A new study published in Nature Human Behaviour has found that speeches in the U.S. Congress have increasingly relied on emotional and subjective language over the past several decades, drifting away from language grounded in facts and evidence. This shift, the researchers say, is linked to rising political polarization, declining legislative productivity, and widening income inequality.
The study was conducted to understand how members of Congress express different conceptions of honesty and truth in their rhetoric. The researchers, based at institutions including the University of Bristol and the University of Konstanz, were interested in what they call a “continuum of truth.” At one end is evidence-based reasoning, grounded in facts and data. At the other is intuition-based reasoning, which draws on feelings, values, and subjective experience.
“There is a basic conundrum about lying politicians and their perceived honesty, which is that some can be disseminating a lot of misinformation despite suffering at the ballot box,” said study author Stephan Lewandowsky, a professor at the University of Bristol.
“For example, Trump made more 30,000 false or misleading claims during his first term, but his supporters nonetheless considered him to be honest by a 3 to 1 margin. This can be explained if people consider someone to be honest simply because they are sincere in expressing their beliefs, irrespective of their accuracy. We extended this to Congressional speeches and we found that since the mid-1970s, political speeches have gradually shifted from evidence-based rhetoric towards rhetoric that is based largely on intuition.”
For their study, the research team analyzed over 8 million congressional speech transcripts from 1879 to 2022. They used computational text analysis techniques to track changes in the types of language used by lawmakers. The researchers developed two dictionaries of keywords—one capturing evidence-based language (such as “proof,” “fact,” and “examine”) and another for intuition-based language (such as “believe,” “feel,” and “opinion”). They then applied a machine learning technique known as Word2Vec to generate word embeddings that represent the semantic meaning of each speech.
The researchers created a measure they called the EMI score, short for “evidence-minus-intuition.” A positive EMI score means a speech used more evidence-based language; a negative score means it leaned more heavily on intuition-based rhetoric. They validated this score by comparing it to human judgments of sample texts, finding a strong correlation.
After computing EMI scores across the full 145-year span of congressional records, the researchers found a clear and sustained drop in evidence-based language starting in the mid-1970s. The EMI score peaked in the 1975–76 legislative session and has trended downward ever since. While both political parties have shifted in this direction, the most recent sessions showed a sharper decline among Republicans.
This decline in evidence-based speech mirrored trends in other areas of American political life. The researchers found that the EMI score was negatively correlated with partisan polarization, as measured by ideological distance between parties in congressional voting records. They also discovered that drops in EMI scores were followed by increases in income inequality two years later, using data from the World Inequality Database. This suggests that intuition-driven rhetoric may be linked to social and economic outcomes.
The researchers also explored whether the decline in evidence-based language affected how productive Congress is. They looked at three measures of legislative productivity: the number of laws passed, the amount of major legislation, and a composite index combining both. In each case, lower EMI scores were associated with reduced legislative output. In other words, when lawmakers used less evidence-based language, they passed fewer and less significant laws.
The researchers accounted for a range of other variables that could influence productivity, such as whether one party controlled both the presidency and Congress, or shifts in public mood toward regulation. Even after controlling for these factors, the relationship between EMI scores and productivity remained.
Interestingly, the link between EMI and income inequality was predictive. Low EMI scores in one session of Congress tended to precede higher income inequality two years later. That wasn’t true of polarization, which moved in step with EMI but didn’t follow or precede it in a clear sequence. This suggests that the rhetorical style of Congress might play a causal role in economic inequality, even if the same cannot be said for polarization.
“Our main finding is that politicians use less evidence-based rhetoric now than they used to, and our study shows that this shift from evidence to intuition has been associated with increasing polarization in Congress and a decline of congressional productivity during that the period,” Lewandowsky told PsyPost.
The findings raise concerns about the health of democratic discourse. As the language of politics becomes more rooted in feeling and belief rather than facts and reasoning, it may become harder for lawmakers to find common ground. Evidence-based arguments can serve as a shared reference point for negotiation, but intuition-based rhetoric is harder to contest or verify, which could contribute to gridlock and partisanship.
The researchers suggest that several forces may be contributing to this shift. Congressional speech is often shaped by party leaders, who control the floor schedule and decide who gets to speak. Increasingly, politicians may also tailor their language to appeal to donors, media, and constituents in a polarized political environment. The rise of televised and later livestreamed congressional sessions may have encouraged performative speech designed to attract attention rather than inform debate.
Broader trends in American politics may also be at play. The expansion of presidential power, particularly through executive orders, has limited the role of Congress in policymaking. If Congress is seen as less powerful or influential, there may be less incentive to engage in detailed, evidence-based debate. The media environment has also shifted, with many politicians using social media platforms where belief-driven statements often attract more engagement than fact-driven ones.
The study does not claim to establish cause and effect, but it does show that changes in political language are strongly correlated with important social and institutional outcomes. The researchers caution that correlation does not imply causation, and they call for future studies to further examine whether and how rhetorical styles influence democratic health.
Still, the evidence points to a troubling pattern: as congressional speeches rely less on facts and more on feelings, Congress passes fewer laws, income inequality grows, and polarization becomes more entrenched. The researchers argue that strengthening a shared commitment to accuracy and evidence could help reverse these trends. They hope to continue exploring the links between political communication and democratic norms.
“I want to explore the connection between misinformation and the welfare of democracy,” Lewandowsky said. “While we have found so far is that when people accept sincerely as a marker of honesty, they become more tolerant of violations of democratic norms by politicians. Conversely, when you remind people accuracy is an important aspect of honesty, they become less tolerant of norm violations. So I am tempted to conclude that honesty and accuracy of politicians are a core component of democracy.”
The study, “Computational analysis of US congressional speeches reveals a shift from evidence to intuition,” was authored by Segun T. Aroyehun, Almog Simchon, Fabio Carrella, Jana Lasser, Stephan Lewandowsky, and David Garcia.
