How to Learn Anything Quickly | Worked Examples — Learn Difficult Problems by Studying Expert Solutions

How to learn anything quickly with worked examples begins with a simple advantage: beginners do not always need to discover every successful method from scratch. A carefully chosen expert solution can show how a difficult problem is represented, which principles matter, why one step follows another, how errors are checked and what a finished performance looks like.

People searching for worked examples, worked example effect, how to solve difficult problems, how to learn faster, example based learning, cognitive load, study techniques, active recall, deliberate practice and learning how to learn are often trying to cross the same gap: moving from explanation to independent problem solving without wasting effort or becoming dependent on solutions.

eduKateSG uses worked examples as scaffolding inside a larger learning loop: inspect the task, predict, study the expert model, explain each decision, hide the solution, reconstruct it, attempt a related problem, diagnose errors, fade support, space another attempt and test transfer. The goal is never to become good at reading solutions. The goal is to become able to produce good solutions independently.

50-Second Router

• Completely new: study one representative worked example and explain every major step.
• Understand while looking but freeze alone: use completion problems and fade the solution.
• Copy methods mechanically: compare examples and explain why each method applies.
• Keep making mistakes: classify the error before studying another example.
• Preparing for exams: move from examples to mixed, delayed and timed independent problems.

1. Why worked examples accelerate early learning

Why worked examples accelerate early learning is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Why worked examples accelerate early learning is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Why worked examples accelerate early learning is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Why worked examples accelerate early learning is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Why worked examples accelerate early learning is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Why worked examples accelerate early learning is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Why worked examples accelerate early learning is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Why worked examples accelerate early learning is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Why worked examples accelerate early learning is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Why worked examples accelerate early learning is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

2. Novice versus expert knowledge

Novice versus expert knowledge is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Novice versus expert knowledge is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Novice versus expert knowledge is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Novice versus expert knowledge is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Novice versus expert knowledge is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Novice versus expert knowledge is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Novice versus expert knowledge is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Novice versus expert knowledge is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Novice versus expert knowledge is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Novice versus expert knowledge is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

3. Working-memory limits

Working-memory limits is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Working-memory limits is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Working-memory limits is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Working-memory limits is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Working-memory limits is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Working-memory limits is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Working-memory limits is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Working-memory limits is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Working-memory limits is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Working-memory limits is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

4. Choose a representative example

Choose a representative example is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Choose a representative example is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Choose a representative example is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Choose a representative example is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Choose a representative example is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Choose a representative example is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Choose a representative example is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Choose a representative example is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Choose a representative example is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Choose a representative example is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

5. Read the problem before the solution

Read the problem before the solution is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Read the problem before the solution is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Read the problem before the solution is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Read the problem before the solution is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Read the problem before the solution is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Read the problem before the solution is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Read the problem before the solution is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Read the problem before the solution is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Read the problem before the solution is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Read the problem before the solution is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

6. Predict the next step

Predict the next step is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Predict the next step is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Predict the next step is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Predict the next step is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Predict the next step is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Predict the next step is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Predict the next step is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Predict the next step is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Predict the next step is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Predict the next step is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

7. Explain why each step exists

Explain why each step exists is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Explain why each step exists is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Explain why each step exists is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Explain why each step exists is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Explain why each step exists is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Explain why each step exists is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Explain why each step exists is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Explain why each step exists is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Explain why each step exists is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Explain why each step exists is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

8. Identify the governing principle

Identify the governing principle is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Identify the governing principle is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Identify the governing principle is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Identify the governing principle is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Identify the governing principle is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Identify the governing principle is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Identify the governing principle is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Identify the governing principle is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Identify the governing principle is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Identify the governing principle is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

9. Separate surface details from structure

Separate surface details from structure is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Separate surface details from structure is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Separate surface details from structure is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Separate surface details from structure is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Separate surface details from structure is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Separate surface details from structure is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Separate surface details from structure is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Separate surface details from structure is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Separate surface details from structure is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Separate surface details from structure is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

10. Compare two solutions

Compare two solutions is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Compare two solutions is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Compare two solutions is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Compare two solutions is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Compare two solutions is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Compare two solutions is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Compare two solutions is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Compare two solutions is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Compare two solutions is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Compare two solutions is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

11. Compare correct and incorrect examples

Compare correct and incorrect examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Compare correct and incorrect examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Compare correct and incorrect examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Compare correct and incorrect examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Compare correct and incorrect examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Compare correct and incorrect examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Compare correct and incorrect examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Compare correct and incorrect examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Compare correct and incorrect examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Compare correct and incorrect examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

12. Completion problems

Completion problems is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Completion problems is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Completion problems is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Completion problems is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Completion problems is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Completion problems is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Completion problems is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Completion problems is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Completion problems is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Completion problems is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

13. Faded worked examples

Faded worked examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Faded worked examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Faded worked examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Faded worked examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Faded worked examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Faded worked examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Faded worked examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Faded worked examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Faded worked examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Faded worked examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

14. Move from example to independent attempt

Move from example to independent attempt is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Move from example to independent attempt is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Move from example to independent attempt is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Move from example to independent attempt is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Move from example to independent attempt is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Move from example to independent attempt is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Move from example to independent attempt is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Move from example to independent attempt is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Move from example to independent attempt is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Move from example to independent attempt is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

15. Retrieval after examples

Retrieval after examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Retrieval after examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Retrieval after examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Retrieval after examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Retrieval after examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Retrieval after examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Retrieval after examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Retrieval after examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Retrieval after examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Retrieval after examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

16. Spacing examples

Spacing examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Spacing examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Spacing examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Spacing examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Spacing examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Spacing examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Spacing examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Spacing examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Spacing examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Spacing examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

17. Interleave problem types

Interleave problem types is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Interleave problem types is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Interleave problem types is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Interleave problem types is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Interleave problem types is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Interleave problem types is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Interleave problem types is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Interleave problem types is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Interleave problem types is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Interleave problem types is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

18. Feedback

Feedback is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Feedback is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Feedback is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Feedback is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Feedback is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Feedback is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Feedback is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Feedback is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Feedback is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Feedback is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

19. Error classification

Error classification is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Error classification is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Error classification is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Error classification is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Error classification is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Error classification is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Error classification is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Error classification is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Error classification is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Error classification is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

20. Vocabulary examples

Vocabulary examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Vocabulary examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Vocabulary examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Vocabulary examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Vocabulary examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Vocabulary examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Vocabulary examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Vocabulary examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Vocabulary examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Vocabulary examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

21. Reading examples

Reading examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Reading examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Reading examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Reading examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Reading examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Reading examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Reading examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Reading examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Reading examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Reading examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

22. Writing examples

Writing examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Writing examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Writing examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Writing examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Writing examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Writing examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Writing examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Writing examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Writing examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Writing examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

23. Mathematics examples

Mathematics examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Mathematics examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Mathematics examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Mathematics examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Mathematics examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Mathematics examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Mathematics examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Mathematics examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Mathematics examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Mathematics examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

24. Science examples

Science examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Science examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Science examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Science examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Science examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Science examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Science examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Science examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Science examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Science examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

25. Language examples

Language examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Language examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Language examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Language examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Language examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Language examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Language examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Language examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Language examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Language examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

26. Technical examples

Technical examples is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Technical examples is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Technical examples is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Technical examples is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Technical examples is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Technical examples is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Technical examples is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Technical examples is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Technical examples is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Technical examples is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

27. Transfer to unfamiliar problems

Transfer to unfamiliar problems is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Transfer to unfamiliar problems is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Transfer to unfamiliar problems is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Transfer to unfamiliar problems is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Transfer to unfamiliar problems is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Transfer to unfamiliar problems is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Transfer to unfamiliar problems is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Transfer to unfamiliar problems is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Transfer to unfamiliar problems is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Transfer to unfamiliar problems is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

28. Exam preparation

Exam preparation is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Exam preparation is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Exam preparation is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Exam preparation is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Exam preparation is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Exam preparation is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Exam preparation is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Exam preparation is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Exam preparation is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Exam preparation is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

29. Seven-day protocol

Seven-day protocol is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Seven-day protocol is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Seven-day protocol is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Seven-day protocol is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Seven-day protocol is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Seven-day protocol is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Seven-day protocol is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Seven-day protocol is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Seven-day protocol is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Seven-day protocol is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

30. Thirty-day protocol

Thirty-day protocol is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Thirty-day protocol is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Thirty-day protocol is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Thirty-day protocol is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Thirty-day protocol is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Thirty-day protocol is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Thirty-day protocol is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Thirty-day protocol is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Thirty-day protocol is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Thirty-day protocol is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

31. Common failures

Common failures is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Common failures is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Common failures is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Common failures is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Common failures is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Common failures is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Common failures is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Common failures is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Common failures is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Common failures is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

32. Teacher implementation

Teacher implementation is most useful when the example changes what the learner can do next. Begin with the performance the learner must eventually produce without assistance. A worked example is not the endpoint; it is temporary instructional support. State what successful independent performance looks like, then use the example to expose the decisions, representations and checks that make that performance possible.

Teacher implementation is most useful when the example changes what the learner can do next. Study the problem before reading the solution. Identify what is given, what must be found or produced, and which features may matter. This prediction step prevents the worked solution from becoming passive reading. Even an incorrect prediction creates useful evidence about the learner’s current model.

Teacher implementation is most useful when the example changes what the learner can do next. Read each step as an answer to a question: why was this move chosen here? Naming the principle behind a step is more useful than copying the step itself. Expert solutions compress many decisions, so the learner should unpack those decisions until the logic can be reconstructed without the page.

Teacher implementation is most useful when the example changes what the learner can do next. Use self-explanation. After a meaningful step, hide the explanation and state the reason in ordinary language, then in precise subject language. If the learner can repeat the procedure but cannot explain why it applies, transfer will remain fragile when the next problem changes its wording or surface features.

Teacher implementation is most useful when the example changes what the learner can do next. Compare examples deliberately. Place two related problems beside each other and ask what remains invariant and what changes. Contrast a correct method with a plausible wrong method. Comparison builds discrimination: the learner begins to recognise which cues matter for selecting a strategy rather than memorising the appearance of one question.

Teacher implementation is most useful when the example changes what the learner can do next. Fade support progressively. A complete worked solution is useful early, but continued dependence can prevent independent generation. Remove the final step, then several intermediate steps, then most prompts. Completion problems create a bridge between studying an expert solution and producing a full solution alone.

Teacher implementation is most useful when the example changes what the learner can do next. Retrieve the method after the example disappears. Close the model and reconstruct the solution path, explanation or structure from memory. Then attempt a new problem. Retrieval distinguishes knowledge that has become available to the learner from information that was only understandable while visible.

Teacher implementation is most useful when the example changes what the learner can do next. Give feedback on the cause of errors. Distinguish missing knowledge, misunderstanding, representation, strategy selection, execution, checking and communication. A learner who selects the wrong method needs different repair from one who selects correctly but makes an arithmetic or language error.

Teacher implementation is most useful when the example changes what the learner can do next. Space and interleave examples after the initial model is understood. Return after a delay and mix related problem types so the learner must decide which approach applies. This makes practice resemble authentic performance, where a chapter heading does not announce the required method.

Teacher implementation is most useful when the example changes what the learner can do next. Measure transfer, not copying speed. The strongest evidence is whether the learner can solve, explain or produce a related but unfamiliar task without the worked model. Track method selection, explanation quality, recurring error types, accuracy after delay and the ability to verify the result.

Seven-Day Worked-Example Protocol

Day 1 studies and explains representative models. Day 2 reconstructs them from memory. Day 3 uses completion problems. Day 4 removes most support and mixes related problem types. Day 5 compares methods and tests transfer. Day 6 revisits fragile cases after a delay. Day 7 completes a cumulative independent performance and uses the error pattern to design the next cycle.

eduKateSG Learning Ecosystem

• How to Learn Anything Quickly — Master Guide
• Deliberate Practice
• Feedback Loops
• Transfer Learning
• Practice Tests
• How X Works Hub

Evidence and Further Reading

• Nature Reviews Psychology — science of effective learning
• The Learning Scientists
• American Psychological Association — learning and memory

Teaching Guide

Select examples that expose important structure, ask learners to predict before revealing steps, require self-explanation, compare related cases, fade guidance and follow the example with independent retrieval. Diagnose errors by cause and return after a delay. Worked examples succeed when support disappears and the learner can still choose, execute, explain and verify the method.