A lack of fact-based reasoning in politics can hurt democracies. (CHUYKO SERGEY/Shutterstock)
In a nutshell
- Congressional language has shifted from evidence to emotion. Since the mid-1970s, lawmakers have increasingly used intuition-based words like “believe” and “feel” instead of evidence-based terms like “fact” and “proof,” marking a historic decline in fact-driven rhetoric.
- This linguistic shift is linked to rising polarization, inequality, and legislative gridlock. The study found that when Congress uses less evidence-based language, partisan divisions deepen, income inequality grows, and fewer significant laws get passed.
- Both parties are part of the trend, but Republicans saw a sharper recent drop. While Democrats and Republicans followed a similar downward path over decades, the steepest decline in evidence-based speech during the 2021–2022 session came from Republicans.
KONSTANZ, Germany — For decades, evidence-based reasoning has been losing ground in the halls of Congress, making way for a more intuition-driven approach to political discourse. According to a new international study, U.S. congressional speeches have undergone a profound linguistic transformation since the mid-1970s, with lawmakers increasingly abandoning factual arguments in favor of gut feelings and personal beliefs.
The study, published in Nature Human Behaviour, analyzed 8 million congressional speeches from 1879 to 2022. It reveals that evidence-based language reached its peak in the 1970s and has been declining ever since. This shift away from fact-driven rhetoric coincides with rising partisan polarization, growing income inequality, and decreasing legislative productivity. Essentially, Congress may be getting less done.
“In many democracies, there is currently much concern about ‘truth decay’: the blurring of the boundary between fact and fiction, not only fueling polarization but also undermining public trust in institutions,” explains lead author David Garcia from the University of Konstanz, in a statement.
The researchers explain that healthy democratic discourse requires a balance between evidence and intuition. While factual reasoning creates a foundation for substantive debate, emotional and experiential perspectives are also valuable for addressing complex societal issues. However, when intuition completely dominates the conversation, evidence can no longer serve as an arbiter between competing political positions, making compromise increasingly difficult.
Researchers developed what they call an “evidence-minus-intuition” (EMI) score to measure the relative prevalence of evidence-based language over intuition-based language in congressional speeches. Words like “fact,” “proof,” “science,” and “analysis” were categorized as evidence-based language, while words such as “believe,” “feel,” “opinion,” and “common sense” were labeled intuition-based.
Congressional speech remained relatively stable from the late 19th century through the early 20th century, with a slight uptick in evidence-based language starting in the 1940s. The researchers observed that evidence-based language peaked in the mid-1970s before beginning its steady decline to present day, when it has now reached its historical minimum.
The decline affected both Democrats and Republicans, with both parties following nearly identical downward trajectories in their use of evidence-based language since the 1970s. In one of the most recent congressional sessions (2021-2022), Republicans showed a substantially steeper drop than Democrats in evidence-based language.
The researchers noted that both parties have moved downward together in their rhetorical patterns since the mid-1970s, suggesting a synchronized shift in messaging strategies across the political spectrum.
But why does this shift matter? Researchers say there may be a link between evidence-based language and several measures of democratic health. The decline in evidence-based language is strongly associated with increasing partisan polarization. When polarization increases, evidence-based language decreases. This suggests that as political discourse becomes more divided, politicians are less likely to appeal to commonly accepted facts or data, instead falling back on subjective interpretations and personal beliefs.
The researchers also discovered that lower levels of evidence-based language predict subsequent increases in income inequality. In other words, when congressional rhetoric shifts away from evidence, income inequality tends to worsen in the following years. This relationship held even after controlling for numerous other factors.
Congressional productivity, measured by the quantity and quality of laws passed, is positively correlated with evidence-based language. When Congress uses more evidence-based language, they tend to pass more significant legislation. This makes sense: when lawmakers ground their arguments in shared facts rather than personal feelings, it’s easier to find common ground and craft effective policy solutions.
Why has this shift happened? Party leadership exerts substantial control over who speaks on the congressional floor, potentially shaping the content and tone of speeches to align with strategic objectives. Members of Congress also face pressure to cater to their base, including constituents, donors, and lobbyists, particularly in a highly polarized environment.
Changes to congressional rules and procedures may have also influenced discourse patterns. For instance, the introduction of the “cloture” rule in the Senate in 1917 provided a mechanism to limit debate time, potentially shortening discussions and limiting the depth of legislative debates.
Media coverage has also shaped congressional rhetoric, especially since the introduction of C-SPAN in the House in 1979 and the Senate in 1986. In an era of increasing polarization, politicians might adopt a perpetual campaign style, transforming congressional speeches into performances aimed at capturing media attention rather than engaging in substantive policy discussions.
When evidence-based language declines, so does the ability to have productive political debates based on a shared understanding of reality. The resulting political impasse makes it increasingly difficult to address pressing societal challenges, from climate change to healthcare reform. This should alarm Americans across the political spectrum. A democracy that cannot engage in factual debate is a democracy in peril.
Paper Summary
Methodology
Researchers analyzed 8.4 million congressional speech transcripts from 1879 to 2022, creating dictionaries of words representing evidence-based language (49 keywords including “fact,” “data,” “science”) and intuition-based language (35 keywords including “believe,” “feel,” “opinion”). They used Word2Vec embeddings to represent speeches and calculate an “evidence-minus-intuition” (EMI) score, measuring the relative prevalence of evidence-based versus intuition-based language. Human raters validated the EMI scores, confirming they accurately distinguished between evidence-based and intuition-based language with an overall accuracy of 79%. The researchers then analyzed how EMI scores correlated with measures of polarization (using DW-NOMINATE scores), income inequality (using share of pretax income of the top 1%), and congressional productivity (using indices measuring legislative output).
Results
The study found that evidence-based language remained relatively stable from 1875 through the early 20th century, increased from the 1940s to a peak in the mid-1970s, and has been declining steadily since. Both Democrats and Republicans showed similar downward trends since the 1970s, though Republicans dropped more steeply in the 2021-2022 session. Statistical analysis revealed strong negative correlations between EMI and partisan polarization, with polarization predicting lower EMI scores. The researchers also found that lower EMI scores predicted higher inequality in subsequent years, and that higher EMI scores were associated with greater congressional productivity in terms of passing significant legislation. These relationships remained significant even after controlling for other variables.
Limitations
The study is correlational, not causal, so it cannot definitively establish cause-and-effect relationships between declining evidence-based language and outcomes like inequality or polarization. The researchers acknowledge that their analysis doesn’t account for every factor that might influence congressional language, polarization, and productivity. Additionally, measuring polarization through voting behavior (DW-NOMINATE scores) captures only one dimension of political divisions and doesn’t account for other aspects like affective polarization or issue polarization.
Funding and Disclosures
The research was supported by the European Research Council, the Humboldt Foundation, the Volkswagen Foundation, and the European Commission, including an ERC Advanced Grant called PRODEMINFO. Researcher David Garcia also received funding from the Deutsche Forschungsgemeinschaft under Germany’s Excellence Strategy. The funders had no role in the study design, data collection, analysis, decision to publish, or preparation of the manuscript.
Publication Information
The study, “Computational analysis of US congressional speeches reveals a shift from evidence to intuition,” was published in Nature Human Behaviour on April 10, 2025. The authors are Segun T. Aroyehun, Almog Simchon, Fabio Carrella, Jana Lasser, Stephan Lewandowsky, and David Garcia, representing institutions including the University of Konstanz (Germany), Ben-Gurion University of the Negev (Israel), University of Bristol (UK), University of Graz (Austria), Complexity Science Hub (Austria), and the University of Potsdam (Germany).
And the authors undermine their own credibility by weaponizing it right at the start with a partisan spin