Shocking headlines spark initial doubt but eventually build belief

When news headlines begin with sensational words like “Shocking,” readers initially tend to doubt the accuracy of the information presented. However, new research suggests that this skepticism fades over time, leading people to eventually believe the claims made in those headlines. These findings were published recently in the journal Communication Research.

This phenomenon suggests that while clickbait-style language might hurt credibility in the short term, it may be effective at planting information that people eventually accept as true. The study highlights a potential mechanism by which misinformation or sensationalized news can bypass critical filters over long periods.

Journalists and content creators frequently use emotional language to capture attention. Previous analyses have shown that posts with highly emotional content are shared more frequently.

Despite this, it was not fully understood how specific high-intensity words affect belief over time. Past research mainly looked at whether words were positive or negative rather than how stimulating they were.

Xiaoyu Zhou from Beijing Normal University and colleagues from Tsinghua University sought to close this gap. They focused on “High-Arousal Sentence Starters,” or HASS. These are short, emotionally charged phrases placed at the beginning of a headline, such as “Warning!” or “Shocking!”

The researchers based their work on a psychological concept known as the sleeper effect. This effect occurs when a persuasive message is paired with a signal that it is untrustworthy, known as a discounting cue.

Initially, the discounting cue causes the audience to reject the message. Over time, however, the audience often dissociates the message from the cue. They remember the information but forget that they originally thought the source was unreliable.

The research team proposed that high-arousal sentence starters act as these discounting cues. They theorized that readers see sensational words and immediately associate them with low-quality journalism or fake news.

To test this theory, the researchers designed a series of five longitudinal experiments. They also performed a single-paper meta-analysis to confirm the consistency of their results across all the experiments.

The first study, Study 1.1, involved 503 participants. The researchers presented them with headlines related to the COVID-19 pandemic. Some headlines included high-arousal starters, while others did not.

For example, one group saw the headline “Shocking! Smoking may bring lower back pain.” The other group saw a plain version that simply read, “Smoking may bring lower back pain.” Participants rated how much they believed the statements.

Fifty-four days later, the participants returned to rate the headlines again. In this second phase, the high-arousal words were removed from the headlines for everyone.

The results showed a clear pattern. In the initial session, participants rated the headlines with sensational starters as less true than the plain headlines. The sensational language acted as a warning sign that lowered their immediate trust.

However, the follow-up session revealed a shift. Belief in the headlines that originally had the sensational starters increased after the 54-day delay. The initial skepticism had worn off.

The researchers conducted a second study, Study 1.2, to replicate these findings with a shorter time interval. They used general health headlines instead of pandemic-specific ones. This time, the delay was only two weeks.

Even with a shorter break, the pattern remained the same. The immediate effect of the sensational language was to suppress belief. Two weeks later, that suppression evaporated, and belief in the content rebounded.

The team then wanted to see if reading a full article would change the outcome. In Study 2.1, participants read short articles accompanied by headlines either with or without the sensational starters.

The participants rated their belief in the headlines immediately after reading the text. Two weeks later, they were asked to recall the content and rate their belief again.

Once again, the sensational titles reduced belief initially. But after two weeks, the participants who had seen the sensational titles showed an increase in belief. This happened even though they had read a full article that explained the concept.

Study 2.2 was designed to test whether the sleeper effect occurred because people simply forgot the sensational headline. The researchers added a “reinstatement” condition.

In the follow-up session, some participants were explicitly reminded of the original headline they had seen, including the sensational words. Others were not given this reminder.

The results showed that the rebound in belief happened regardless of the reminder. Even when participants were reminded of the “Shocking!” label, their belief in the underlying information had still grown since the first viewing.

This suggests the effect is robust. It implies that the cognitive separation between the “untrustworthy” cue and the information happens deeply. Simply seeing the cue again does not necessarily reverse the belief that has formed.

Finally, Study 3 examined whether this effect extended beyond the specific headline to related beliefs. The researchers wanted to know if a sensational headline made people believe other, connected ideas.

Participants read an article with or without a high-arousal title. They then rated their agreement with statements that were relevant to the topic but not explicitly in the headline.

The data showed that the effect did generalize. The sensational titles initially lowered belief in related arguments. Over time, however, belief in those related arguments increased for the group that saw the sensational titles.

A mini meta-analysis of all five studies confirmed the reliability of these trends. The immediate suppression of belief was a medium-sized effect across the board. The delayed rebound in belief was also consistent and of medium magnitude.

The researchers concluded that high-arousal sentence starters function as discounting cues. They signal to the reader that the news might be fake or low quality. This triggers a defensive skepticism right away.

But as time passes, the association between the information and that skepticism weakens. The content of the headline remains in memory, while the “low credibility” tag fades away.

This leads to a situation where the reader eventually believes the information they once doubted. The initial sensationalism might actually help the information stick in memory, even if it is doubted at first.

These findings have implications for how media is consumed and regulated. It suggests that fact-checking or warning labels might function similarly to these sensational starters.

If a warning label acts as a discounting cue, it might work immediately. But there is a risk that the warning itself is forgotten over time while the false information is remembered.

The authors note that their study focused specifically on health news. It is possible that political or financial news might trigger different reactions.

They also point out that they used true information for ethical reasons. They did not want to implant false beliefs in participants. However, the psychological mechanism suggests false information would behave similarly.

Future research is needed to understand the exact cognitive processes at play. For instance, it is unclear if the rebound happens because the skepticism fades or because the sensationalism makes the content more memorable.

The researchers also suggest looking at different types of arousal words. Words like “Urgent” might have a different impact than words like “Shocking.”

Cultural differences could also play a role. The study was conducted with Chinese participants. Western cultures might respond differently to high-arousal language due to different cultural norms regarding emotion.

The study, “The Immediate and Delayed Beliefs in Headlines With High-arousal Sentence Starters,” was authored by Xiaoyu Zhou, Zhang Tan, Danjun Wang, Fei Wang, and Kaiping Peng.