Conspiracy beliefs and IQ: a weak link, and what each study measured

0 The short answer

People who endorse conspiracy theories score slightly lower on intelligence tests on average, but the link is weak, about minus 0.16 in the largest meta-analysis, and in the studies that tested it the association with reasoning skills appeared for implausible theories and not for plausible ones. The strongest evidence concerns reflective thinking, which correlates at about minus 0.19 across 145 samples, and only a small part of the literature gives a measured ability test. A correlation of minus 0.16 explains about 2.6 percent of the differences between people (our arithmetic), so it describes a tilt in group averages and says nothing reliable about any one person.

1 What Counts as a Conspiracy Belief, and What Counts as IQ?

The question joins two loosely defined things, and every study answers it with its own pair of definitions. In this literature a conspiracy theory is an explanation of events as the work of a secret plot by powerful actors. Researchers measure belief in three main ways. A generic scale asks about the general idea that secret groups shape events. A specific scale asks about named events. A conspiracy mentality questionnaire measures a standing suspicion of the powerful, whatever the event. The three can give different answers for the same person, and, as later sections show, they give different correlations with ability.

The definition of the theory matters more than it first appears. In a December 2024 commentary in the Harvard Kennedy School Misinformation Review, Imhoff and Bertlich note that political leaders, oppositional forces and businesspeople have often coordinated in secret to the public's disadvantage, so conspiracy theories can describe the world accurately. They add that most research nonetheless treats conspiracy theories as irrational beliefs that contradict everyday knowledge. If a study samples mainly theories that are unlikely to be true, part of its finding is built into its choice of items. The authors argue that the cognitive association may be partly trivial for this reason, since people with less sophisticated reasoning accept wrong statements more readily in general.

The IQ side needs the same care. An IQ score in the strict sense is a result on a standardized battery scaled to a mean of 100 and a standard deviation of 15, as the page on how IQ scores are normed and the page on the standard deviation of 15 explain. Very few conspiracy studies administer such a battery. They use short proxies, such as a 42-item vocabulary test, a matrix-type reasoning test, a handful of numeracy items, the cognitive reflection test, an open-ended critical thinking test, or a question asking people to place themselves on a distribution of intelligence. The page on what IQ measures and the page on the g factor describe what a full battery captures, and the page on the CHC model names the broad abilities involved.

The meta-analysts who coded these measures show how mixed they are. Bowes, Costello and Tasimi sorted the intelligence measures in their data into general intelligence (total scores on performance-based measures, from 5 studies), matrix reasoning (3 studies), numeracy (7 studies), verbal reasoning (3 studies), base-rate neglect (1 study) and self-reported intelligence (4 studies). That is the evidence base for the sentence "conspiracy believers have lower IQs": a mixture of ability proxies, mostly short, mostly collected online.

Three neighboring questions belong elsewhere. The page on IQ and political orientation covers a different trait, and the page on IQ and religion covers another belief domain. This page does not rank any particular belief as true or false, does not cover how to correct misinformation, and offers no diagnostic reading of any person.

2 Intelligence, Reflective Thinking and Education: Three Variables, Three Different Sizes

Headlines often merge at least three variables that the research keeps apart: measured ability, reflective thinking style and education, and each one has its own correlation with conspiracy belief. Measured ability is how well a person solves reasoning, vocabulary or numeracy problems under standard conditions. Reflective thinking is a disposition to check an intuitive first answer, usually measured by the cognitive reflection test or by self-report questionnaires. Education is years or level of schooling, which reflects ability but also opportunity, field, income and social position, as the page on average IQ by education level and the page on intelligence versus knowledge explain.

The boundary between the first two is blurry. The authors of the reflective thinking meta-analysis discussed below note earlier work suggesting that the cognitive reflection test behaves more like a cognitive ability test than like a measure of thinking disposition. The page on the cognitive reflection test describes the format, and the page on critical thinking versus intelligence covers why judgment is not the same thing as IQ.

The table lines up the pooled correlations from the three meta-analyses that report them. Every value is a correlation with conspiracy belief or conspiratorial ideation as the authors define it, so a negative sign means higher scores go with weaker belief. The last column squares each correlation, which is our arithmetic, to show the share of differences accounted for.

Two points stand out. First, every ability or reflection measure sits between about 0.13 and 0.19 in absolute size, which is small. Second, the larger values belong to thinking dispositions and to a gullibility measure, not to ability. The table also shows why a headline that says "intelligence" can rest on any row. Education has no row because the studies reviewed here report it through mediation analyses, covered in the section on direction and cause, and not as a pooled correlation.

3 What Do the Two Large Meta-Analyses Say About Measured Intelligence?

Two meta-analyses that include measured ability agree on a small negative correlation, minus 0.13 and minus 0.16, and both warn that the heterogeneity behind the average is large. Neither is a dedicated IQ meta-analysis. We found no meta-analysis devoted only to IQ and conspiracy belief in the sources we could verify, so the figures come from reviews of many correlates in which ability is one row.

Bowes, Costello and Tasimi published the larger review in Psychological Bulletin in 2023. It spanned 170 studies, 257 samples, 52 variables, 1,429 effect sizes and 158,473 participants. The strongest correlates, the authors report, concerned perceiving danger and threat, relying on intuition and having odd beliefs and experiences, and being antagonistic and acting superior. For intelligence the figures, read in the authors' accepted manuscript on the Open Science Framework on October 13, 2026, are 22 studies, 26 samples, 39 effect sizes and 12,276 participants, with a pooled correlation of minus 0.16 and a 95 percent interval of minus 0.21 to minus 0.11. The authors' standard deviation of true effects across observations is 0.10. That means that, on their model, the true correlation in a given setting plausibly runs from near zero to about minus 0.26 (our reading of their figure, not their sentence). By their level 3 index, 83.2 percent of the variation among results is between samples, not within them.

The moderator analysis is informative. Correlations were larger for verbal reasoning (slope of minus 0.21) and for general intelligence measured as total scores on performance-based tests (minus 0.22) than for self-reported intelligence (minus 0.06). In plain terms, when people are tested the link is about 0.2, and when they are asked how smart they are it nearly disappears. That is consistent with the well-documented gap between self-rated and measured ability, which the page on estimating your own IQ discusses. The authors also probed whether restricted range, for example in college samples, hid larger effects. They found limited evidence that relations were stronger in community than in college samples and essentially no differences by education level, and they say these checks need replication.

Stasielowicz published a Bayesian three-level meta-analysis in the Journal of Research in Personality in 2022: 686 correlations from 127 independent samples, covering 12 personality correlates of conspiracy beliefs. It is a personality review, and cognitive ability is one of the twelve. Bowes and colleagues tabulate its cognitive ability result as 15 samples, a correlation of minus 0.13 and a credible interval of minus 0.18 to minus 0.07. In the author's posted preprint, read on October 13, 2026, that estimate rests on 28 correlations and 6,254 participants, and it is smaller than the correlations for pseudoscientific beliefs (0.46), paranoia (0.34), schizotypy (0.30) and narcissism (0.28), and similar in size to religiosity (0.14, opposite sign). The author concludes that intelligent people are less likely to believe in conspiracy theories but are certainly not immune.

The preprint also splits the result by how conspiracy belief was measured. The negative correlation with cognitive ability was largest with the Belief in Conspiracy Theories Inventory, which asks about specific events (minus 0.20), and smaller with the Conspiracy Mentality Questionnaire (minus 0.09) and the Generic Conspiracist Beliefs Scale (minus 0.08), with other instruments at minus 0.14. The author cautions that only 15 studies were available and that the pseudo-R-squared suggested the instrument explained heterogeneity within studies but not between them. The pattern fits a reading the author proposes, that people with higher ability may reject specific theories more than they reject the general idea that conspiracies exist, and it foreshadows the plausibility results below.

The two reviews draw on the same published literature, so they are not two independent confirmations of one number. They are two analyses of one body of evidence, and they agree that the correlation is small.

4 What Does the Reflective Thinking Meta-Analysis Add, and What Does It Not Contain?

The largest synthesis in this field measures reflective thinking, not IQ, and its pooled correlation of minus 0.189 is the best-supported number on the page. Yelbuz, Madan and Alper pooled published and unpublished correlational data from 145 samples and 181 effect sizes, drawn from 64 articles of which 48 were published and 16 were not. They used a random-effects model and reported a pooled correlation of minus 0.189 with a 95 percent interval of minus 0.206 to minus 0.172. In their data 94 percent of the individual estimates were negative. The authors describe this as a medium-sized effect, a label that depends on the convention used. It is slightly larger than the ability correlations in the table above.

The study splits the result by how reflective thinking and conspiracy belief were measured. The authors present these as subgroup analyses, not formal moderator tests, and the subgroup counts they print add up to 178 and not 181.

Two things follow. The correlation holds whether reflective thinking is a questionnaire answer or a task score, which argues against a shared-method artifact. And the strongest cell, performance tests against specific theories at minus 0.219, is the one with the most samples and the narrowest interval. The authors' performance-based measures consist mainly of the cognitive reflection test, which is why the line between thinking style and ability matters. The authors themselves note, citing earlier work, that this test may be closer to a cognitive ability test than to a thinking disposition measure.

The meta-analysis has clear limits for the question on this page. The authors report that no moderators could be assessed beyond publication status, because most studies reported only the basic correlation. Intelligence was not tested as a moderator, and no intelligence measure entered the analysis. Education was not tested either. The authors flag country-level corruption as a candidate moderator for future work and coded a corruption index but did not analyze it.

On publication bias the authors report a mixed picture and then a conclusion. Egger's regression test (p = 0.038) flagged funnel plot asymmetry, and the rank correlation test did not (p = 0.155). The authors note that the asymmetry runs opposite to what publication bias would predict, because small, low-powered studies showed smaller or even positive correlations, and a p-curve analysis was right-skewed, which they describe as strong support for a true effect. Their conclusion is that there is no evidence of publication bias. A skeptical reader can fairly say that no test settles this, and that the pattern is at least not the one selective publication usually produces.

5 Studies With a Measured Ability Test, One by One

The studies that measured ability with a real test find correlations that are small and often fragile, and the closer the measure is to pure ability, the smaller the link tends to be. The table lists the individual studies this page could verify, with the design and the measure in each. Results are quoted as the authors report them in their abstracts or papers.

Stieger and colleagues used a 42-item German vocabulary test, so their ability measure is crystallized, in the sense the page on fluid versus crystallized intelligence explains, and the page on the verbal IQ test describes what such a score represents. The correlation with belief in the kidnapping theories was minus 0.12 in the first sample and minus 0.07 in the second. A correlation of that size is hard to tell apart from zero in a sample of fewer than 300 people, which is why one sample reached significance and the other did not.

Jastrzębski and Chuderski separate fluid reasoning (Gf, solving novel problems) from analytic thinking style. Across three studies and 827 participants they report that fluid reasoning predicted about 11 percent of the variance in the rejection of pseudoscience, 4 percent for paranormal beliefs and less than 2.5 percent for conspiracist beliefs, while analytic thinking explained about 37 percent for pseudoscience and around 20 percent for paranormal and conspiracist beliefs. They report that the relationships were stronger for males than for females and conclude that analytic thinking is more important than cognitive ability for adopting an epistemically supported worldview.

Rogoza and colleagues tested a Polish sample of 1,143 and a British sample of 388 and report consistent negative relationships between intelligence and both national narcissism, which they define as a belief in the nation's greatness that requires outside recognition, and conspiracy beliefs. Lower national narcissism partly accounted for the association between higher intelligence and weaker conspiracy belief. The authors call this initial evidence. We read their paper only at the level of its published abstract, so the specific intelligence test and the correlation values are not reported here.

Ståhl and van Prooijen add an interaction. In their first study, analytic thinking was associated with lower belief in various conspiracy theories and in paranormal phenomena, but only among individuals who strongly value epistemic rationality, meaning the motivation to form beliefs on rational grounds. They replicated the effect for paranormal belief in a second study but not for conspiracy beliefs, and they report evidence that general cognitive ability, more than analytic style, is the facet responsible. Lantian and colleagues found, in two studies of 86 and 252 participants, that critical thinking ability measured by an open-ended test was negatively associated with conspiracy belief, while self-reported critical thinking was not. Swami and colleagues found in a British sample of 990 that stronger belief went with lower analytic thinking and open-mindedness and with more intuitive thinking.

A pattern runs through the table, and it is our synthesis, not a tested claim in any one paper. Vocabulary and fluid reasoning tests give the smallest links, performance-based critical thinking gives somewhat larger ones, and thinking dispositions and gullibility measures give the largest. This matches the meta-analytic ordering in the earlier table, where intelligence sits at minus 0.16 and active open-minded thinking at minus 0.25.

6 How Large Is a Correlation of Minus 0.16?

A correlation of minus 0.16 means that knowing a person's IQ shifts the expected belief score by about a sixth of a standard deviation per standard deviation of IQ, which is a tilt in averages and not a prediction. Every conversion in this section is our arithmetic on the published correlations, assumes a straight-line relationship and approximately normal scores, and is labeled as such. None of it comes from the authors.

Start with variance. Squaring a correlation gives the share of differences between people that one variable accounts for in the other. For minus 0.16 that is 0.0256, or about 2.6 percent. For Stasielowicz's minus 0.13 it is about 1.7 percent, and for Yelbuz and colleagues' minus 0.189 it is about 3.6 percent. The remaining 96 to 98 percent of differences in belief is unrelated to the ability measure in a straight-line sense. Jastrzębski and Chuderski reached a similar number from the other direction, with fluid reasoning predicting less than 2.5 percent of the variance in conspiracist beliefs.

Next, translate into IQ points. A correlation equals the slope when both variables are in standard deviation units. So a person 15 points above the mean, at 115 and near the 84th percentile (the page on IQ percentiles converts any score), is expected to sit about 0.16 standard deviations below the average belief score. A person at 130, near the 98th percentile, is expected to sit about 0.32 below. Those are expected values for groups of people with that IQ, around which individual scores spread widely. The page on IQ scores versus percentiles explains why equal point gaps do not mean equal percentile gaps.

A third way to feel the size is the concordance probability. For two people drawn at random from a population with a bivariate normal relationship of minus 0.16, the one with the higher IQ is also the one with the weaker belief about 55 percent of the time, against 50 percent for no relationship. At minus 0.13 the figure is about 54 percent, and at minus 0.189 it is about 56 percent. The formula is one half plus the arcsine of the correlation divided by pi. A coin flip would be right 50 percent of the time, and a rule based on IQ would be right about 55 percent of the time.

The spread in true effects adds a caution. In the Bowes meta-analysis the standard deviation of true effects is 0.10, so some settings plausibly show a correlation near zero and others near minus 0.26. A single pooled value hides that range.

Finally, compare with the other correlates in the same meta-analysis. In the authors' narrative, conspiratorial ideation was related to perceiving outgroup members as threatening at 0.56, delusion-proneness at 0.42, a belief that the world is dangerous at 0.39, anomie at 0.34 and collective narcissism at 0.34, against 0.16 for intelligence. Squaring 0.56 gives about 31 percent of the differences, twelve times the share for ability. These are correlations at the group level, and they do not make any of those traits a cause or a label for a person, but they show that ability is a minor correlate among many. The page that ranks what IQ correlates with gives the scale for other outcomes, and the page on IQ and success shows correlations of similar size elsewhere.

One more factor is measurement error. Short tests and self-reports are noisy, and noise lowers an observed correlation below the correlation between the underlying abilities. The sources opened for this page do not give a corrected value, so the size of that gap is unknown. The page on reliability and validity explains why a correlation between two imperfect measures understates the link between what they aim to measure.

7 Does the Plausibility of the Theory Change the Result?

In the studies that tested it, reasoning and reflection were linked to belief in implausible conspiracy theories and not to belief in plausible ones, which means the choice of theories can create or erase the association. This is the controversy that gives the literature its methodological interest, and it is the reason the page title carries a catch.

Hattersley and colleagues framed it as an overfitting hypothesis. Theories that fit conspiracy-related data at the expense of wider evidence should create dissonance, and reasoning biases may partly reflect the need to reduce it. The hypothesis predicts that reasoning biases should be more closely tied to implausible theories, such as the claim that the moon landings were faked, than to plausible ones. In two preregistered studies, endorsement of implausible theories, but not plausible ones, was associated with reduced information sampling in an information-foraging task and with less reflective reasoning. The authors conclude that the link between conspiracy belief and reasoning is not homogeneous and is not tied to the conspiracy aspect itself.

Frenken, Reusch and Imhoff tested plausibility in two preregistered studies with 563 participants in total. In one, plausibility was set by pretest ratings of fictitious theories. In the other it was manipulated experimentally by changing the supporting information in a belief updating task. A general suspicious mindset, conspiracy mentality, went with greater perceived plausibility for plausible theories and markedly more so for implausible ones. Cognitive variables, by contrast, were negatively associated with plausibility judgments only for theories deemed implausible. The authors read this as limited cognitive skills and rational thinking dispositions going with biased plausibility judgments and with underweighting evidence that disconfirms a theory.

The Imhoff and Bertlich commentary summarizes their own German studies in the same direction. Lower cognitive reflection, greater receptivity to pseudo-profound statements and a more experiential thinking style went with implausible theories, and none of the three went with plausible ones. A second study reached a similar pattern when evidence made a theory plausible or implausible. The commentary also reports that conspiracy mentality had somewhat weaker links to the cognitive variables than conspiracy beliefs did. The authors list their own limits: a pilot with fictitious theories, German samples only, and conspiracy mentality still showing some link to cognitive variables. The commentary names no intelligence test, and it reports no coefficients.

Inside the meta-analyses there is a related signal. Bowes and colleagues found that cognitive reflection was a stronger correlate of belief in specific conspiracy theories (minus 0.19) than of general conspiratorial ideas (minus 0.14), and the instrument split in Stasielowicz's preprint points the same way for cognitive ability. Neither review tested plausibility directly, so both point the same way as the plausibility studies without confirming them.

Three limits keep this section honest. First, "plausible" in these studies means rated plausible by pretest participants or made plausible by the experimenter's evidence, and it does not mean true. Second, none of the studies above used a full IQ battery. They tested reflection, reasoning and receptivity to pseudo-profound statements, so the claim that the IQ correlation depends on plausibility is an extrapolation from adjacent measures. Third, a correlation that is absent in one cell is weaker evidence than a present one, because a small true correlation can fail to reach significance in a few hundred participants, as the Stieger results show.

Still, the lesson for reading a headline is concrete. A study that asks only about theories most readers consider far-fetched will find a clearer link with reasoning than one that also asks about theories with documented real-world cases. Neither design measures a person's capacity to judge a particular theory, and neither can tell you which of a person's beliefs is true.

8 Direction and Cause: What a Correlation Cannot Settle

Every study above is observational, so none shows that lower ability leads to conspiracy belief, that belief lowers test performance, or that a third factor drives both. Adults cannot be randomly assigned an IQ, so the question of direction has to be approached through adjacent designs, and those have so far been inconclusive.

The experimental attempts target thinking style, not ability. Swami and colleagues reported three experiments, with 112, 189 and 140 participants, in which tasks meant to elicit analytic thinking, a verbal fluency task and a difficult-to-read font, reduced belief in conspiracy theories. Većkalov and colleagues then ran two high-powered direct replications of two of those studies as a Registered Report. With 302 participants for the scrambled sentence task and 488 for the difficult fonts, they found no evidence that either manipulation elicited more analytic thinking or reduced belief. They conclude that effective ways of inducing analytic thinking remain to be identified before its causal role can be gauged. The correlation between analytic thinking and belief is well established. A causal effect of briefly priming it is not.

Education offers a different clue. van Prooijen notes that people with more education are less likely to believe conspiracy theories, then asks why. In the first study, three independent mediators accounted for the link: belief in simple solutions for complex problems, feelings of powerlessness and subjective social class. A nationally representative sample replicated the result except for subjective social class, and variation in analytic thinking statistically accounted for the path through belief in simple solutions. The author concludes that the education link cannot be reduced to a single mechanism. For this page the point is narrow. Education carries ability, but it also carries a sense of control and social position, so an education gap is not an IQ gap.

Mediation findings need the same caution. Rogoza and colleagues report that lower national narcissism partly accounted for the link between intelligence and conspiracy belief, and they call it initial evidence. A statistical mediation in cross-sectional data fits one causal story, but it fits others as well.

Three broad stories remain open, and the sources opened for this page test none of them. Ability may help people evaluate evidence and notice weak arguments, which would make lower ability a contributing cause. Third factors such as schooling, trust, social position and a sense of control may shape both test performance and belief. And belief systems may change what information a person meets, which could feed back on knowledge, though no source here tests that. The accurate summary is that the correlation is replicated and its explanation is not.

A final design problem is shared method. Bowes and colleagues note in their discussion that conspiratorial ideation is nearly always measured by self-report, and so are most of the correlates. Where both sides of a correlation are questionnaires, a shared response style can create an association. Performance-based ability tests avoid that problem, which is one reason the minus 0.16 for measured intelligence carries more weight than correlations among questionnaires.

9 Why Smart People Still Believe: Motivation, Threat and Trust

Ability explains little of who endorses conspiracy theories, because the strongest correlates are threat, trust, intuition and motivation, and a high score does not remove any of them. The 2023 meta-analysis puts the correlates in three groups: perceiving danger and threat, relying on intuition and having odd beliefs and experiences, and being antagonistic and acting superior. Among the motivational constructs, the authors' discussion singles out trust as the largest negative correlate and social threat perception as the largest positive one.

The motivation result in Ståhl and van Prooijen shows the mechanism in miniature. Analytic thinking went with less belief only among people who strongly valued forming beliefs on rational grounds. If ability is the tool, motivation decides whether it is picked up. Jastrzębski and Chuderski reached a compatible conclusion from a different design: analytic thinking style outran fluid reasoning, and fluid reasoning explained less than 2.5 percent of the variance in conspiracist beliefs. The author of the personality review reaches the same conclusion: higher ability lowers the odds a little and confers no immunity.

Personality adds little. In the Bowes data the Big Five traits were weak correlates, while general humility and intellectual humility were small to medium negative correlates, particularly when the measures were comprehensive. The page on IQ and personality places such trait correlations in context. The authors also list schizotypy, paranoia, unusual experiences, psychoticism and hostility among the strongest correlates in the abnormal range. These are population-level associations, and they do not make any belief a symptom or any believer a patient. The page on IQ and mental health covers the separate evidence on cognition and clinical conditions.

One finding bears on confidence more than on ability. Pennycook, Binnendyk and Rand report eight studies with 4,181 U.S. adults in which conspiracy believers consistently overestimated their performance on numeracy and perception tests, even after actual performance was taken into account. The pattern held when analytic thinking, need for uniqueness and narcissism were controlled, and it was strongest for the most fringe theories. Claims were believed by a majority of participants only 12 percent of the time, yet believers thought themselves in the majority 93 percent of the time. Confidence in one's own reasoning and tested reasoning are different variables, which fits the weak self-report result in the meta-analysis.

The social correlates matter for tone as much as for science. Perceiving outgroups as threatening (0.56), collective narcissism (0.34) and anomie (0.34) sit far above ability in the same review. Threat, low trust and powerlessness go with endorsement far more strongly than ability does, so a reading that reduces belief to low ability misses most of what the data show. Ability is the correlate that is easiest to measure and to talk about. It is not the largest.

10 Online Samples, Short Tests and Self-Report: How the Measures Limit the Answer

The conspiracy literature measures ability with short, mixed and mostly online instruments, so the pooled correlation describes those instruments and not full-battery IQ. Bowes and colleagues observe that most existing research has focused on college and online participants, and they call for work on who conspiracy theorists are in the real world. Three features of the measures matter for the number.

First, the content differs. In CHC terms, the vocabulary and verbal reasoning tests tap crystallized ability (Gc), matrix tasks tap fluid reasoning (Gf), and numeracy items tap quantitative knowledge (Gq), the mapping being ours and not the authors'. A pooled "intelligence" correlation averages across these. The page on the CHC model lists the broad abilities, and the page on quantitative reasoning shows why numeracy is its own ability. Stasielowicz's preprint says only a few specific abilities have been examined, so comparisons between abilities cannot yet be made meaningfully.

Second, the mode differs. Self-reported intelligence correlated at minus 0.06 in the Bowes moderator analysis, against minus 0.21 to minus 0.22 for tested verbal reasoning and general intelligence, so a study that asks people to rate themselves will find almost nothing. Unsupervised online testing adds its own noise. The pages on data quality in online cognitive testing and on intelligence measures for online studies cover the problem, and the page on whether online IQ tests are accurate addresses the consumer version.

Third, the samples differ. The meta-analysts note that restricted range, in college or highly educated samples, could attenuate relations. Their own checks found limited evidence of larger relations in community samples and essentially none across education levels, which is reassuring but not decisive, since these checks used coarse categories. The reflective thinking meta-analysis also includes unpublished work, which reduces its reliance on what gets published. The page on reporting IQ in research lists the details a score hides.

11 What This Does Not Say About You, a Relative or a Friend

A group correlation of minus 0.16 says nothing reliable about the ability of anyone who holds a particular belief, and nothing about the beliefs of anyone with a particular score. The arithmetic earlier shows why. About 97 percent of the differences in belief are unrelated to ability, and a rule based on IQ would pick the right person about 55 times in 100 against 50 by chance.

The finding also does not say that skepticism of powerful institutions is a cognitive error. The commentary by Imhoff and Bertlich is explicit that coordinated secret action by the powerful has occurred and that theories can describe it. The cognitive association appears most clearly for theories that are implausible, and the studies that tested plausibility did not find it for plausible ones.

It does not say that anyone who holds a belief is unintelligent, or that anyone who rejects one is intelligent. The studies with measured ability report small correlations, and the studies of motivation show that people with high scores endorse theories as well. It does not say that a particular belief is true or false, and this page takes no position on any specific theory. It does not give a way to change anyone's mind, and the one experimental route tested here did not replicate.

For a reader who is worried about someone close, one practical reading follows from the evidence. A relative's belief is not evidence about their ability, and their test score is not evidence about their beliefs. The strongest correlates in the data are feelings of threat, low trust and powerlessness, not ability.

An IQ test measures how a person performs on reasoning, vocabulary, numeracy, memory and speed tasks under standard conditions. It does not score beliefs, and no part of an ACIS report is about conspiracy belief. ACIS is an online, unsupervised test, it is not a clinical or diagnostic instrument, and its report describes abilities, not opinions.

12 What the Evidence Supports, Stated Narrowly

The evidence supports a weak negative correlation between measured ability and conspiracy belief, a slightly larger one for reflective thinking, a dependence on which theories are asked about, and no conclusion about cause or about any individual. Stated claim by claim:

• Across meta-analyses that include measured ability, the pooled correlation with conspiracy belief is minus 0.13 (Stasielowicz, 15 samples) and minus 0.16 (Bowes and colleagues, 26 samples), with large heterogeneity.
• Across 145 samples, reflective thinking correlates at minus 0.189 with conspiracy belief, and the correlation held for both questionnaires and performance tasks.
• Self-reported intelligence shows almost no link, while tested ability shows slopes near minus 0.2, in the one review that compared them.
• In several studies, reasoning and reflection related to belief in implausible theories and not plausible ones. None of those studies used a full IQ battery.
• A priming experiment that supported a causal role for analytic thinking did not replicate in a high-powered Registered Report.
• Threat, trust, intuition and motivation correlate with conspiracy belief much more strongly than ability does.

What the evidence does not support is any use of a belief to estimate someone's ability, any use of a score to estimate someone's beliefs, any claim that ability causes or protects against belief, and any statement that believers as a group are less intelligent in a sense that applies to a member of the group. The designs are correlational, the ability measures are short, and the effects are small.

The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) define validity as the degree to which evidence and theory support the interpretations of test scores for proposed uses. By that definition, using one person's IQ score to infer their beliefs, or using their beliefs to infer their IQ, is an interpretation that nothing reviewed here supports. The American Psychological Association's Ethical Principles of Psychologists and Code of Conduct, in the version effective January 1, 2017, requires in Standard 9.02(a) that assessment instruments be used in a manner and for purposes appropriate in light of the research on their usefulness. Standard 9.06 requires psychologists interpreting results to take into account the purpose of the assessment and the test factors, test-taking abilities and situational, personal, linguistic and cultural characteristics of the person assessed that might reduce the accuracy of their interpretations.

For a reader who wants to see their own ability profile, the ACIS online test reports a Full Scale IQ and six index scores, scaled to a mean of 100 and a standard deviation of 15, with a 95 percent confidence interval. Prices were read on October 13, 2026 and can change: Quick is 15 dollars, Optimized is 30 dollars and Full Scale is 50 dollars, each a one time payment, and the free trial needs no card. ACIS is online and unsupervised, not clinical or diagnostic, not for hiring, school accommodations or admission to high IQ societies, and available in English only. The technical manual documents the instrument, and the page on how IQ is normed explains what the scale means.

13 Sources Behind This Page

Every figure above is traceable to one of the following, and each is linked at the point where it is used. Correlations, sample counts and intervals are quoted as the authors report them, and any squaring, conversion into standard deviations or concordance probability is labeled as our arithmetic. Several papers sit behind publisher paywalls. For Rogoza and colleagues, Jastrzębski and Chuderski, Ståhl and van Prooijen, Swami and colleagues, Lantian and colleagues, Pennycook and colleagues, Hattersley and colleagues, Frenken and colleagues, van Prooijen and Većkalov and colleagues, figures come from the published abstracts. For Bowes and colleagues, the intelligence figures come from the authors' accepted manuscript on the Open Science Framework, and for the instrument split in Stasielowicz, from the author's posted preprint, both read on October 13, 2026. The wording of the published versions may differ.

• Bowes S, Costello T and Tasimi A. The conspiratorial mind: A meta-analytic review of motivational and personological correlates. Psychological Bulletin, 2023, volume 149, issue 5-6, pages 259 to 293; accepted manuscript at osf.io/cusq8.
• Stasielowicz L. Who believes in conspiracy theories? A meta-analysis on personality correlates. Journal of Research in Personality, 2022, volume 98, article 104229; preprint at osf.io/pfw3x.
• Yelbuz B, Madan E and Alper S. Reflective thinking predicts lower conspiracy beliefs: A meta-analysis. Judgment and Decision Making, 2022, volume 17, issue 4, pages 720 to 744.
• Imhoff R and Bertlich T. Are conspiracy beliefs a sign of flawed cognition? Reexamining the association of cognitive style and skills with conspiracy beliefs. Harvard Kennedy School Misinformation Review, December 12, 2024.
• Stieger S, Gumhalter N, Tran U, Voracek M and Swami V. Girl in the cellar: a repeated cross-sectional investigation of belief in conspiracy theories about the kidnapping of Natascha Kampusch. Frontiers in Psychology, 2013, volume 4.
• Jastrzębski J and Chuderski A. Analytic thinking outruns fluid reasoning in explaining rejection of pseudoscience, paranormal, and conspiracist beliefs. Intelligence, 2022, volume 95, article 101705.
• Rogoza M, Marchlewska M, Molenda Z, Szczepańska D and Rogoza R. Intelligence is negatively related to national narcissism and conspiracy beliefs. International Journal of Psychology, 2025, volume 60, issue 5.
• Ståhl T and van Prooijen J. Epistemic rationality: Skepticism toward unfounded beliefs requires sufficient cognitive ability and motivation to be rational. Personality and Individual Differences, 2018, volume 122, pages 155 to 163.
• Lantian A, Bagneux V, Delouvée S and Gauvrit N. Maybe a free thinker but not a critical one: High conspiracy belief is associated with low critical thinking ability. Applied Cognitive Psychology, 2021, volume 35, issue 3, pages 674 to 684.
• Swami V, Voracek M, Stieger S, Tran U and Furnham A. Analytic thinking reduces belief in conspiracy theories. Cognition, 2014, volume 133, issue 3, pages 572 to 585; and Većkalov B, Gligorić V and Petrović M. No evidence that priming analytic thinking reduces belief in conspiracy theories: A Registered Report of high-powered direct replications of Study 2 and Study 4 from Swami and colleagues (2014). Journal of Experimental Social Psychology, 2024, volume 110, article 104549.
• van Prooijen J. Why education predicts decreased belief in conspiracy theories. Applied Cognitive Psychology, 2017, volume 31, issue 1, pages 50 to 58.
• Hattersley M, Brown G, Michael J and Ludvig E. Of tinfoil hats and thinking caps: Reasoning is more strongly related to implausible than plausible conspiracy beliefs. Cognition, 2022, volume 218, article 104956.
• Frenken M, Reusch A and Imhoff R. "Just because it's a conspiracy theory doesn't mean they're not out to get you": Differentiating the correlates of judgments of plausible versus implausible conspiracy theories. Social Psychological and Personality Science, 2025, volume 16, issue 6, pages 684 to 696.
• Pennycook G, Binnendyk J and Rand D. Overconfidently conspiratorial: Conspiracy believers are dispositionally overconfident and massively overestimate how much others agree with them. Personality and Social Psychology Bulletin, 2026, volume 52, issue 8, pages 2509 to 2528 (first published online May 24, 2025).
• American Educational Research Association, American Psychological Association and National Council on Measurement in Education. Standards for Educational and Psychological Testing, 2014; and American Psychological Association, Ethical Principles of Psychologists and Code of Conduct, Standards 9.02 and 9.06, version effective January 1, 2017.

14 Frequently Asked Questions