Dual coding: why images make vocabulary stick

Quick test. Think of the last hotel room you stayed in. You can probably still walk through it: bed on the left, weird lamp, bathroom door that stuck, the particular gloom of the hallway. Now try to recall the confirmation number from the same booking.

One of those memories cost you zero effort to keep. The other was gone before checkout. Your visual memory is a warehouse; your verbal memory is a shoebox. Dual coding is the trick of storing your vocabulary in both.

Two systems, two routes

Allan Paivio spent his career on this idea and formalized it as dual coding theory (the 1991 paper is the standard reference). The claim: the mind runs separate verbal and visual systems, and information encoded in both leaves two traces instead of one.

Two traces mean two retrieval routes. When you reach for la playa and the word itself won't come, the image of the beach can drag it out. Either code can cue the other. A text-only card has no backup.

Paivio's own experiments kept finding the same ordering: pictures remembered better than concrete words, concrete words better than abstract ones. The gradient tracks imageability, how easily a thing throws a picture, which is exactly what the two-systems account predicts. The more naturally a word recruits the visual warehouse, the more durable its storage.

Mayer's multimedia learning research says the same thing from the education side: people learn better from words plus pictures than from words alone, reliably enough that he calls it the multimedia principle and has replicated it across dozens of experiments.

The picture superiority effect

How big is the visual system's advantage? Comically big. In 1973, Lionel Standing showed participants 10,000 photographs over 5 days, each for around 5 seconds. On a later recognition test, they correctly picked the seen picture about 2 times in 3, against pairs of lookalikes. From 1 glance each, across 10,000 items.

Standing 1973 · memory for pictures

0

photographs viewed once, about 5 seconds each

0 days

of viewing sessions to get through them all

0%

later recognized, roughly, from a single exposure

The same asymmetry shows up in ordinary lab tasks: shown a mixed list of words and pictures, people consistently remember more of the pictures. That's the picture superiority effect, and it's one of the sturdier findings in memory research.

Standing wasn't even the first to find the ceiling missing. Shepard (1967) had run a smaller version with 612 pictures and got recognition near 98% immediately after. The visual system appears to have been built for a world where remembering places, faces, and things was survival, and it brings that overqualified hardware to your flashcards for free.

Tested on foreign vocabulary

Fine, but does it work for learning words? Carpenter and Olson (2012) tested it directly: English speakers learned Swahili nouns paired either with English translations or with pictures. Across experiments, picture-paired words were recalled better than translation-paired ones. The materials were deliberately ordinary (common nouns, simple photos), which is the encouraging part; no memory-palace heroics required, just a picture where a translation would have been.

There's also a subtler benefit. A picture of a beach carries things a translation can't: the glare, the sand, the specific beach-ness of it. Pairing la playa with your memory of a real beach hooks the word into episodic memory, the system that stores your life. Translations hook it into your English, which keeps English in the loop as a middleman. The image lets you eventually skip the middleman and go straight from beach to playa.

Learners feel this difference within weeks. Translation-trained words tend to surface with a beat of lag, the sound of English being consulted. Image-trained words surface the way your own language does: the thing appears in mind and the word comes with it. That lag matters at conversation speed, where half a second per word is the difference between talking and translating.

Making the unpicturable picturable

The obvious objection: fine for beaches and dogs, but what about la verdad (truth), aunque (although), el éxito (success)? You can't photograph although.

You can usually photograph its neighborhood, though. Three moves cover most abstract words:

• Borrow a scene. La verdad can live on a courtroom photo, a pinky-promise, a lie-detector needle. The image doesn't define the word; it just gives the visual system a handle to grab.
• Use a personal episode. El éxito over a photo from the day you finished the marathon beats any stock image, because the memory arrives pre-installed.
• Let the sentence carry it. Grammar words like aunque don't want images at all. A vivid cloze sentence from your own lesson ("aunque llueva, vamos a la playa") is the hook, and the picturable words around it do the visual work.

Wyner's rule of thumb is worth keeping: spend the effort where it pays. Concrete words get direct images cheaply; abstract words get scenes or episodes when one presents itself; function words get good sentences and no guilt.

The catch in the title

Carpenter and Olson's paper has one of my favorite titles in the literature: "Are pictures good for learning new vocabulary in a foreign language? Only if you think they are not."

The catch: pictures make words feel so easy that learners get overconfident, slack off, and give back some of the gain. In their studies, the picture advantage grew when participants were warned against assuming pictures made words easy. The image helps the memory and inflates the self-assessment at the same time.

So use images, and stay humble about what feels easy. The feeling of knowing is exactly the thing pictures distort, and the fix costs nothing: grade the retrieval, never the glow.

Putting images on cards, properly

A few rules make the difference between decoration and dual coding:

• Pick images that mean something to you. Your own photo of a beach beats a stock photo, because it comes with an episode attached. Wyner's Fluent Forever method is built on this, and it's why Verbamor has you choose each card's image rather than assigning one.
• One image, one meaning. A cluttered image gives the visual system nothing crisp to store. The word correr wants a runner, mid-stride, and nothing else in frame.
• Skip the translation when the image can carry the meaning. Beach photo plus la playa, no English on the card. Concrete nouns and many verbs can go fully direct.
• Still test yourself. An image doesn't excuse the card from being a retrieval. The blank stays; the picture is a second hook, a hint that lives in another system.

Abstract words are the honest limitation, and the section above is the workaround, not a cure. When no scene fits, the sentence context does the heavy lifting and the image sits that card out, which is fine. Roughly 400 of Verbamor's base 630 are concrete and picturable on purpose; the frequency list was chosen that way.

The payoff compounds quietly. Every picture-backed word is a word with a spare key under the mat. Months from now, mid-conversation, the verbal route will jam on something you know perfectly well. And a beach will surface, and the word will follow it out.

---

Sources

Every load-bearing claim Verbamor makes is traced to its paper on the research page.