What This Article Is And Is Not
This article is about how a psychological construct is measured and what a body of research does and does not establish. It is not medical, psychological or clinical advice, and nothing in it is a diagnosis or an assessment of any reader.
We are not going to tell anyone how they feel or whether their self-doubt is warranted. The subject here is the evidence and the instruments, which is a question a business publication can address and a question about anyone's inner life is not.
If self-doubt is affecting your work, your health or your life, that is worth discussing with a doctor or a qualified professional rather than with a website, and this article is not a substitute for either.
One further note on why we are writing about this at all, given that framing. Business owners are told a statistic about this constantly, usually in a leadership context and usually without a source, and whether that statistic means anything is exactly the sort of question this publication exists to check.
Key Takeaway
A systematic review of 62 studies and 14,161 participants reports that "prevalence rates of impostor syndrome varied widely from 9 to 82% largely depending on the screening tool and cutoff used to assess symptoms." It also reports that "no published studies evaluated treatments for this condition."[1] On our own arithmetic, that entire range is reproduced by moving a cutoff 2.26 standard deviations.
The Verdict, Stated First
Five claims, in descending order of confidence.
One. The experience is real and widely reported. Nothing in this article questions that people feel this way, and the review documents it across genders, age groups and many professional settings.
Two. The prevalence figures are a property of the instruments. A ninefold range, attributed by the reviewers themselves to the screening tool and cutoff used.
Three. On our own arithmetic that range requires no disagreement about people at all. It is reproduced by moving a threshold from 1.34 standard deviations above a mean to 0.92 below it.
Four. There was no treatment evidence. The review states plainly that no published studies evaluated treatments, which is a startling gap for something discussed as a condition.
Five. It is not a clinical diagnosis, and the review's own background section describes it as something increasingly presented in media and lay literature as a behavioural health condition, which is a careful way of noting the gap.
Our Grades For These Claims
Applying the scheme from the first article in this series.
Grade A for the review's findings, from an abstract obtained verbatim from a national medical library database and corroborated word for word by the publisher and by an open-access archive.
Grade A for the 1978 citation, confirmed by four independent reference lists.
Grade A for our own arithmetic, which is standard, on an assumption about the distribution that no source states and which we flag at every use.
Grade F for the figure most readers have heard, which we could not trace to any study at all.
Our position: the experience is not in doubt and the numbers attached to it are not measurements of a population, they are consequences of where a line was drawn.
A Note On Method
Everything here is verified to August 2026.
We obtained the 2020 review's abstract verbatim from a national medical library database[1], corroborated word for word by the publisher's record and an open-access archive copy[2], together with passages of its background section[2]. We did not obtain the full review, so we report none of its tables, its list of instruments, or its analyses of predictors.
We did not obtain the 1978 founding paper and report it from four independent reference lists[3].
We did not trace the widely quoted prevalence figure that circulates in business writing to any study, and we report that as a failure to find rather than as a finding.
All arithmetic is ours, rests on an assumption about the shape of the distribution that no source we obtained states, and our first version of one table was wrong, which we describe in the body.
This article is not medical, psychological or clinical advice.
What It Describes
The construct, in the review's own words.
"Impostor syndrome (also known as impostor phenomenon, fraud syndrome, perceived fraudulence, or impostor experience) describes high-achieving individuals who, despite their objective successes, fail to internalize their accomplishments and have persistent self-doubt and fear of being exposed as a fraud or impostor."[2]
And the mechanism it proposes: "People with impostor syndrome struggle with accurately attributing their performance to their actual competence (i.e., they attribute successes to external factors such as luck or receiving help from others and attribute setbacks as evidence of their professional inadequacy)."[2]
Four observations, ours.
The construct has five names in its own definition, which is usually a sign that a field has not settled what it is describing.
The mechanism is an attribution asymmetry: successes go to luck, setbacks go to inadequacy. That is a specific, testable claim about how a person explains outcomes.
It is also the exact mirror of the self-serving attribution the seventy-first article covered, where successes go to skill and failures to circumstance. The two describe opposite errors in the same machinery.
And the definition begins with "high-achieving individuals", which is a selection built into the construct rather than a finding about it.
That last point is worth one more sentence, because it has a consequence for every prevalence figure that follows. If the construct is defined as applying to high achievers, then prevalence depends on who counts as one, and no study can measure that.
It also produces a claim that cannot be tested as stated. The definition requires objective success and persistent doubt about it, so somebody whose doubts are warranted is by definition not an instance, and separating the two requires knowing whether each person is actually competent.
The 1978 Origin
Where it comes from.
Clance, P. R., and Imes, S. A. (1978), The imposter phenomenon in high achieving women: Dynamics and therapeutic intervention, Psychotherapy: Theory, Research and Practice, 15(3), 241–247[3].
The review records the history: "Psychologists Clance and Imes first described impostor phenomenon in 1978, and it came to widespread public attention after Clance's 1985 book. Clance originally identified the syndrome among high-achieving professional women, but more recent research has documented these feelings of inadequacy among men and women, in many professional settings, and among multiple ethnic and racial groups."[2]
We did not obtain the 1978 paper.
Four observations, ours.
The title's subtitle is "Dynamics and therapeutic intervention", which locates it in clinical practice. This originated as a description of something clinicians were seeing, not as a survey of a population.
That matters for the prevalence question and is easy to miss. A construct identified among people who sought therapy is not a random sample of anyone, and generalising it to a population is a later and separate step.
The review notes it reached the public through a 1985 book, which is a route this series has learned to watch, because the seventy-second and seventy-eighth articles both traced figures that travelled through popular writing and changed on the way.
And the expansion beyond the original population is the review's own claim, supported by its citations. The construct has been extended well past where it was first described, which is normal and worth stating.
The 2020 Systematic Review
The source, and it is a good one.
Bravata, D. M., Watts, S. A., Keefer, A. L., Madhusudhan, D. K., Taylor, K. T., Clark, D. M., Nelson, R. S., Cokley, K. O., and Hagg, H. K. (2020), Prevalence, Predictors, and Treatment of Impostor Syndrome: a Systematic Review, Journal of General Internal Medicine, 35(4), 1252–1275, DOI 10.1007/s11606-019-05364-1[1].
Its stated motivation: "Impostor syndrome is increasingly presented in the media and lay literature as a key behavioral health condition impairing professional performance and contributing to burnout. However, there is no published review of the evidence to guide the diagnosis or treatment of patients presenting with impostor syndrome."[1]
Four observations, ours.
The first sentence is a careful piece of scholarly phrasing. "Increasingly presented in the media and lay literature as" describes how something is talked about without endorsing it.
The second sentence states the gap the review exists to fill, and it is a large one. As of the search closing in 2018, nobody had reviewed the evidence at all, despite the term being in general circulation for over thirty years.
The journal is a general internal medicine journal rather than a psychology one, which places the review in the clinical literature and explains its framing around diagnosis and treatment.
And the author list spans a university medical school and three health companies, per the publisher's affiliation record[2], which we note neutrally.
Sixty-Two Studies
The base, verbatim.
"In total, 62 studies of 14,161 participants met the inclusion criteria (half were published in the past 6 years)."[1]
The search covered "Medline, Embase, and PsycINFO (January 1966 to May 2018) and bibliographies of retrieved articles."[1]
Four observations, ours.
62 studies over 52 years of searching, which is a modest literature for a term this widely used.
The average study has about 228 participants, on our own division, which is a normal size for survey research and small for a prevalence estimate.
The parenthesis is the most striking part. Half the literature appeared in the six years before the search closed, so the research followed the popular attention rather than preceding it.
And a search from 1966 for a term coined in 1978 is thorough by construction, which is the correct way to do it and worth noting as a mark of quality.
One implication of the sixty-two figure deserves stating, because it cuts against the impression the topic gives. This is a small literature. Several of the effects this series has covered rest on meta-analyses pooling well over a hundred studies, and the seventy-second article's rested on fifty-five with five thousand participants for a single narrow claim.
Here sixty-two studies are covering prevalence, predictors, comorbidities and treatment across every population studied, which spreads them thin. That is not a criticism of the review, which can only pool what exists.
Nine To Eighty-Two
The finding this article is named after.
"Prevalence rates of impostor syndrome varied widely from 9 to 82% largely depending on the screening tool and cutoff used to assess symptoms and were particularly high among ethnic minority groups."[1]
Four observations, ours.
The reviewers attribute the range themselves, and they attribute it to the instrument, not the population. That is the clause the whole article turns on.
A ninefold range is not a noisy estimate. Nine percent and eighty-two percent are not two attempts at the same number; they are answers to different questions that share a name.
The review also reports that impostor syndrome "was common among both men and women and across a range of age groups (adolescents to late-stage professionals)"[1], which is a claim about breadth rather than prevalence and does not depend on the cutoff in the same way.
And the clause about ethnic minority groups is a separate finding that we note and do not develop, because we did not obtain the underlying analyses and it deserves better treatment than a sentence.
Where That Range Comes From
Putting arithmetic to it. Our own calculation, resting on an assumption no source states: that the underlying feeling is a continuous trait, roughly normally distributed across a population, and that a screening instrument reports the share above some cutoff.
On that assumption, the share screening positive is entirely a function of where the cutoff sits:
At 1.34 standard deviations above the mean: 9.0 percent.
At 1.00: 15.9. At 0.75: 22.7. At 0.50: 30.9. At 0.25: 40.1. At the mean: 50.0.
At 0.25 below: 59.9. At 0.50 below: 69.1. At 0.92 below: 82.1 percent.
Three observations.
The reported range of 9 to 82 percent is reproduced exactly by moving the cutoff across 2.26 standard deviations.
Nothing about the population changes across that table. The same people, measured with different thresholds, yield the entire published range.
And the assumption is ours and is doing real work. If the underlying trait is not continuous, or not roughly normal, the specific figures move, though the general point that a cutoff on a continuous scale determines prevalence would survive any reasonable distribution.
Two ways to check whether the assumption is doing too much, and both point the same way. The reviewers themselves attribute the range to the tool and cutoff, which is the substantive claim; our arithmetic only shows that the claim is quantitatively sufficient to explain the whole range.
And if the underlying distribution were bimodal, with a distinct group of affected people separated from everyone else, cutoffs in the gap would give stable prevalence and the range would be small. The observed range is itself weak evidence against a distinct group, which is our inference and not the review's.
We Got The Table Backwards
Continuing this series' practice, because the error is instructive.
Our first version of the table above printed the cumulative distribution function, being the share below each cutoff, and labelled the column as the share above it. Every interior row was inverted.
Four observations, ours.
The two endpoints were correct anyway, because we had solved for them separately by asking which cutoff produces 9 percent and which produces 82. The middle of the table was wrong and the ends were right.
That is the most dangerous shape an error can take. A table that agrees with the figures you were checking against, and is wrong everywhere else, passes exactly the check most people would run.
We caught it by reading the direction of the trend and finding it ran the wrong way: a higher cutoff was producing a higher prevalence, which cannot be right.
And the general lesson is one this series keeps arriving at from different directions. Check the sign and the direction before checking the numbers, because a plausible-looking table with the correct endpoints will otherwise survive review.
Two further notes on why we publish these rather than quietly fixing them. An article arguing that a literature's numbers cannot all be right has an obligation to show its own working, and a correction is part of the working.
And the specific failure mode is common enough to be worth naming for anyone who builds spreadsheets. Solving for the endpoints separately and then generating the middle by a different route is how you get a table that reconciles at the edges and is wrong throughout, which is a familiar way for a financial model to go bad too.
How Little Movement It Takes
The sensitivity, which is the practically useful part. Ours, same assumption.
Moving the cutoff upward by a quarter of a standard deviation at each step, prevalence falls:
From the mean to 0.25 above: 50.0 to 40.1 percent.
From 0.25 to 0.50: 40.1 to 30.9. From 0.50 to 0.75: 30.9 to 22.7. From 0.75 to 1.00: 22.7 to 15.9. From 1.00 to 1.25: 15.9 to 10.6.
Four observations.
Near the middle of the distribution, a quarter of a standard deviation is worth about ten percentage points of reported prevalence.
The sensitivity falls as the cutoff rises, which means disagreements about strict thresholds matter less than disagreements about lenient ones.
To report a prevalence of 70 percent requires a cutoff 0.52 standard deviations below the population mean. To report 82 percent requires 0.92 below.
And that last point is the one we would put to anyone quoting a high figure. A cutoff below the mean means most people screen positive by construction, which is a decision about an instrument rather than a discovery about people.
What That Does To A Headline Number
The practical consequence. Ours.
In a room of ten people, the number who would screen positive, across the published range:
At 9 percent: 0.9 people. At 20: 2.0. At 40: 4.0. At 60: 6.0. At 82: 8.2.
Four observations.
One person or eight, in the same room, depending entirely on which questionnaire was handed out. That is the practical meaning of a ninefold prevalence range.
Business writing on this topic quotes a single figure, usually a high one, without naming an instrument. A prevalence claim without a named instrument and cutoff is not a claim about a population.
We could not trace the figure most readers will have encountered to any study, and we report that as a failure to find rather than as evidence it does not exist.
And we would apply the same standard to ourselves. We are not offering a prevalence figure either, because on the evidence we obtained nobody is in a position to.
Two things that framing does not license, and both matter. It does not mean the figures were made up: every study in that review presumably applied its instrument correctly and reported what it found.
And it does not mean prevalence is unknowable. It means prevalence is undefined until somebody specifies a threshold and defends it, which is a solvable problem and a different complaint from measurement being impossible.
No Published Studies Evaluated Treatments
The sentence in the abstract we found most surprising.
"No published studies evaluated treatments for this condition."[1]
Four observations, ours.
That is a complete absence, as of a search closing in May 2018, for something being written about as a workplace problem with solutions.
It sits oddly beside the volume of advice in circulation. A great deal is written about overcoming this, and as of that review none of it had been tested in a published study.
The gap is about treatment specifically, not about the phenomenon. The review documents associations at length and reports nothing on whether anything helps.
And things may well have changed since May 2018, which is when the search closed. We did not search for treatment studies published since, and a reader interested in that question should.
Two things that absence does and does not mean, since it is the most quotable sentence in the review and the easiest to misuse. It does not mean nothing helps. Plenty of things may help and none had been tested in a published study the review's search could find.
And it does not mean the general therapeutic literature is silent. The associated conditions the review names have substantial treatment literatures of their own, and someone whose difficulty runs through anxiety or low mood is not in an evidential vacuum, whatever the label attached to it.
The Review's Own Framing
What the authors conclude, verbatim, because it is more measured than either the popular version or a debunking would be.
"Clinicians and employers should be mindful of the prevalence of impostor syndrome among professional populations and take steps to assess for impostor feelings and common comorbidities."[1]
And its own stated limitations: "Studies were heterogeneous; publication bias may be present."[1]
Four observations, ours.
The conclusion recommends assessment rather than any particular response, which follows correctly from having found no treatment evidence.
It addresses clinicians and employers jointly, which is worth noticing: the review treats this as sitting across both domains rather than in one.
The limitations sentence is short and does substantial work. Heterogeneous means the studies were not measuring the same thing the same way, which is the prevalence range restated.
And naming possible publication bias is standard practice and correct. Studies finding no association are less likely to be published, which inflates apparent associations across any literature of this shape.
Not A Diagnosis
A point worth making plainly, and the review's own framing supports it. Ours.
Four observations.
The review's background section describes this as "increasingly presented in the media and lay literature as a key behavioral health condition"[1], which is a description of how it is discussed rather than a statement of what it is.
The review exists precisely because there was no evidence base to guide diagnosis or treatment, which is not the position of an established clinical category.
We would draw the practical conclusion carefully. Recognising the description does not mean you have a condition, and treating a common experience as a disorder is not obviously helpful to anyone.
And the opposite error matters too. Someone whose self-doubt is genuinely interfering with their life should talk to a doctor, and the fact that a construct is loosely measured is no reason to dismiss what they are experiencing.
What The Review Says About Comorbidity
Reported carefully, because it is the part of the abstract with the most weight.
"Impostor syndrome is often comorbid with depression and anxiety and is associated with impaired job performance, job satisfaction, and burnout among various employee populations including clinicians."[1]
Four observations, ours.
These are associations, and the review does not claim a direction. Self-doubt and low mood travel together, and which precedes which is not established here.
The associations with job performance, satisfaction and burnout are the reason employers appear in the conclusion, and they are the commercially relevant part.
We did not obtain the effect sizes for any of these associations, so we report the direction only, and a reader should not infer magnitude from the confidence of the sentence.
And the practical implication we would draw is narrow and cautious. Where these feelings appear alongside depression or anxiety, the appropriate response is professional support rather than a workplace intervention, and the review's own recommendation is assessment for exactly that reason.
One further point on reading associations of this kind, which applies well beyond this topic. The direction of causation is genuinely open here and the popular framing usually assumes one. Self-doubt might contribute to burnout; burnout might produce self-doubt; and a third thing might produce both.
The review does not claim to have separated them, and we would not either. An association between two things that plainly influence each other is weak ground for any intervention aimed at only one of them.
Half In Six Years
An observation about the shape of the literature. Ours.
Four observations.
The review records that half the qualifying studies appeared in the six years before its search closed[1], from a literature spanning four decades.
That means the term became popular first and was studied afterwards, which is the reverse of the usual order and has a predictable consequence.
The consequence is the prevalence range. A rapidly expanding literature using several instruments with different cutoffs will produce incompatible numbers, because nobody agreed on the measure before the work started.
And it explains why the figure in circulation predates the evidence. The number people quote was available before the studies that would have tested it, which is the pattern the sixty-ninth article documented in a different literature.
Is It One Thing?
A question the prevalence range raises and which we can address only partially. Ours.
Four observations.
A construct measured by several instruments with different cutoffs might be one thing measured inconsistently, or several things sharing a name. The prevalence range cannot distinguish those.
The review's own limitations sentence gestures at it: "Studies were heterogeneous"[1], which is a statement about the studies and is compatible with either reading.
There is a reason to suspect the second, and it is in the definition itself. Five names appear in the review's own opening sentence for the construct, and fields that have settled what they are measuring do not usually carry five.
And this series has recorded the same shape repeatedly. The sixty-sixth article found one name covering two distinct mechanisms; the sixty-first found a single label spanning two separate fourfold patterns. A ninefold prevalence range is the kind of evidence that should at least raise the question.
What Would Settle It
The study we would want, ours, since criticising a literature without saying what would improve it is cheap.
Four observations.
Administer several instruments to the same sample. If they are measuring one thing, the people scoring high on one should score high on the others, and the correlations would say so directly.
That study may well exist. We did not obtain the review's full text, which is where a table of instruments and their agreement would appear if it were reported, so this may be a gap in our sourcing rather than in the literature.
The second thing we would want is a defensible cutoff rather than a conventional one, anchored to something outside the questionnaire: an outcome, an impairment, anything that gives the threshold a meaning beyond a score.
And the third is the obvious one the review names by its absence. A trial of anything, since the volume of advice in circulation rests on no published test as of the search date.
What Actually Survives
Our reading, stated directly.
Five statements.
The experience is real and widely reported, across genders, ages and professional settings, and nothing here disputes it.
The prevalence figures are properties of instruments, on the reviewers' own attribution, and range ninefold.
On our own arithmetic that range is 2.26 standard deviations of where a line is drawn, requiring no disagreement about people.
As of a 2018 search, no published study had evaluated any treatment, despite the volume of advice in circulation.
And the review's own recommendation is to assess rather than to act, which is the correct conclusion from an evidence base with associations and no trials.
If You Recognise Yourself In This
The section a business owner came for. Ours, and not psychological advice.
Four observations.
The description is broad enough that a great many capable people will recognise it, and recognising a description is not a diagnosis. That is the first thing the arithmetic above should do for a reader.
The prevalence figure you have been quoted was probably chosen by whoever wrote the article, from a range spanning nine to eighty-two percent. It is not a measurement of how many people around you feel this way, and it cannot tell you whether you are unusual.
We would also note what the construct's own mechanism implies, since it is testable in a way feelings are not. The claimed error is attributional: successes to luck, setbacks to inadequacy, and that is something a written record can check.
And if this is affecting your health or your life rather than merely your confidence, the evidence base offers you nothing and a doctor may, which is a genuine conclusion from a review that found no treatment studies.
One more thing the arithmetic offers, and it is the reason we thought this article worth writing for this audience. Being told that most people feel this way is meant as reassurance and can function as the opposite. It says the feeling is ordinary and it also says nothing about you.
The honest version is less tidy and more useful. Nobody knows how common this is, the range spans ninefold, and the question of whether you are unusual cannot be answered from the literature. Which means it is not a question worth spending effort on, and the attributional check below is.
The Reframe We Would Offer
Ours, and offered as reasoning rather than as a finding.
Four observations.
The seventy-fourth article established that most business decisions occur in environments that cannot supply enough feedback to learn from, and the seventy-first that introspection cannot detect your own accuracy.
Put together, those say something specific about a founder's uncertainty. If you cannot tell whether your judgment is good, that is frequently a correct assessment of your evidence rather than a distortion of it.
That is not the same as saying the feeling is always warranted, and we would not claim that. It is saying that uncertainty about your own competence is the appropriate state in a wicked environment, and confidence there would be the thing requiring explanation.
And it points at the same procedural remedy this series keeps arriving at. Not more confidence, but a record: what you predicted, when, and what happened, which converts an unanswerable question about your worth into an answerable one about your calls.
Two limits on that reframe, because it could be read as dismissing something real. It applies to uncertainty about judgment in genuinely uncertain domains, which is a specific claim, and not to every form of self-doubt a person might have.
And it is our own synthesis of two findings rather than anything the impostor literature says. Neither the review nor the 1978 paper makes this argument, and we present it as reasoning a reader can accept or reject rather than as evidence.
A Different Question Worth Asking
The practical version, ours and untested.
Four observations.
The construct's mechanism is an attribution asymmetry, and asymmetries are measurable. Write down, for your last ten significant outcomes, what you think caused each one.
If the good ones are attributed to conditions and the bad ones to yourself, that is the pattern described. If the attributions are mixed, the description does not fit however the feeling presents.
That exercise has an advantage the feeling does not. It produces evidence you can look at next year, and the seventy-first article's finding is that behavioural evidence about yourself is the only kind that works.
And it can come out either way, which is what makes it worth doing. Someone who finds their attributions balanced has learned something; someone who finds them skewed has too, and neither answer requires a prevalence figure.
One caution on the exercise, which the sixty-sixth article's arithmetic supplies. Ten outcomes is a very small sample, and a pattern in ten cases could easily be chance. It is a prompt for attention rather than a measurement, and reading it as the latter would repeat the error this whole article is about.
What Not To Conclude
Ours, and this matters more than usual here.
Four points.
Do not conclude the experience is not real. Nothing in this article questions that, and a measurement problem in a literature says nothing about what anyone feels.
Do not conclude it is universal either. The reported range starts at nine percent, and the high figures require cutoffs below the population mean.
Do not treat the label as a diagnosis, in yourself or in anyone who works for you. It is not a clinical category, and the review exists because there was no evidence base to guide diagnosis.
And do not use a measurement critique to dismiss someone's distress. The review reports associations with depression, anxiety and burnout, and those are matters for a professional rather than for an argument about instruments.
If You Employ People
The organisational application, ours, untested, and nothing here bears on the lawfulness of any employment practice.
Four points.
The review's conclusion addresses employers directly and recommends being mindful and taking steps to assess, which is deliberately modest and stops short of prescribing a programme.
Given that no treatment studies existed as of the search, we would be sceptical of any workshop or programme sold as addressing this, on the same grounds as the seventy-fifth article's scepticism about a pricing tactic.
What the evidence does support is unglamorous and free. The associations run through job satisfaction and burnout, and those have their own literatures with actual interventions in them.
And we would keep the sixty-fifth and seventy-first articles' caution in view. An employer assessing staff for a psychological construct with a ninefold prevalence range is measuring their instrument, and the results should not follow anyone into a personnel file.
Two safer things an employer can do that require no construct at all, ours. Make the basis for decisions explicit, since a person who cannot see why they were promoted has no evidence to weigh against their own doubt, and the seventy-first article's finding is that behavioural evidence is the only kind that works.
And notice who is asked to justify themselves and who is not. That is observable, costs nothing to record, and does not require any assessment of anyone's inner state.
Two Spellings, One Literature
The bibliographic entry, since this series keeps a count.
The 1978 founding paper's title spells it "imposter". The 2020 review spells it "impostor" throughout, including in its own title[1][3]. A published commentary on that review spells the review's title with "Imposter"[4].
Three observations, ours.
Both spellings are acceptable English and neither is an error. The problem is search: a literature split across two spellings is harder to find completely, which is why the review's own method notes it used both terms.
The commentary misspelling the title of the paper it is commenting on is the ordinary kind of slip this series has now recorded two dozen times. It changes nothing and it would defeat an exact-title search.
That brings the running count of bibliographic variants across this series to twenty-five.
One observation about the pattern at this count, ours. The variants cluster in the citing literature rather than in the papers themselves: wrong pages, shifted volumes, altered author orders, misspelled titles, all in reference lists compiled by people working from other reference lists.
Which suggests where the checking effort belongs. The published papers are largely accurate about themselves and the apparatus around them is not, so a reader who goes to the source is buying more reliability than the extra minute suggests.
What To Do
Ask which instrument produced any prevalence figure. The published range is 9 to 82 percent, attributed by the reviewers to the screening tool and cutoff used.
Treat a figure above about 70 percent with particular suspicion. On our own arithmetic that requires a cutoff below the population mean, which makes most people positive by construction.
Do not treat recognition as diagnosis. The description is broad, it is not a clinical category, and the review was written because no evidence base existed to guide diagnosis.
Check the attribution pattern rather than the feeling. Write down what you think caused your last ten significant outcomes, because the claimed mechanism is measurable in a way the feeling is not.
Be sceptical of programmes sold to address this. As of a search closing in May 2018, no published study had evaluated any treatment.
Note that the evidence may have moved. That search closed years ago, and a reader interested in treatment specifically should look for work published since.
Separate the measurement question from the person. A loose construct is no reason to dismiss anyone's distress, and the review reports associations with depression, anxiety and burnout.
Where it is affecting health or life rather than confidence, see a doctor. That is a genuine conclusion from a review that found associations and no trials.
The Limits Of This Analysis
Several caveats matter, and the framing ones come first. This article is not medical, psychological or clinical advice, contains no diagnosis or assessment of any reader, and nothing here bears on the lawfulness of any employment practice; the applications are our own reasoning and untested. Everything is verified to August 2026. We did not obtain the full 2020 review, only its abstract, obtained verbatim from a national medical library database and corroborated by the publisher and an open-access archive, plus passages of its background section. That means we report none of its tables, none of the individual instruments it identified, none of its analyses of predictors, and no effect sizes for any of the associations it describes. The finding about ethnic minority groups is noted and not developed, because we did not obtain the underlying analyses and it deserves fuller treatment than we can give it. We did not obtain the 1978 founding paper and report it from reference lists. We could not trace the prevalence figure most commonly quoted in business writing to any study at all, and report that as a failure to find rather than as evidence of anything. All arithmetic is ours and rests on an assumption no source states: that the underlying feeling is continuous and roughly normally distributed and that instruments report the share above a cutoff. If the trait is not distributed that way the specific figures change, though the general point about cutoffs would survive. Our first version of the central table was wrong, printing the cumulative distribution and labelling it as its complement, with correct endpoints and every interior row inverted; we describe the error and its correction in the body. And the review's search closed in May 2018, so its finding that no treatment studies existed describes the literature as of that date and not today, which is the single most likely respect in which this article is out of date.
Frequently Asked Questions
What does the research actually establish?
Why is the prevalence range so wide?
Is the figure I have been quoted reliable?
Is it a clinical diagnosis?
Does anything help?
What can I actually do with this?
Does this mean the feeling is not real?
References
- Bravata, D. M., Watts, S. A., Keefer, A. L., Madhusudhan, D. K., Taylor, K. T., Clark, D. M., Nelson, R. S., Cokley, K. O., & Hagg, H. K. (2020). Prevalence, Predictors, and Treatment of Impostor Syndrome: a Systematic Review. Journal of General Internal Medicine, 35(4), 1252–1275, DOI 10.1007/s11606-019-05364-1, epub 17 December 2019. National medical library database record reproducing the abstract: on impostor syndrome being increasingly presented in the media and lay literature as a key behavioral health condition impairing professional performance and contributing to burnout, with no published review of the evidence to guide diagnosis or treatment; on the purpose being to evaluate the evidence on prevalence, predictors, comorbidities and treatment; on data sources being Medline, Embase and PsycINFO from January 1966 to May 2018 plus bibliographies of retrieved articles; on 62 studies of 14,161 participants meeting the inclusion criteria, half published in the past six years; on prevalence rates varying widely from 9 to 82 percent largely depending on the screening tool and cutoff used to assess symptoms, and being particularly high among ethnic minority groups; on impostor syndrome being common among both men and women and across a range of age groups from adolescents to late-stage professionals; on it often being comorbid with depression and anxiety and associated with impaired job performance, job satisfaction and burnout among various employee populations including clinicians; on no published studies having evaluated treatments for this condition; on the limitations that studies were heterogeneous and publication bias may be present; and on the conclusion that clinicians and employers should be mindful of the prevalence among professional populations and take steps to assess for impostor feelings and common comorbidities. Note: a national medical library database record. We obtained the abstract verbatim and not the review, so no tables, instruments, predictors or effect sizes are reported here. The search closed in May 2018 and the treatment finding describes the literature as of that date. pubmed.ncbi.nlm.nih.gov
- Publisher record and open-access archive copy of the same review, corroborating the abstract word for word and reproducing passages of its background section: that impostor syndrome, also known as impostor phenomenon, fraud syndrome, perceived fraudulence or impostor experience, describes high-achieving individuals who despite their objective successes fail to internalize their accomplishments and have persistent self-doubt and fear of being exposed as a fraud or impostor; that people with impostor syndrome struggle with accurately attributing their performance to their actual competence, attributing successes to external factors such as luck or receiving help from others and attributing setbacks as evidence of their professional inadequacy; that psychologists Clance and Imes first described impostor phenomenon in 1978 and it came to widespread public attention after Clance's 1985 book; that Clance originally identified the syndrome among high-achieving professional women but more recent research has documented these feelings among men and women, in many professional settings and among multiple ethnic and racial groups; and that the authors' study is the first published synthesis of the peer-reviewed evidence. The same records give the author affiliations as a university school of medicine's centre for primary care and outcomes research, three health companies, and a second university. Note: the publisher's record plus an open-access archive copy, used to corroborate the abstract independently and to obtain the background passages. We did not obtain the full review. link.springer.com
- Four independent reference lists confirming Clance, P. R., and Imes, S. A. (1978), The imposter phenomenon in high achieving women: Dynamics and therapeutic intervention, Psychotherapy: Theory, Research and Practice, 15(3), 241–247; together with Clance, P. R. (1985), The Impostor Phenomenon: When Success Makes You Feel Like a Fake, Atlanta: Peachtree Publishers; and Harvey, J. C., and Katz, C. (1985), If I'm So Successful Why Do I Feel Like a Fake?, New York: St. Martin's Press. Note: reference lists; citations only. We obtained none of the works named, including the 1978 paper that defines the construct. Recorded also as the source of a spelling variant: the 1978 title uses "imposter" where the 2020 review uses "impostor" throughout. clinicaltrials.gov
- Repository record for a published commentary on the 2020 review, which renders the review's title with the spelling Imposter Syndrome where the published title uses Impostor Syndrome, and which carries a reference list confirming the review's citation as Journal of General Internal Medicine, 2020, 35(4), 1252–1275, alongside the 1978 Clance and Imes paper and further works on impostorism in specific populations. Note: a repository record for a commentary. Recorded solely as a bibliographic variant: the commentary misspells the title of the paper it comments on, which would defeat an exact-title search and changes nothing else. researchgate.net
This article is not medical, psychological or clinical advice and contains no diagnosis or assessment of any reader; nothing here bears on the lawfulness of any employment practice. The 2020 review was obtained as an abstract and background passages only, so no tables, instruments, predictors or effect sizes are reported. The commonly quoted prevalence figure could not be traced to any study. All arithmetic is the authors' own and rests on a distributional assumption no source states; the first version of the central table was wrong and the error is described in the body. The review's search closed in May 2018, so its finding of no treatment studies describes the literature as of that date. If self-doubt is affecting your health or your life, that is worth discussing with a doctor.