Learning how to learn means learning to control the loop
Learning how to learn, effective learning strategies, retrieval practice, spaced practice, interleaving and metacognition are often presented as separate study tips. They make more sense as parts of one system: choose a capability, inspect the starting state, learn what is missing, retrieve it, practise the right decision, receive feedback, return after time and test whether the knowledge travels.
Cornell’s Learning Strategies Center prominently organises effective study strategies around retrieval practice, interleaving and spaced practice. Those are strong mechanisms, but no technique is a universal answer. Retrieval cannot replace first instruction when knowledge is absent; spacing cannot rescue a misconception; interleaving cannot help if every component is still unstable. The strategy has to fit the state.
Return to How Learning Works →
1 · Start with the future performance
“Study Chemistry” is an activity category. “Explain why the temperature changes and apply the model to a new apparatus” describes a capability. Effective learners increasingly define the performance they are trying to build before choosing a method.
Alicia may need to select evidence, Beatrice to identify a changing reference quantity, Ciara to distinguish the requested quantity from a familiar one, Denise to reconstruct a logical condition, Emily to select an algebraic method, and Faith to test the boundary of a shortcut. The same study technique cannot be assumed to repair all six.
2 · Diagnose before adding volume
A short attempt can reveal whether the bottleneck is knowledge, retrieval, representation, selection, execution, checking or transfer. More work is useful only when the work contains the operation that needs strengthening.
This is why a marked paper can be more valuable than another unanalysed paper. The score locates a region; the working reveals the route.
3 · Retrieval practice turns memory into an action
Close the source and reconstruct. Explain the concept. Solve without the worked example. Draw the process. State the rule and its conditions. Retrieval both tests access and can strengthen later access.
The important distinction is between actual retrieval and recognition. Looking at an answer and thinking “yes, I knew that” is not the same operation as producing it before the answer appears.
Use How Retrieval Practice Works for the mechanism owner.
4 · Spacing makes retrieval necessary again
Massed practice can create rapid within-session fluency. Spacing changes the memory problem by allowing time and context to intervene. The learner must reconstruct rather than continue the same episode.
There is no single perfect interval. Return while the knowledge is still recoverable but after enough time that retrieval has meaningful work to do. Let the result change the next interval.
Use How Spacing Works in Learning.
5 · Interleaving trains method selection
Blocked practice can help acquire a new procedure. Mixed practice becomes valuable when the learner must decide which procedure applies. The heading should eventually stop making the first decision for the student.
Emily may factorise accurately on a factorisation worksheet yet hesitate in a mixed set. That is not contradiction. Execution and selection are different capabilities.
6 · Feedback must change a fresh attempt
Feedback is not complete when a comment is delivered. It becomes educational when the learner can use it. Correct, compare, explain, then attempt again without the answer doing the thinking.
“Be careful” is weak feedback because it does not identify a controllable decision. “You compared final temperatures, but the question asks for the decrease” gives Ciara a next action.
7 · Metacognition calibrates feeling against evidence
Familiarity feels like knowledge. A difficult retrieval attempt can feel like failure. Metacognition improves when learners compare their predictions with actual performance and update strategy accordingly.
Cornell’s teaching resources describe metacognitive strategies as ways to help learners recognise what they know, notice errors and take greater control of learning. The practical version is compact: predict, attempt, inspect, adjust.
8 · Self-explanation exposes hidden links
Ask why a step is valid, what changed between two cases, what evidence supports a claim or what assumption makes a shortcut safe. Explanation reveals whether knowledge is connected or merely familiar.
But explanation is not magic. A learner cannot explain accurately from knowledge they do not possess. Teach first when necessary.
9 · Comparison builds boundaries
Put a case where a rule works beside one where it fails. Ask what changed. Faith’s average-speed shortcut becomes more intelligent when equal-distance and equal-time journeys are compared. The learner acquires a condition, not merely a prohibition.
10 · Worked examples should fade
Worked examples reduce unproductive search while a novice builds a schema. Their job is not to remain permanently beside every problem. Remove steps, delay hints, ask for method selection, then test a fresh problem.
Support is successful when more of the next performance belongs to the learner.
11 · Transfer is the real test of flexible learning
The National Academies describes transfer to new problems and settings as an important index of flexible learning. Transfer is often difficult because learners bind knowledge to the contexts and cues in which it was acquired.
Change the numbers, representation, wording, context or combination of ideas. Ask what survives. Transfer should be designed rather than assumed.
12 · Sleep and recovery protect the learning conditions
Study time cannot be evaluated independently of attention and recovery. An extra late-night hour may add exposure while degrading the next day’s learning conditions. Good learning plans consider the quality of the cognitive operation, not only minutes accumulated.
13 · Notes are infrastructure, not proof
Notes can organise external memory. Their value increases when the learner later reconstructs without them. A perfect notebook can become a hidden scaffold if headings, examples and definitions permanently supply decisions that future performance requires internally.
14 · Flashcards are good servants and poor curricula
Flashcards are excellent for compact retrieval targets. They become weak when a complex subject is shrunk into disconnected prompts. Retrieve components, then reconnect them to problems, explanations, writing and transfer.
See How Studying With Flashcards Works.
15 · AI assistance should end in independent performance
AI can explain, quiz, compare, generate examples and give feedback. It can also perform the exact thinking the learner needs to practise. Attempt before assistance when feasible, request targeted help, verify important claims and then close the tool and perform again.
See How AI-Assisted Study Works.
16 · Motivation is part of the control system
Effort needs value, expectancy and a workable route. “Try harder” supplies none of these. Visible progress, meaningful goals, proportionate choice and reduced avoidable friction can make return more likely.
17 · Learning strategies are domain-sensitive
General strategies do not float free of subject knowledge. Critical thinking in biology requires biology. Mathematical problem solving requires mathematical concepts and representations. Reading comprehension depends partly on vocabulary and background knowledge.
The National Academies’ work on transferable skills emphasises that problem-solving and metacognitive competencies should be taught within disciplines rather than treated as content-free mental powers.
18 · Alicia: learn to ask what the evidence proves
Alicia’s strategy is not “do more comprehension.” It is to identify the claim, select evidence, explain the connection and test the same decision in a new passage. The strategy is attached to a decision.
19 · Beatrice: learn to recover the reference quantity
Beatrice practises labelling the changing whole, retrieves the relation after delay, then uses it under time. Concept repair and examination control are trained separately before being recombined.
20 · Ciara: learn where speed needs a checkpoint
Ciara does not need to become generally slower. She needs high-value checks attached to decisions that repeatedly produce error. Automaticity remains an asset; control protects it.
21 · Denise: learn to create a smaller doorway
When language overloads the task, Denise externalises the relation in two cases, then returns to the required written form. Adaptation provides access without silently removing the target capability.
22 · Emily: learn to remove the hidden decision-maker
Emily’s worked examples are gradually faded until she must choose the method. The notebook becomes a reference rather than a co-pilot.
23 · Faith: learn the conditions of the shortcut
Faith’s strong pattern recognition becomes more expert when every shortcut carries a boundary question: what has to be true for this to work?
24 · A practical learning-how-to-learn loop
Define the capability. Attempt. Locate the first useful uncertainty. Learn what is missing. Retrieve. Practise. Receive feedback. Reattempt. Wait. Return. Mix. Transfer. Calibrate. Reduce support. Repeat where evidence says the system remains fragile.
The loop is deliberately correctable. A failed transfer task may send the learner back to prior knowledge. A retrieval failure may reveal shallow encoding. A repeated execution error may call for focused practice. Learning how to learn is the ability to update the route rather than worship the routine.
25 · The destination is not independence from people
Independent learning does not mean refusing teachers, books, peers or tools. It means increasingly knowing what resource is needed, why, how to judge it and how to return to performance after using it.
The learner becomes better at running the loop.
