Learning Principles: Turn Effort into Evidence

Before choosing an app or textbook, decide why you are learning, what task you need to complete, how you will judge it, and how to train without sacrificing health.

Separate Three Kinds of Claim

• Research evidence reports effects observed under stated conditions, not a guarantee for every learner.
• Personal experience explains why I made a choice; it is a clue for your experiment, not a universal rule.
• Hypothesis lacks enough evidence and should become a question to test.

When advice omits the population, conditions, outcome, and source, treat it as a hypothesis.

Why Learn English?

English connects you to technical documentation, research, courses, cultural work, and international collaboration. That importance is still not an actionable goal. Name the context:

• an exam with a date, target score, and task types;
• study that requires discussion, academic reading, or writing;
• work that requires documentation, email, meetings, interviews, or presentations;
• life that requires travel, reading, or communication with particular people.

My primary context is ______. In 12 weeks I will complete ______ under ______ conditions. The evidence will be ______.

The context determines material and feedback. Do not distribute time evenly in pursuit of a vague wish to “improve English”.

Remove the Learning-pyramid Percentages

Earlier editions repeated claims such as “people retain 10% of reading and 90% of practice”. Those precise numbers have no reliable research basis and are not what Edgar Dale's Cone of Experience proposed. This edition removes the old diagram and percentages.

Keep a testable principle instead: repeated exposure can create familiarity, while retrieving, explaining, applying, and receiving feedback reveals whether knowledge is usable.

Two broadly supported strategies are:

• retrieval practice: close the source, recall, answer, retell, write, or teach;
• distributed practice: return at multiple points instead of cramming once.

Neither offers universal percentages or a universal calendar. Difficulty, prior knowledge, delay, and assessment method all matter.

A Complete Learning Loop

Each 25–45 minute session should contain:

1. Task: “retell a two-minute news clip”, not “practise listening”.
2. First attempt: produce a baseline without excessive lookup.
3. Targeted input: review only the information needed for current errors.
4. Active output: close the material and perform again.
5. Feedback and record: compare versions and update Learning State.

Finishing a video or feeling that it made sense is not the endpoint. Performance after the source is closed is stronger evidence.

Output Is Not Performance Theatre

Active output can be small:

• write a three-sentence summary from memory;
• list a main point and three details after listening;
• explain a concept without notes;
• use a new chunk in two settings;
• record sixty seconds, replay it, and mark pauses.

Output exposes gaps. Its job is not to prove intelligence.

Let Performance Set the Interval

A schedule such as days 1, 2, 4, 7, and 15 can be a reminder, but it is not Ebbinghaus's universal prescription for you.

• Easy and accurate recall: lengthen the interval or raise difficulty.
• Effortful success: keep a similar interval.
• Frequent failure: shorten the interval, reduce material, or repair prerequisites.
• Recognition without use: add production and varied contexts.

Record performance and adjust. Missing an imagined “perfect review time” is not a reason to abandon the system.

Sleep and Recovery Are Inputs

Personal experience

At school I once confused visible effort with learning: waking at 5:20, skipping midday rest, and reading after lights-out. Time rose while performance fell. I interpreted this as insufficient effort and pushed harder.

That story does not prove all early rising is ineffective. It shows that for me, chronic sleep loss, no error diagnosis, and mechanical hours combined badly.

Evidence boundary

Sleep contributes to memory consolidation, while sustained sleep loss can affect attention and performance. Needs and health conditions differ. This guide gives no medical diagnosis; persistent sleep, mood, or physical problems deserve qualified professional care.

Budget sleep, movement, food, and unassigned time. Recovery is a condition for the next session, not a prize after it.

Do Not Match Fixed “Learning Styles”

People may prefer pictures, sound, or text, but evidence does not support classifying people as visual or auditory learners and improving outcomes by matching instruction to that label.

Choose representation by content and task:

• pronunciation needs sound, articulation, and acoustic feedback;
• spelling needs visual recognition and production;
• spatial relationships often benefit from diagrams;
• communication integrates listening, speaking, reading, and writing.

Multiple representations can help. A preference is not a fixed identity or ability ceiling.

Compare with Your Baseline

A useful goal includes a real context, observable task, quality threshold, deadline, and comparable evidence. Use the CEFR self-check for four separate baselines. An average label cannot describe B2 reading and A2 speaking.

Order Feedback by Impact

Treat only one to three high-impact problems at a time:

1. task completion and meaning;
2. organisation and logic;
3. recurring vocabulary, grammar, or pronunciation errors;
4. low-impact details last.

AI can classify errors and create parallel practice, but it can misjudge facts, pronunciation, and usage. Check important work against dictionaries, corpora, teachers, peers, or real audiences.

Seven Days, Thirty Days, Twelve Weeks

Seven days: baseline

• save short samples in all four skills;
• choose one real context;
• complete three full learning loops;
• change only one variable from the evidence on day seven.

Thirty days: rhythm

• keep at least one comparable output each week;
• classify errors instead of calling everything “careless”;
• replace exercises gradually with authentic material;
• repeat the baseline conditions on day thirty.

Twelve weeks: transfer

• save parallel samples in weeks 1, 4, 8, and 12;
• add time pressure, unfamiliar topics, or a real audience;
• complete an exam simulation, meeting, presentation, long reading, or writing deliverable;
• let the result define the next cycle.

Action

1. Complete the English Diagnostic.
2. Create a Learning State.
3. Delete one task that produces effort but no evidence.
4. Schedule the first 25–45 minute learning loop.

Sources

• Dunlosky et al. (2013), Improving Students' Learning With Effective Learning Techniques
• Roediger & Karpicke (2006), Test-Enhanced Learning
• Cepeda et al. (2006), Distributed Practice in Verbal Recall Tasks
• Pashler et al. (2008), Learning Styles: Concepts and Evidence

Previous: CEFR Goals | Next: Vocabulary