A textbook page shows a paragraph on the left and a labelled diagram on the right. The student reads the paragraph, glances at the picture and moves on.
The page contains two representations. The learner may have built only one idea.
Visuals can support learning because some relationships are easier to show spatially than verbally: parts inside a system, movement through a cycle, forces acting in different directions, layers, sequences, proportions and mappings. Text can name, qualify and explain relationships that a picture cannot make explicit by itself. The educational advantage appears when the learner connects the two forms. If the labels are far from the objects, if the text describes features the student cannot locate, if the diagram contains decorative clutter, or if no prompt directs comparison, the learner can process text and visual separately—or ignore one entirely.
The What Works Clearinghouse practice guide Organizing Instruction and Study to Improve Student Learning gives moderate evidence to combining graphical representations with verbal descriptions. Institute of Education Sciences research on multiple representations, spatial alignment, virtual manipulatives and textbook design reinforces the mechanism: correspondence, proximity, signalling and translation matter because learners must know which element in one representation maps to which element in another.
This article owns one narrow canonical job on eduKateSG: integration of verbal and visual representations through mapping, alignment, signalling and learner translation so that text and diagrams form one coherent mental model. It does not own generic “dual coding,” visual note-taking, concrete–pictorial–abstract progression, multimedia learning as a whole or diagramming as an isolated skill.
The useful question is not, “Did we add a picture?” It is: can the learner identify what each visual element corresponds to in the words, explain the relationship, and move between the representations without losing the underlying idea?
The 50-second answer
Text-and-diagram integration works when the two representations carry complementary information and are easy to map.
- Choose a relationship that benefits from a visual form.
- Remove decorative detail that does not serve the learning goal.
- Place labels and explanations near the relevant visual elements where possible.
- Use consistent terminology, symbols and colour meaning.
- Signal the parts learners should compare.
- Ask students to point, trace, label or explain correspondences.
- Translate in both directions: text → diagram and diagram → text.
- Check whether the learner can reconstruct the relation without the original layout.
- Vary diagrams so students learn the structure, not one picture.
- Keep accessibility and alternative representations in view.
The shortest principle is: a visual helps when learners know what to look at, what it stands for and how it connects to the words.
1. Visual presence is not visual learning
Students can look at a diagram without extracting its structure.
Eye contact with an image is not evidence of mapping.
Teachers need tasks that require the learner to use the visual information.
2. Text and visuals should do complementary jobs
If the diagram simply repeats every word, it may add little. If it introduces a new relationship without explanation, it may confuse.
Strong combinations divide labour: text explains causal or conditional meaning; visuals display spatial or structural relations.
The two forms should meet around one concept.
3. Spatial relationships are natural visual candidates
Anatomy, circuits, geography, geometry and systems often depend on where parts are relative to one another.
A diagram can display these relationships at a glance.
Text should direct attention to the relationships that matter rather than describe every visible feature.
4. Process diagrams show sequence and branching
Cycles, algorithms and biological processes can be represented through arrows and stages.
Learners need to understand what arrows mean: movement, causation, sequence, transformation or information flow.
Never assume arrow semantics are obvious.
5. Graphs are visual representations with their own grammar
Axes, scales, legends, data points and slopes carry meaning.
Text explaining a graph should connect claims to visible features: “the line steepens here,” “the groups overlap,” “the value peaks at…”
Students need graph literacy, not merely exposure.
6. Labels reduce search when placed close to the feature
If students must look repeatedly from numbered diagram to distant key, they spend working memory on locating correspondences.
Direct labels can reduce unnecessary search when the layout permits.
Proximity is an instructional design choice.
7. Legends are useful when repetition would clutter the visual
Maps and complex diagrams cannot label every element directly.
When a legend is necessary, keep symbols distinctive and the mapping simple.
Teach students to use the legend rather than assuming they will.
8. Spatial alignment can clarify correspondence
Place related text and images near one another. Align equation steps with diagram stages. Put captions under the exact panel they describe.
Distance creates search and increases the risk that learners connect the wrong elements.
9. Alignment is conceptual as well as physical
A paragraph may use “left ventricle” while the diagram says “LV.” A graph may use “velocity” while the text alternates with “speed.”
Inconsistent terms can make mapping harder.
Use consistent language or explicitly teach the equivalence.
10. Colour should carry stable meaning
If red represents oxygenated blood in one diagram and deoxygenated blood in the next, colour becomes misleading.
Use colour sparingly and consistently.
Never rely on colour alone because of colour-vision accessibility.
11. Decorative colour can compete with signal
Bright backgrounds, icons and illustrations can attract attention to features unrelated to the learning goal.
Beauty and clarity can coexist, but decoration should not outrank information.
Ask what each visual element is doing educationally.
12. Signalling tells learners where to allocate attention
Arrows, highlights, bold labels or numbered steps can direct attention to critical relations.
Signalling is especially useful for novices who do not yet know what experts notice automatically.
Fade some signals later so learners practise independent selection.
13. Animation is not automatically better than static diagrams
Movement can show dynamic processes and can disappear before the learner has inspected it.
Allow pause, replay or step-through control.
For some processes, a sequence of static panels supports comparison more effectively.
14. Learner control matters in dynamic representations
Students should be able to revisit difficult stages.
Continuous animation can overload attention when several elements change at once.
Segment complex processes.
15. Diagrams simplify reality deliberately
A heart diagram may distort size and location to make chambers visible. A circuit diagram replaces physical wires with abstract symbols.
Students need to know what the representation preserves and what it changes.
This is part of representational competence.
16. Photographs and diagrams serve different purposes
A photograph preserves surface appearance. A diagram can remove detail to reveal structure.
Use both when students need to connect real appearance with conceptual organisation.
Do not assume one is inherently more “authentic.”
17. Concrete images can distract from abstract relations
A realistic picture may draw attention to irrelevant features.
When the goal is mathematical structure, a simpler schematic may be better.
Representation should fit the learning target.
18. Novices need help reading diagrams
Experts know where to look and which conventions matter.
Novices may read diagrams like pictures.
Teach a routine: title, purpose, labels, legend, direction, relation, then explanation.
19. Pointing can make mappings explicit
Teachers can physically or digitally point between text and diagram.
“This phrase—‘pressure increases’—corresponds to this upward section of the graph.”
Gestures externalise the mapping.
20. Learners should point too
Ask students to locate the feature described in a sentence or trace the pathway while explaining it.
Student action reveals whether mapping is genuine.
Observation alone can hide confusion.
21. Label completion can test correspondence
Remove selected labels and ask students to restore them.
This is useful when the goal is identifying parts.
Follow with relation questions so labelling does not become the endpoint.
22. Diagram annotation can reveal reasoning
Ask students to add arrows, notes, causes or values.
The diagram becomes a workspace rather than a finished illustration.
Annotations should serve the concept, not become decorative note-taking.
23. Text-to-diagram translation is a powerful integration task
Give a verbal description and ask students to construct a diagram.
They must decide what entities and relations are essential.
The task exposes whether the text has been organised mentally.
24. Diagram-to-text translation tests explanatory language
Ask students to explain a graph, process or system in sentences.
Good responses identify relations, not merely list labels.
This builds the ability to move from spatial representation to academic prose.
25. Bidirectional translation strengthens flexibility
Learners who can only recognise a familiar diagram may fail when information arrives as text.
Practise both routes.
Transfer often requires reconstructing the same model from a new representation.
26. Multiple diagrams can reveal invariants
Show the same concept with different layout, colour and scale.
Ask what stays the same.
This helps students separate deep structure from one memorised picture.
27. One canonical diagram can become a trap
Students may memorise the appearance of the water cycle or cell rather than the relationships.
Rotate, simplify, relabel or redraw examples.
Knowledge should survive visual change.
28. Comparing two representations can reveal errors
Present a diagram and text that disagree in one place.
Ask students to find the inconsistency.
This forces active mapping rather than passive acceptance.
29. Mathematics representations need explicit correspondence
Tables, graphs, equations and verbal rules can represent the same function.
Ask which table change corresponds to graph slope or equation coefficient.
Students learn the relation across forms.
30. Geometry diagrams can be misleading if not drawn to scale
Students may infer properties from appearance.
Teach that markings and stated relationships, not visual impression alone, determine mathematical facts.
Representations have conventions.
31. Science models often represent invisible processes
Particle diagrams, force arrows and field lines are conceptual tools, not photographs.
Students should know what each symbol represents and where the model has limits.
32. History visuals need source analysis
Maps, photographs and propaganda posters are not neutral illustrations.
Students need provenance, purpose and context.
Visual literacy includes epistemic judgement as well as mapping.
33. Geography maps require scale and symbol literacy
A map compresses space and selects information.
Teach legend, scale, orientation and projection where relevant.
Text should help students interpret rather than merely name locations.
34. Reading comprehension can benefit from useful diagrams
Expository texts often describe systems that a diagram can organise.
Students should learn to integrate across paragraphs and images.
Questions can require both sources so neither is optional.
35. Captions should not carry critical information students are never told to read
Many learners skip captions because school tasks rarely use them.
If captions matter, teach students to treat them as part of the text.
Assessment should reinforce that expectation.
36. Textbook layout is instructional design
Information split across pages, boxes and sidebars can increase search.
Teachers can compensate by guiding sequence and pointing to relations.
When creating materials, layout deserves the same care as wording.
37. Slides can overload through simultaneous text and speech
A dense paragraph on a slide while the teacher speaks a different explanation can divide attention.
Use concise text and visuals that support the spoken message.
Do not make students choose which stream to follow.
38. Reading identical text aloud adds little visual integration
If students can read the words, teacher narration should add explanation, emphasis or connection rather than simply duplicate every line.
The principle is meaningful coordination among representations.
39. Accessibility requires alternative forms
Blind and low-vision learners may need alt text, tactile diagrams, verbal description or data tables. Colour-vision differences require non-colour cues.
Accessible alternatives should preserve the relation, not merely mention that an image exists.
40. Alt text should explain purpose, not every pixel
For an instructional diagram, describe the relationships needed for the task.
A long decorative inventory can bury meaning.
Accessibility is representational design.
41. Student-created diagrams can reveal misconceptions
Ask learners to draw how they think a system works.
Missing arrows, misplaced parts or incorrect scale can expose the mental model.
Use the drawing diagnostically, not artistically.
42. Drawing quality should not distort conceptual assessment
A student can understand a process and draw poorly.
Allow simple boxes, arrows and labels when the target is conceptual relation.
Do not grade artistic skill unless that is the objective.
43. Diagrams can become external memory
A good visual lets students hold relations outside working memory while reasoning.
This is useful in complex systems and multi-step problems.
The aid should support thinking, not become a cue students cannot work without.
44. Fade completed visuals toward reconstruction
Start with a fully labelled diagram, move to partial completion, then ask students to recreate the structure.
This tests whether the mental model has become internal.
45. Generative AI can create attractive wrong diagrams
AI-generated visuals can contain spatial or factual errors.
Teachers and learners need to verify labels, direction and relationships.
Visual polish is not evidence of correctness.
46. Students should critique visual quality
Compare two diagrams and ask which better communicates the target relation and why.
This develops representational judgement.
Criteria might include accuracy, relevance, legibility and correspondence with text.
47. Integration should eventually become self-directed
Expert learners naturally compare graph and equation, caption and image, diagram and description.
Teach the habit explicitly, then reduce prompts.
The goal is not dependence on arrows drawn by the teacher.
48. The endpoint is one coherent model expressed in several forms
Students should be able to recognise that the words and picture are not two lessons.
They are two windows on the same relation.
Integration succeeds when the learner can move among them flexibly.
Worked case 1 — The heart diagram
Students memorise chamber labels but confuse blood flow. The teacher places short explanatory text next to each stage, uses arrows with stable meaning and asks students to trace the route while explaining oxygenation changes.
Later, the diagram is rotated and relabelled. Students reconstruct the pathway.
The learning moves from picture memory to system relation.
Worked case 2 — Equation and graph
A class can solve linear equations but struggles to interpret slope graphs. The teacher aligns equation coefficient, table differences and graph slope in one display.
Students point to corresponding features, then receive a new graph and write the equation.
Translation builds the connection.
Worked case 3 — The cluttered textbook page
A Science page contains four images, two sidebars and a central explanation. Students attend to the most colourful picture and miss the key process diagram.
The teacher crops the page for instruction, signals the relevant panels and later returns to the full layout after the relation is understood.
Temporary simplification reduces search without permanently avoiding authentic materials.
Worked case 4 — Accessible representation
A student with low vision cannot use the colour-coded map provided to the class. The teacher supplies a high-contrast version, tactile boundary cues and a table of the same values.
The student performs the same reasoning task through accessible representations.
Equity preserves the cognitive demand, not the identical visual channel.
Practical route for teachers
Choose visuals because they represent something useful. Place related words and features close together. Signal critical correspondence. Teach conventions such as arrows, legends and axes. Ask students to translate, annotate and reconstruct.
Use varied versions to test whether the concept survives layout change. Remove decorative elements that compete with the target.
Practical route for learners
When you see a diagram, ask: What is it representing? What do the arrows, colours or positions mean? Which sentence explains each important feature? Can I describe the diagram without looking? Can I redraw the relationship in a simpler way?
Do not treat visuals as decoration beside the “real” text.
Practical route for parents and families
When supporting study, ask the learner to point from a sentence to the relevant part of a diagram or explain a picture in words. Avoid judging drawing quality when the purpose is understanding.
Encourage the child to verify attractive online diagrams against reliable sources.
Common failure modes
- Adding pictures without a learning job.
- Placing related labels and text far from visual elements.
- Using inconsistent terminology across representations.
- Relying on colour alone.
- Allowing decorative visuals to capture attention.
- Assuming arrows and legends are self-explanatory.
- Using continuous animation with no learner control.
- Failing to explain how diagrams simplify reality.
- Using one canonical picture until students memorise appearance.
- Asking students to label without explaining relationships.
- Never translating between text and visual forms.
- Using graphs without teaching graph grammar.
- Letting slide text compete with spoken explanation.
- Ignoring captions.
- Making diagrams inaccessible to disabled learners.
- Grading artistic drawing quality when conceptual mapping is the target.
- Using AI-generated visuals without verification.
- Giving students completed visuals forever.
- Assuming visual preference means visual instruction is universally superior.
- Forgetting that the endpoint is a coherent mental model, not multiple pretty representations.
Frequently asked questions
Do diagrams always improve learning?
No. They help when they represent relevant relations and learners can connect visual features to the explanation. Poorly designed visuals can distract or overload.
What does the WWC recommend?
The WWC practice guide gives moderate evidence to combining graphical representations with verbal descriptions as part of effective instruction.
Is this “dual coding”?
The ideas overlap, but this article uses the more specific mechanism of mapping between verbal and visual representations rather than broad popular claims about “learning styles” or simply using two formats.
Should labels be on the diagram?
Direct labels often reduce search, but dense diagrams may need legends. The design should minimise unnecessary mapping effort.
Are animations better than static images?
Not automatically. Dynamic processes can benefit from animation, but learners need control, segmentation and enough time to inspect important states.
Should students draw their own diagrams?
Yes when drawing helps reveal or construct relationships. Use simple schematics and judge conceptual accuracy rather than artistic quality unless art is the objective.
How do I know integration occurred?
The learner can explain correspondences, translate between forms and reconstruct the relation when layout or surface features change.
What about accessibility?
Provide high-contrast, tactile, verbal or tabular alternatives as needed while preserving the same conceptual relationships.
Can visuals reduce cognitive load?
They can externalise relations and reduce search when designed well. They can also increase load through clutter, split attention or unfamiliar conventions.
What is the most important design question?
What relation does the visual show more clearly, and what must the learner connect between the visual and the words to understand it?
Evidence boundary and caveats
Multiple-representation effects depend on subject matter, prior knowledge, layout, signalling and the learner’s familiarity with representational conventions. More representations are not necessarily better. A learner can be overloaded by redundant or poorly aligned formats.
WWC and IES evidence supports deliberate coordination of graphics and verbal description, not a generic rule that every lesson needs a picture. The mechanism is integration.
Sources and further reading
- What Works Clearinghouse, Organizing Instruction and Study to Improve Student Learning: https://ies.ed.gov/ncee/wwc/PracticeGuide/1
- Institute of Education Sciences research on multiple representations, spatial alignment and educational visual design: https://ies.ed.gov/
Continue exploring on eduKateSG
- How X Works Hub
- How Self-Explanation Prompts Work
- How Contrasting Cases Work
- How Transfer of Learning Works
- How Checks for Understanding Work
- How Metacognition Works
The final idea
A diagram is not a shortcut around thinking. It is another language for the idea.
Students have to learn that language: what position means, what an arrow means, what colour means, what has been simplified, and how those choices correspond to the words beside it.
When the mapping is clear, text and visual can share the burden of explanation. When it is not, the learner receives two sources of information and builds no common model.
The design goal is simple to say and demanding to achieve: make the representations meet in the learner’s mind.