The note you can still see

Think back to something you learned years ago and never lost. For a lot of people it isn't a sentence — it's a picture. The shape of the water cycle with its looping arrows. A diagram of the heart with its four chambers colored in. The map of a country you once had to label. The words came and went, but the image stayed, and the words came back when you pictured it.

That experience isn't sentimental. It points at one of the better-supported ideas in cognitive psychology: your mind doesn't store everything in a single format. It has at least two cooperating systems for holding information, and when you feed both at once, memory gets noticeably stronger. The idea is called dual coding theory, and once you understand it, you stop treating images as decoration and start using them as a second set of handles on the same fact.

Two channels, not one

Dual coding theory comes from the psychologist Allan Paivio, who spent decades arguing that human cognition runs on two distinct but linked subsystems. One is verbal — the system that handles language, words, and the sounds of speech. The other is nonverbal, or imagery-based — the system that handles pictures, shapes, spatial relationships, and the mental images you can conjure with your eyes closed.

The key claim is that these two systems encode information in different codes, and that the codes can be stored independently and retrieved independently. A concrete word like apple can be coded twice: once as the word itself, and once as the image it evokes. An abstract word like justice tends to get coded mostly verbally, because it doesn't summon an easy picture.

This matters for memory because of a simple arithmetic of retrieval. If a fact lives in two codes instead of one, you have two routes back to it. If the verbal trace fades, the image may still be there to reconstruct it, and the other way around. A single cue can activate the whole network. You are, in effect, storing a backup in a different language.

Why pictures get remembered

There's a closely related and very robust finding that falls naturally out of this framework: the picture superiority effect. Across many experiments, people who study pictures of objects tend to remember them better than people who study the words for those same objects. Pictures, the theory goes, are more likely to be coded both ways — you see the image and you spontaneously name it — while words are more likely to be coded only verbally unless you make the effort to visualize them.

That asymmetry is the whole practical lesson in miniature. Words don't automatically become images. Pictures often do automatically become words. So if you want the double coding that makes memory durable, the move is to add the missing channel deliberately — to give your verbal material a picture, and to make sure your pictures get named.

It's worth being honest about the limits here. Dual coding is not a claim that we think in photographs, and it doesn't mean a wall of clip art will save a bad set of notes. The benefit comes from two meaningfully related codes pointing at the same idea. Which brings us to the part people get wrong.

The trap of decorative images

If images help memory, more images should help more — that's the intuition, and it's wrong. Decades of multimedia learning research, much of it from Richard Mayer, draw a sharp line between images that carry meaning and images that merely sit on the page looking relevant.

A picture that depicts the actual relationship you're trying to learn — how the parts of an engine connect, where a country sits relative to its neighbors, what a cell's organelles look like in arrangement — gives the visual system real structure to encode. A picture chosen because it's pretty, or vaguely on-theme, does the opposite. It pulls attention away from the words, forces your mind to split its limited working memory across competing things, and can leave you remembering the decoration instead of the point. Researchers call these distracting extras seductive details, and they tend to hurt.

So the rule isn't "add images." It's "add the image that is the idea." A good visual for dual coding is one you could point at and explain. If you can't say what it teaches, it's furniture.

How to actually do it

The encouraging part is that you don't need to be an artist, and you don't need a perfect picture for everything. You need to engage the nonverbal system honestly. A few ways that work:

Draw it badly. The act of sketching a process — even a crude diagram with boxes and arrows — forces you to decide how the parts relate spatially. That decision is the encoding. The ugliness is irrelevant; nobody's grading the drawing, and the messy version is often more memorable because you made it.

Name what you see. If you're studying a diagram, photo, or map, say or write the labels in your own words. This closes the loop in the other direction, attaching a verbal code to the visual one so the picture superiority effect actually fires.

Build a concrete image for an abstract idea. Abstract concepts resist imagery, which is exactly why they're hard. Inventing a small scene — a metaphor, an analogy you can picture — gives an abstraction a borrowed visual handle. Inflation is hard to see; a balloon slowly stretching while the same dollar buys less is not.

Pair, don't duplicate. The point isn't to write a caption that repeats the picture word for word. It's to let each channel do something the other can't: the words carry the precise claim, the image carries the structure or the feel.

Notice what all of these have in common. They make you generate the second code yourself rather than passively receive it. The effort is the mechanism, not a side cost of it.

Two codes, then time

Dual coding does one specific job: it gets a fact in more richly, with more routes back to it. What it doesn't do is fight forgetting on its own. Even a beautifully dual-coded memory fades if you never revisit it — encoding strength and retrieval over time are different problems, and you need both.

The combination is where it gets powerful. A fact encoded through both channels and then retrieved at spaced intervals isn't just learned once well; it's reinforced along two paths every time you bring it back. The image cues the words, the words cue the image, and each successful recall makes the next one easier.

Where Recall fits

This is the thinking behind how Recall handles cards. Because it's built for fast, clean flashcards that hold images alongside text, you can put the diagram and the definition on the same card — the picture that is the idea, paired with the words that pin it down — instead of choosing between them. Then its FSRS-based spaced repetition schedules each card to come back right as it's about to slip, so both codes get exercised together over time rather than once and forgotten. You can bring decks over from Anki or Quizlet and add the missing channel to cards that are all text, and because it works fully offline, the practice goes wherever you do.

You don't need the app to use any of this — sketch your notes, name your diagrams, and you'll remember more starting today. But if you want the two codes and the timing working together without managing the schedule yourself, that's the tool we built. You can try it at https://recall.lumenlabs.works.