Participants were American individuals from MTurk who volunteered in exchange for $2.50USD (they were not required to be citizens or voters). This sample was from a larger dataset of 409 individuals from which quantitative results have been previously published (Harmon-Jones et al., 2020). Of these, 217 rated their support for Trump a 1 (not at all favorable) and were excluded. Another 64 gave a nonsensical response to the open-ended questions (i.e., examples, one-word answers, text copied from the Internet that did not answer the questions). In this and the other studies, these exclusion criteria were based on the guidelines from Chmielewski and Kucker (2020). The remaining 128 (75 men, 53 women; M_Age = 39.05, SD = 11.61) rated their favorability towards Trump at least a 2 (slightly favorable) and provided an intelligible answer to at least one of the open-ended questions on why they support Trump. Their reported ethnicities were White 75.0%, Black 9.4%, Asian 3.1%, Hispanic 7.8%, and Other 4.7%. They participated in this study on October 31, 2019 (start and end date), at a time when support was building in the US legislature to call for Trump’s impeachment. All 3 studies in this article were approved by the Human Research Ethics Approval Panel – C: Behavioral Sciences at the University of New South Wales.

After participants provided written informed consent, participants answered demographic questions (gender, age, ethnicity, first language, religious affiliation). They also reported how religious they considered themselves (1 = not at all to 7 = very), their employment status (1 = unemployed to 6 = employed greater than full time), and their yearly household income (1 = < $20,000 to 7 = > $200,000). Then, they answered the following questions: 1) How much do you try to follow the rules/stick to the rules/respect authority; 2) How much do you care about justice; 3) How much do you believe in one true religion; 4) How often do you tend to vote for politically liberal candidates? These questions were answered on 1 (not at all) to 7 (very much) scales. Next, they were asked to indicate their city and state of current residence; which social media platforms they used (Facebook, Twitter, Instagram, Snapchat); how often they used social media sites (1 = never to 8 = several times per day); how often they shared memes on social media (1 = never to 8 = several times per day); how often they read online or paper newspapers (1 = never to 7 = very often).

Then, they answered the following questions: 1) How favorable are you toward Donald Trump as president of the United States at the current moment (1 = not at all favorable to 7 = very favorable); 2) The US House of Representatives recently voted to approve two articles of impeachment against President Trump. To what extent do you agree with this statement? The House of Representatives made the correct decision (to approve impeachment) (1 = disagree completely to 7 = agree completely).

After they responded to these questions, they were instructed to read the following news article carefully and that they would be asked “questions about its content and about your thoughts and opinions on the article and the events that it reported upon.”

They were randomly assigned to read one of two articles. One article was titled, “Donald Trump sexual misconduct allegations.” It was presented on 3 separate screens and it contained a total of 495 words. It summarized the allegations of sexual misconduct by Trump. It was based heavily on the summary provided in Wikipedia at the time we created the study. The other article was titled, “Space Exploration.” It was also presented on 3 separate screens and it contained a total of 502 words. It was a neutral article about the history of space exploration involving multiple countries, and it was also based on a summary provided in Wikipedia at the time we created the study. See Supplementary Materials for exact wording. After reading the article, participants were asked to report how likely they were to share a meme (these results were previously published).

Next, participants were informed: “We are interested in the reasons why people do or do not support Donald Trump as president.” And they were instructed to answer one of the two following open-ended questions based on whether they were favorable or unfavorable toward Trump: 1) “If you are favorable toward Donald Trump as president, what are the most important reasons for your support for his presidency”; 2) “If you are unfavorable toward Donald Trump as president, what are the most important reasons why you do not support his presidency?” We included these questions to provide context to why Trump was supported. We also thought participants might give more honest and valid responses if they were asked both positive and negative questions about Trump, as recommended in guidelines for methods (Harmon-Jones, Harmon-Jones, et al., 2025).

After that, participants were informed: “The media has included many reports that Donald Trump has engaged in illegal or sexually immoral activities. We are interested in how people feel about this.” And they were instructed to answer one of the two following open-ended questions based on whether they were favorable or unfavorable toward Trump: 1) “If you support Trump, why do you support him?” and 2), “Trump has been accused of sexual misconduct; if you support him, how do you justify your support given these allegations?” The second question was more specifically relevant to dissonance because it explicitly asked participants to justify their support given allegations of misconduct, whereas the first question asked for their reasons for support without mentioning possible misconduct.

Then, within each article condition, participants were asked 4 multiple-choice questions related to their article, to assess memory for the article’s contents. Following those questions, participants were asked the following: How accurate do you believe the information presented in the newspaper article is? (1 = not at all accurate to 7 = very accurate).

The open-ended responses were coded by a judge (first author) to identify the themes expressed therein. This single rater read all of the responses to get an overall sense of the themes they included. She then created categories that reflected these themes. She then reread the responses and assigned a 1 to the category if the participant’s response included the theme and a 0 if it did not. Responses were allowed to fit multiple categories. A second judge independently coded the responses and reliability was good for the categories of economic reasons and not being a career politician (ICC > .75; Koo & Li, 2016). Reliability was acceptable for the other categories (ICC > .5), except for good character which was poor (ICC < .5), suggesting that this category may be difficult to categorize. Experimental condition did not influence open-ended responses.

In answer to the question of why participants support Trump, the most frequent category was economic reasons. This was followed by Trump’s communication style, his competence, Trump's good character, that he is not a career politician, that he cares about Americans and is committed to making America great, and his specific policies/issues. An additional 18.8% of the responses could not be placed in any of these categories (see Table 1).

Of the participants, 113 answered the second question regarding how they justify their support for Trump given the reports of his sexual misconduct. In response to this question, three non-exclusive categories were identified (see Table 2). The most common category of responses was explicitly disbelieving the accusations, and the other two categories were stating that they support Trump based on his policies not his personal life, and pointing out that other politicians also behave immorally. In addition, 9.7% provided responses that could not be placed into these categories. A second judge coded the responses and reliability was good (ICC > .75) for all categories.

Similar to Study 1, American participants were recruited from MTurk and paid $2.50 USD for their participation (they were not required to be citizens or voters). Of the 713 individuals who participated in this study, 309 rated their favorability towards Trump as a 1 and were excluded, another 10 did not answer this question and were excluded. Another 221 participants gave non-sensical responses (examples, one-word answers, text copied from the Internet that did not answer the questions). The remaining 173 (105 men, 68 women, age M = 37.05, SD = 12.28) rated their favorability towards Trump at least a 2 (slightly favorable) and provided an intelligible answer to at least one of the open-ended questions on why they support Trump. Their reported ethnicities were White 76.3%, Black 4.0%, Asian 4.0%, Hispanic 8.7%, and other 6.9%. The open-ended responses were coded by two judges, similarly to Study 1.

Participants completed the same MTurk study that was used in Study 1 and they were prevented from completing it if they had already participated in Study 1. They were given the same instructions and questions that were used in Study 1.

In answer to the question of why participants support Trump, similar themes were identified to those categorized in Study 1, plus one additional category (see Table 3). In this sample, the most frequent category was economic reasons, followed closely by specific policies/issues. The other categories were that Trump cares about America and her people, his competence, his good character, his communication style, and that he is not a politician. The new category was that Trump has been treated unfairly by his enemies. An additional 6.9% of the responses could not be placed in any of these categories.

The second judge (who also coded Study 1) independently coded the responses and reliability was good for the categories of economic reasons, caring about America, and unfair treatment (ICC > .75). It was acceptable for the other categories (ICC > .5; Koo & Li, 2016).

Of the participants, 164 answered the second question regarding how they justify their support for Trump given the reports of his sexual misconduct. In response to this question, three non-exclusive categories that were found in Study 1 were again identified (see Table 4). The most common was explicitly disbelieving the accusations, followed by stating that they support Trump based on his policies not his personal life, and pointing out that others also behave immorally. A new category also emerged: explicitly stating that they do not care about the allegations (“I really don’t care about these activities”, “I don’t care about his sex life in the past”). In addition, 17.1% provided responses that could not be placed into these categories. A second judge coded the responses, and reliability was good (ICC > .75) for all categories except stating that Trump had been treated unfairly, which had adequate reliability (ICC > .5).

In both Studies 1 and 2, participants indicated that they supported Trump because of economic issues. To unpack the meaning of economic issues, we collected examples of some direct quotations from the open-ended responses: 1) “The economic policies of Donald Trump, which were outlined in his campaign pledges, include trade protectionism, immigration reduction.” 2) “I would support Donald Trump do (sic) to his views on Trade Deals, Taxes, and Economic Issues.” 3) “Almost 4 million jobs created since election. More Americans are now employed than ever recorded before in our history, created more than 400,000 manufacturing jobs.”

This study was conducted on October 17, 2022 (start and end date), two days after Trump was arraigned for election interference. There were initially 192 participants, and 3 were excluded for rating their support for Trump as a 1. Another 2 were excluded for not providing intelligible answers to the questions, leaving 187 for analyses. Of these individuals, 173 (96 men, 91 women, age M = 45.27, SD = 13.66) responded to at least one of the open-ended questions. Their reported ethnicities were White 85.0%, Black 3.2%, Asian 2.7%, Hispanic 5.9%, and other 0.5%. These participants were recruited from Prolific with the pre-screening condition being that they indicated they had voted for Trump in 2020. Participants provided written informed consent. As in the first two studies, the open-ended responses were coded by two judges (the first author and a new second coder was added to this study). The second open-ended question was slightly rephrased: “If you were favorable toward Donald Trump as president, how do you reconcile your support for his presidency with the reports that he engaged in illegal activities related to the 2020 election and January 6?”.

As in Studies 1 and 2, prior to responding to the open-ended questions, participants answered demographic questions (gender, age, ethnicity, first language, religious affiliation). They also reported how religious they considered themselves (1 = not at all to 7 = very), their employment status (1 = unemployed to 6 = employed greater than full time), and their yearly household income (1 = < $20,000 to 7 = > $200,000). Then, they answered the following questions: 1) How much do you try to follow the rules/stick to the rules/respect authority; 2) How much do you care about justice; 3) How much do you believe in one true religion; 4) How often do you tend to vote for politically liberal candidates? These questions were answered on 1 (not at all) to 7 (very much) scales. Next, they were asked to indicate their city and state of current residence; which social media platforms they used (Facebook, Twitter, Instagram, Snapchat); how often they used social media sites (1 = never to 8 = several times per day); and how often they shared memes on social media (1 = never to 8 = several times per day); and how often they read online or paper newspapers (1 = never to 7 = very often). Then, they answered the following questions: 1) How favorable were you toward Donald Trump when he was president of the United States (1 = not at all favorable to 7 = very favorable); 2) The US Congress is holding hearings to investigate the protest that occurred on January 6, 2021 at the US Capitol. To what extent do you agree with the following statement. Congress made the correct decision to investigate these events (1 = disagree completely to 7 = agree completely).

After they responded to these questions, they were instructed to read the following news article carefully and that they would be asked “questions about its content and about your thoughts and opinions on the article and the events that it reported upon.” The article was titled, “Summary of the Hearing on the Events of January 6, 2021.” It was presented on 3 separate screens and it contained a total of 761 words. It summarized the public hearings of the United States House Select Committee on the January 6 attack that occurred between primarily in June and July 2022. It was based heavily on the summary provided in Wikipedia at the time we created the study. See Supplementary Materials for exact wording.

After this article, participants were asked to indicate whether or not they had read the article. Then, they were asked the following: 1) How accurate do you believe the information presented in the newspaper article is? (1 = not at all accurate to 7 = very accurate); 2) Thinking back to the article you read, to what degree did it contradict or dispute what you had previously believed? (1 = It did not contradict what I believe at all to 7 = It contradicted what I believe an extreme amount); 3) Thinking back to the article you read, if it contradicted or disputed what you had previously believed, to what extent did this cause you to feel bothered, uncomfortable, or upset? (1 = The article did not contradict what I had previously believed to 8 = The article contradicted what I previously believed and I felt extremely bothered). This last question was included to measure dissonance discomfort (Elliot & Devine, 1994; Harmon-Jones, Willardt, & Harmon-Jones, 2025).

Next, participants were asked 4 multiple-choice questions to assess their memory for the article they read. One example question was: According to the news article, in which states did election officials say they had been pressured to "find votes" for Donald Trump and change the results in their jurisdictions? 1) Massachusetts and Connecticut; 2) Georgia and Arizona; 3) Oregon and California; 4) Florida and Virginia.

After these questions, participants were informed that we were interested in the reasons why people did or did not support Donald Trump as president. They were asked to answer only 1 of the 2 open-ended questions: 1) If you were favorable toward Donald Trump as president, what are the most important reasons for your support for his presidency; 2) If you were unfavorable toward Donald Trump as president, what are the most important reasons why you did not support his presidency?

Next, they were informed that: The media has included many reports that Donald Trump has engaged in illegal activities related to the 2020 election and January 6. We are interested in how people feel about this. Please select only one of the two questions below to answer, regarding these issues. Select the question depending on whether you were favorable or unfavorable towards Donald Trump's presidency. Please write, in detail, your thoughts and feelings about these activities. You will have one minute to write your response before the button to advance to the next page becomes visible.

The questions were: 1) If you were favorable toward Donald Trump as president, how do you reconcile your support for his presidency with the reports that he engaged in illegal activities related to the 2020 election and January 6? 2) If you were unfavorable toward Donald Trump as president, how do the reports that he engaged in illegal activities related to the 2020 election and January 6 affect your opinion regarding his presidency?

Then, participants were to “Please select only one of the two questions below to answer. Select the question depending on whether you were favorable or unfavorable towards Donald Trump's presidency and whether your feelings have changed over time. Please write, in detail, your thoughts and feelings about why your views changed. If your favorability toward Donald Trump's presidency has not changed, simply write, "My views have not changed." You will have one minute to write your response before the button to advance to the next page becomes visible.”

The questions were: 1) If you were previously favorable toward Donald Trump as president, but are now unfavorable, why did your views change? When did your views change and how have they changed? 2) If you were previously unfavorable toward Donald Trump as president, but are now favorable, why did your views change? When did your views change and how have they changed?

Finally, participants were asked to provide any comments they wished to share about the study. Then, they were provided a short explanation of the reasons for conducting the study.

The categories of reasons given for supporting Trump were similar to those in Studies 1 and 2. The most common categories were policies and issues and economic reasons, followed by that Trump cares about America. Communication, competence, good character, that he is not a politician, and that he had been treated unfairly were less common. Agreement between the raters was good (ICC > .75) for economic reasons, not a politician, cares about America, and treated unfairly. Agreement was acceptable (ICC > .5) for communication, competence, and policies and issues. Agreement was poor (ICC < .5) for unable to categorize and for good character (see Table 5).

In response to the second question, on why they continue to support Trump despite allegations against him, the categories were also similar to Studies 1 and 2, with the addition of a category for those who said they supported Trump in the past but no longer support him. Agreement between the raters was good (ICC > .75) for disbelieving the allegations and no longer supporting him. Agreement was acceptable (ICC > .5) for others doing similar misdeeds. Agreement was poor (ICC < .5) for policies matter not his personal life and not caring about the allegations (see Table 6).

Across all 3 studies, the most common justification for continuing to support Trump despite the allegations against him was that participants disbelieved the allegations. Past research on belief disconfirmation from a dissonance perspective has typically examined responses of individuals who believed the belief disconfirming information was accurate (presumably because of researchers’ interest in counterintuitive responses). However, it is also possible that the current participants believed the disconfirming information enough to experience dissonance, but then denied the information as a way to reduce dissonance. This method of reducing dissonance would be conceptually equivalent to subtracting a dissonant cognition.

When individuals state that other politicians engage in the same immoral/illegal behaviors, or when they state that they “don’t care” about the behaviors, they may be reducing dissonance by reducing the importance of the dissonant cognitions, conceptually speaking. This mode of dissonance reduction has been referred to as “trivialization” in past research (Simon et al., 1995).

When individuals stated that Trump’s policies mattered and his personal life did not, they may be reducing dissonance by compartmentalizing information about him. In conceptual terms, they may be making the information about his immoral/illegal personal life an irrelevant cognition, which is an unexplored and novel way of reducing dissonance.

Dissonance theory predicts that Trump supporters may feel dissonance discomfort upon considering Trump’s illegal and immoral behaviors. Correlational evidence in Study 3 supports this prediction by showing that the more they favored Trump, the more discomfort they felt after reading about his illegal behaviors. In addition, the more they felt discomfort, the more they indicated that they did not believe the accusations (and discomfort mediated the effect of Trump favorability on disbelief). Moreover, in the current studies, participants were asked questions that could have reduced dissonance (e.g., why they supported him) prior to the open-ended responses about why they supported Trump given his misdeeds. This procedure could have reduced the dissonance and open-ended justifications. The current results are also impressive because it is likely that this study was not the first time they encountered this dissonant information, and participants had thus had plenty of opportunities to reduce this dissonance prior to taking part in our studies. Thus, these results suggest that dissonance may persist for long periods of time after individuals have had several chances to reduce it. In contrast, most past research had assumed that once dissonance discomfort is reduced, it stays low (Elliot & Devine, 1994).

Although each of the specific dissonance reduction methods individuals may have used can be placed into one of the conceptually abstract ways of reducing dissonance, we believe that this research suggests some previously untested ways of reducing dissonance, at both concrete and abstract levels of analysis. Most of the concrete ways are relatively new to dissonance theory research, because no previous research has measured them, with the exception of research showing that individuals respond to the misdeeds of a respected ingroup member by pointing out how opponents also misbehave (Harmon-Jones et al., 2020).

With regard to the more conceptually abstract ways of reducing dissonance, making cognitions irrelevant to the dissonant relationship has not been empirically examined, to our knowledge. Dissonance theory posited that cognitions are either relevant or irrelevant to each other; that is, relevant cognitions are psychologically connected or related to each other. If they are relevant to each other, then they are consonant or dissonant with each other. And then those cognitions are weighted for their importance, to determine the magnitude of dissonance discomfort and reduction. Past research has explored dissonance reduction via adding consonant cognitions, subtracting dissonant cognitions, or altering the importance of the relevant cognitions. It is possible that statements that were coded as “policies matter, not his personal life” could be viewed as reduction in the perceived importance of dissonant cognitions. However, we suspect that some participants may have instead been making those cognitions irrelevant, which is a dissonance reduction strategy that was not formally mentioned by Festinger (1957), but that is theoretically possible. Good examples of statements seeming more like making cognitions irrelevant and not reducing their importance were shown in Tables 2 and 4: “…what he does in private is none of the public’s business, since I feel he is running the country well” and “As for immoral actions, I don't really like Trump as a person, but I didn't vote for him to be my best friend. I voted for him to by my president, and as long as he does good at that, his person issues and morality are none of my business anymore than anyone else's.”

Although the goal of the present research was to explore the wide variety of ways individuals might reduce dissonance when exposed to belief disconfirming information, we should note that the samples were not representative of the US population, and results might differ with a more representative sample. They might also differ during the midst of an election, where partisans’ passions may further fuel dissonance discomfort and dissonance reduction when their dear leader acts in ways that violate standards (e.g., morals, laws). Also, their leader’s opponent may provide additional consonant cognitions that bolster support for their preferred candidate (e.g., the opponent is old and incoherent).

In addition, researchers have noted that some MTurk participants might provide low quality data (Hauser et al., 2019). Studies 1 and 2 used MTurk participants but Study 3 used Prolific participants, who generally provide higher quality data (Douglas et al., 2023). The results from Study 3 were very similar to those obtained in Studies 1 and 2. Also, we carefully read the open-ended responses from participants and all participants (included in the analyses) provided sensible answers, suggesting that their responses were not of low quality.

When this research was initiated, we assumed that supporters of Trump would perceive his sexual assaults, abuses of power, and incitement of insurrection as immoral and/or illegal. However, this may not be the case, and these studies do not provide any direct evidence of this. Our assumption, however, is supported by other research that has found that over 74% of Americans regard the issue of sexual assault as very important (Pew Research, 2017). In addition, many of Trump’s supporters are Christians (73% of our sample), and 2 of their 10 commandments (i.e., adultery, coveting neighbor’s wife) concern sexual behaviors that seem much less immoral/harmful than sexual assault. With regard to Trump’s illegal behaviors, we trust that most Americans find acting in illegal ways to be dissonant with their beliefs. Indeed, for the question we asked about following rules, the mean was 5.92 (SD = 1.12) on a 7-point scale, suggesting that most participants believe in following the rules and laws. Future research should more directly test this assumption.

Continuing to follow and to idealize a person despite being confronted with belief-disconfirming information about that person is not limited to Trump followers. Another recent US President, Bill Clinton, was also accused of sexual misconduct that led to his impeachment (Waxman & Fabry, 2018). We suspect that his supporters may have also reduced dissonance about his misconduct. To our knowledge, no research examined these responses.

We hope these results shed light on the multiple ways that people respond to belief disconfirmation. The results from the current studies suggest that individuals respond to disconfirming information in a variety of ways when they are not constrained to a single dissonance reduction opportunity. Further research is needed to verify the robustness of these phenomena and whether different responses emerge in other circumstances.