But what if the conventional wisdom is wrong—and that using attention-grabbing language could help scientists gain support for innovative research?
“This goes against the way scientists are trained,” says Brian Uzzi, Kellogg professor of management and organizations and co-director of the Northwestern Institute on Complex Systems (NICO) and the Ryan Institute on Complexity. When communicating with peers, he says, “you shouldn’t even use adjectives.”
The right kind of adjectives, however, can make or break a researcher’s chance of getting their work funded. Uzzi calls it “promotional language”—and after studying nearly 17,000 scientific grant applications, Uzzi and his colleagues (Hao Peng and Huilian Sophie Qiu of NICO and Henrik Barslund Fosse of the Novo Nordisk Foundation) found that scientists who used it twice as likely to win financial support for their research.
But that was no Angry men– style publicity stunt, confusing hapless funders into cutting checks for bad science. Instead, Uzzi and colleagues found that promotional language tended to appear in more inherently innovative research proposals—and that the science supported by these successful grants was more likely to be published in high-quality journals. The research team also used machine learning to analyze how advertising language affected the “feel” of successful grant applications – that is, whether a reviewer would respond positively or negatively to them. When promotional language in an app was replaced with neutral-sounding synonyms, positive sentiment for that app dropped—indicating that an actual reviewer would be less likely to endorse it.
“Grants are the gateway to doing bigger and better science,” says Uzzi. “This work suggests that promotional language is important not only for securing funding, but also for effectively communicating the benefits of good ideas.”
Definition of “Promotion”
Some research has already been done on what makes a successful scientific grant application — and the results suggest that truly innovative work can be difficult. In fact, says Uzzi, “more new grant proposals tend to be rejected.” However, the study showing this association analyzed only 30 applications. Uzzi and his colleagues, meanwhile, were able to draw from a much larger sample size of 16,730 grant applications—both successful and unsuccessful—submitted to the National Institutes of Health (NIH), the National Science Foundation (NSF), and Novo Nordisk Foundation. “This gave us a much stronger basis for drawing our conclusions,” says Uzzi.
To measure advertising language in this data set, Uzzi and colleagues relied on a 139-word dictionary compiled in 2022 by linguist Neil Millar. The use of these words—such as “unique,” “revolutionary,” and “groundbreaking”—increased significantly in successful NIH grant applications between 1985 and 2020. By analyzing their frequency in both successful and unsuccessful applications, Uzzi and colleagues they hoped to determine whether those buzzwords moved the needle on a grant’s ability to secure funding.
However, researchers were also interested in the ability of advertising language to successfully convey the importance of new scientific ideas. To measure the intrinsic novelty of a grant proposal, Uzzi relied a measure he published in Science in 2013which searches for unconventional citation pairs in scientific papers.
“Scientists generally combine ideas from the past,” he explains, “but many papers combine this past knowledge in very familiar ways”—for example, citing Albert Einstein and Isaac Newton in the same physics paper. Innovative papers, meanwhile, contain more unusual citation pairs. Uzzi likens it to the ingredients in a recipe: “Tomato sauce and garlic is a combination that will appear in many different recipes, but garlic and ice cream would be unusual.”
Since grant proposals also contain references to prior scientific work, Uzzi’s metric could provide a similar measure of grant innovation in their data set. And comparing this inherent innovation to the frequency of buzzwords in a grant proposal—and to the success or failure of the grant—would shed light on whether the pitch language helped communicate the value of innovative science.
Believing the hype
Uzzi and his colleagues found a clear link between advertising language and successful grants. On average, one out of every hundred words in our sample of grant applications was advertising—about one advertising word for every four proposals. In the majority of grants, increasing ad word frequency from 1 to 2 percent was associated with a 46 percent increase in the likelihood of funding success.
But this wasn’t just hype at work: the researchers also found that promotional language was associated with more innovative research proposals. By Uzzi’s measure, grant applications with two advertising adjectives for every hundred words—about twice as many as usual—were 54 percent more innovative than the average proposal.
Furthermore, the science that actually emerged from these proposals also tended to be highly valued. By tracking the scientific papers that came out of these successfully funded grants, Uzzi and his colleagues showed that the same amount of promotional language in a sentence—two words out of a hundred—led to subsequent publications that increased their journal’s predicted impact factor (measure how often the typical article is cited in this journal) 51 percent higher than average.
“People tend to say, ‘Well, isn’t advertising language just a lot of window dressing?’ Isn’t it just promoting a lot of science that doesn’t deserve it, and weakening research as a whole as a consequence?” says Uzzi. “What this research shows is that scientists use advertising language in a judicious way that reflects the true innovation of their work.”
Of course, as any good scientist knows, correlation does not imply causation— so Uzzi and his colleagues needed a way to show that the use of advertising language could actually cause the effects they observed. To do this, Uzzi and his colleagues devised a “what if” experiment to simulate how removing promotional language would affect the success of a grant proposal.
First, the researchers compiled a list of neutral synonyms for the 139 advertising words in their dictionary: for example, replacing the word “unique” with the more neutral-sounding “specific,” or swapping the word “rebel” for “different.” They then replaced the promotional language in the grant proposals with these neutral words.
The researchers then used a machine learning model to categorize the “sentiment” of each grant release. (Sentiment analysis, which measures the positive or negative emotional tone of a text, is a common task in natural language processing.) There is no way to know for sure how real grant reviewers would have reacted to alternative versions of the sentences. But previous research has shown that positive sentiment is linked to grant approval—so by measuring the difference in sentiment between promotional and non-promotional versions of the proposals, Uzzi and his colleagues could assess how the change in wording might have affected their chances of success .
When all of the ad words were replaced with neutral terms, the positive sentiment dropped significantly for more than 80 percent of the proposals—suggesting that they would be less likely to secure funding. Furthermore, Uzzi and colleagues found that advertising language was actually a better prediction of success rather than general positive feeling.
“Once you have the right measure, the previous one ceases to be relevant,” explains Uzzi.
Hidden pieces
For Uzzi, these findings puncture the long-held notion that cold, hard data is the only thing that matters for good science. Instead, using language that helps highlight the importance of that science—“hype,” in less charitable terms—can actually be a critical part of advancing scientific knowledge, especially when it comes to supporting innovative research, the importance of which may not be immediately obvious even to experts.
“It’s not about changing perception – it’s that something important might not even be exists at you until your mind is ready to understand it,” says Uzzi. “If you and I were walking in the woods, we might step over all kinds of animal tracks and never know, not because they’re not there, but because we’re not trained to see them in the first place. Promotional language helps to stimulate and direct your attention to things that you may not have realized were worth it.”
So should scientists resist their traditional “just the facts, ma’am” impulses and deliberately use more promotional language when seeking support for innovative research? Uzzi’s answer is clear: “I would say yes.”
The reason, he argues, is not to tout innovations where they don’t exist—it’s to make sure that real breakthroughs actually have a chance to happen. “What we see in the data is that even if the grant has great ideas, the absence of promotional language reduces the chances of getting funding,” says Uzzi. “And that slows things down for the rest of us, because good science isn’t recognized and adopted.”