’s understanding?
• Where does the paragraph move rather than repeat?

Self-explanation becomes editorial analysis.

Later Jonas can apply the same reasoning to his own draft.

Science Self-Explanation: Why Does This Evidence Support the Claim?

Nadia sees a data table.

She gives the correct conclusion.

Now ask:

Why is this conclusion justified by the design and the observations?

She needs to identify:

• what changed;
• what was controlled;
• what was measured;
• what pattern appeared;
• why the conclusion is no stronger than the evidence.

Now scientific reasoning is visible.

Self-Explanation and Vocabulary

Vocabulary self-explanation should go beyond definitions.

Ask:

• Why does this word fit here?
• Why would the near-synonym be worse?
• What tone does the word carry?
• What grammatical form is required?
• What context would make this word inappropriate?

Jonas may know that “frugal” and “stingy” both relate to spending little.

Explaining why one is more approving and the other more critical develops lexical precision.

Self-Explanation and Training Perturbation

After a learner explains why a step works, change one feature.

This connects to Training Perturbation.

Ask:

Does your explanation still hold after this change?

If yes, the explanation may have captured an invariant.

If no, the learner must identify which condition controlled the rule.

This is one of the best ways to prevent verbal explanations from becoming memorised slogans.

Self-Explanation and Training Invariants

Ask the learner to explain what stayed the same across varied examples.

This links self-explanation with Training Invariants.

Mira solves ratio problems about recipes, maps and scale drawings.

Her explanation should name the multiplicative relationship that remains invariant.

Nadia examines several experiments and explains the common logic of change, control, measure and inference.

Jonas reads several genres and explains why evidence still constrains interpretation.

The explanation becomes abstraction.

Self-Explanation and Training Granularity

Do not ask learners to explain an entire subject at once.

Choose the right-sized explanation target.

One critical algebra transition.

One inference-evidence relationship.

One experimental conclusion.

One paragraph-development move.

This is the link to Training Granularity.

The explanation should be small enough to be precise and large enough to contain a meaningful relationship.

Self-Explanation and Training Independence

At first, the tutor asks the questions.

Later, the learner should ask them internally.

• Why am I using this method?
• What evidence supports this?
• What changed from the last line?
• Does this conclusion go too far?
• What assumption am I making?
• How could I check?

When self-explanation becomes self-questioning, it begins to support Training Independence.

Do Not Turn Self-Explanation Into Performance Theatre

Students can learn to produce impressive-sounding explanations that do not guide action.

“Because of the distributive law.”

“Because the author uses language.”

“Because it is a fair test.”

These phrases may be correct but too vague.

Ask for the relationship.

Which distribution?

Which language feature?

Which variable is controlled and why does that matter?

The explanation must do explanatory work.

Failure Mode: Explaining Too Much

Self-explanation can become a second workload.

The learner spends more time narrating than practising.

The systematic review of student-generated explanation in mathematics identifies material complexity and low prior knowledge as meaningful boundary conditions. The prompt itself can add demand.

Use selective explanation.

Target the step where understanding matters most.

Failure Mode: Explaining Before a Correct Model Exists

A novice may generate an elaborate explanation of a misconception.

That does not mean self-explanation should be avoided.

It means feedback matters.

Let the explanation reveal the model.

Then correct the model.

Then ask the learner to explain again.

Failure Mode: Tutor Explains the Learner’s Explanation

The tutor asks, “Why?”

The learner hesitates for two seconds.

The tutor answers.

The self-explanation never occurs.

Protect a short thinking window.

If the learner is stuck, reduce the prompt:

• What changed?
• What rule do you know that might apply?
• Which part are you certain about?

Support the generation without replacing it.

Failure Mode: Self-Explanation Never Becomes Performance

A learner explains beautifully but still cannot solve.

That tells us something important.

The conceptual model may be stronger than procedural execution.

Return to Training Repetition.

The learner now needs correct executions under progressively less support.

Explanation is a training mechanism.

It is not a substitute for performance.

The Three-Student Room and Self-Explanation

Mira, Jonas and Nadia can benefit from hearing one another explain because different explanations reveal different representations.

But peer explanation should not become confident transmission of error.

The tutor remains responsible for checking the mathematical, linguistic or scientific validity.

A useful sequence is:

• student explains;
• peer restates or questions;
• tutor checks the critical relationship;
• student reattempts independently.

The group becomes a reasoning surface rather than a lecture audience.

The Parent Self-Explanation Audit

• Can my child explain why a method works?
• Can the child connect the answer to evidence?
• Does the explanation use their own understanding rather than memorised phrases?
• Can the child explain a correction after feedback?
• Can the learner explain what would make the method fail?
• Does the explanation improve the next attempt?

The Tutor Self-Explanation Audit

• Which step contains the generalisable relationship?
• Is explanation worth the cognitive cost here?
• What prompt will elicit reasoning rather than recall of teacher wording?
• How will I check the explanation?
• What fresh problem will test whether the explanation guides performance?
• When can the prompt fade?

The Deeper Idea: Explanation Turns Procedure Into Model

A procedure says what to do.

A model says why the action makes sense.

That difference matters when the problem changes.

If Mira remembers only a sequence, an unfamiliar equation can break it.

If she understands the relationship, she can reconstruct a valid move.

If Jonas remembers only a model answer, a new passage can break it.

If he understands how evidence constrains inference, he can rebuild the answer.

If Nadia memorises only a Science sentence, unfamiliar apparatus can break it.

If she understands the causal relationship, she can construct a new explanation.

Self-explanation is valuable when it helps the learner carry the reason behind the step into a future situation where the original example is gone.

Research Foundations

Useful current sources include the 2026 British Journal of Educational Psychology replication on worked examples and self-explanation, the 2025 systematic review of student-generated explanation in mathematics and statistics, the 2025 Educational Research Review meta-analysis of prompting in digital learning, and the 2025 systematic review of erroneous examples. Together they support a conditional view: learner-generated explanation can deepen processing and support example-based learning, but benefits depend on prompt design, prior knowledge, complexity, feedback and whether the learner actually engages in the intended reasoning.

Continue Through How Training Works

Read this with Training Analogies, Training Contrast, Training Nonexamples, Training Invariants, Training Perturbation and Training Repetition.

Next: How Training Works | Training Generation — Learn by Producing Examples, Questions and Counterexamples.