How Completion Problems Work | Finish the Missing Steps Between Worked Examples and Independent Practice

Quick Read: The Completion Strategy

Cognitive Load Theory treats completion problems as an important bridge between fully worked examples and conventional problem solving. In the 2019 review Cognitive Architecture and Instructional Design: 20 Years Later, Sweller, van Merriënboer and Paas describe completion problems as partial solutions learners must finish. They also describe the completion strategy as a forerunner of guidance fading: begin with almost-complete worked solutions, require progressively more learner-generated steps, and end with conventional problems.

Worked example → one missing step → several missing steps → mostly learner-generated solution → independent problem.

This progression matters because the learner’s knowledge is changing during instruction. What is essential guidance for a novice can become redundant once the underlying schema is established.

Why the Jump to Independence Fails

Suppose a mathematics tutor demonstrates a five-step algebraic transformation. The student follows every line and says it makes sense. The tutor then gives a fresh problem with no support.

The student freezes.

Nothing mysterious has happened. During the example, the route-selection problem was already solved. The student could focus on understanding each step because the sequence was visible. During independent practice, the learner must now identify the problem type, choose the first operation, preserve constraints, remember the sequence, execute accurately and monitor progress.

The cognitive job changed abruptly.

A completion problem reduces the discontinuity. It says: “You have seen the whole route. Now take responsibility for the last part. When that is stable, take responsibility for more.”

Completion Is Not Copying

A poor completion task leaves one trivial arithmetic blank. The learner can fill it without understanding the structure.

A strong completion task removes a meaningful decision point.

In algebra, the missing step may require selecting the inverse operation. In writing, it may require constructing the reasoning that connects evidence to a claim. In science, it may require completing the causal link between two states. In coding, it may require writing the function body while the inputs and outputs are supplied.

The learner must understand enough of the supplied structure to continue it correctly.

Why Backward Fading Often Makes Sense

One common design is backward fading: remove the final step first, then the final two, then the final three, until the whole solution is generated by the learner.

There is a practical reason. The beginning of a solution often establishes the problem representation and activates the relevant schema. If the learner can study those early decisions and then finish the route, the task remains anchored.

As expertise grows, more of the early representation can be handed over too.

Research on guidance fading has repeatedly linked this gradual transition with cognitive-load management. A financial-services training study reported stronger delayed and transfer performance for a fading condition than for fixed example–problem pairs or pure problem solving. The important design idea is not that one exact fading pattern always wins, but that support should shrink as competence grows.

The Expertise Reversal Boundary

Instructional support has an expiry date.

For a novice, a fully worked example reduces unnecessary search. For a learner who already understands the structure, the same example can become redundant. The student is forced to process information that is already available internally.

This is the expertise reversal effect: instructional methods helpful at low expertise can lose value or become counterproductive as expertise increases.

Completion problems give teachers a practical response. Do not switch from maximum guidance to zero guidance on a calendar date. Reduce guidance in response to evidence.

What Evidence Says the Learner Is Ready for Less Help?

Look for changes in the route, not only final accuracy.

• The learner identifies the problem type quickly.
• The learner chooses the first step without prompting.
• Errors move from structural to local execution mistakes.
• The learner can explain why the supplied steps are valid.
• The learner begins checking work independently.
• Completion time falls without a loss of accuracy.
• The learner can solve a near-transfer problem after studying an example.

When these signals appear, remove a layer of support. If performance collapses structurally, restore enough support to rebuild the route.

Completion Problems for Mathematics

Mathematics is the most obvious domain because solutions can be decomposed into visible steps.

For solving an equation, the first item may be fully worked. The next may provide the setup and first transformation but require the final isolation. The next may provide only the setup. The final item may contain no worked steps.

The learner is not merely doing more. The locus of decision-making is moving from the example to the learner.

Good completion problems also vary surface features so students cannot survive by copying a visual pattern. The invariant method should remain, while numbers, contexts or representation change.

Continue through the Mathematics Learning Hub.

Completion Problems for English

English solutions are not always algorithmic, but partial completion still works.

Give a paragraph with claim and evidence supplied but leave the reasoning blank. Give a situational-writing plan with audience and purpose identified but leave content selection to the learner. Provide an introduction and ask the student to construct the counterargument. Supply a comprehension answer with the quotation selected and ask the learner to explain what the quotation proves.

The support exposes architecture while preserving a meaningful act of judgement.

As the learner becomes stronger, remove more of the architecture.

Continue through the English Learning Hub.

Completion Problems for Science

Science explanations often fail because students know facts but cannot assemble a causal chain.

A completion problem can provide the first two links of the mechanism and ask the learner to finish the chain. Or provide an experimental design with variables identified but require the student to complete controls, prediction or justification.

Later tasks remove more of the structure until the learner constructs the whole explanation or investigation independently.

Continue through the Science Learning Hub.

Completion Problems for Coding

The completion strategy has deep roots in programming instruction because incomplete programs make the bridge unusually visible.

A novice receives most of a function and completes one missing line. Later the learner completes the loop body. Later still the learner writes the whole function from a specification.

Because the surrounding code constrains the task, the learner can focus on one structural decision at a time while still operating inside an authentic program.

Do Not Fade on a Fixed Calendar

“Week 1 examples, Week 2 completion, Week 3 independent practice” is administratively tidy but educationally crude.

Different learners reach readiness at different times. Different problem types have different complexity. A student may be independent in linear equations and still need completion support in trigonometric proof.

Fade by state, not merely by date.

Forward Fading, Backward Fading and Mixed Fading

Backward fading removes later steps first. Forward fading removes early steps first. Mixed fading removes steps according to instructional purpose.

Backward fading is often attractive because the learner keeps seeing how an expert begins. Forward fading can be useful when the first decision is the target skill. Mixed fading is useful when one particular transition is the bottleneck.

The principle outranks the pattern: remove the support that the learner is ready to generate while preserving the support still needed for productive performance.

Completion Problems and Self-Explanation

Partial solutions can become passive if students merely fill blanks. Add reasoning prompts.

• Why is this supplied step valid?
• What information made this operation appropriate?
• What would break if this step were omitted?
• What invariant is being preserved?
• How does your missing step connect to the line above?

These prompts make the learner inspect the architecture rather than treat the provided work as decoration.

Completion Problems and Retrieval

Once the learner has successfully completed a partial solution, close it and ask for reconstruction later.

This connects completion with retrieval and successive relearning. The immediate goal is to finish the route with support. The longer-term goal is to preserve the route when support disappears and time passes.

Use How Successive Relearning Works for the maintenance layer.

Completion Problems and Error Diagnosis

Partial solutions produce unusually clear diagnostic evidence.

If the student understands every supplied step but fails exactly where responsibility begins, the boundary of independent competence is visible.

If the learner cannot continue because an earlier supplied step is not understood, the problem is not the missing step at all. The first weak link lies inside the model.

This is why completion tasks can be more informative than either full examples or full problems. They create a controlled handoff point.

The Tutor Handoff

Tutors often fade support verbally without realising it.

At first: “Read the question. This is a rate problem. Set up rate × time = distance. Substitute the values. Now solve.”

Later: “What type of relationship is this?”

Later still: silence.

Completion problems make that fading visible and reproducible. They allow the tutor to decide exactly which part of the route the student owns today.

The Parent Trap: Helping at the Wrong Step

At home, parents often provide too much help or too little.

Too much help turns homework into guided following. Too little help can leave a child stuck at a structural decision they are not yet equipped to make.

A completion mindset asks: Which step can the child currently own?

Supply only enough of the route to reach that step, then stop talking.

Why the Last Step Is Not Always the Best Missing Step

Some tasks have a trivial ending. If the final step is just arithmetic, removing it teaches little about strategic problem solving.

In those cases, remove a meaningful middle step or the first classification decision.

The missing work should correspond to the learning objective.

How Completion Problems Fail

• The blank is mechanically obvious.
• The supplied steps contain unexplained reasoning.
• Too many steps disappear at once.
• Support remains long after competence is established.
• Students can imitate the surface pattern without understanding the invariant.
• Completion tasks never progress to independent problems.
• The teacher fades support by schedule rather than evidence.
• Errors are marked but not traced to the first weak link.

A Practical Five-Level Ladder

• Level 1 — Model: full worked example with reasoning.
• Level 2 — Finish: final meaningful step removed.
• Level 3 — Extend: several later steps removed.
• Level 4 — Start and finish: only a strategic scaffold remains.
• Level 5 — Independent: conventional problem with no solution steps supplied.

Move forward when the learner shows structural control. Move backward when the route collapses. There is no shame in restoring a scaffold. The scaffold is a tool, not a grade.

Completion and Transfer

A learner can become excellent at completing familiar partial solutions and still fail novel problems.

Transfer requires changing surface features, representations and contexts while preserving the underlying structure. Completion should therefore fade across varied examples, not across copies of the same template.

The end state is not “fill the blank correctly.” It is “recognise and execute the structure when no blank tells you where to think.”

The Deep Principle: Independence Should Be Engineered

Schools often speak about independence as a trait: some students are independent, others are not.

Completion problems suggest a more useful view. Independence can be constructed through carefully managed handoffs.

First the expert carries most of the route. Then the learner carries one part. Then several parts. Then the learner carries the whole route while the expert watches. Eventually the learner operates without the expert in the room.

The tutor disappears by design, not by hope.

Research and Further Reading

• Sweller, van Merriënboer & Paas — Cognitive Architecture and Instructional Design: 20 Years Later
• The consequences of fading instructional guidance on delayed performance
• How Worked Examples Work for Performance
• How Fading Works
• Study & Learning Methods Hub

---

eduKateSG Learning Node Series · 0021. Continue through the Study & Learning Methods Hub.