The more you believe in conspiracy theories, the less consistent your beliefs become

There are two competing claims about Princess Diana's death: that she was killed by British intelligence, and that she faked her own death and is still alive. Both cannot be true at the same time. Yet nine out of 10 people who strongly believe in conspiracy theories appear to accept both claims simultaneously, a new study has found.

The findings were published in PNAS Nexus, Volume 5, Issue 8, by a research team led by Alessandro Miani of the University of Fribourg in Switzerland and Stephan Lewandowsky of the University of Bristol in the United Kingdom.

Holding mutually contradictory beliefs has long been identified as a hallmark of conspiracy theorists. The pattern was first reported using the Diana case in 2012, and similar results followed — but a 2023 study challenged those findings, arguing that flawed measurement methods meant fewer than 3 percent of people actually held such incoherent beliefs, leaving the academic debate unresolved.

The research team argued that both sides of the earlier debate had methodological problems. Previous approaches divided responses into simple agree-or-disagree categories and counted how many people agreed with both contradictory claims. This meant someone who slightly agreed was counted the same as someone who was fully convinced, making it impossible to assess the strength of a belief.

The new research team converted responses into probabilities. Answers ranging from "strongly disagree" to "strongly agree" were mapped onto a scale from 0 to 100 percent, and the two numbers assigned to a pair of mutually exclusive claims were then added together.

A combined total exceeding 100 percent was treated as incoherent. If a respondent said there was an 80 percent chance Diana was murdered and a 65 percent chance she faked her death, the sum would be 145 percent.

The research team used a pizza analogy: no matter how many slices you cut it into, you cannot have more than one whole pie.

The team measured incoherence by the amount by which the combined total exceeded 100 percent. Respondents who gave low scores to both claims would not exceed 100 percent and would score zero — meaning that being uncertain about both is not itself a problem.

The team also refined the terminology. Academic literature had loosely used the word "contradiction" to describe incompatible statements, but in formal logic a contradiction means that if one statement is true, the other must necessarily be false.

The claims that Diana was murdered and that she faked her death cannot both be true — but they can both be false.

Using the new method, the research team reanalyzed responses from 8,590 participants across eight existing studies. Those studies had included a separate question, independent of agreement with any specific conspiracy theory, asking whether respondents believed the world is controlled by a small group of powerful actors — a measure of general conspiratorial thinking.

The team used scores on that question as a measure of conspiratorial disposition and ranked respondents accordingly. Assessing incoherence required a separate data source from the one used to measure disposition, so the two could be meaningfully compared.

Among those with the lowest conspiratorial disposition, only 3 in 100 gave combined scores exceeding 100 percent. Among those with the highest disposition, the figure was 91 in 100.

In other words, the stronger a person's belief in conspiracy theories, the more likely they are to give incoherent answers.

Explanations for why this happens diverged.

One view holds that people believe because they want to. For conspiracy theorists, what matters is not how Diana died but that the official account is false. Whether she was murdered or faked her death, both narratives reject the official version — and so both get accepted.

To identify the precise reason, the research team conducted a new experiment with 469 participants.

Instead of real events, the team created 16 fictional scenarios set in Europe to strip out any familiarity effect that might come from prior exposure to well-known conspiracy theories.

One fictional scenario involved the daughter of a European Parliament member who disappeared while skiing and returned three days later.

Respondents were asked to choose between pairs of statements that could not both be true but also could not both be false — for example, "her father had her detained to embezzle public funds" versus "she was not detained."

The higher a respondent's conspiratorial disposition, the more likely their combined scores exceeded 100 percent. They chose incoherent answers regardless of whether the explanation involved a cover-up or an accident.

The team also noted research showing that simply pointing out an incoherence did not reduce belief. Even when shown the contradiction directly, respondents generally found ways to reconcile the two claims.

However, the team acknowledged that the fictional experiment incorporated multiple variables at once, making it difficult to isolate the precise cause of the results.

DOI: 10.1093/pnasnexus/pgag250

Alessandro Miani, Nicole Cruz, Stephan Lewandowsky, "Still very much dead and alive: Incoherence in conspiracism as a departure from Bayesian rationality," PNAS Nexus, Volume 5, Issue 8, August 2026, pgag250.