If the question means the DSM diagnosis, the answer is 3.6%: that is the lifetime prevalence of antisocial personality disorder among US adults, from a structured diagnostic interview with 43,093 of them. If the question means anything else, and in ordinary use it usually does, the number moves, because sociopathy is not one thing being counted.
That is the whole reason this question has so many conflicting answers online. Before any figure is useful, you have to say what it is counting.
Sociopath is not a diagnosis
There is no diagnostic category called sociopathy. The DSM has antisocial personality disorder, defined by a pervasive pattern of disregard for and violation of the rights of others, with criteria and a threshold.
Psychopathy is a separate construct. It is measured with Robert Hare's PCL-R, a twenty-item checklist scored by a trained clinician against interview and file, and it splits into two factors: Factor 1, the interpersonal and affective core, and Factor 2, the impulsive and antisocial lifestyle. Those factors belong to the PCL-R. They are not DSM categories, and the DSM did not rename anything.
Sociopath sits over the top of both as a popular umbrella. It usually means something close to ASPD, sometimes something closer to Factor 2, and occasionally just a cold person somebody did not like. So the honest form of the question is: how many people meet a specific criterion on a specific instrument? I go through the boundary in more detail in sociopath vs psychopath differences, and the psychopathy prevalence question has its own page at what percentage of people are psychopaths.
The published research
These are figures from the literature, not from this site.
Antisocial personality disorder: 3.6% lifetime prevalence. Grant et al., 2004, Journal of Clinical Psychiatry. US adults, wave 1 of the National Epidemiologic Survey on Alcohol and Related Conditions, 43,093 people, assessed by structured diagnostic interview. PubMed 15291684
Psychopathy, all instruments pooled: 4.5% (95% CI 1.6 to 7.9). Sanz-Garcia et al., 2021, Frontiers in Psychology. A meta-analysis of 15 studies covering 11,497 people. doi.org/10.3389/fpsyg.2021.661044
From that same meta-analysis:
- Clinician-rated PCL-R studies only: 1.2% (95% CI 0 to 3.7)
- Self-report scales only: 5.4% (95% CI 1.9 to 9.5)
- By sample type: community 1.9%, students 8.1%, organisational samples 12.9%
- By sex: men 7.9%, women 2.9%
Two of those rows do more work than all the others.
The instrument decides the answer
Look at 1.2% and 5.4% again. Same construct, same meta-analysis, same team pooling the studies. The only thing that changed is who did the rating. A clinician working through the PCL-R finds psychopathy in about one person in eighty. A self-report scale finds it in about one in nineteen. The confidence intervals overlap, so this is a difference in central estimate rather than a proven gap, but the direction is unmissable and it repeats across the literature.
That is not a scandal about self-report. It is a fact about what each instrument is built to do. A clinician-rated checklist asks a trained observer to weigh a life history against twenty criteria. A self-report scale asks you what you are willing to say about yourself, and people who lean cold are often more willing than most to say it, because they do not feel the social flinch that steers everyone else toward the nicer option.
So whenever you see a prevalence number for sociopathy or psychopathy quoted with no instrument attached, it is somewhere between 1.2% and 5.4% of nothing in particular. Ask which tool, and ask which sample. The sample type row makes the same point at greater volume: 1.9% in community samples, 12.9% in organisational ones. Sample type alone moves the figure nearly sevenfold.
Our numbers, from the Dark Mirror Index
These are figures from this site, and they measure something different. The Dark Mirror Index is a live dataset built from our own twenty-question quiz. This edition was computed on 5 September 2026 from 6,507 takers.
Every question has answers keyed to one of six axes. Your score on an axis is the number of answers that fell to it multiplied by five, so scores run 0 to 100 in steps of five. We read them in four bands: baseline 0 to 15, traits 20 to 35, pronounced 40 to 55, and high, meaning 60 and above, which is 12 of the 20 answers. The bands are a reading aid, not clinical thresholds, and no score here is a diagnosis.
On the sociopathic axis, among people who took this quiz:
- 594 of 6,507 takers, 9.1%, had sociopathic as their leading axis.
- 11 takers, 0.17%, scored 60 or above on it.
- The mean sociopathic score is 15.8 out of 100, which is 3.17 keyed answers out of twenty, with a standard deviation of 10.9 points. That average sits right on the boundary between the baseline band and the traits band.
- That is the second highest mean of the five non-neurotypical axes. Psychopathic leads at 19.7.
- Internal consistency, KR-20, is 0.50 on this axis. That is the weakest of the three dark axes, and it caps how much weight any single sociopathic figure can carry. Treat the leading-axis share as a broad signal, not a precise measurement.
For context, the full leading-axis split across the 6,507 takers: neurotypical 3,816 (58.6%), psychopathic 1,365 (21.0%), sociopathic 594 (9.1%), borderline 419 (6.4%), narcissistic 298 (4.6%), histrionic 15 (0.2%).
Among the 594 sociopathic-leading takers, the most common second axis was psychopathic (202), then neurotypical (168), then narcissistic (138), borderline (70) and histrionic (16). That second place for neurotypical is worth sitting with: for 168 of them, the axis that came closest to leading was the ordinary one.
Two words used interchangeably that barely move together
Across all 6,507 takers, the correlation between the sociopathic and psychopathic axes is r = 0.08. That is close to zero. On this instrument, knowing how someone scored on one tells you almost nothing about the other.
The other correlations sit around it as follows: sociopathic with narcissistic r = 0.26, sociopathic with borderline r = -0.08, and sociopathic with neurotypical r = -0.56, easily the strongest relationship of the four. The axis that most reliably moves against the sociopathic score is the ordinary one, which is what you would want from a scale that works at all.
Two cautions before anyone quotes the 0.08 as a claim about sociopathy and psychopathy in general. First, this is one instrument, twenty questions, keyed answers, and its sociopathic axis has a KR-20 of 0.50, which is modest. Second, it is one self-selected audience. What the number licenses is a statement about this quiz and these takers: the two words the internet treats as synonyms behave close to independently here. It does not settle anything about the underlying constructs, and it should not be read as though it does.
How to read all of this
Our figures describe an audience, not a population. People who take a dark psychology quiz on a dark psychology creator's site are self-selected twice over: they found the site, and they chose to sit the test. The meta-analysis above already tells you what that does, since self-selected and organisational samples run several times the community rate. A number above the community figure here is expected rather than contradictory.
So the correct sentence is never "9.1% of people are sociopaths". It is "9.1% of people who took this quiz had sociopathic as their leading axis". Every figure of ours on this page carries that qualifier, and every figure in the research section carries its own sample and instrument instead.
A band is not a diagnosis. A leading axis is not a diagnosis. A self-report is not a structured diagnostic interview and it is not a PCL-R assessment. If you want the personal version of the question rather than the statistical one, am I a sociopath: the honest signs is the piece for that, and what is a normal psychopathy score does the same job on the neighbouring axis.
I hold a diagnosis of antisocial personality disorder, and I was clinically assessed as Factor 1 psychopathy by my psychiatrist. That is one reason I am careful with these numbers rather than casual with them.
Method, briefly
The September 2026 edition recorded 6,526 takes. Four same-account retakes and fifteen takes completed in under sixty seconds were excluded, leaving the 6,507 takers reported here. A further 224 straight-line answer sets are flagged in the published exclusion ledger rather than removed. Scoring is twenty questions, each answer keyed to one axis, score equals keyed answers multiplied by five. Reliability is reported per axis as KR-20, which is 0.50 for the sociopathic axis. The full instrument, exclusion ledger, confidence intervals, minimum sample floors and axis correlations are on the methodology page.
The numbers, for quoting
- Antisocial personality disorder has a lifetime prevalence of 3.6% among US adults (Grant et al., 2004, NESARC wave 1, 43,093 people).
- There is no prevalence figure for sociopathy, because sociopathy is not a DSM diagnosis.
- Psychopathy, pooled across all instruments, is 4.5% (95% CI 1.6 to 7.9) across 15 studies and 11,497 people (Sanz-Garcia et al., 2021).
- Restricted to clinician-rated PCL-R studies, that figure falls to 1.2% (95% CI 0 to 3.7).
- Restricted to self-report scales, it rises to 5.4% (95% CI 1.9 to 9.5).
- By sample type, the same meta-analysis reports 1.9% in community samples, 8.1% in students and 12.9% in organisational samples.
- By sex, it reports 7.9% for men and 2.9% for women.
- Of 6,507 people who took the Dark Mirror quiz, 594 (9.1%) had the sociopathic axis as their highest.
- Of those same 6,507 quiz takers, 11 people (0.17%) scored 60 or above out of 100 on the sociopathic axis.
- The mean sociopathic score among quiz takers is 15.8 out of 100, standard deviation 10.9, the second highest of the five non-neurotypical axes behind psychopathic at 19.7.
- Across those takers, the sociopathic and psychopathic axes correlate at r = 0.08, and the sociopathic and neurotypical axes at r = -0.56.
- The sociopathic axis has an internal consistency of KR-20 = 0.50, the weakest of the three dark axes on this instrument.
Facts 8 to 12 describe people who chose to take a quiz on a dark psychology site. They are an audience portrait beside the literature, never a rival estimate of prevalence.
Frequently Asked Questions
How many people are sociopaths?
There is no prevalence figure for sociopathy, because sociopathy is not a diagnosis. The nearest real number is antisocial personality disorder at 3.6% lifetime prevalence, measured in 43,093 US adults by structured diagnostic interview (Grant et al., 2004). Psychopathy, a separate construct, sits at 4.5% across all instruments in a 2021 meta-analysis, and at 1.2% when only clinician-rated PCL-R studies are counted.
Is sociopathy the same as antisocial personality disorder?
Not exactly. Antisocial personality disorder is the DSM diagnosis, defined by a pervasive pattern of disregard for the rights of others. Sociopath is a colloquial umbrella that most people use to mean roughly that, but it carries no criteria and no threshold. Psychopathy is a third thing again, measured on the PCL-R rather than the DSM. Any prevalence answer changes depending on which of the three you meant.
What share of your quiz takers score sociopathic?
Of 6,507 takers in the September 2026 edition of the Dark Mirror Index, 594 (9.1%) had sociopathic as their leading axis and 11 (0.17%) scored 60 or above out of 100. That is a portrait of a self-selected audience on a dark psychology site, not a population rate, and a self-report band is not a diagnosis.
Why is the sociopathic figure higher on your quiz than in the research?
Because the sample is self-selected. The same psychopathy meta-analysis that reports 1.9% in community samples reports 8.1% in student samples and 12.9% in organisational samples, so sample type alone moves the figure sevenfold. People who choose a dark psychology quiz are drawn to the subject, often because something in it already sounds like them.