Dual Coding Theory: What It Is and When Visuals Backfire

Dual coding means pairing words with a visual that represents the same idea or relationship. A labelled diagram can show how parts connect while a short explanation names the process. A timeline can show sequence while spoken words explain why one event led to another.

The two forms must work together. An attractive photograph beside unrelated text is not dual coding. A diagram on one slide and its labels on the next also makes learners search for the connection. The practical question is not, “Have I added a picture?” It is, “Does this visual make the important relationship easier to see and explain?” Clark and Paivio (1991) provide the main educational account.

‌

What is dual coding?

Allan Paivio's dual coding theory proposes linked verbal and non-verbal systems for representing knowledge. Words can cue images, and images can cue words. Concrete ideas are often easier to picture than abstract ones, so they may gain more than one useful route for recall.

Clark and Paivio's educational review (1991) connects this account with comprehension, vocabulary, mathematics and the concreteness of learning materials. The theory does not mean that one half of the brain handles words and the other handles pictures. It also does not mean that every sentence needs an image.

Teachers usually use the term more broadly than Paivio did. It now covers well-designed combinations of words and pictures in explanations, slides, diagrams and learner-generated work. That practice also draws on multimedia learning research.

The two traditions overlap, but they are not identical. Our Visual Learning guide places these ideas beside graphic organisers and other visual methods.

‌

Paivio's theory and multimedia learning

Paivio's account concerns verbal and non-verbal representations in long-term memory. Richard Mayer's multimedia learning account concerns how learners select, organise and connect words and pictures while learning. It adds a sharper focus on limited working memory and instructional design.

Mayer's programme of multimedia research (2003) found that well-designed words and pictures can support deeper learning than words alone. The same research also found that design details matter.

Relevant words should sit near the corresponding visual. Extraneous material should be removed. More media is not automatically better teaching. The Cognitive Load Theory guide explains how this design problem relates to limited working memory.

‌

How can words and visuals support learning?

A useful visual carries information that prose alone makes hard to hold. It can preserve a sequence, layout, hierarchy or causal relation on the page. Words can then name the parts, qualify the relation and direct attention to the important feature.

The learner still has work to do. They must select the relevant parts and connect them with what they know. Mayer and Moreno (1998) found that the arrangement of narration and animation affected problem-solving transfer. This is evidence for careful coordination, not proof that every learner benefits equally from every visual.

Prior knowledge changes the task. A novice may need a labelled, step-by-step representation. An expert may find the same labels repetitive. The purpose is to make the relationship available, then remove support when it is no longer useful.

‌

Five common dual coding mistakes

The most common mistakes are decoration, duplicate wording, remote labels, a visual form that does not match the knowledge and independent drawing without enough guidance. Each one forces learners to search, ignore or invent a connection that the explanation should make clear.

‌

1. Adding a picture that carries no teaching meaning

Decorative images can pull attention towards an interesting detail that does not explain the topic. In four experiments, Harp and Mayer (1998) found that seductive details could reduce learning from a science explanation. Delete the image if you cannot state what relation it shows.

‌

2. Reading dense on-screen text aloud

Spoken and printed versions of the same long explanation can compete. Use a brief label, a visual and a spoken explanation. Keep the full text for reading before or after the explanation.

‌

3. Separating labels from the diagram

A key at the bottom of a page makes learners switch between two places. Put each label next to the relevant line, part or stage. Use colour only as a secondary cue. Repeat the meaning through labels, symbols, patterns or position, so the diagram still works in greyscale and for colour-blind learners.

‌

4. Choosing the wrong visual form

A mind map is poor at showing a strict sequence. A timeline is poor at comparing criteria. The shape should match the knowledge, not the teacher's favourite template.

‌

5. Asking learners to draw without guidance

Drawing can reveal and build understanding, but novices may focus on surface detail or copy without thinking. Model how to select the key parts. Give a partial structure, then ask learners to explain each link.

Research on generative learning shows that such activities depend on the learner, material and support provided (Fiorella, 2023). The Scaffolding guide shows how to reduce that support after success.

‌

Choose the visual from the knowledge

Start with the relationship learners need to understand. Then choose the simplest visual grammar that can show it. The visual should preserve a sequence, location, comparison, cause or hierarchy that would otherwise be difficult to hold and inspect.

‌

A worked classroom example

Imagine a Year 7 science lesson on the water cycle. A photograph of rain may attract attention, but it does not show the process. A useful dual-coded explanation needs the changes of state and the movement of water.

1. State the learning goal. Learners should explain how water moves between the surface and atmosphere.
2. Draw only the necessary system. Add sea, land, cloud and sun. Leave out decorative buildings and animals.
3. Add relational arrows. Use arrows for evaporation, condensation, precipitation and run-off.
4. Place words beside the process. Say “liquid water gains energy and becomes water vapour” while pointing to the evaporation arrow.
5. Make learners connect the forms. Hide one label. Ask a learner to restore it and explain the change using the diagram.
6. Test a new case. Ask what changes on a colder day, or why water can return to the sea by more than one route.

The picture is not a second copy of the paragraph. It holds the system in view while the words explain the relations.

‌

Practical dual coding strategies

Useful strategies make the connection visible and ask learners to use it. Build representations in stages, label them directly, align comparisons and move between a visual and subject-specific prose. Retrieval can then test whether the relationship is remembered. This is the same evidence-to-classroom test used in our guide to Rosenshine's principles.

• Build diagrams live. Draw and explain one relation at a time instead of revealing a finished poster.
• Label directly. Put the word beside the feature, not in a remote key.
• Signal the current point. Use a pointer, box or brief colour cue while speaking.
• Compare aligned examples. Keep corresponding parts in the same position so differences are visible.
• Turn prose into structure. Convert a causal paragraph into a labelled chain, then check it against the text.
• Turn structure back into prose. Ask learners to explain a diagram in complete subject language.
• Complete, then create. Let novices finish a partial diagram before asking for an independent one.
• Retrieve the visual. After a delay, ask learners to redraw the structure from memory and correct it in another colour. Use the checking sequence in our Retrieval Practice guide.

‌

Dual coding and adaptive teaching

A clear representation can reduce avoidable barriers without changing the learning goal. Keep vocabulary visible. Break a complex diagram into stages. Describe the visual aloud and give meaningful alternative text where digital access requires it.

Do not assume that a learner with a special educational need is a “visual learner”. Check the specific access demand. A learner with low vision may need a tactile or spoken representation.

A learner with weak prior knowledge may need a worked example before an independent drawing. A learner who already understands the representation may need less labelling.

Dual coding is also not evidence for tailoring lessons to learning styles. Pashler and colleagues (2009) found no adequate evidence for matching instruction to a diagnosed visual or verbal style. Choose the representation that fits the content and the access need. The Learning Styles Myth guide gives the fuller evidence test.

‌

Plan one word-picture explanation

Use this mini app before you make a slide, worksheet or board explanation. It asks for one concept, one relationship and one check. The output is a printable design brief, not a score.

Mini app · visual explanation · lesson planning

‌

Plan one word-picture link

Choose the relationship first. Then decide what words and visual form will make it visible.

Concept Key relationship Essential words Visual formChoose oneLabelled diagramTimeline or numbered flowCausal chainComparison tableHierarchy or nested diagramWorked example with annotations Remove or avoid Learner check Build the plan Clear

Your entries stay in this browser and are not saved or sent.

‌

Dual coding across subjects

The best representation depends on the subject knowledge. A generic visual template can hide the relation a learner needs to see. Mathematics may need quantities aligned, history may need sequence separated from cause, and science may need observations separated from a model.

• Mathematics: align a bar model with the quantities and operation it represents. Ask learners to explain what each part means. A subject-specific graphic organiser should preserve the same mathematical relation.
• Science: use arrows and labels to show a process, system or change. Separate an observable event from the particle model used to explain it.
• History: use a timeline for chronology and a separate causal map for influence. Time order alone does not prove cause.
• English: map the sequence of an argument or the relation between a quotation, method and interpretation. Do not replace close reading with icons.
• Geography: combine a map with a process diagram when location and change both matter. Make the scale and key explicit.

‌

A five-minute slide audit

Check one slide rather than redesigning a whole presentation. Name the relationship it should teach, remove anything that does not help learners see it, place each label beside its feature, and rehearse the explanation. Then ask a colleague to identify the relationship without your narration.

• Does every visual carry subject meaning?
• Are labels beside the parts they explain?
• Is colour repeated through words, symbols, patterns or position?
• Can a learner trace the sequence or relation?
• Does the spoken explanation add meaning rather than repeat dense text?

‌

Limitations and critiques of dual coding

The evidence supports carefully designed combinations of words and pictures for some learning tasks. It does not support a universal percentage improvement, a fixed number of memory pathways or the claim that visual material always reduces cognitive load. The Information Processing Theory guide explains why attention, encoding and retrieval must still be checked.

Results depend on the content, the learner's prior knowledge, the relation between words and pictures and the assessment used. Mayer and Sims (1994) found that spatial ability affected who gained most from coordinated animation and narration. Boers and colleagues (2017) found that attention to multimodal annotations may help explain some vocabulary gains. These findings warn against treating one mechanism as settled fact.

Use dual coding as a testable design choice. Decide what relationship the visual should reveal. Check whether learners can explain that relationship later and in a new example.

‌

Frequently asked questions

These answers separate dual coding from learning styles and set boundaries around memory claims. They also explain when teacher drawing and learner-generated diagrams are useful, and when an additional visual is more likely to distract than teach.

‌

What is dual coding in simple terms?

Dual coding pairs words with a visual that represents the same idea or relationship. The learner should connect the two, not inspect two separate sources.

‌

Is dual coding the same as learning styles?

No. Dual coding chooses a representation to fit the content. Learning-styles teaching tries to match instruction to a diagnosed personal preference, which lacks adequate evidence.

‌

Does dual coding always improve memory?

No. Relevant and well-aligned words and visuals can help. Decorative, redundant or poorly placed material can distract and increase needless demand.

‌

Should teachers draw while explaining?

Often, yes. Building a simple diagram in stages can direct attention to each relation. A finished diagram may be better when drawing live would be slow or unclear.

‌

Should learners make their own diagrams?

They can, when the task is guided. Novices may need a partial structure, a worked example and a prompt to explain each link before drawing independently.

‌

Further reading and references

These eight sources cover the educational theory, multimedia design, split attention, seductive details, learning styles, attention to annotations and learner-generated representations. They support the article's conditional claims and its decision not to promise a universal memory gain.

• Clark, J. M., & Paivio, A. (1991). Dual coding theory and education.
• Mayer, R. E., & Sims, V. K. (1994). For whom is a picture worth a thousand words?
• Mayer, R. E., & Moreno, R. (1998). A split-attention effect in multimedia learning.
• Harp, S. F., & Mayer, R. E. (1998). How seductive details do their damage.
• Mayer, R. E. (2003). The promise of multimedia learning: Using the same instructional design methods across different media.
• Pashler, H., McDaniel, M., Rohrer, D., & Bjork, R. (2009). Learning styles: Concepts and evidence.
• Boers, F., Warren, P., He, L., & Deconinck, J. (2017). On the benefits of multimodal annotations for vocabulary uptake from reading.
• Fiorella, L. (2023). Making sense of generative learning.