There is a size problem hidden inside almost every training decision.

The task can be too large.

Or it can be too small.

Too large, and the learner fails in so many places that nobody knows what to fix first.

Too small, and the learner becomes excellent at a fragment that has lost its connection to the real performance.

Training granularity is the choice of the right-sized unit of practice for the learner’s current state and the capability we are trying to change.

It is the answer to a deceptively practical question:

How much of the task should we practise at once?

A full examination paper is one size.

A fifteen-minute mixed section is another.

One Mathematics problem is smaller.

One algebraic transformation is smaller still.

One arithmetic fact is smaller again.

None of these sizes is universally correct.

The right size is the one that makes the intended learning mechanism visible while preserving enough context for the skill to remain meaningful.

---

Quick Read: The Goldilocks Problem of Practice

A useful training unit is:

• small enough to diagnose;
• small enough to repeat;
• small enough to receive focused feedback;
• large enough to preserve the relevant decision;
• large enough to retain meaning;
• large enough to reconnect naturally to the whole task.

In compressed form:

Too Large → Noisy Failure | Right Size → Trainable Signal | Too Small → Meaningless Fragment

Granularity Comes After Decomposition

Training Decomposition asks how a complex performance can be broken into meaningful components.

Granularity asks where to stop breaking it down.

Suppose Mira repeatedly loses marks during algebraic rearrangement.

“Mathematics” is too large.

“Algebra” may still be too large.

“Sign control during multi-step rearrangement involving brackets” may be an effective training unit.

“Recognising the minus symbol” is too small.

The correct granularity sits where the actual error mechanism becomes repeatable and observable.

Granularity Comes Before Repetition

Before asking how many repetitions a learner needs, decide what is being repeated.

If the unit is too broad, each repetition may contain different failure causes.

If the unit is too narrow, repetition can automate a fragment that never appears by itself in real performance.

Granularity therefore controls the quality of repetition.

That is why the next article in this batch is Training Repetition.

The Largest Useful Unit: Whole Performance

Sometimes the correct training unit is the entire task.

A full composition.

A complete comprehension section.

A full Mathematics problem that combines several topics.

An unfamiliar Science investigation.

A full paper.

Whole performance is the right unit when integration itself is the target.

For example, a Secondary 4 student may know all the component Mathematics methods but lose marks because switching, timing and prioritisation break under full-paper conditions.

Here the whole paper is not too large.

It is exactly the capability being trained.

When Whole Performance Is Too Large

A whole task becomes too large when failure loses diagnostic resolution.

The learner is wrong, but nobody can tell which part of the chain deserves attention.

A two-hour Mathematics paper contains twenty errors.

Some are conceptual.

Some are algebraic.

Some are timing.

Some come from not finishing.

Another full paper may reveal the same fog.

Reduce the unit.

Perhaps one mixed algebra section is enough to see the first weak link.

The Middle Unit: A Meaningful Subtask

A subtask often gives the best balance between context and precision.

One developed paragraph.

One inference question with its evidence.

One multistep Mathematics problem.

One Science experiment with data and explanation.

The subtask is large enough to preserve relationships but small enough to repeat several times inside one session.

This makes it one of the most useful granularities for tuition.

The Decision Unit

Sometimes the thing that needs training is not a procedure but a decision.

Which method?

Which evidence?

Which relationship?

Which word?

Which representation?

If decision-making is the bottleneck, the training unit should preserve choice.

For Mira, instead of solving ten full equations, a tutor might present ten short problem statements and ask only:

Which method would you choose, and why?

Now ten method-selection repetitions can occur in the time required for two full solutions.

The granularity matches the training target.

The Procedural Unit

Sometimes the learner chooses correctly but execution fails.

Then the decision can temporarily be removed so the procedure can receive concentrated practice.

Mira knows substitution is required but repeatedly mishandles negative signs.

Train the transformation sequence itself.

Nadia knows what a causal explanation needs but omits the measured evidence.

Train several short cause-evidence chains.

Jonas knows the point he wants to make but repeats it in three different sentences.

Train the development move.

Then restore choice and whole-task integration.

The Atomic Unit

The smallest useful unit is sometimes a basic fact, symbol, word or micro-operation.

Multiplication facts.

Spelling patterns.

High-frequency vocabulary.

Common unit conversions.

Formula recall.

These units can benefit from focused retrieval because later tasks require them repeatedly and often with low time cost.

But atomic practice is easy to overuse.

Knowing the atom is not the same as knowing what to build with it.

Granularity and Cognitive Load

The size of the training unit changes how many elements the learner must coordinate at once.

This makes granularity a cognitive-load decision.

A 2026 study in the European Journal of Psychology of Education examined task complexity and pre-training, reinforcing the idea that pre-training relevant components can reduce cognitive demands during later complex problem solving.

The practical implication is not that difficult tasks should always be simplified.

It is that task size should match the learning job.

If the learner needs to understand how three components interact, keep them together.

If one routine component is consuming all available attention, reduce the task until that component can be trained or pre-trained.

Granularity and Worked Examples

A worked example changes granularity without necessarily changing the problem itself.

The learner no longer has to generate every step.

Attention can be directed toward one decision, relationship or transition.

Then support can fade.

A 2025 Learning and Instruction study found benefits when learners were prompted to retrieve upcoming steps inside worked examples before those steps were revealed. This illustrates a useful granularity principle: the training unit can be one step within a larger worked problem while the whole context remains visible.

The learner practises a small unit without losing the map.

Granularity and Transfer

The smaller the practice unit, the more important later recombination becomes.

A child who practises one vocabulary item must later recognise or use it inside real language.

A student who trains one algebraic procedure must later recognise where it belongs.

A learner who practises one Science explanation structure must later adapt it to different experimental contexts.

Small granularity improves practice density.

Large granularity improves contextual fidelity.

Training needs both at different moments.

Granularity and Feedback

Feedback becomes easier to use when the training unit is appropriately sized.

A full composition can generate twenty feedback points.

One paragraph can generate three.

One development move can generate one.

The correct size depends on what the learner can process and reattempt.

If feedback is too broad to change the next attempt, reduce granularity.

If feedback becomes so microscopic that the learner loses the whole writing purpose, increase it again.

Granularity and Time

Smaller units increase repetition density.

A learner may make ten method-selection decisions in ten minutes if calculation is removed.

The same ten decisions might require an hour if every problem is solved fully.

This can make small-unit practice extremely efficient when the decision itself is the bottleneck.

But efficiency must not be confused with completeness.

The learner eventually has to perform the whole solution.

Granularity allows us to concentrate repetitions where they have the highest value, then return to authentic scale.

Granularity for Mathematics

Mathematics offers a natural ladder of sizes.

• fact;
• operation;
• procedure;
• method-selection decision;
• single problem;
• mixed set;
• timed section;
• full paper.

Use the smallest level that still contains the current failure.

If Mira forgets the formula, fact-level retrieval may help.

If she remembers the formula but cannot choose it, method-selection granularity is better.

If she chooses and executes correctly but cannot finish the paper, full-section or paper-level granularity is required.

The learner’s error tells us where to work.

Granularity for English Reading

Reading can be trained at several scales.

• word;
• sentence;
• sentence relationship;
• paragraph;
• passage;
• question set;
• complete paper section.

If Jonas misreads one key vocabulary item, word-level repair may be enough.

If he understands every sentence but loses the author’s argument across paragraphs, passage-level training is required.

If he reads well but cannot manage time across the comprehension section, the unit must grow again.

Granularity for English Writing

Writing can be trained at:

• word choice;
• sentence;
• sentence pair;
• paragraph;
• section;
• whole composition;
• timed paper.

If Jonas’s vocabulary is imprecise, word-level contrast may help.

If sentences are individually good but the paragraph repeats, paragraph granularity is better.

If paragraphs are strong but the composition loses direction, whole-composition granularity is necessary.

The right unit follows the failure.

Granularity for Science

Science training units can include:

• term;
• concept;
• causal relationship;
• variable decision;
• graph interpretation;
• single open-ended question;
• investigation;
• topic set;
• mixed paper.

Nadia may know the concept but fail at variable recognition.

Do not reteach the whole topic.

Train several variable decisions, then return them to investigations.

Granularity for Vocabulary

Vocabulary is often trained at the word level because words are convenient units.

But meaning lives in context.

A useful vocabulary granularity ladder is:

Word → Contrast Pair → Sentence → Paragraph → Reading/Writing Context

Retrieve the word at small scale.

Then increase the unit until the learner can select it naturally in language.

Granularity for Examination Training

Full papers are expensive.

They are useful when the target is integrated examination performance.

They are inefficient when the target is one narrow recurring error.

A good exam-training system moves across granularities.

Full Paper → Identify Error Class → Narrow Repair → Mixed Section → Full Paper

This prevents students from using two-hour papers to practise ten-minute weaknesses.

Granularity and Training Readiness

The right size changes with readiness.

A novice may need a smaller or more supported unit.

An expert learner may need a larger, more authentic unit because small drills no longer create useful challenge.

This is why Training Readiness changes granularity.

The practice unit should grow as the learner can coordinate more of the performance independently.

Granularity and Training Constraints

If the whole task contains too many possible failure points, decrease granularity until the constraint becomes visible.

If the learner succeeds at every smaller unit but still fails the whole, increase granularity.

The missing capability may be integration itself.

Granularity and Training Signals

Smaller units create sensitive signals.

You can see whether one error recurrence is falling.

Larger units create ecological signals.

You can see whether the capability survives realistic complexity.

Use both.

Training Signals should be interpreted relative to the size of the task that generated them.

Granularity and Training Maintenance

Maintenance often uses larger authentic units than acquisition.

Once Mira’s sign control is stable, there is little need to maintain it through isolated sign drills.

Let normal Mathematics tasks maintain it.

Once Jonas’s paragraph development is stable, full writing can maintain the skill.

Maintenance allows the granularity to expand because the part no longer needs concentrated attention.

Granularity and Training Independence

As learners become more independent, they should learn to resize their own practice.

A mature student looks at a weak full paper and does not automatically decide to do another full paper.

The student asks:

What is the smallest useful thing I need to repair before I return to the whole?

This is a sophisticated self-regulation skill.

It reduces wasted effort.

Failure Mode: The Unit Is Too Large

• too many errors occur at once;
• feedback becomes overwhelming;
• the learner cannot tell what changed;
• reattempt requires too much time;
• the same narrow weakness is repeatedly buried inside large tasks.

Reduce granularity.

Failure Mode: The Unit Is Too Small

• the learner never chooses when to use the skill;
• context disappears;
• the drill becomes mechanical;
• transfer remains weak;
• whole-task performance does not improve despite high drill accuracy.

Increase granularity.

Failure Mode: The Unit Never Changes

A learner begins with one-sentence practice.

Months later, the learner is still doing one-sentence practice.

A student begins with isolated algebra.

Months later, algebra has still never returned to mixed work.

Training granularity should change as capability grows.

That is progression.

Mira’s Granularity Ladder

Mira’s current Mathematics weakness is sign control during multi-step algebra.

Too large: full examination paper.

Useful first size: short algebraic transformations containing signs and brackets.

Next size: complete equations.

Next: mixed algebra.

Next: algebra embedded inside geometry and trigonometry.

Final: full paper.

The training unit grows as the repaired capability becomes more robust.

Jonas’s Granularity Ladder

Jonas’s current writing weakness is paragraph development.

Too large initially: full timed composition.

Useful first size: one claim with one explanation and example.

Next: full paragraph.

Next: two connected paragraphs.

Next: full composition.

Final: timed composition with self-editing.

The skill is not left microscopic.

It is grown back into writing.

Nadia’s Granularity Ladder

Nadia’s current Science weakness is linking evidence to conclusion.

First size: one data statement and one inference.

Next: one short experiment.

Next: unfamiliar experiment with irrelevant information.

Next: mixed Science section.

Final: full paper under realistic timing.

Again, the unit grows with readiness.

The Granularity Decision Tree

Is failure too noisy to diagnose?
Make the unit smaller.

Does the learner know exactly what skill is being tested every time?
Consider making the unit larger or mixing alternatives.

Does feedback contain too many corrections to reattempt?
Make the unit smaller.

Is drill accuracy high but whole-task performance weak?
Make the unit larger.

Is the learner overwhelmed by interacting elements?
Reduce or support the unit.

Is the learner bored and over-supported?
Increase the unit, variation or independence.

The Family-Life Value of Correct Granularity

Correct granularity saves time.

A child does not need another two-hour paper to repair a ten-minute error class.

A student does not need to write a full essay every time paragraph development needs practice.

A Science learner does not need to revise an entire chapter when the weak link is evidence construction.

At the same time, a learner should not spend months on tiny drills that never return to real school demands.

Granularity protects both precision and life.

It concentrates effort where it matters, then gives the time back.

A Parent’s Granularity Audit

• Is the current task so large that we cannot tell why the child is failing?
• Is the current drill so small that it has lost the real decision?
• Can the learner receive focused feedback and reattempt?
• Does the practice unit match the actual weak link?
• Is the unit growing as capability improves?
• Does the repaired part eventually return to school-like work?
• Are full papers being used for integration rather than every small repair?

A Tutor’s Granularity Audit

• What is the whole performance?
• What component is failing?
• What is the smallest unit that still contains the relevant decision?
• How many meaningful repetitions does this unit allow?
• What context is lost at this size?
• When will the unit grow again?
• What whole-task test will prove transfer?

The Deeper Idea: Good Training Can Zoom

Complex learning is easier to manage when the training system can change scale.

Zoom out to see the whole performance.

Zoom in to find the failure.

Zoom further if the mechanism is still hidden.

Stop when the unit becomes trainable.

Then zoom back out as capability stabilises.

The right training unit is not the smallest thing we can isolate. It is the smallest thing worth isolating for the change we need.

Research Foundations

This article draws on whole-task instructional design, cognitive-load research, worked examples and transfer. Useful recent sources include the 2025 systematic review of Four-Component Instructional Design, the 2026 4C/ID review in higher education, the 2026 study of task complexity and pre-training, and recent work on variability and generalisation. Together, these support a flexible design principle: complex capability should remain connected to meaningful whole tasks, while component practice is used at a granularity that reduces unnecessary load without stripping away the relationships learners ultimately need to recognise and coordinate.

Continue Through How Training Works

Read Training Decomposition and Training Recombination alongside Training Readiness, Training Signals and Training Review.

The next article asks what should happen once the training unit is correctly sized: How Training Works | Training Repetition — Repeat the Decision, Not Just the Worksheet.