Learning Science · Questions that learning science still needs to answer

Competing explanations of learning effects

Use competing predictions to distinguish a learning effect from an explanation of how it occurs.

Use the examples to connect this lesson with the earlier topics. No specialist equipment is needed.

You can return to this lesson from the Learning Science map.

After reading, you can

Separate an observed comparison from a proposed mechanism.
Derive different predictions from two explanations.
Recognise when evidence leaves several explanations possible.

An effect does not explain itself

Suppose learners who explain a worked example later solve more related problems than learners who read it silently. The difference, if supported by the design, establishes something about those arrangements. It does not automatically show why the difference occurred. Explanation might direct attention to relationships, expose gaps, increase time on task or change later study.

A mechanism is an account of the processes connecting an activity to an outcome. A useful account should say more than that the activity causes “deeper learning.” It should identify a process and a prediction that could be checked.

Find a comparison on which accounts disagree

Imagine two hypotheses. One says the benefit comes entirely from extra time; another says that generating a relationship matters beyond that time. A design could compare explanation with an equal-time activity that does not require generating the same relationship.

The design will still need accurate implementation and an appropriate outcome. Matching time does not match every mental activity. The point is to reduce a particular ambiguity and make the remaining assumptions visible.

Constructed hypotheses about revision: one attributes the benefit entirely to receiving correct information; another assigns an additional role to explicit old–new comparison.
Account	Both groups receive correction	Prediction for adding explicit comparison
Correct information alone is sufficient in this setting	Relevant ingredient is present in both	No added benefit from comparison, under this account
Comparison adds something beyond receiving correction	Relevant common information is held available	Added benefit predicted on the chosen outcome

Use measurements as evidence, not labels

Recording a learner’s explanation can show which relationships were expressed. It cannot reveal every internal process. A self-report of effort does not prove that a particular memory mechanism occurred. A neural measure adds another observation with its own interpretation requirements.

Several processes can operate together. A teaching method need not have one exclusive mechanism. The scientific question may concern which processes matter under particular conditions and how their contributions can be distinguished.

Keep a useful explanation revisable

A theory becomes more informative when it predicts a failure as well as a success. If an account predicts that a benefit should disappear when a crucial opportunity is removed, that boundary is testable. If every possible result is described as support, the account contributes little.

Do not confuse unresolved mechanism with absence of useful evidence. A practical comparison can be informative before its full explanation is settled. State the effect evidence and mechanism evidence separately so that each can improve.

Worked example

Two explanations for a revision benefit

How could a next study distinguish receiving a correction from explicitly comparing old and new explanations?

In a hypothetical lesson, revising an incorrect explanation improves a later answer. One account emphasises receiving correct information; another emphasises actively comparing the old and new explanations.
A next design can give both groups the correction while varying the explicit comparison activity. Its result would address that distinction, subject to the design, rather than settle every possible account of revision.

Give both groups the correct information while varying the comparison activity, with appropriate timing and outcome checks. The contrast addresses that particular distinction without settling every possible revision mechanism.

Try it yourself

Think through each question before opening the answer.

Why is “it works because it activates the brain” not a sufficient mechanism?
Show the answer and reasoning
Every relevant cognitive activity involves the brain. The statement does not identify a discriminating process or a testable prediction.
An account needs to discriminate among possible activities or outcomes. Naming the brain, which is involved in every relevant activity, supplies neither the specific process nor a differing prediction.
Can a method be useful while its mechanism remains debated?
Show the answer and reasoning
Yes. Evidence for a bounded outcome comparison and evidence for a mechanism are related but distinct.
A well-supported practical contrast can inform a bounded decision before every process is identified. Mechanism claims still need additional evidence rather than borrowing certainty from the outcome comparison.

What to carry forward

Separate effect evidence from mechanism evidence.
Seek predictions that distinguish plausible accounts.
Use boundaries to improve an explanation.

References

Sources supporting the explanations and learning checks in this lesson.