Picture Superiority Effect: Why Images Outlast the Words for Them

The effect is a difference in what a single encounter leaves behind. A picture of a hammer and the word hammer refer to the same thing, yet the picture is the version people can pick out of a crowd of alternatives later, and the version they are more likely to write down in a free recall test. The size of the gap depends on the materials and the test, and it is largest for concrete things that have an obvious picture: a kite, a drum, a lemon.

Allan Paivio's dual coding theory is the long-standing explanation. On that account a picture of a concrete object is coded twice, once as an image and once as its name, while a printed word is reliably coded only verbally, so the picture leaves two routes back to the item instead of one. Paivio and Csapo tested this directly in 1973 by asking participants to form a mental image of each word's referent as it appeared. That instruction gave the words a second code too, and free recall for words and pictures came out about the same.

Dual coding is not the only account on offer, and the current evidence does not settle the question. Work on distinctiveness argues that pictures win because they are physically more distinctive than printed words, which all look alike in the same font. Higdon, Neath, Surprenant and Ensor concluded in 2025 that distinctiveness rather than dual coding explains the pattern, and earlier work had shown that making words visually distinctive can wipe out the picture advantage in recognition. The effect itself is one of the more reliable findings in memory research; what it is caused by is still argued over.

Two boundaries matter for anyone planning to use it. The advantage is documented for recognizing and recalling items, not for every kind of memory, and a picture only carries the meaning a viewer reads off it. A drawing of a cell is remembered as a shape; whether the shape comes back with the right name attached depends on whether the name was learned with it.

Roger Shepard published "Recognition memory for words, sentences, and pictures" in the Journal of Verbal Learning and Verbal Behavior in 1967. Separate groups of participants studied one kind of material: single words, short sentences, or pictures. In the picture condition, 612 pictures were presented one at a time, each for roughly six seconds, with no special memorizing instructions, and each item was seen once.

Memory was then tested by two-alternative forced choice. Each test item paired something from the studied set with something that had not been shown, and the participant chose the one they had seen. Tested immediately, the picture group was correct about 98 percent of the time. The word group was correct about 90 percent of the time and the sentence group about 88 percent. Accuracy fell as the delay between study and test grew, as it does for all three materials, but the ordering held.

Because the three materials were studied by different groups, Shepard's design compares conditions rather than tracking one person across both. That is a limit on the inference: the picture figure and the word figure come from different people doing a very similar task.

Allan Paivio and Kalman Csapo followed in 1973 with "Picture superiority in free recall: Imagery or dual coding?" in Cognitive Psychology, which moved the question from recognition to free recall and asked what produced the advantage. Their key manipulation was to tell participants to generate a mental image of the thing each word named. Doing so left free recall for words and pictures about equal, which is what dual coding predicts: the words had been given a second code. Lionel Standing then pushed the scale of the recognition finding in 1973 by presenting 10,000 pictures and still measuring high recognition accuracy, which is the study usually cited for the sheer capacity involved.

The effect works for you when you choose the picture and against you when someone else does. The rows below are illustrations, not findings from the studies cited on this page.

The effect is an advantage when you control the picture and a cost when the picture carries something you did not mean to transmit, so the practical rules are about who is choosing the image.

• Choose the image to be the message. Whatever is pictured is what will come back, so pick the thing worth keeping rather than decoration.
• Label the picture. A drawing is remembered as a shape; the name has to be learned alongside it or the shape comes back without it.
• Keep pictures for concrete things. The documented advantage is for items that have an obvious picture; an abstract claim drawn as a generic icon has nothing to gain.
• When you are the audience, ask what was in the image and what was only in the text. The part only in the text is the part you will have to work to keep.
• Do not let the effect stand in for study method. Retrieval practice and spacing do more for a whole set of material than picture format does for single items, and they can be combined with it.
• Watch for images doing argument. A vivid photograph makes a case memorable, not true, which is the same mechanism working against you.

Several memory effects produce an advantage for one kind of item over another, and the picture superiority effect is the one about format.

• It is not the von Restorff effect, in which one item that differs from a uniform list is remembered better. That one needs the rest of the list to be the same; the picture advantage shows up when every item is a picture.
• It is not the bizarreness effect, the better memory for strange material. A picture of an ordinary kettle beats the word kettle without being odd at all.
• It is not the availability heuristic, the judgment of frequency by how easily examples come to mind. The picture effect is about what gets stored; availability is about what is then read off the ease of recall.
• It is not a learning style. Pashler, McDaniel, Rohrer and Bjork reviewed the evidence in 2008 and found almost no studies with the design needed to show that matching teaching to a supposed visual or verbal type helps; the picture advantage shows up across people rather than sorting them into types.