Dual coding is a theory of how verbal and nonverbal representations can support one another in memory. For a learner, the useful question is not whether a page contains both words and pictures. It is whether the learner can connect the representations accurately and use either one to recover the intended meaning. This guide separates the theory from the popular instruction to make everything visual. It then develops practical ways to build, check and use a word–image pair without mistaking a decorated notebook for understanding.
Begin with the larger How Learning Works mechanism map when the difficulty is not yet clear. This article has a narrower job: deciding what a visual adds to a verbal explanation, and how to find out whether the connection belongs to the learner.
The notebook can look finished before the learning is finished
Consider an illustrative revision page about fractions. The definition is highlighted. Three colourful circles sit beside it. The heading is attractive and the examples are neatly boxed. Asked what three quarters means, the student points to the circle with three shaded sectors. Asked whether three quarters of a small cake must be more cake than one half of a large cake, the student hesitates.
The page has words and images. The missing relationship is the reference whole. The student has learned to recognise a familiar picture without necessarily knowing what its proportions represent. Adding another colour will not settle the question. Comparing equal and unequal wholes might.
This distinction gives dual coding a practical standard. A visual should carry something that the learner needs to understand or retrieve. The learner should be able to say what it carries, identify what it leaves out, and check that the verbal account and the visual account agree. The standard is meaning, not the amount of visual material.
Throughout this article, the classroom activities are original worked illustrations, not reports of measured outcomes in eduKate classes. They show how to inspect a representation. They do not establish that one particular notebook design will produce a particular examination result.
What the theory proposes—and what it does not prove
In their account of dual-coding theory, James Clark and Allan Paivio describe interconnected verbal and nonverbal representations. The theory concerns how information is represented and associated, not merely which sensory organ receives it. A printed word is seen, but its linguistic meaning is not thereby the same kind of representation as the shape or spatial arrangement of a pictured object. Clark and Paivio’s theoretical account is the foundation for this distinction.
That distinction rules out several shortcuts. Reading a sentence and hearing the identical sentence does not automatically create the same representational contrast as explaining a relationship in words and showing that relationship spatially. Two copies of a paragraph are not two different models. A photograph beside a definition is not useful simply because it occupies another part of the page.
Nor does the theory imply that a child belongs permanently to a visual category and should therefore avoid verbal work. A claim about possible representations does not logically establish a fixed type of learner. The practical decision remains tied to the concept, the task, the learner’s current knowledge and the accessibility of the material.
The mechanism is also not settled by observing that pictures are sometimes remembered well. In experiments reported by Kate Higdon and colleagues, manipulations of physical distinctiveness challenged the dual-coding explanation of picture superiority. The authors favour a distinctiveness account for that effect. Their work does not show that useful diagrams should be abandoned; it does show why a familiar memory result should not be presented as decisive proof of one theory. Read the original study and its argument.
For teaching, separate three statements: a theory proposes a mechanism; an experiment reports an outcome under specified conditions; an educator proposes a classroom application. These statements can support one another, but they are not interchangeable. The activities below are applications to test, not demonstrations that the brain has stored two independent copies of a lesson.
Give the visual a job before giving it a place
Before selecting a picture, finish this sentence: “Without this representation, the learner may miss…” The answer might be the part–whole relationship, the order of events, the location of a quantity, the distinction between two cases, or the connection between a term and its referent. A useful answer tells the teacher what to build and what to test afterwards.
A drawing of a shopping bag beside the word “percentage” does little by itself. A strip divided into one hundred equal parts can make a proportion inspectable. A second strip representing a different original quantity can expose why equal percentages need not mean equal absolute amounts. Both are visual, but only the latter designs directly address the conceptual question.
The verbal part needs a job too. It can identify the reference whole, state a condition, explain a symbol, or specify that an arrow represents a possible influence rather than a proven cause. Pictures can leave these matters ambiguous. A carefully chosen sentence can prevent a visually persuasive but incorrect reading.
Work from the relationship towards the format. For order, a short sequence may be appropriate. For part and whole, a partitioned shape may help. For a comparison, two aligned examples may be clearer than one elaborate illustration. For exact numerical values, a table can be more useful than a stylised picture. Sometimes the best decision is to retain a precise sentence and add no image at all.
This is a design argument, not a prescription that every lesson must use the same visual form. The representation should solve a specific explanatory problem. Otherwise the extra material asks the learner to process more without making the target meaning more accessible.
Worked example: fractions, equal parts and the reference whole
Use the statement: “Three quarters of a whole means three of four equal parts of that whole.” Now inspect what a matching drawing must preserve. There must be an identifiable whole. It must be divided into four equal parts. Three parts must be selected. The picture must not quietly change what counts as the whole halfway through the explanation.
First show an upright rectangle with three of four equal strips selected. Next describe the same rectangle rotated. The orientation changes; the fraction does not. Ask the learner to explain why. The task checks whether the learner is reading the part–whole relationship rather than remembering a familiar position on the page.
Then show a tempting non-example: a rectangle cut into four unequal regions, three of them shaded. The claim “three shaded regions out of four” is not sufficient to establish three quarters of the area. The missing condition is equality of the parts. The verbal definition can now correct a misleading interpretation of the picture.
Finally, compare three quarters of an 8-unit strip with one half of a 20-unit strip. The selected amounts are 6 units and 10 units. Three quarters is the larger fraction of its own whole, but 6 units is the smaller selected amount. The result follows from the stated quantities; no memory theory is needed to prove it. The teaching value of the paired representation lies in making the changing reference whole difficult to overlook.
A learner who can label the first drawing has passed a recognition check. A learner who rejects the unequal-parts example has shown something more specific: attention to a defining condition. A learner who explains the unequal-wholes comparison has shown that the visual and verbal representations can be coordinated under a changed condition. These are different observations and should be recorded separately.
For a follow-up, remove the pictures and ask the learner to describe what an accurate drawing would need to show. Later, provide a new drawing without a label and ask for the fraction and the reference whole. The two directions expose different gaps. A student may produce the familiar image from the phrase but be unable to interpret an unfamiliar image correctly.
Worked example: a timeline must preserve whose time is being described
Consider the sentence: “By the time the bus arrived, Lina had already returned the library book.” A learner can recognise every word yet misread the sequence. A minimal timeline can mark two events: returning the book, followed by the bus arriving. Place the sentence beside the timeline and ask which phrase tells us the order.
The point is not to teach grammar through a decorative clock. It is to connect a linguistic relationship to an ordered representation. “Had already returned” places the completed return before the reference event. The diagram should not introduce an exact clock time because none is stated. Nor should it imply that returning the book caused the bus to arrive.
Now change the sentence: “After the bus arrived, Lina returned the library book.” Ask the learner to revise the timeline rather than merely repeat the original drawing. The order changes. A copied timeline would now disagree with the language. The revision reveals whether the learner is using the words to control the representation.
For a harder comparison, use two descriptions that preserve the same event order but change the sentence order. The learner should distinguish the order in which a writer mentions events from the order in which those events occur. This is an original language exercise, not a claim that timelines are always the best way to teach tense.
Finish by asking for an explanation without the timeline: “Which happened first, and which words tell you?” The visual has served its purpose when it helps the learner inspect the sentence more accurately. It has not served its purpose when the learner simply remembers that the book icon always belongs on the left.
Arrows are claims, not ornaments
Many revision pages use arrows without specifying what they mean. An arrow can represent sequence, movement, dependence, a proposed causal influence, a mathematical mapping or simply “read this next”. These are not the same relationship. If one arrow style is used for all of them, a tidy page can conceal several incompatible meanings.
In a planning example, “task assigned → draft written → feedback received” can indicate an intended sequence. It does not prove that receiving feedback always improves the draft. A second arrow labelled “may inform revision” makes a different, conditional claim. The words prevent the sequence from being mistaken for a guarantee.
For a student’s explanation map, ask them to read every arrow as a sentence. “This happened before that.” “This quantity is included in that total.” “This observation supports, but does not prove, this interpretation.” If the learner cannot supply the sentence, the visual relationship may be undefined.
Also inspect missing arrows. A page can contain all the right nouns while omitting the relationship that the question asks the learner to explain. Conversely, a page can connect every box to every other box and become impossible to interpret. A small number of labelled, defensible connections is preferable to an impressive network of unspecified associations.
This inspection is especially useful when the student says that making a mind map helped them study. Ask what a connection means, not just whether a connection exists. The educational question is whether the visual organisation makes the reasoning available for inspection and later use.
A practical test: translate, explain, change and check
The following four checks are a proposed classroom routine, not a standardised diagnostic instrument. They help separate the presence of materials from the learner’s ability to use them.
| Check | Task | What the response can reveal |
|---|---|---|
| Translate | Construct a simple representation from the explanation, or explain the representation in words. | Whether the correspondence is understood in the tested direction. |
| Explain | State what one part, mark or connection means. | Whether the learner understands the visual convention rather than only its appearance. |
| Change | Alter one condition and revise the representation. | Whether the representation responds to meaning or remains a copied template. |
| Check | Find and repair a mismatch between the words and the image. | Whether one representation can constrain an error in the other. |
Do not combine these observations into a sweeping judgement about intelligence or memory. They concern particular tasks under particular conditions. A learner may understand the relationship but struggle to draw it neatly, or may know the language but not the diagram convention. Ask a follow-up before deciding which support is needed.
For example, allow a student to place labelled cards instead of drawing boxes. If the ordering is now correct, the initial difficulty may have involved production rather than the sequence itself. Alternatively, provide the boxes but remove the labels. The changed task helps locate what was carrying the earlier success.
One successful attempt should lead to another question, not a permanent label of mastery. Use a changed example and a later return. The purpose is to gather converging evidence that the representation is usable beyond the original page.
When adding an image makes the task worse
Mayer and Moreno’s account of multimedia learning distinguishes useful processing from demands created by presentation. Their analysis explains why related information that is hard to coordinate can impose unnecessary work. This supports a design question—not a universal ban on complexity: how much effort is the learner spending locating and holding the explanation rather than understanding it? See their discussion of processing demands.
Imagine a geometry explanation whose letter labels sit on one page, measurements on another and task instructions beneath an unrelated illustration. Before reasoning, the learner has to assemble the problem. A more integrated arrangement can remove this avoidable search without making the geometry itself less demanding.
Decoration introduces a different risk. Harp and Mayer’s experiments found poorer recall of main ideas and fewer transfer solutions when certain science passages included interesting but irrelevant details. The finding concerns those instructional designs; it is not a claim that all stories, colour or interest are harmful. It is a reason to ask whether an addition serves the explanation or competes with it. The original research abstract is available through ERIC.
A useful practical distinction is between an image that helps answer the question and an image that merely reminds the reader of the topic. A picture of a train may establish a context. A labelled distance–time representation can help answer a rate question. Context can have a legitimate role, but it should not be credited with work that only the representation performs.
Do not remove necessary information merely to create a sparse page. A legend, unit label or boundary condition can be essential. Simplicity should mean fewer unnecessary demands, not less truth. A beautiful diagram with an omitted condition may be easier to read and easier to misunderstand.
Make the connection accessible, not just visible
A learning activity should not depend on a distinction that the learner cannot reliably access. Do not make red and green the only difference between two categories. Add text labels or another distinguishable marking. Do not place the only essential explanation inside a tiny image. Provide the relationship in readable language as well.
These are practical access decisions, not tests of effort. When a learner cannot identify a symbol, the next move is to explain the convention. When a learner cannot inspect a small image, the next move is to provide an accessible form. Repeating “look more carefully” does not repair an inaccessible representation.
Likewise, mental imagery should not become a compulsory report of an inner experience. Ask what the learner can explain, organise or produce. A student who does not describe a vivid internal picture may still reason accurately with a table, a tactile arrangement, a verbal description or a diagram kept in view.
The objective is not to prove that every learner processes the lesson in the same way. It is to make the target relationship available and inspect whether the learner can use it. Offer an alternative response format when the original format introduces an irrelevant obstacle.
Use the method differently as the learner becomes more knowledgeable
For an early learner, a useful task may be to connect a short phrase to one clear part of a representation. “One half” can be paired with a whole divided into two equal parts. The discussion should establish the whole, the equality of the parts and the selected amount before adding several competing examples.
For an upper-primary learner, the task can require coordination. Ask the student to explain why two differently shaped wholes can each show three quarters, or why the same shaded count can represent different fractions when the total number of equal parts changes. The image now supports comparison rather than only naming.
For a secondary learner, give responsibility for choosing and criticising the representation. Which aspect is better expressed by a sentence, an equation or a diagram? What assumption does the diagram conceal? Which interpretation would become invalid if the scale changed? The student should not merely consume a teacher-selected picture.
These are suggested progressions, not age-based ceilings or a universal curriculum sequence. Select the entry point from the learner’s actual response. A younger learner may handle a sophisticated comparison in a familiar domain; an older learner may need the visual convention taught explicitly in a new subject.
A study session that produces evidence rather than a prettier page
Choose one relationship from the topic. Write it accurately in a sentence. Produce the smallest representation that preserves it. Then label only what is needed to explain the relationship. This prevents the session from becoming an open-ended artwork project before the learning question has been settled.
Next, read the representation back into words. Compare this explanation with the original sentence. Check whether the whole, quantities, event order, conditions and connections still agree. Repair a discrepancy immediately enough that the final study object does not preserve a known error.
Now change one feature. Rotate the shape, alter the reference whole, reverse an event order, remove an irrelevant detail or introduce a non-example. Ask whether the representation must change. The answer matters less than the justification: which part of the meaning controls the decision?
End with a task that does not simply reproduce the page. Explain a new example, reject a misleading image or describe a representation from a fresh sentence. Record the exact point that still needs support. “Could label the diagram but could not explain the unequal parts” gives a useful starting point for the next session.
On the next return, begin from the question rather than immediately rereading the finished page. The aim is to inspect what is available now. A delay and a changed prompt make the test different from the original supported performance; they do not guarantee a particular improvement by themselves.
Questions worth asking before adopting the technique
Must every definition have a picture?
No. Some definitions are best clarified by precise language, an example and a non-example. A forced illustration may introduce ambiguity. Add a visual when it has a defensible explanatory or retrieval purpose, not to satisfy a quota.
Does a memorable image prove understanding?
No. Remembering a picture, explaining its meaning and applying the concept are different tasks. Ask for the relationship and a changed case. A student can remember the appearance while misunderstanding the condition it was meant to represent.
Should the teacher supply the image or the learner create it?
They serve different purposes. A supplied image can introduce a convention or a structure the learner does not yet know. A generated image can expose what the learner believes the explanation means. Choose deliberately, and do not confuse copying with independent construction.
How should a parent respond to a confusing diagram?
Ask the child to explain one mark at a time. What is the whole? What does this arrow mean? Which sentence supports that connection? When the meaning is unclear, help locate a trustworthy explanation rather than praising the presentation and moving on.
Can a correct explanation remain useful without the picture?
Often that is an appropriate test, but not always the final performance requirement. A map-reading task legitimately includes a map. A graph-interpretation task legitimately includes a graph. Remove temporary teaching hints, not information that belongs to the authentic task.
What this guide claims, and what it leaves open
Dual coding offers a theoretical language for considering verbal and nonverbal representation. The practical proposals here ask learners to make correspondences explicit and check their accuracy. They do not promise doubled memory, establish a fixed learning style, or prove a neural mechanism from a worksheet result.
A sound evaluation should specify the goal before comparing methods. Is the goal remembering a term, explaining a relationship, reading a representation or transferring to a new problem? Keep the content and available study time reasonably comparable, inspect more than one attempt, and avoid interpreting a small classroom comparison as a controlled causal study.
When the paired representation does not help, diagnose before adding more. The concept may be inaccurate, the image may be ambiguous, the learner may not know the notation, or the task may never require the connection. Each possibility points to a different repair. “Make it more visual” is not a sufficiently precise diagnosis.
The most useful final question is simple: What can the learner now explain or do because these representations were connected? An answer grounded in a fresh task is more informative than the number of colours, icons or pages produced.
Research and further reading
Clark and Paivio (1991), Dual Coding Theory and Education presents the theoretical account. Higdon and colleagues (2025 journal issue; first published online in 2024) tests a competing explanation of picture superiority. The theoretical disagreement should not be hidden when discussing why pictures may be remembered.
Mayer and Moreno (2003), Nine Ways to Reduce Cognitive Load in Multimedia Learning considers processing and presentation demands. Harp and Mayer (1998), How Seductive Details Do Their Damage reports experiments on interesting but irrelevant additions to science explanations. These sources inform the design questions; the worked fraction and language activities in this guide are original educational illustrations.
Continue through Working Memory when the learner loses track of the task, Self-Explanation when the connection cannot be explained, and Concept Boundaries when a familiar picture triggers the wrong classification. Return to How Learning Works to choose the next mechanism from the actual difficulty.
