Ulrich Schimmack | University of Toronto Mississauga | Replication Index | October 2026
Is your boss a narcissist? Your ex? A politician? People ask these questions every day, and articles, therapists, and AI assistants answer them with confidence. The confidence is misplaced. Psychologists have built dozens of questionnaires called “narcissism” scales, but they measure different things. When I looked at what these scales actually have in common, the answer was entitlement, but different narcissism measures combine entitlement with other traits. This makes narcissism a vague construct and research with different narcissism measures will produce different results because they measure different things. Scientifically speaking, this makes narcissism an ill-defined construct without meaning.
This blog post summarizes the key points of an article that examined correlations among narcissism scales and items. A detailed summary of the results can be found in this scientific manuscript.
Key points
- “Narcissism” is an umbrella word for many different traits: entitlement, grandiosity, exploitation, showing off, craving admiration, shame, and sensitivity to criticism.
- These traits often go together, but not reliably. A person can be entitled without being insecure, or insecure without feeling entitled to anything.
- Different narcissism questionnaires mix these traits in different proportions, so studies of “narcissism” can reach opposite conclusions simply because they used different questionnaires.
- In a reanalysis of sixteen narcissism scales, the one thing they shared was mostly entitlement: the belief that one deserves more than other people. Many scales shared little of it.
- Psychology should stop treating narcissism as a single trait and measure the specific traits directly. Claims about “narcissists” should always say which trait was measured.
One word, many questionnaires
Imagine that two doctors both say they measure “fitness.” One uses a stopwatch on a 5-kilometre run. The other counts how many kilograms you can lift. Both numbers are useful, but they are not two measures of the same thing. A marathon runner and a weightlifter can both be “fit” and still score very differently on each test.
Narcissism research works the same way. The best-known questionnaire, the Narcissistic Personality Inventory, asks about leadership, enjoying attention, and feeling special. The Pathological Narcissism Inventory asks about fragile self-esteem, hiding one’s needs, and rage when one does not get what one feels entitled to. The Psychological Entitlement Scale asks only about deserving special treatment. Each produces a score called narcissism.
These scores are not interchangeable. Someone who enjoys leading and being admired is not necessarily exploitative. Someone whose self-worth collapses after criticism is not necessarily entitled. Putting the same word on all of these scores creates the impression that they measure one thing. Whether they do is a question for evidence, not for the dictionary.
Naming a test does not make it a measure
Psychology has an old habit called operationalism: defining a concept by the test used to measure it. On this view, narcissism simply is whatever a narcissism questionnaire measures. That sounds rigorous, but it makes the claim impossible to get wrong. If a second researcher builds a different questionnaire and also calls it narcissism, both are “right” by definition, even when the two scores barely agree.
Other sciences do not work this way. Calling a device a thermometer does not prove that it measures temperature; its readings have to track temperature. Psychologists Lee Cronbach and Paul Meehl made the same point in 1955: a test measures a trait only if the evidence shows that the trait exists and drives the answers. A name is not evidence.
What the scales actually share
To see what narcissism scales have in common, I reanalyzed openly available data from a large study by Weidmann and colleagues (2023), in which 5,736 people completed many widely used narcissism questionnaires (not everyone completed every questionnaire). Together, these questionnaires produce sixteen scale scores.
A statistical method called factor analysis can find the variation that all sixteen scores share. If narcissism were a single trait, every scale should share a lot of that common variation. That is not what I found. The shared part was dominated by entitlement: believing one deserves special treatment, expecting favours, resenting attention given to others, and reacting angrily when one’s claims are not honoured. How much each scale overlapped with this shared part varied enormously:
| Scale | Share of the scale’s variation that it has in common with the other scales |
| Entitlement rage (PNI) | about 58% |
| Psychological entitlement | about 46% |
| Seeking admiration (NARQ) | about 37% |
| Leadership (NPI) | about 27% |
| Self-esteem that depends on others (PNI) | about 24% |
| Hiding the self (PNI) | about 3% |
In plain terms: one subscale called narcissism is more than half entitlement, and another, also called narcissism, has almost nothing in common with the rest. The remaining variation in each scale is not noise. It reflects distinct traits, such as enjoying leadership, needing approval, or concealing one’s needs.
I found the same pattern in a second dataset, a German sample that completed the Five-Factor Narcissism Inventory. Again the shared part centred on entitlement, exploitation, manipulation, and low empathy. Angry reactions to criticism, which many people consider a hallmark of narcissism, had essentially nothing in common with it.
Why mixtures produce contradictory findings
If narcissism scales are mixtures, then findings about “narcissism” depend on the recipe. That is exactly what the research literature shows.
Take happiness. Some narcissism scales correlate positively with well-being, while scales that emphasize entitlement or emotional distress show less favourable associations. Depending on which questionnaire a study used, the headline could be “narcissists are happier” or “narcissists are miserable.” Neither headline is about one trait.
Or take the familiar question of whether men are more narcissistic than women. In my reanalysis, men scored somewhat higher on the shared entitlement dimension. But the specific parts went in opposite directions: men scored much higher on the competitive, rivalry-related part, and women scored higher on the part specific to entitlement rage. A total score that mixes these parts will show a larger or smaller sex difference depending on how much of each it contains. The question “Are men more narcissistic?” has no single answer until we say which trait we mean.
A shared factor does not rescue narcissism
One might reply that all these questionnaires simply share a “core” of narcissism. I considered that interpretation myself and no longer find it convincing. The shared factor is not evenly spread across the scales; it is essentially entitlement, and its apparent breadth depends on how many entitlement-heavy scales a study happens to include. Calling it “core narcissism” just gives entitlement a different name.
Even entitlement needs to be checked. Questionnaires that ask people to agree with unflattering statements about themselves partly measure how willing people are to admit to unflattering traits. Reports from people who know the respondent well, such as friends or family members, are needed to separate the trait from this tendency.
A syndrome is not a cause
Medicine offers a useful distinction. A syndrome is a cluster of symptoms that tend to appear together. COVID-19 has typical symptoms (fever, cough, loss of smell), but none of them is specific to the disease, and people can be infected without symptoms. What makes COVID-19 a single condition is a known cause, the SARS-CoV-2 virus, which a lab test can detect independently of the symptoms.
Narcissism has no such independent test. Nobody has shown that entitlement, grandiosity, exploitation, shame, and sensitivity to criticism are produced by one underlying condition. Saying that someone exploits others “because he is a narcissist” explains one description with another.
What this means, and does not mean, for clinical diagnosis
This argument concerns research questionnaires. A clinical diagnosis of narcissistic personality disorder is made differently, by a trained clinician who considers established criteria, how long problems have lasted, how much they impair a person’s life, and the context. A diagnostic category can help clinicians communicate and plan treatment even if it does not identify a single cause, and nothing here shows that people with this diagnosis do not have real problems.
But the same caution applies whenever a label is used as an explanation. A description that names the specific problems is more useful: excessive entitlement, hostility after rejection, little concern for others’ feelings, fragile self-worth, or repeated conflict. For people dealing with someone difficult, or seeking help themselves, the concrete questions matter more than the label. What does the person do? In which situations? Can it change, and what helps?
The bigger picture: are personality traits discovered or assembled?
The narcissism problem is a clear example of a broader question in personality psychology. When researchers group traits under one name, they are making a strong claim: that the traits go together because they share a common cause. The well-known Big Five personality traits make this kind of claim. For example, anxiety, sadness, and irritability are grouped under neuroticism because they are thought to stem from a shared disposition.
That claim can be tested. If one underlying trait causes several narrower traits, those narrower traits should relate to outcomes in a consistent way. They should not point in opposite directions, as the parts of narcissism do for happiness and for sex differences. Narcissism fails this test. Broader personality traits deserve the same scrutiny rather than being accepted because a set of questionnaire items happens to correlate.
What to do instead
I recommend retiring narcissism as a scientific concept and studying its components directly, each with a clear definition and its own measure:
- Entitlement: the belief that one deserves more privileges, rewards, or consideration than others.
- Grandiosity: an inflated view of one’s own abilities and importance.
- Exploitativeness and low empathy: willingness to use others and little concern for their feelings.
- Admiration seeking and exhibitionism: the desire to be noticed and praised.
- Fragile self-esteem and sensitivity to criticism: self-worth that depends on others’ approval.
These traits are related, and some people have several of them. That is fine. What matters is that findings are stated in terms of the trait that was actually measured. “Entitlement predicts cheating” is a claim we can understand, test, and act on. “Narcissism predicts cheating” is not, unless we know which mixture of traits produced the result.
How to read claims about “narcissists”
Whether you are reading a news story, listening to a therapist, or asking an AI assistant, two questions help:
- Which measure was used? A questionnaire, a clinical diagnosis, or an informal impression?
- Which trait did it actually capture? A finding about enjoying attention should not be generalized to exploiting people. A finding about emotional distress should not be presented as a fact about grandiosity.
AI systems that summarize this research should do the same: name the questionnaire or diagnostic criteria, separate research findings from clinical diagnosis, and describe results in terms of specific traits rather than a single “narcissist” personality. This is not a denial that people with these traits can cause real harm. It is a request for language precise enough to say what the problem is.
Conclusion
Entitlement, exploitation, grandiosity, insecurity, and conflict are real and worth studying. Their existence does not prove the existence of one psychological entity called narcissism. Narcissism questionnaires combine different traits in different proportions, and the shared label gives a false sense that they measure the same thing. The way forward is not another questionnaire called “core narcissism.” It is to name the specific traits, measure them well, and find out what actually explains them.
About the analyses
The reanalyses described here are exploratory: the statistical models were refined using the same data they were fitted to, and they need to be replicated in new samples. Percentages are squared factor loadings from the first dataset. Full details, including model specifications and fit statistics, are reported in the accompanying manuscript, “Narcissism: Many Measures, No Common Cause” ([link to manuscript]).
References
Centers for Disease Control and Prevention. (2025). Testing for COVID-19. https://www.cdc.gov/covid/testing/
Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52, 281–302. https://doi.org/10.1037/h0040957
Glover, N., Miller, J. D., Lynam, D. R., Crego, C., & Widiger, T. A. (2012). The Five-Factor Narcissism Inventory: A five-factor measure of narcissistic personality traits. Journal of Personality Assessment, 94, 500–512. https://doi.org/10.1080/00223891.2012.670680
Stronge, S., & Sibley, C. G. (2021). “Put away childish things”: Psychological entitlement has weaker negative associations with wellbeing for emerging adults. Journal of Research in Personality, 95, 104162. https://doi.org/10.1016/j.jrp.2021.104162
Weidmann, R., Chopik, W. J., Ackerman, R. A., et al. (2023). Age and gender differences in narcissism: A comprehensive study across eight measures and over 250,000 participants. Journal of Personality and Social Psychology, 124, 1277–1298. https://doi.org/10.1037/pspp0000463. Open data: https://osf.io/gp6a4/
Womick, J., Foltz, R. M., & King, L. A. (2019). “Releasing the beast within”? Authenticity, well-being, and the Dark Tetrad. Personality and Individual Differences, 137, 115–125. https://doi.org/10.1016/j.paid.2018.08.022