How Coherence Works in Learning | Remove What Competes With the Lesson

eduKateSG Learning Node Series · 0010

A lesson can become worse by becoming more interesting.

That sounds absurd until you watch what attention actually has to do.

A science slide explains the greenhouse effect. The diagram is accurate. The causal chain is important. Then someone adds a dramatic photograph, a cartoon, a decorative fact about the hottest recorded day, a soundtrack, three quotations and an animated icon because the slide feels too empty.

The learner now has more to look at, more to interpret and more possible organising themes. Nothing added is necessarily false. Some of it may even be fascinating. But if it competes with the structure the learner needs to build, it creates processing without creating learning.

The coherence principle in multimedia learning asks a disciplined question: does this element help the learner build the target model, or does it merely occupy attention?

Quick Read: Coherence Is About Relevant Structure

The 2021 Cambridge Handbook of Multimedia Learning summarises coherence as a principle for reducing extraneous processing: people tend to learn better when extraneous words, pictures and sounds are excluded from multimedia lessons rather than added.

The important word is extraneous. Coherence does not mean “make every lesson short.” It does not mean “remove all stories.” It does not mean “never use humour, colour or examples.” It means that every element should have a defensible relationship to the learner’s job.

Coherence is not minimalism for its own sake. It is protection of the model the learner is trying to construct.

The Museum Gift-Shop Problem

Imagine visiting a museum exhibition about one ancient civilisation. You enter expecting a clear route through agriculture, trade, government, writing and war. Instead, every room contains unrelated objects because each one is beautiful.

The objects may all deserve attention. The exhibition still fails if the visitor cannot tell what story the collection is trying to build.

Learning materials can fail the same way. A lesson becomes a collection of interesting things rather than a designed route through a concept.

Coherence is the editorial discipline that keeps the route visible.

Why Irrelevant Material Is Expensive

Attention is not free. Every element can create at least one demand: notice me, interpret me, decide whether I matter, connect me, suppress me or remember me.

If a learner is already near working-memory limits, those demands can displace processing that should have gone to the essential relationships.

Extraneous material can hurt in several ways. It can divert attention from the important element. It can interrupt organisation. It can suggest the wrong organising theme. It can encourage the learner to remember a vivid anecdote while missing the mechanism the anecdote was meant to illustrate.

This is one reason memorable teaching and memorable learning are not identical. A learner can remember the joke and forget the law of conservation.

Interesting Is Not the Same as Instructionally Relevant

Teachers often add material because it increases engagement. Engagement matters, but engagement has to be attached to the learning target.

An unusual fact can be useful if it exposes a principle. A story can be useful if it creates a causal model. A photograph can be useful if it makes an abstract relationship visible. Humour can lower tension or improve attention.

The same elements become extraneous when they pull the learner into a different model.

The question is not “Is this interesting?” It is “What thinking does this cause, and is that thinking part of the instructional goal?”

The Seductive Details Effect

Research on so-called seductive details examines material that is interesting but irrelevant to the learning goal. The problem is not that the learner enjoys the detail. The problem is that vivid, emotionally salient or unusual information can attract processing disproportionate to its instructional value.

A paragraph about lightning may include a dramatic account of a famous historical strike. If the learning target is charge separation and discharge, the story should either illuminate that mechanism or move elsewhere.

Good teaching can still be vivid. Coherence asks vividness to serve structure.

The Decorative Image Problem

Images are often added because pages look unfinished without them.

But a decorative image may consume visual attention without carrying explanatory information. Worse, it may suggest a relationship that does not exist.

A meaningful diagram should answer a question: Where is the component? What changes? What connects to what? What is larger, earlier, faster or closer? A decorative image merely occupies space.

Design is strongest when the learner can explain why the visual exists.

The Music Problem

Background music can make educational media feel polished. But polish is not evidence of better learning.

If music competes with narration, carries lyrics, changes arousal or adds an additional stream the learner must suppress, it may consume resources without supporting the target model.

This does not imply that all music is harmful in all studying. Different tasks place different demands on language, attention and memory. Coherence asks whether the sound serves the current job.

Coherence and Longform Writing

Long articles are not automatically incoherent.

A 20,000-word guide can be coherent if every section earns its place, the hierarchy remains visible, repetition performs a useful return function and examples deepen the same central model. A 700-word article can be incoherent if it jumps between loosely related ideas.

Coherence is therefore not word count. It is structural discipline.

The useful test is compression: if a section were removed, what capability, distinction, evidence or route would the learner lose? If the answer is “nothing important,” the section may be decoration.

Coherence in Mathematics

A mathematics worked example should make the mathematical structure easier to see, not bury it beneath commentary.

Suppose a solution contains every arithmetic detail, three historical facts about the formula, a sidebar about the mathematician who developed it and four alternate notations before the learner understands the basic relationship. The material may all be true, but the instructional sequence is overloaded.

Coherent mathematics teaching protects the invariant. It makes clear what changed, what stayed fixed, which rule authorised the step and what decision the learner must eventually make alone.

Coherence in Science

Science benefits from rich context, but context should sharpen causality rather than obscure it.

When teaching osmosis, the essential model concerns particles, concentration, a partially permeable membrane and net movement. A story about a dramatic medical case may motivate the topic, but the lesson still needs to return cleanly to the mechanism.

If students leave remembering the story but cannot explain the direction of net water movement, engagement did not complete the learning job.

Coherence in English

English lessons can become incoherent when too many objectives are stacked into one text.

A single passage becomes vocabulary, grammar, comprehension, punctuation, literary analysis, current affairs and composition inspiration simultaneously. Every objective is legitimate. The learner may no longer know what the current decision is.

A coherent lesson can still reuse one passage, but it separates jobs. First infer meaning. Then inspect sentence structure. Then analyse tone. Each pass has a clear target.

Coherence in Revision Notes

Students often respond to uncertainty by adding more notes.

The notebook grows. Every teacher comment is copied. Every textbook detail enters. Every online explanation is appended. The learner now possesses more information but less navigability.

Coherent revision notes have hierarchy. Core relationships are visually dominant. Supporting examples are subordinate. Exceptions are attached to the rules they modify. Definitions sit near their use.

The notes should make the next retrieval route easier to see.

Coherence and Signaling Are Different

Coherence removes what should not compete. Signaling highlights what should matter.

A cluttered slide can be signaled beautifully and still remain cluttered. A perfectly concise slide can still be hard to parse if its structure is invisible.

Use both principles deliberately: remove unnecessary processing, then guide attention through the necessary structure.

See How Signaling Works in Learning.

Coherence and Segmenting Are Different

Segmenting changes the size and pacing of the essential material. Coherence changes how much nonessential processing surrounds it.

A lesson can be perfectly segmented and still contain irrelevant detail in every segment. Or it can be coherent but delivered too quickly.

See How Segmenting Works.

Coherence and Prior Knowledge

An expert can ignore distractions that consume a novice.

Experts already possess schemas that identify what matters. They can often classify an aside as irrelevant and return to the main model. Novices may not know which detail is central and which is optional.

This makes coherence especially important early in learning. The less the learner knows, the less reliably the learner can perform the filtering that the material failed to perform.

The Expertise Boundary

As knowledge grows, richer context can become useful rather than distracting.

An expert historian can profit from a tangential comparison because it activates a larger network. An introductory learner may lose the central chronology. An advanced mathematics student may benefit from alternative proofs. A novice may need one clean route first.

Coherence therefore depends partly on learner state. The same detail can be scaffolding for one learner and noise for another.

The Motivation Boundary

There is a legitimate concern: remove every story, image and surprise, and the lesson may become sterile.

Motivational value is real. But motivation should be treated as part of the learning system, not as a free decoration layer.

A story earns its place if it creates curiosity about the mechanism, gives the learner a memorable application, or provides a meaningful context for transfer. The design failure occurs when the story becomes the main memory and the concept becomes background.

The Personalisation Boundary

Conversational language can support learning even though extra words can also create load. These ideas are not contradictory.

A concise conversational sentence may be easier to process than a formal but dense paragraph. The relevant question is whether the words create useful social and explanatory structure or unnecessary elaboration.

Coherence removes words that do not earn cognitive rent. It does not require robotic prose.

A Teacher Editing Protocol

State the one learning job of the lesson.
Mark every element that directly supports that job.
Mark examples that clarify a boundary, mechanism or application.
Remove decorative elements that create no instructional action.
Move useful but nonessential enrichment to a later section.
Reduce duplicated explanations.
Keep labels and definitions close to their use.
Check whether the central relationship is still visible after the edit.

A Student Editing Protocol

Students can apply coherence to their own notes.

Circle the three to five relationships that the topic depends on.
Move examples underneath the relationship they illustrate.
Delete copied sentences that repeat the same meaning.
Separate “must know” from “interesting extension.”
Turn long descriptive passages into diagrams when spatial structure matters.
Keep exceptions beside the rule they modify.
Test the notes by closing them and reconstructing the model.

Coherence in AI-Assisted Learning

AI makes it exceptionally easy to add explanation.

Ask for more examples, more analogies, more context, more definitions, more history and more elaboration, and the system can generate all of it instantly.

The new scarcity is not information. It is attention and judgement.

A useful AI learning workflow therefore asks for the smallest explanation that resolves the current gap, then expands only where the learner remains uncertain. More content should be triggered by need, not by the model’s ability to produce it.

The Search-Engine Boundary

Online articles face another tension. Writers want comprehensive coverage. Search systems reward pages that answer many related questions. Learners need a route that remains coherent.

The answer is not thin content. It is architecture.

Longform pages should separate the quick route from the deep route, preserve clear headings, attach tangents to the right section and link outward when another page owns the deeper topic. Comprehensive does not have to mean everything mixed together.

What Recent Synthesis Adds

The research base behind multimedia-learning principles is substantial, but contemporary synthesis cautions against treating effect sizes as universal constants. A 2025 meta-analysis of Mayer’s multimedia-learning research examined variability across principles, media, domains and learner characteristics.

The practical lesson is stronger than a slogan: use coherence as a default hypothesis, then test it against the actual task. If an apparently “extra” element improves understanding, motivation, transfer or accessibility for the learner in front of you, it may not be extraneous after all.

The Deeper Principle: Attention Should Have a Job

Every second of attention spent interpreting an irrelevant element is a second that cannot be spent on something else.

This does not mean learning should be joyless. It means joy, curiosity, imagery, story and explanation should point into the model rather than away from it.

The best lesson can be rich and coherent at the same time.

Its richness has structure.

Use This Tomorrow

Take one page of notes, one slide or one explanation. Ask of every element: What learning job does this perform? Keep what clarifies the model, exposes a boundary, supports retrieval or enables transfer. Move the rest out of the main route.

The goal is not less learning material. It is less competition for the learning that matters.

Research and Further Reading

eduKateSG Learning Node Series · 0010 of the continuing series. Previous: 0009 — How Modality Works in Learning.