Psychopathy, narcissism and IQ: what the meta-analyses actually found

0 The short answer

Across the meta-analyses, psychopathy, narcissism and the other Dark Triad traits correlate with measured intelligence at close to zero, so the claim that dark traits come with a higher IQ is not supported, and neither is the opposite claim that they come with a much lower one. O'Boyle and colleagues pooled 48 independent samples in 2013 and found no consistent relation with any trait, so the evil genius hypothesis was not supported, and Michels pooled 143 studies in 2022 and found none of the traits meaningfully related to intelligence, apart from a small negative correlation with the impulsive and antisocial side of psychopathy. The 2019 meta-analysis by Sánchez de Ribera and colleagues put total psychopathy at r = -.07. The reliable finding about narcissism concerns self-image: high grandiose narcissism goes with believing oneself intelligent, not with scoring higher on a test.

1 What Do Psychopathy, Narcissism and the Dark Triad Mean, and How Are They Measured?

The words psychopathy and narcissism name different things depending on whether a study uses a clinician-style checklist in a forensic setting or a self-report scale in a community sample, and that choice shapes what the intelligence results can mean. The term Dark Triad comes from Paulhus and Williams, who described Machiavellianism, subclinical narcissism and subclinical psychopathy as the three most prominent offensive yet non-pathological personalities in the literature. In a sample of 245 students they found the three measures moderately inter-correlated but not equivalent, with disagreeableness as the only Big Five correlate common to all three, and they concluded that the traits are overlapping but distinct constructs. Subclinical here means traits measured in ordinary samples, not a diagnosis.

Psychopathy has an older and more clinical history. According to the introduction of the Sánchez de Ribera meta-analysis, the construct dates to Cleckley's 1941 clinical descriptions, it is commonly described as a cluster of relatively stable personality traits, and one of its most popular measures, the Hare Psychopathy Checklist, was designed for use with prisoners. The studies below use the Psychopathy Checklist-Revised (PCL-R), its youth version (PCL-YV) and its screening version (PCL:SV), alongside self-report scales used in community samples. A checklist applied in a prison and a questionnaire completed by a student are not interchangeable measuring devices, and the same meta-analysis notes that one self-report scale failed to show measurement invariance between incarcerated and student samples. This page keeps the two apart wherever a study allows it.

Narcissism has its own split. Zajenkowski and colleagues distinguish grandiose narcissism from vulnerable narcissism, and they measured the two forms separately. The two behave differently in the intelligence studies, as later sections show. Machiavellianism, the third trait, appears here mainly inside the pooled results.

Intelligence is measured in several ways across these studies. Some report a Full Scale IQ (FSIQ), others split verbal IQ (VIQ) from performance IQ (PIQ), and others use a single reasoning test such as the Advanced Progressive Matrices. In the vocabulary of the CHC model, verbal measures lean on crystallized knowledge (Gc), while matrix and block tasks lean on fluid reasoning (Gf) and visual processing (Gv). That mapping is our reading, since the authors work in the older verbal and performance vocabulary. The page on what IQ measures and the page on the g factor explain why a total score and its parts can behave differently, which turns out to matter here.

The scope is narrow on purpose. The meta-analyses also examine antisocial personality disorder and conduct disorder, which are diagnoses, and offending, which is behavior. This page concerns traits. For the disorders and the behavior, see the pages on IQ and mental health and IQ and crime, and nothing here is a statement about any diagnosis.

2 The Studies at a Glance

Read as a set, the studies point the same way for psychopathy and narcissism: pooled correlations with measured intelligence are small or near zero, and the clearest narcissism finding is a gap between self-rated and measured intelligence. The table lists each study with its design, sample, measure and what its abstract or text printed. Figures appear only where the source printed them, and where a source printed no coefficient in the part we could open, the cell says so.

Three cautions apply to every row. First, the checklist studies sample people who are detained, sentenced or hospitalized, so they describe a narrow slice of the population and cannot be generalized to everyone who scores high on a psychopathy questionnaire. Second, the intelligence tests differ, so correlations from a brief screening test and from a full battery do not stack. Third, a population correlation is a summary of many people, and the page on reliability and validity explains why such a summary does not predict one person.

3 What Did O'Boyle and Colleagues Find About the Evil Genius Hypothesis?

The 2013 meta-analysis tested two opposite claims at once, that dark traits come with high intelligence and that they come with low intelligence, and found support for neither. O'Boyle, Forsyth, Banks and Story set out to determine whether people who display socially exploitative qualities tend to be more intelligent or less intelligent. The abstract defines the first claim as the evil genius hypothesis, that highly intelligent individuals tend to display socially exploitative personality traits. It defines the second as the compensatory hypothesis, that less intelligent individuals compensate for their cognitive disadvantages by adopting manipulative behavioral tendencies.

Across 48 independent samples, the authors report that general mental ability showed no consistent relation with any Dark Triad trait, and that the effects were not sufficient to support either hypothesis. They add that the relations were moderated, to some extent, by the sex and age of the participants, the type of sample studied and the measure of general mental ability. The abstract prints no coefficients, and we could not open the full text, so this page does not repeat the effect sizes that circulate for it online. A later paper summarizes the design: according to the Sánchez de Ribera team, the 2013 review excluded psychiatric, child and incarcerated samples because the Dark Triad was conceived for non-clinical populations, and it examined four moderators, age, sex, sample type and measure of intelligence.

That design choice is the single most important fact for reading the result. The 2013 meta-analysis answers a question about traits in non-clinical samples. It was not built to answer a question about people rated on a checklist in prison, which is why the psychopathy-specific meta-analysis came afterward and why the offender studies get their own section below. The page on average IQ by personality type deals with a different trait system, and the page on IQ and personality covers the broader personality literature.

The two-sided test matters for a second reason. A single positive or negative correlation could be read as a story about compensation or cleverness. A null result in both directions removes the need for either story at the level of the population. It does not remove the possibility that particular facets, settings or individuals differ, which is the subject of the sections that follow.

4 What Did the 2022 Meta-Analysis of 143 Studies Add?

The 2022 meta-analysis pooled roughly three times as many sources as the 2013 one, split intelligence into general, verbal and nonverbal parts, and again found that none of the three dark traits is meaningfully related to intelligence. Michels begins from a plausible idea: manipulating a social environment effectively seems to call for high cognitive ability, so the Dark Triad might be an adaptive package. To test it, the paper identified 143 studies estimating the strength of relationships between the Dark Triad and general, verbal and nonverbal intelligence. The counts are not strictly comparable with 2013, because that review counted 48 independent samples and this one counts studies, so "about three times" is our rough reading, not an author claim.

The abstract reports four things. The results indicate that none of the Dark Triad constructs is meaningfully related to intelligence. There was, however, a small negative correlation between intelligence and Factor 2 psychopathy, the impulsive and antisocial behavior component that the next section defines. The heterogeneity in effect sizes was substantial and, for the most part, meta-regression could not explain it. And there was no evidence for publication bias. The abstract closes by saying the results challenge the notion that the Dark Triad is an adaptive set of personality traits that enables individuals to effectively manipulate their social surroundings. As with 2013, the abstract prints no coefficients, so this page reports direction and size as the authors worded them.

Two of those findings deserve a plain translation. Heterogeneity means that the studies disagree with each other by more than chance would produce, so the pooled average conceals samples where the correlation was somewhat positive and samples where it was somewhat negative. It does not signal a hidden large effect. Publication bias means that studies with null or unwelcome results go unpublished, which can inflate a pooled average. Finding no evidence of it is reassuring for a null result, because a null average built from a filtered literature would otherwise be suspect.

The step from 2013 to 2022 also shows the field maturing. The earlier review asked whether general mental ability related to the three traits. The later one added verbal and nonverbal intelligence and analyzed psychopathy by factor, which is where the one reliable signal appears. The next section follows that signal into the psychopathy-specific meta-analysis.

5 What Does the Psychopathy Meta-Analysis Show, Factor by Factor?

The 2019 psychopathy meta-analysis found a small negative correlation between total psychopathy and intelligence of r = -.07, driven by the impulsive and antisocial factor, with essentially none for the interpersonal and affective factor. Sánchez de Ribera, Kavish, Katz and Boutwell set out to evaluate the direction and magnitude of the association between intelligence and global psychopathy, its factors and facets, and related disorders. Their abstract reports a small, significant, negative relationship between intelligence and total psychopathy (r = -.07, p = .001), differential associations across factors and facets, no significant evidence of publication bias or outsized effects of outliers, and significant heterogeneity across studies for most effects.

In the preprint of the same paper, the authors report that 105 correlations underlie the total psychopathy and FSIQ result. Factor 1 is the interpersonal and affective component of psychopathy, and Factor 2 is the impulsive and antisocial behavior component, as the paper describes them. The table prints the pooled results by factor and by type of intelligence score.

Three readings follow from the table, and the authors support the first two in their text. First, the association is concentrated in the behavioral component: the authors write that the association between intelligence and psychopathy seems to be driven by measures, or domains of measures, that index risky and antisocial behavior. Second, verbal IQ carries more of it than performance IQ, with a total-psychopathy correlation of -.12 for verbal IQ against -.05 for performance IQ, and in the authors' structural equation modeling the two together accounted for a small share of the variance in psychopathy (R squared of .01). Third, the confidence interval for Factor 1 straddles zero, which is what the label "no association" means here: the data are compatible with no relation, not proof that every sample shows none.

Heterogeneity was high, with an I squared of 83 percent for FSIQ and total psychopathy, so the pooled number is an average over studies that disagree. The authors examined ten potential categorical moderators across the whole review and, for total psychopathy and FSIQ, only one was significant after correcting for multiple comparisons: the sex composition of the sample. Female samples showed r = -.19, male samples r = -.03, and mixed samples r = -.07. Age group, psychopathy measure, intelligence measure, recruitment site, inclusion of covariates, region and publication type were not significant moderators. The authors describe their interpretations of these moderator results as highly speculative and urge waiting for replication, and this page does the same. One of their suggested explanations is relevant to the next section: institutional samples of incarcerated males are overrepresented in the literature, which could act like range restriction on the male effect. They add a caveat: their moderator analyses did not show a significant difference between institutional and other sample types for most effect sizes, though that test did not control for sex.

The published abstract also reports the associations of intelligence with two diagnoses that overlap with psychopathy: antisocial personality disorder (r = -.13, p = .001) and conduct disorder (r = -.13, p = .001). Those are disorders, not traits, and the authors treat them as related but not identical constructs. They belong on the pages about IQ and mental health and IQ and crime, where the designs and the stigma risks are discussed separately.

6 Checklist Studies in Offender Samples Versus Self-Report Studies in Community Samples

A psychopathy score from a checklist in a jail or prison and a psychopathy score from a questionnaire in a student sample are different measurements of different populations, so their intelligence correlations should be read separately and never merged into one claim about people. The offender studies below used checklist ratings, the PCL:SV, PCL-R or PCL-YV, in detained or sentenced groups. The community studies used self-report scales in ordinary samples. According to the Sánchez de Ribera team, the 2013 meta-analysis excluded psychiatric, child and incarcerated samples, so it covered non-clinical samples, and the 2019 meta-analysis pooled forensic, clinical and community samples.

Vitacco, Neumann and Wodushek studied 100 males detained in a county jail using the PCL:SV. In their structural model, the interpersonal factor had a strong positive relationship to the full scale IQ latent variable, and the antisocial factor was moderately positively associated with it, while IQ was negatively related to both the affective and lifestyle factors. Overall, the psychopathy factors accounted for 25 percent of the variance of FIQ. That positive antisocial result runs against the pooled direction for impulsive and antisocial scores in the larger meta-analysis. One jail, one screening instrument and a model that estimates overlapping factors together can each produce a pattern that does not match a pooled average, and the heterogeneity of 83 percent in the meta-analysis tells the same story from the other side.

Johansson and Kerr tested a more specific idea in 370 men sentenced for violent, nonsexual crimes. They note that earlier PCL-R studies had shown no intelligence differences between people classified as psychopathic and those not, while Cleckley had argued that people described as psychopathic often show superior intelligence. Their hypothesis was that intelligence relates to the severity of criminal development in opposite ways for the two groups. For those not classified as psychopathic, higher total IQ and particularly verbal intelligence meant a later start in violent crime. For those classified as psychopathic on the checklist, the association was reversed. This is a pattern in one sample of sentenced men, and the authors present it as a way of explaining the gap between Cleckley and later data, not as a general law.

Salekin, Neumann, Leistico and Zalot examined 122 detained children and adolescents with the PCL-YV and two kinds of intelligence measure, the traditional K-BIT and a triarchic test of analytic, creative and practical thinking. Traits reflecting a superficial and deceitful interpersonal style were positively related to verbal skills and to the triarchic measure, while traits reflecting disturbances in affective processing were inversely associated with verbal ability. The authors conclude that Cleckley's hypothesis was partially supported when the facets of psychopathy are separated and intelligence is viewed through more than one model.

DeLisi, Vaughn, Beaver and Wright used 840 cases from the MacArthur Violence Risk Assessment Study. Regression models showed inverse relationships between verbal intelligence and psychopathy for eight of twelve items of the disorder. The authors call the study exploratory and warn that the uniqueness of the MacArthur sample limits how far the result generalizes.

The point of the table is restraint. A pooled r of -.07 does not say that offender samples show -.07, and a null result in a student sample does not say anything about prison populations. What the studies share is the absence of the large positive link that the evil genius idea predicts.

7 Do the Facets of Psychopathy Point in Opposite Directions?

Yes, in the 2019 meta-analysis: the facets of psychopathy relate to intelligence in different directions, which is why a total score near zero can hide a positive link for one facet and a negative link for another, though the positive interpersonal result rests on 11 correlations and an interval that reaches zero. Cleckley's 1941 description, as the meta-analysis authors summarize it, included good intelligence among its key attributes, and facet-level research began as an attempt to see whether any part of that description survived. The authors report facet results from small sets of correlations, six to twelve for each facet with FSIQ, so each number below is fragile.

The pattern is that the interpersonal facet, which includes the superficially charming and deceitful style, has a small positive FSIQ correlation of .14 whose confidence interval just reaches zero, the affective facet is reliably negative, and the lifestyle and antisocial facets are negative for verbal and performance IQ. The authors also analyzed fearless dominance and boldness scales, a controversial operationalization that some measures include, and found a small positive FSIQ correlation of .15. When they combined impulsive antisociality with meanness and disinhibition scales, the correlation across intelligence types was negative .06 (p = .050). Our reading is that this is why the Factor 1 total sits at .01: a positive interpersonal piece and a negative affective piece can offset each other inside it. That is our arithmetic on printed numbers, not an author claim.

One longitudinal study moves from association toward sequence. Kavish and colleagues linked intelligence at age 8 to psychopathy scores at age 48 in an analytic sample of 292 White, urban male individuals from London, using the PCL:SV. They found a significant but weak inverse association between age 8 intelligence and the affective, lifestyle and antisocial facets, and a nonsignificant association with the interpersonal facet. They conclude that psychopathy, as conceptualized in most modern models, is either very weakly inversely related to intelligence or simply not a correlate of it. This is a predictive association in one sex, one city and one cohort, and it does not show that early cognitive ability causes later traits or the reverse.

8 Are People High in Narcissism Measurably More Intelligent?

No: in four studies with 1,141 participants, neither grandiose nor vulnerable narcissism was related to objectively measured intelligence, and the pooled Dark Triad meta-analysis agrees. Zajenkowski, Czarna, Szymaniak and Dufner tested the associations of both forms of narcissism with subjectively and objectively assessed intelligence, basic personality traits, test-related stress, beliefs about intelligence and well-being. Their abstract states that both forms were unrelated to objective intelligence. The same abstract reports that grandiose narcissism went with reduced distress in the context of IQ testing and low engagement in cognitive performance in Study 2, while vulnerable narcissism went with increased distress in that context. The page on stress and IQ covers how test-related stress and performance relate more generally.

An earlier study pointed the other way, slightly. In the Paulhus and Williams sample of 245 students, narcissism showed small positive associations with cognitive ability, though the abstract prints no coefficient. The two results are not hard to reconcile. A single sample of 245 students is a thin base for a stable estimate, and the 2022 meta-analysis of 143 studies later found none of the traits meaningfully related to intelligence. Our reading is that the early positive signal was small and sample-bound, and the larger pooled evidence does not carry it forward, although the abstract of that meta-analysis prints no narcissism-specific coefficient to compare.

The distinction between grandiose and vulnerable forms matters for self-perception more than for performance. On measured ability, the evidence in these studies is the same for both: no relationship. On self-perceived ability, the forms diverge, which the next section covers. Readers who want the Big Five and the broader personality literature, where intelligence relates to traits through small and specific links, can read the page on IQ and personality.

9 Do People High in Narcissism Overestimate Their Intelligence?

People high in grandiose narcissism rate their own intelligence higher, the trait explained more of that self-rating than measured intelligence did, and measured intelligence was unrelated to the trait, but this is a statement about averages in research samples, not about any person. Gabriel, Critelli and Ee reported an early version of the finding in a sample of 146 college students. According to the abstract of their 1994 paper, narcissism predicted positive illusions in self-evaluations of intelligence and of physical attractiveness for both males and females, self-esteem predicted the intelligence illusion for males, and both sexes overestimated their own intelligence on average, with males showing greater positive illusions, an effect the authors attributed at least partly to gender differences in narcissism.

The Zajenkowski team reported the self-perception link across several studies. Grandiose narcissism was associated with high self-perceived intelligence in Studies 1 to 3 and explained more variance in self-perceived intelligence than objective intelligence and the Big Five personality traits did. Individuals with high grandiose narcissism based their well-being partly on intelligence (Study 3) and considered intelligence important for success in different life domains, especially for social relations (Study 4). Vulnerable narcissism was unrelated to self-perceived intelligence in Studies 1 to 3.

Does the self-rating cause or follow the trait? A separate set of experiments helps. Zajenkowski, Sedikides, Gignac and colleagues set out to examine the causal direction of the relation between grandiose agentic narcissism and self-assessed intelligence by manipulating narcissism with a validated procedure. In Studies 1 and 2, high versus low narcissism increased self-assessed intelligence, and in Study 2 self-assessed intelligence mediated the effect of narcissism on academic goal-pursuit, expected academic achievement and psychological well-being. That is experimental evidence that a temporary rise in narcissistic feeling raised self-rated intelligence in those samples. It supports a causal reading in that direction for a momentary state, and it does not show how a lifelong trait shapes a self-image or that the reverse path never operates.

Any claim about overestimation needs a baseline, because almost everyone's self-estimate is imperfect. Patzl, Oberleiter and Pietschnig pooled 278 effect sizes from 115 independent samples (N = 36,833) and found a moderate positive correlation of r = 0.30 (95 percent confidence interval 0.27 to 0.33) between self-assessed intelligence and psychometric IQ. If the typical self-estimate tracks measured ability at about .30, the question for narcissism is how much additional inflation the trait adds beyond ordinary error, and the studies above are consistent with an upward shift in self-ratings without any matching shift in test scores. The page on estimating your own IQ covers the general accuracy of self-ratings in more detail.

A caution about method applies. Gignac and Zajenkowski studied 929 community participants who completed a self-assessment of intelligence and the Advanced Raven's Progressive Matrices. They argued that the common way of showing that low scorers overestimate themselves, comparing means across quartiles, is confounded by the better-than-average effect and regression toward the mean. They found no significant heteroscedasticity and an essentially linear association between measured and self-assessed intelligence. That paper concerns the Dunning-Kruger hypothesis, not narcissism, and its conclusions have been debated, so it is cited here only as a reminder that "people with trait X overestimate" claims depend on how the gap is computed.

10 Where Does the Smart Psychopath Come From, and Why Is Near Zero Not the Same as Equal?

The research literature traces the image of the genius psychopath to a clinical description and to fictional portrayals, not to the correlations, and a correlation near zero does not mean that people high and low in a trait are alike in every respect. The Sánchez de Ribera authors note that Cleckley's 1941 description included good intelligence, and that the public has often pictured the psychopath as an evil genius or criminal mastermind. DeLisi and colleagues used the character Hannibal Lecter as their example, a fictional figure framed as having a superior intelligence that enhanced his ability to manipulate and victimize others, and they tested the idea in their 840-case sample. A plausible reason such stories persist, which none of the cited studies tests, is that vivid cases are remembered and the many unremarkable ones are not. For a case where a circulating number has no test behind it, see the page on the Adolf Hitler IQ claim.

The statistics need the opposite caution. Squaring a correlation of -.07 gives about .005, so total psychopathy and FSIQ share roughly one half of one percent of their variance, which is our arithmetic on the printed figure. That is why the correlation is called small. But a small correlation across a population is silent about several things. It does not describe the tails, since a correlation summarizes the whole spread of both variables. It does not describe any subgroup, and the heterogeneity of 83 percent warns that studies differ. It does not describe the facets, which offset each other in the total. And it does not describe settings, since offender and community samples are different populations. "Unrelated" means that knowing a person's score on one measure tells you almost nothing about the other. It does not mean that every group is identical on every dimension.

For the other end of the scale, where the same caution about group averages applies, see the page on what a low IQ score means.

11 What the Evidence Supports, Stated Narrowly

The evidence supports a handful of narrow statements about group averages in research samples, and nothing about any individual.

• Across meta-analyses of 48 samples in 2013 and 143 studies in 2022, the Dark Triad traits are not consistently or meaningfully related to measured intelligence. The evil genius hypothesis and the compensatory hypothesis are not supported at the level of the pooled data.
• Total psychopathy correlates with intelligence at about r = -.07 in a 2019 meta-analysis, a small negative association concentrated in the impulsive and antisocial factor, with an interval around the interpersonal and affective factor that includes zero.
• The facets point in different directions in small sets of studies: a small positive association for the interpersonal facet, whose interval reaches zero, and a negative one for the affective facet, which is a reason to avoid reading the total as the whole story.
• Checklist studies in detained or sentenced samples and self-report studies in community samples measure different things in different populations, and this page does not merge them. Some individual studies in offender samples report patterns that differ from the pooled average, and heterogeneity is high.
• A longitudinal study of 292 men from London found at most a weak inverse association between intelligence at age 8 and psychopathy facets at age 48, which is a predictive association and not a causal estimate.
• Grandiose narcissism goes with higher self-rated intelligence and is unrelated to objectively measured intelligence in four studies totaling 1,141 participants. An experimental manipulation raised self-rated intelligence temporarily, which supports a causal path for a momentary state.
• Self-rated and measured intelligence correlate at about .30 in the general literature, so imperfect agreement between the two is the norm before any trait is considered.

Everything beyond these statements is outside what the cited designs can show. Neither the pooled correlations nor the individual studies say that a person with a given trait has a given IQ, that a given IQ implies a trait, or that either can be inferred from behavior.

12 What This Does Not Say About You, a Colleague or Anyone Else

This page reports group findings about research constructs, and it gives no way to judge, label or diagnose any person, including you, from an IQ score or from behavior. Words such as psychopathy and narcissism name measured traits in studies and clinical constructs in a professional setting. They are not labels for people in someone's life, and several of the studies here used checklists applied in clinical or forensic settings, which an online test and a reader's own impressions cannot replicate.

The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014), published jointly by the three organizations at testingstandards.net, define validity as the degree to which evidence and theory support the interpretations of test scores for proposed uses of tests. A cognitive score has evidence behind it for describing performance on cognitive tasks under standard conditions. The research reviewed on this page gives no support for reading it as a sign of a personality trait, and reading it that way is a different use from the one it was built for. The APA's Ethical Principles of Psychologists and Code of Conduct, Standard 9.06, says that when interpreting assessment results psychologists take into account the purpose of the assessment, test factors, test-taking abilities and other characteristics of the person assessed, and that they indicate significant limitations of their interpretations. Both documents point the same way: the meaning of a score depends on what it is used for.

ACIS measures cognitive abilities, not personality. It reports a Full Scale IQ and six index scores, with percentiles and a 95 percent confidence interval, and it is online and unsupervised, so it is not a clinical or diagnostic instrument and it cannot detect, screen for or rule out any personality trait or condition. A reader curious about their own profile can start with the page on what an IQ score means for you, the comparison of online IQ tests or the page on the Full Scale IQ test. If a concern about someone's behavior or about your own is the real question, it is a matter for a qualified professional and not for a cognitive score.

13 Sources Behind This Page

Every figure on this page comes from one of these documents, and the list separates peer-reviewed meta-analyses and studies from the two standards documents that govern how scores are read. Abstract-only figures for the 2013 and 2022 meta-analyses and the details of the 2019 meta-analysis from its preprint are described in the note near the start of the page. The numbers cited were opened or confirmed on October 6, 2026, and the arithmetic statements on this page are labeled as ours.

• Sánchez de Ribera O, Kavish N, Katz I and Boutwell B. Untangling Intelligence, Psychopathy, Antisocial Personality Disorder, and Conduct Problems: A Meta-Analytic Review. European Journal of Personality, 2019, volume 33, issue 5, pages 529 to 564.
• O'Boyle E, Forsyth D, Banks G and Story P. A meta-analytic review of the Dark Triad-intelligence connection. Journal of Research in Personality, 2013, volume 47, issue 6, pages 789 to 794.
• Michels M. General Intelligence and the Dark Triad. Journal of Individual Differences, 2022, volume 43, issue 1, pages 35 to 46.
• Paulhus D and Williams K. The Dark Triad of personality: Narcissism, Machiavellianism, and psychopathy. Journal of Research in Personality, 2002, volume 36, issue 6, pages 556 to 563.
• Zajenkowski M, Czarna A, Szymaniak K and Dufner M. What do highly narcissistic people think and feel about (their) intelligence? Journal of Personality, 2020, volume 88, issue 4, pages 703 to 718.
• Zajenkowski M, Sedikides C, Gignac G, Górniak J and Maciantowicz O. Induced narcissism increases self-assessed intelligence: implications for academic goal-pursuit, expected academic achievement, and psychological well-being. Journal of Research in Personality, 2022, volume 101, article 104311.
• Gabriel M, Critelli J and Ee J. Narcissistic Illusions in Self-Evaluations of Intelligence and Attractiveness. Journal of Personality, 1994, volume 62, issue 1, pages 143 to 155.
• Gignac G and Zajenkowski M. The Dunning-Kruger effect is (mostly) a statistical artefact: Valid approaches to testing the hypothesis with individual differences data. Intelligence, 2020, volume 80, article 101449.
• Patzl S, Oberleiter S and Pietschnig J. Mirror, Mirror on the Wall: A Meta-Analysis on the Validity of Self-Assessed Intelligence through the Lens of the Multiverse. Journal of Intelligence, 2024, volume 12, issue 9, article 81.
• DeLisi M, Vaughn M, Beaver K and Wright J. The Hannibal Lecter Myth: Psychopathy and Verbal Intelligence in the MacArthur Violence Risk Assessment Study. Journal of Psychopathology and Behavioral Assessment, 2010, volume 32, issue 2, pages 169 to 177.
• Salekin R, Neumann C, Leistico A and Zalot A. Psychopathy in Youth and Intelligence: An Investigation of Cleckley's Hypothesis. Journal of Clinical Child and Adolescent Psychology, 2004, volume 33, issue 4, pages 731 to 742.
• Vitacco M, Neumann C and Wodushek T. Differential Relationships Between the Dimensions of Psychopathy and Intelligence. Criminal Justice and Behavior, 2008, volume 35, issue 1, pages 48 to 55.
• Johansson P and Kerr M. Psychopathy and Intelligence: A Second Look. Journal of Personality Disorders, 2005, volume 19, issue 4, pages 357 to 369.
• Kavish N, Bergstrøm H, Narvey C, Piquero A, Farrington D and Boutwell B. Examining the association between childhood cognitive ability and psychopathic traits at age 48. Personality Disorders: Theory, Research, and Treatment, 2021, volume 12, issue 1, pages 81 to 85.
• American Educational Research Association, American Psychological Association and National Council on Measurement in Education. Standards for Educational and Psychological Testing, 2014, open-access PDF at testingstandards.net, validity definition checked in the file on October 6, 2026.
• American Psychological Association. Ethical Principles of Psychologists and Code of Conduct, Standard 9.06, Interpreting Assessment Results, text checked on October 6, 2026 in a copy of the 2017 amended code, because the apa.org page returned no text to our tools.

14 Frequently Asked Questions