New research suggests that the way poll questions are phrased can blur the line between association and causation, especially when public health claims are politically charged.
Research: Support for increased risk claims regarding Tylenol and autism may not reflect a causal relationship. Image credit: Chemical Industry / Shutterstock
Recent reports published in magazines scientific report suggest that linguistic frameworks can influence the interpretation of scientific claims, with the clearest wording effects observed among participants with a university degree.
Researchers investigated whether differences in the wording of recent poll questions about the association between Tylenol (acetaminophen) use during pregnancy and autism risk influenced public support for this claim.
This study did not evaluate the biological relationship between acetaminophen and autism. Instead, we tested how people interpret the wording of public health claims in different ways. They found that “associated with” and “increased risk” were endorsed by similar proportions.
Participants, especially college-educated respondents, strongly endorsed “associated with” over “caused,” and the wording appears to influence how strongly people endorse the statement, rather than simply agreeing or disagreeing. The higher the education level, the greater the overall skepticism towards this claim.
Although Republicans, Democrats, and independents differed in their overall likelihood of support, they responded similarly across the three expressions. These findings indicate that representation has only a small effect on public interpretation.
Causal interpretations of correlational findings can mislead scientific information. The language used in press releases and other communications can obscure the original results. People may also form beliefs based on causal conclusions drawn after reading media, institutional, or research-based information.
On September 22, 2025, at a White House press conference, officials said that Tylenol use during pregnancy may be associated with increased risk of autism. This statement contributed to broader public debate and concern.
In a subsequent KFF poll, many respondents said it was probably or definitely true that Tylenol use during pregnancy increases the risk of autism, although the poll did not directly ask whether they believed Tylenol causes autism. In that context, the KFF poll and subsequent coverage raised the question of whether support for high-risk language reflects a belief in direct causation or a recognition of a possible association.
About the report
For the new report, researchers conducted an experiment with 1,957 people to see how people interpreted poll questions written in different words. These participants were recruited through CloudResearch’s Connect platform. The researchers compared three expressions that appeared in media coverage of the poll. These phrases included “cause,” “increased risk of,” and “associated with.”
Participants first indicated their awareness by responding yes or no to whether they had heard or read about the use of Tylenol by pregnant women or people with autism. They then responded whether they believed the association was definitely true, probably true, probably false, or definitely false.
The team ensured participant attention by including a question asking which two topics were covered in the survey. Participants chose from a list of opioids, vaccines, Tylenol, and autism. The team then excluded 43 participants who failed this attention check, leaving 1,957 participants for the primary analysis. Seven participants without education data were excluded from analyzes regarding education only.
The researchers also examined the effects of political party and educational background on the interpretation of the expression and support for health claims. They used logistic regression to estimate the odds ratio (OR) of endorsement across the three terms. We also performed an exploratory analysis of response options to the phrases using ordinal regression to assess the level of support.
result
Nearly 90% of study participants were aware of this theme. Their perceptions were similar across expressions. People were less likely to strongly support a link between Tylenol and autism when the word “cause” was used than when the word “increased risk of…” was used.
In the main binomial analysis, this effect was only marginally significant, but in the exploratory four-point ordinal analysis, the probability of stronger support was shown to be significantly lower. Specifically, the odds of receiving stronger support were approximately 22% lower (OR, 0.78). Participants responded similarly to the phrases “associated with” and “increased risk.” They were more willing to support the statement that Tylenol is associated with autism than the statement that Tylenol causes autism.
Perceptions of Tylenol Autism Claims by Causal Language, Political Party, and Education. The value is the observed sample proportion.
Overall, the wording had only a limited effect on how strongly participants endorsed the statement. This wording influenced not only whether participants agreed or disagreed, but also their movement between “probably” and “definitely” responses. However, the answers vary widely depending on the characteristics of the respondents.
Support for the stated relationship was high among Democrats, Republicans, and independents. Specifically, compared to Democrats, Republicans had approximately 9.8 times the odds of endorsing the statement, and Independents had 3.25 times the odds of endorsing the statement. However, they all responded to different phrases in similar patterns.
Educational level influenced the effectiveness of phrases, with the clearest expression-related differences found among participants with a university degree. Highly educated participants expressed weaker support for a claim if the language used “cause” instead of “increases the risk of” or “associated with.” They used the latter two phrases to indicate a similar likelihood of support.
Participants with a four-year college degree were 34% less likely to support the claim than participants with less education. The sample was recruited online and was not nationally representative, so results may not be fully generalizable to the broader U.S. population.
conclusion
Based on our findings, even small changes in the language used to describe research findings can somewhat shape the way people express support for scientific claims, and in particular the strength of support.
This result indicates the need for caution when drawing conclusions about the public’s beliefs about causal relationships. Although subtle differences in the wording of studies may go unnoticed by many, higher education levels may help distinguish correlation from causation more clearly.
In the future, researchers should employ measures that can more clearly distinguish between beliefs about association and beliefs about causation among broader populations.
Want to read it later? Click here to download a PDF copy.
Reference magazines:
- J. S. Downs, C. Drummond-Otten, SM Anglin (2026). Support for increased risk claims regarding Tylenol and autism may not reflect a causal relationship. scientific report, Doi: https://doi.org/10.1038/s41598-026-56863-1
