A Canadian psychologist's 1970s insight — that the brain runs two distinct memory systems — turns out to explain almost everything about why some lessons stick and others evaporate.
Somewhere in a lecture hall right now, a professor is reading bullet points aloud from a slide. Students are furiously copying the same text onto their notebooks. And almost none of it will survive to the next morning.
This is not a motivation problem. It is a neuroscience problem — one that Allan Paivio, a psychologist at the University of Western Ontario, began solving in 1971 with a deceptively simple observation: the human brain processes verbal information and visual information through separate, independent channels. And when both channels are engaged simultaneously, memory formation becomes dramatically more powerful.
Half a century later, dual coding theory has grown from a niche academic proposition into one of the most robustly supported frameworks in cognitive science. It shapes how textbooks are designed, how surgical trainees learn anatomy, and how elite athletes rehearse performance. Understanding it changes how you study, teach, and absorb anything.
Paivio's central claim was that human cognition operates via two distinct symbolic systems: a verbal system, which processes language in sequential, rule-governed ways, and a nonverbal system, which handles imagery in a more holistic, spatial manner. These systems are interconnected — words can evoke mental images, and images can prompt verbal labels — but they store and retrieve information independently.
The practical implication is profound. When you learn something using only words, you activate one storage pathway. When you pair those words with a corresponding image — whether a diagram, a photograph, a mental picture, or a hand-drawn sketch — you create two independent memory traces for the same concept. During recall, either trace can cue the other, dramatically improving retrieval success.
Paivio demonstrated this through a series of elegant experiments. When participants were asked to memorise lists of words, concrete nouns — words like "elephant" or "bicycle" that easily conjure a mental image — were recalled significantly better than abstract nouns like "justice" or "virtue." The difference, he argued, was that concrete words activate both verbal and imagery codes simultaneously, while abstract words tend to remain anchored in the verbal system alone.
"The power of dual coding is not that images are better than words or words better than images. The power is in the redundancy — two independent routes to the same concept mean that forgetting requires both routes to fail simultaneously." — Dr. Rosamund Kiely, cognitive psychologist, University of Edinburgh
Since Paivio's foundational work, hundreds of studies have tested dual coding predictions across remarkably diverse domains. The pattern is consistent enough to be called a law rather than a tendency.
A landmark 1994 analysis by Mayer and Anderson compared learning outcomes across different presentations of scientific processes — the same content delivered as text only, animation only, narration only, narration plus animation, and text plus animation. The combination of narration and animation — engaging both channels without overwhelming either — produced the best transfer performance. Text plus animation, by contrast, tended to cause cognitive overload when the text duplicated what the animation showed, supporting a key refinement to Paivio's original theory.
That refinement — what Richard Mayer would later formalize as the Cognitive Theory of Multimedia Learning — established a crucial nuance: dual coding works when the verbal and visual channels carry complementary information. When they carry identical information (reading words on a slide while a speaker says the same words), the redundancy creates interference rather than reinforcement. This is the "redundancy effect," and it is precisely why narrated bullet-point presentations are neuroscientifically suboptimal.
Research published in the British Journal of Educational Psychology found that students who drew concept maps while reading passages outperformed both passive readers and those who took conventional linear notes, with effect sizes averaging around 0.7 — educationally significant by any standard. The act of converting verbal content into a spatial visual representation forces a kind of generative processing that dramatically deepens encoding.
Modern neuroimaging has lent biological specificity to what Paivio observed behaviourally. When participants process words, left-hemisphere language networks (Broca's area, Wernicke's area, and associated cortices) dominate. When they process images, bilateral occipital, temporal, and parietal regions associated with visual and spatial processing become active.
Crucially, when participants process concrete words that evoke imagery — or directly process image-word pairings — both networks activate together. Functional MRI studies show increased connectivity between language and visual cortices during dual-code learning, and this connectivity correlates with subsequent recall performance.
The hippocampus, which binds disparate cortical representations into unified memories, also shows stronger activation and greater synaptic consolidation when encoding happens across multiple channels. This is consistent with the broader neuroscientific principle that memories encoded in richer, more distributed cortical patterns are more resistant to forgetting — the brain has more "anchors" from which to retrieve the experience.
The educational applications are extensive. The most direct is the use of complementary visual-verbal pairings: diagrams annotated with brief labels rather than dense paragraphs; worked mathematics examples that show both the algebraic steps and a geometric or graphical interpretation; history lessons that pair primary source text with photographs, maps, or contemporaneous illustrations.
Self-generated visuals appear to be even more powerful than provided ones. A 2018 meta-analysis in Educational Psychology Review examined drawing-to-learn interventions across 28 studies and found a mean effect size of 0.63 in favour of drawing conditions over control conditions. When learners translate their own understanding into a sketch — however crude — they must actively reorganise the material, identify what they do and do not understand, and make spatial decisions about how concepts relate. Each of these processes deepens encoding.
Retrieval practice further amplifies dual coding benefits. Testing yourself by trying to reconstruct a diagram from memory — rather than simply re-reading it — engages both the visual recall pathway and the verbal explanation pathway simultaneously, producing what researchers call "elaborative interrogation" effects. Students who combine self-testing with sketch reconstruction consistently outperform those who use either strategy alone.
The principles extend well beyond academic settings. In medical education, anatomical learning has long combined verbal description with cadaveric and imaging study — a form of dual coding that clinicians intuitively understood before the theory was formalised. Recent research from Johns Hopkins has shown that surgical trainees who supplement technical reading with annotated operative video (matching verbal narration to specific visual moments) demonstrate significantly faster skill acquisition than those who use either medium alone.
In aviation, crew resource management training combines verbal checklists with spatial visualisation of cockpit layouts and procedural flows. The combination ensures that pilots can retrieve correct procedures under conditions of stress — when purely verbal memory often degrades — because the visual spatial memory trace remains accessible through a different neural pathway.
Elite sports science has applied dual coding through the concept of mental imagery training. Numerous controlled trials have shown that athletes who combine physical practice with structured visualisation — vividly imagining the exact sensory and kinematic experience of executing a skill — perform better than those who use physical practice alone. The visualisation engages the nonverbal imagery system to encode motor patterns that physical repetition encodes through procedural channels, creating a richer combined trace.
Dual coding is not a universal learning cure, and its limits deserve acknowledgment. Abstract conceptual material — philosophical arguments, mathematical proofs at high levels of abstraction, complex causal reasoning — often resists simple visual representation. Forcing a diagram onto content where genuine imagery equivalents do not exist can produce a superficial schematic that misleads rather than aids.
The theory also does not address motivation, prior knowledge, or metacognitive awareness — all of which significantly moderate learning outcomes. A student who does not understand a concept will not necessarily understand it better by drawing a confused picture of it. Dual coding amplifies understanding that already exists; it does not substitute for it.
Finally, there is the individual differences question. A small but consistent strand of research suggests that learners with stronger spatial abilities may benefit more from visual components than those with weaker spatial skills. However, the evidence does not support the popular notion of rigid "visual" versus "verbal" learning styles — these are not fixed traits, and virtually all learners benefit from appropriately designed dual-code instruction.
There is a certain irony in the fact that the information environments we now inhabit — feeds of text, video thumbnails, infographics, podcasts — are, in a superficial sense, already multimodal. Yet most of this content is designed for passive consumption rather than active encoding. Scrolling past an animated infographic is not dual coding; watching a documentary is not dual coding; reading a thread illustrated with photographs is not dual coding. What activates both systems meaningfully is the effortful combination of verbal and visual processing in the service of genuine understanding.
The best learning environments have always known this intuitively. Renaissance masters had their apprentices draw before they could paint, ensuring that visual knowledge was built through active construction rather than passive observation. Japanese design education traditionally requires students to sketch thousands of objects before designing anything, building a dual-coded library of visual-verbal knowledge about material, proportion, and function.
Paivio gave this ancient wisdom a formal theoretical home. The two channels in your skull are waiting to be engaged together. Bring a pencil.