What exactly is dual coding theory?

It is Allan Paivio's proposal that human memory uses two partly separate systems, one verbal and one for nonverbal imagery. Material encoded in both systems creates two independent retrieval paths, which is why information presented as words plus a relevant picture is generally remembered better than words alone.

Is dual coding the same as being a visual learner?

No, and the difference matters. Learning styles theory claims each person has one preferred channel and should be taught through it, a claim that fails when researchers test for the predicted matching effect. Dual coding says combining verbal and visual information helps learners generally, so the choice depends on the material rather than the person.

Do I need to be able to draw to use dual coding?

No. Crude sketches work as well as polished ones and often better, because the benefit comes from deciding what the parts are and how they connect rather than from the finished drawing. Boxes, arrows and stick figures are sufficient, and redrawing a diagram from memory doubles as a retrieval test.

Can adding images ever make learning worse?

Yes. Decorative images that carry no information add load without benefit, and presenting an animation with on-screen text at the same time forces the visual channel to handle both, which competes rather than helps. Richard Mayer's multimedia principles cover these cases: keep images relevant, place labels inside the picture, and prefer narration over text when a visual is already busy.

Learning Science

Dual Coding Explained: Why Pictures Help You Remember More

Ingrid Larsen

Learning Science Writer

Aug 16, 2026

Last updated: August 2026

7 min read

TL;DR

Dual coding theory, proposed by Allan Paivio in the 1970s, says that the mind handles verbal and visual information through two partly separate systems. Material encoded in both leaves two retrieval routes instead of one, which is why a diagram paired with an explanation is remembered better than either alone. The evidence is solid and has held up across decades of multimedia learning research, most notably in Richard Mayer's work. It is emphatically not the learning styles myth: dual coding says combining channels helps everyone, whereas learning styles claims each person has one preferred channel, a claim that repeatedly fails when tested. To apply it, sketch rough diagrams rather than admiring finished ones, put labels inside the picture, use timelines and maps for anything sequential or spatial, and cut decorative images that carry no information. Disclosure: MindSnap is our app, and its illustrated story lessons are built on this principle.

There is a small experiment you can run on yourself. Read a description of how a lock and a key interact, in words only, carefully. Then look at a cross-section drawing of the same mechanism for four seconds. Come back in a week and see which one you can reconstruct.

For most people it is not close. That asymmetry is the starting point for one of the more durable findings in cognitive psychology, and unlike a lot of educational folklore, this one has survived being tested.

The theory in one paragraph

Allan Paivio, working at the University of Western Ontario, proposed dual coding theory in the early 1970s. His claim was that human cognition operates two functionally independent but interconnected systems: a verbal system that handles language, and an imagery system that handles nonverbal, perceptual representations. Information can be encoded in one system or in both. When it is encoded in both, you have created two routes to the same memory, and if one route fails the other may still deliver.

This explains a finding that predates the theory and motivated it: concrete nouns are remembered better than abstract ones. Elephant is easier to recall than justice, and Paivio's answer is that elephant gets coded twice, once as a word and once as a picture, while justice only really gets the word.

The evidence, briefly and honestly

Three lines of support are worth knowing.

First, the picture superiority effect. Across many experiments, pictures are recognised and recalled better than their corresponding words, and the gap tends to widen as the delay before testing gets longer.

Second, concreteness and imageability effects in word learning, which is where Paivio started. Words that easily evoke a mental image are learned faster and retained longer, and the effect survives when you control for frequency and familiarity.

Third, and most useful for anyone designing lessons, Richard Mayer's multimedia learning research from the 1990s onwards. Across a long programme of experiments Mayer found consistent advantages for presenting words together with relevant graphics, and derived design principles from it: keep related words and pictures next to each other in space and time, cut material that does not serve the explanation, and prefer spoken narration over on-screen text when a complex animation is playing, because reading and watching compete for the same visual channel.

That last point is a real constraint. The two channels help because they are partly separate, but each has limited capacity, so you can overload the visual channel just as easily as the verbal one. This is where dual coding meets working memory limits, which we covered in cognitive load explained.

It is not the learning styles myth

This needs saying plainly, because the two get conflated constantly and it makes teachers rightly suspicious of the good idea.

The learning styles claim is that individuals have a preferred sensory modality, visual, auditory or kinaesthetic, and that teaching matched to that preference produces better learning. When researchers test the part that matters, the meshing hypothesis, they look for an interaction: visual learners should do better with visual instruction than auditory learners do, and vice versa. Reviews such as Pashler and colleagues in 2008, and later replication attempts, keep failing to find that interaction, even though people are perfectly willing to report a preference.

Dual coding makes a different claim in a different shape.

	Learning styles	Dual coding
Claim	Each learner has one best channel	Everyone has two channels and both can be used
Prescription	Match teaching to the person	Combine words and relevant images for all learners
Unit of variation	The individual	The material
Evidence when tested	The matching effect does not reliably appear	Combining relevant words and pictures reliably helps

Two ideas that get confused

So the correct conclusion is not that some students need diagrams. It is that diagrams paired with explanation help almost everyone, and that the right choice depends on what the content is rather than who is reading it.

Applying it, in order of return on effort

Draw badly, on purpose

The largest gain comes from generating the image yourself rather than viewing one. Sketching forces you to decide what the parts are and how they relate, which is an act of retrieval and organisation at the same time. Quality is irrelevant. Boxes, arrows, stick figures. If you can draw a system from memory a week later, you understand it.

Pick the diagram type that matches the structure

Sequence or causal chain: a flow diagram with arrows, one step per box.
Comparison: a two-column table, which is a visual structure even though it is made of words.
Hierarchy or category: a branching tree.
Change over time: a timeline, which is why history is unusually well served by drawing.
Anything geographical: a rough map, however crude. Borders and chokepoints are almost impossible to hold verbally.
Cycles and feedback: a loop, drawn as a loop, so the return path is visible.

Put the words inside the picture

Label parts directly on the diagram rather than in a key underneath. Mayer's spatial contiguity principle is one of the most consistent effects in the set: making the reader shuttle between a legend and an image wastes working memory on bookkeeping.

Cut decoration

A stock photograph of a smiling student next to a paragraph about mitosis is not dual coding. It adds visual load without adding information, and in Mayer's terms it is extraneous material. If the image does not carry part of the explanation, remove it.

Then test yourself on the image

Cover your diagram and redraw it. This combines dual coding with the single most effective study technique there is, and it converts a nice-looking note into a genuine memory check. The mechanism and the evidence are in our retrieval practice guide.

Why illustrated micro-lessons work

Short lessons have an awkward constraint: very little time to establish context. A well-chosen image solves that faster than any sentence can, because a painting, a map or a cross-section delivers setting, scale and relationship in a single glance, leaving the words free to do the explaining.

There is a compounding effect with narrative too. Stories already encode information in a rich, imageable form, which is part of why they stick, and adding a picture gives the same content a third handhold. We looked at the narrative half of that in why your brain loves stories.

How we use it, with a disclosure

Disclosure: MindSnap is our app, so this is a design note from an interested party.

Our lessons pair a two-minute piece of writing with a single carefully chosen image, usually a painting, artefact, map or chart that is part of the argument rather than wallpaper. In the flagship story collections the image often is the evidence: a self-portrait that shows a face aging across decades, a chart of a market at the moment it broke, a map of a strait a fleet had to pass. Alongside those five collections there are unlimited topics, and the same rule applies throughout, which is that if the picture is not doing work we do not include it.

None of that requires an app, though. The version you can start today costs nothing: keep a plain notebook, and after anything you read that you want to keep, draw it in thirty seconds. Ugly is fine. Ugly and yours beats beautiful and someone else's.