A particular source of dread for politicians is how to respond to negative campaigning or other information impugning their character. By responding, they might only bring attention to an issue that voters hadn’t even recognized: “Contrary to my opponent’s claims, I have stopped beating my wife, and I haven’t consumed more than a fifth of hard liquor in a single sitting.”

Worse, many studies have found that even unequivocal denials fail to register in memory. In one study, participants read a report about the possible cause of a fire: a room full of oil paint and pressurized gas cylinders. Later in the report an addendum indicated the room was actually empty, but when questioned, most respondents still believed that the flammable materials in the room had caused the fire.

Clearly there is considerable power in being the first to assert the “truth.” But surely people are capable of revising their opinions based on factual evidence. Some of the greatest works of literature in history hinge on readers recognizing that an early impression of a character turned out to be false, from Oedipus learning that he has married his mother to Elizabeth Bennet realizing that Mr. Darcy isn’t actually a rat and a scoundrel. If we go on thinking Darcy is proud and prejudiced, we miss the whole point of the novel.

So why is it that sometimes we go on thinking Senator so-and-so is a wife-beater despite his denials but other times we re-evaluate our position based on the evidence? Are we just paying more (or less) attention? David Rapp and Panayiota Kendeou had student volunteers read 24 different “stories” involving a character who demonstrates a trait like sloppiness or laziness, which then may be immediately contradicted in the story. The stories were all just 13 sentences long, and each sentence was displayed one at a time on a computer screen. Here’s the beginning of a sample story:

Albert was listening to the radio. He had finished getting ready to meet his friends at the movies. They were going to see a new comedy that was getting rave reviews. He pulled a sweater over his head. Then he began to look for his shoes. They were buried under old candy wrappers, crumpled magazines, and some dirty laundry. Albert cared about the condition of his room, even though it currently wasn’t up to par.

So the idea that Albert is sloppy is refuted by the assertion that he “cared” about his room. The next part of the story then offered a situation where Albert’s sloppiness (or lack thereof) is put to the test:

Albert had to take the bus to go to the movies. He bought a newspaper to read during the ride to the theater. Albert had finished leaving through the paper when his stop was announced. Albert put the newspaper on the seat next to him. As he waited for the bus to stop, he noticed a sign asking riders not to leave garbage on the bus. Albert ignored the sign and got off the bus.

Readers were asked if this ending was plausible given the details they had learned in the story. The researchers actually created eight different versions of each story, and while each reader only saw one version of each story, eight of the 64 participants read each different version.

To see how readers were affected, the researchers systematically changed the last sentence of each part of the story. There were four different endings to the first part, and two endings to the second part of the story. The endings to the first part contradicted or confirmed the trait demonstrated in the story in different ways. The example above is a “simple refutation.” Another type was a “refutation with explanation”:

Albert cared about the condition of his room, but had only moved into the apartment yesterday.

Another ending confirmed the trait:

Albert didn’t care about keeping his room clean, and this is how it usually would look.

The final version (actually affecting the final two sentences) was a control didn’t describe his trait at all.

Finally, there were two possible endings to the second part, one which was consistent with the trait of sloppiness (as above), and the other which was inconsistent:

Albert picked up the newspaper to throw away later.

That’s a total of eight different versions of the whole story. Remember, each individual reader only saw one version. Over the course of the entire study, they saw every possible combination of refutations/confirmations/controls/endings, just focused around different stories. Here are the results:

There’s a lot of data in this graph, but the take-home point is relatively simple, especially when we consider it in the context of our example. The trait presented in the story is “sloppiness.” When the ending of the story is consistent with the notion that Albert is sloppy, and nothing in the story contradicts that notion, then nearly 80 percent of readers say the ending is plausible. But when the trait is refuted in the story, as in our example above, the portion of readers agreeing with a “sloppy” ending decreases significantly. When Albert’s sloppiness is refuted with the explanation that he just moved, most readers believe it’s implausible that Albert would have left his newspaper on the bus.

The researchers also found that readers spent more time reading the sentences that contradicted the trait in the story, again suggesting that readers do pay attention to refutations.

In two additional experiments with the same stories, Rapp and Kendeou looked at the impact of different sets of instructions, and found one additional curious result:

When they gave no instructions on how to read the stories, then readers behaved differently with simple refutations of the trait compared to full explanations of why the trait wasn’t true. With simple refutations and no instructions, when the outcome of the story was consistent with the trait (Albert is sloppy and left the newspaper on the bus), readers spent less time reading the outcome then when it was inconsistent (Albert is sloppy and picked up the newspaper). Even though the trait was refuted, their reading behavior followed the identical pattern as when the trait was confirmed by the story. Only when the refutation was accompanied by an explanation (Albert had just moved in to his apartment) did readers spend more time reading the unexpected outcome (Albert isn’t actually sloppy but left the newspaper on the bus).

But when readers were instructed to read carefully, this difference between simple refutation and refutation with explanation disappeared. As you might expect, more time overall was spent reading, but readers also spent more time reading unexpected outcomes (Albert isn’t actually sloppy but left the newspaper on the bus) compared to expected ones (Albert isn’t actually sloppy and he picked up the newspaper).

This result may offer some insight into why politicians have so much difficulty refuting those nasty campaign ads: when the public isn’t paying much attention, they may be less likely to allow the refutation of an attack to replace the original memory of the attack itself.

Rapp, D.N., Kendeou, P. (2007). Revising what readers know: Updating text representations during narrative comprehension. Memory & Cognition, 35(8), 2019-2032.