Abstract
The question this article answers is: why does the quiet, easily wounded colleague end up doing more damage than the visibly difficult one, and why does reassuring them change nothing?
Vulnerable narcissism is the least researched and most confidently described construct in this literature, and the gap between those two facts is the reason for this article. It appraises what the trait predicts in ordinary working behaviour, and separates that from what the popular treatment of covert narcissism asserts without evidence.
It adopts the distinction that has organised the field since the middle of the last decade, treating narcissistic grandiosity and narcissistic vulnerability as two higher-order dimensions sharing a common antagonistic core (Crowe, Lynam, Campbell & Miller, 2019; Miller, Back, Lynam & Wright, 2021), and separates the trait from ordinary anxiety, from introversion, and from clinical diagnosis.
Its central contribution is to apply a recent three-study programme in the workplace to a practical question. Across supervisor samples in Germany and the United Kingdom, leaders higher in vulnerable narcissism reported more hostile treatment of the people reporting to them; the pattern did not replicate for grandiose narcissism, and the route ran through a tendency to attribute failure to the self and the shame that followed (Braun, Schyns, Zheng & Lord, 2025). This contrasts with evidence that managers, as a population, score lower on this form of the trait than people in general, while it predicts worse work performance when present (Dåderman & Kajonius, 2024).
The mechanism is traced through converging designs: a pooled analysis of 437 studies establishing that the aggression link is not confined to the loud form and is far stronger under provocation (Kjærvik & Bushman, 2021), a diary study finding daily shame elevated in people higher on the trait (Di Sarno, Zimmermann, Madeddu, Casini & Di Pierro, 2020), and an experiment finding that only people high on it responded to disappointing feedback with shame (Freis, Brown, Carroll & Arkin, 2015).
A dedicated late chapter then sets against the popular account three findings that correct it: that most of what the standard measure captures is ordinary emotional instability (Miller et al., 2018), that the commonest short measure splits into a sensitive half and a self-centred half of which only the second behaves like a dark trait (Starlinger, Voracek & Tran, 2022), and that vulnerability rises sharply at the very top of grandiosity rather than sitting cleanly opposite it (Jauk et al., 2022).
That chapter also sets out the five kinds of person most often wrongly labelled with this trait, because here, more than anywhere else in this series, the diagnostic error runs in the direction of accusing the anxious.
The practical section offers four moves and three refusals. The conclusion is that this trait is better evidenced at the level of mechanism than at the level of what to do about it, and that the honest version of the advice protects people downstream rather than trying to fix the person upstream.
Keywords: vulnerable narcissism; covert narcissism; abusive supervision; shame; attribution of failure; workplace behaviour; personality assessment; organisational psychology.
PART I. WHAT ACTUALLY HAPPENS
Chapter 1: The One Who Never Quite Says It
1.1 Three Rooms
A senior analyst has been at the firm for four years and has not once raised her voice in a meeting, which is part of why nobody can work out what is happening. When her recommendation is passed over, she agrees pleasantly that the other approach is probably better. Then, over the following fortnight, the work she owes the person whose approach was chosen arrives late, arrives thin, and arrives with a note explaining that she did not want to overstep.
A department head takes the blame for a lost bid in front of his own director, which everybody present agrees is admirable, and which two of his team later describe as the worst afternoon of their year. He said the failure was his. He then spent the rest of the day making it very clear, in a series of individually reasonable conversations, whose it actually was.
A programme manager sends an email at eleven at night, copying two people who didn't need to be copied, correcting a small factual error in a document the recipient wrote nine days earlier. There is nothing in it you could take to anybody. The recipient reads it four times, looking for what is wrong with it, cannot find it, and stops volunteering for that programme.
What all three of those people have in common, and what makes them so hard to describe to anybody who was not in the room, is that each behaved in a way which would look entirely reasonable if you saw it once, which is only recognisable as a pattern after the fourth or fifth occurrence, and which by then has produced a set of colleagues who have quietly adjusted their own behaviour in order to avoid a conversation that none of them could explain to a third party if they were asked to. The adjustment is the cost.
1.2 What the Three Have in Common
None of those three people did anything you could put in a grievance, and that is not incidental to the pattern; it is the pattern, because every one of those behaviours was chosen for its deniability and would stop working the moment it became explicit enough to name. What they share is not aggression in the ordinary sense of the word. It tends to be a persistent, low-grade sense that something is owed and has not been given, together with a conviction that saying so out loud would be humiliating.
The result is a form of score-keeping that never gets settled because nobody states it. The popular literature calls this the covert narcissist. The research literature calls it vulnerable narcissism, and the two bodies of work overlap far less than the shared vocabulary suggests.
What makes the pattern expensive is not any single incident, since every incident is small enough to absorb, but the fact that incidents keep arriving and each one costs the person on the receiving end a small amount of certainty about where they stand. Over a year that can accumulate into something that looks like a difficult team rather than a difficult person, and it is usually described that way in the exit interview.
This article focuses on the second. The popular account rests largely on the testimony of people who have been badly hurt, which is a real thing and deserves its own treatment somewhere else. This one rests on what has been observed in working adults, which is a narrower claim and a more useful one for anybody who has to sit next to the person in question on Monday morning.
1.3 Why Reaching for the Label Makes It Worse
There is a reason to be careful here that does not apply with the same force to the loud form of the trait, and it is worth stating before any of the evidence, because a reader who takes only one thing from this article should take this one. What I have described above is also what ordinary anxiety looks like from the outside. It is also what somebody looks like when a person with more power than they do manages them badly.
The behaviour on its own does not tell those three apart. Somebody who withdraws after criticism, who takes correction hard, who apologises more than the situation calls for, and who is difficult to read, has told you almost nothing about which of the three is happening to them. Reaching for the label then amounts to a decision to stop investigating, dressed up as insight.
I come back to this at length in Chapter 7, which sets out the five people most commonly mislabelled here. It sits late in the article deliberately, because a caution delivered before the reader has been given a reason to care gets skipped.
Chapter 2: The Failure They Take Personally
2.1 The Bid Nobody Won
Picture the situation the researchers put in front of their participants, because it is deliberately ordinary and does all the work. You lead a sales and marketing department of five teams; one of those teams has just failed to win a significant piece of business, and an email has arrived from the unhappy client.
In one version, the email makes clear that the failure was yours. In another, it makes clear that the failure belonged to the team leader who reports to you. In a third, it says nothing about whose fault anything was, and leaves you to work that out for yourself.
The question the researchers were asking is what you would then do to the team leader sitting downstream of you. The answer depended almost entirely on which of those three emails you had received, and on where you sat on one particular set of personality questions.
2.2 The Route Nobody Expects
Almost everybody assumes that the manager who lashes out is the manager who refuses the blame. It is an assumption with a certain moral tidiness to it, since it lets the aggression and the arrogance be one failing rather than two. The evidence points the other way. What one research team proposed, and then tested three times, is that the leaders who treat their people worst are the ones who take the failure inward first, decide that it says something about them rather than about the situation, feel the shame that follows from that conclusion, and then relieve the shame by turning it outward onto somebody who cannot easily object (Braun, Schyns, Zheng & Lord, 2025).
Shame is the specific emotion doing the work here, and it is worth separating from guilt, because the two get used as synonyms and behave completely differently. Guilt is about what you did, and it points to repairing it. Shame is about the kind of person the thing proves you are, and it points at getting out of the room.
There is not much anybody can do with shame except ignore it, deny responsibility for it, avoid everybody who witnessed it, or attack whoever seems to be holding it up in front of them. Only the last of those four is available to somebody who has to chair the same meeting again on Thursday.
Which produces a sequence that is very fast, mostly invisible, and almost impossible to interrupt from outside once it has started, running from a piece of ordinary bad news, through a private conclusion about what that bad news proves, into an emotional state that the person experiencing it would not describe out loud to anybody at work, and out again as a series of small hostile acts directed at whoever in the immediate vicinity is least able to make a complaint about them. None of the four steps looks like the one before it.
2.3 What the Three Studies Actually Did
The programme ran three times across two countries with three different designs, which matters, because a single study of this kind is a hypothesis and three converging ones are closer to a finding. The first surveyed 320 supervisors in Germany at two points about ten days apart, asking them first about the trait and about their general tendency to explain workplace failures by reference to themselves, and then about how they had actually been treating their people. Leaders higher on the quiet form reported treating their people worse, and their tendency to explain failure by reference to themselves carried part of that relationship.
The second moved 326 German supervisors from measurement to experiment, using the lost-bid scenario above, so that the researchers controlled who was to blame rather than merely asking. The link between the trait and the intention to treat a subordinate badly was strongest when blame was left open. That condition also most closely reflects ordinary working life, where almost nobody sends an email apportioning fault.
The third repeated the experiment with 292 supervisors in the United Kingdom and replaced the imagined situation with a real one, asking each participant to recall a recent mistake involving one of their own people and describe it in detail before answering anything else. This is the one to read first, because it produced the cleanest result. The link between the trait and hostile treatment appeared when the participant was left to work out the blame for themselves, and vanished when the instruction told them the failure was theirs.
Across all three, the researchers repeated every analysis for grandiose narcissism, the loud form covered in the previous article in this series, and the pattern did not hold. Whatever is going on here, it is not narcissism in general.
2.4 What This Does Not Show
Four limits, stated plainly, because this is early work and overstating it now would be expensive later. The hostile treatment was reported by the leaders themselves rather than by the people on the receiving end of it, which is a deliberate choice, since the researchers were studying what goes on inside the leader rather than what the team sees. This means we do not know what any of this looked like from below, and no study has yet answered that.
The second study did not cleanly block the route it was designed to block, reducing it rather than removing it, and the third supported the first part of the proposed route while failing to support the second part as the researchers predicted. They say so themselves, at length, which is a good sign about the work and a reason to hold the mechanism loosely. The trait and its loud counterpart moved closely together in these particular groups of supervisors, which is not what usually happens and not what happened among the managers in Chapter 5. Some caution about how these supervisors were recruited is reasonable.
And two of the three used an imagined situation. People can be poor judges of what they would do to a colleague in a scenario they read on a screen, though the third study’s use of a real remembered event goes some way toward answering that.
2.5 Why Reassurance Is the Wrong Lever
Here is the part that changes what a reader does, and it follows from the shape of the mechanism rather than from any tested intervention, which I flag now and flag again in Chapter 8. When somebody says this is all my fault in a tone that invites contradiction, the natural and generous response is to contradict them. Most of us reach for some version of don’t be silly, it was nobody’s fault, which is kind, cheap, and aimed at the wrong target.
Nothing here suggests that these leaders are short of reassurance. It suggests they are running a private accounting exercise in which a failure has already been converted into a verdict about themselves, and a colleague’s warm generality does not touch a verdict, because a verdict answers only to specifics.
What would touch it, on this evidence, is a specific and public separation of the failure from the person, and the reason to do it publicly is that the shame is about being seen, so a correction delivered only in private leaves the audience intact. That has never been tested. I am reasoning ahead of the data when I say it, which I would rather admit here than let the sentence borrow authority from the citation beside it.
Chapter 3: Who Is Actually Keeping Score
3.1 Aggression Is Not a Speciality of the Loud Kind
The most useful number in this article comes from a meta-analysis, a study that pools the results of many earlier ones to see what survives across them. This one pooled 437 separate studies covering 123,043 people (Kjærvik & Bushman, 2021).
Narcissism was related to aggression across the board, and to violence nearly as much as to milder forms, which is itself worth knowing. The finding that belongs in this article is narrower and more surprising. All three of the dimensions the researchers examined, including the quiet vulnerable one, were related to aggression.
That is not what the popular account predicts. In the popular account, the quiet one suffers and the loud one harms, which makes a tidy story and survives contact with 437 studies.
3.2 Provocation Is the Switch
The second result in that same analysis is the one to carry into a working week. The link between narcissism and aggression was roughly twice as strong when the person had been provoked as when they had not, though it did not disappear entirely in the absence of provocation.
Provocation, in the studies pooled here, does not have to be dramatic. Being ignored counts. Being insulted counts. So does being left out, and the previous article in this series set out at length how readily people high in the loud form read an ambiguous situation as exclusion.
The practical picture is therefore not a person who is dangerous at rest. It is a person whose response to a fairly ordinary workplace slight runs larger than the slight, and who is for that reason unusually sensitive to how the slight was delivered.
That distinction is worth holding on to, because a great deal of the popular writing on this subject treats the person as a constant hazard whose behaviour is a property of who they are rather than of what has just happened to them, and the evidence pooled across those hundreds of studies says something considerably more specific and considerably more actionable, which is that the behaviour is a response, that it is roughly twice as likely to appear once something has been read as a slight, and that a very large share of what gets read as a slight at work was never intended as one by anybody. The trigger might be the part you can change.
3.3 What Provocation Looks Like at Work
Almost nobody provokes a difficult colleague deliberately, and that is exactly why this matters, because the commonest provocations in an office are administrative rather than personal and are usually nobody’s decision in particular.
The meeting invitation that goes to seven people and not to the eighth. The decision taken in a corridor and announced in a room. The document that circulates with somebody’s contribution folded silently into a section under another name. The reply that comes back in four words when the message took twenty minutes to write.
Every one of those is ambiguous by construction; none of them was intended as a message, and each is the sort of thing that the evidence in this chapter suggests may be read as one. Nobody sent it. It arrived anyway.
This is the point at which most practical advice on difficult colleagues quietly stops being useful, because the advice assumes a provocation that somebody chose to deliver and can therefore choose to withhold, when what the evidence describes is a steady background of small ambiguities generated by the ordinary administrative machinery of an organisation that has no view about anybody. You cannot stop scheduling meetings. What you can change is how much interpretive work a colleague is left to do about why they were not in one.
Consider what a single ordinary Tuesday supplies: a calendar invitation issued by somebody’s assistant on the basis of who happened to be free, a decision reached in a lift by two people who then forgot to say so, a document circulated in a version that silently dropped one contributor’s comments during a merge, a reply written in four words by a person who was standing in an airport, and a piece of praise delivered to a team in general rather than to anybody in particular. Not one of those was about anyone.
A daily diary study, meaning a design in which people report on their day every day for a period rather than recalling it afterwards, found that those higher on the quiet form reported more shame in ordinary daily life than other people did (Di Sarno, Zimmermann, Madeddu, Casini & Di Pierro, 2020). The raw material for the mechanism in Chapter 2 is therefore not scarce. Most days, a working environment generates it without knowing it.
PART II. WHY IT RUNS THIS WAY
Chapter 4: Two Things Wearing One Label
4.1 Almost Unrelated, and Then Not
Ask a room of managers to describe a narcissist, and they will describe the loud one, and if you then tell them that the research treats the quiet one as a version of the same thing, they will reasonably want to know in what sense the two are related at all. It is a fair question, and the answer is awkward. When 344 managers answered questions about both forms, how somebody scored on one told you next to nothing about how they scored on the other (Dåderman & Kajonius, 2024). Two patterns sharing a name, sitting at right angles to one another, and predicting different outcomes is not a comfortable position for a field to be in, and the researchers working on this are not comfortable with it either.
The two do meet in one place, and it is the place the popular account has long insisted on, which makes it worth being exact about what the evidence supports. Pooling the individual responses of more than eighteen thousand people across forty-five separate groups, a preregistered analysis found that the relationship between the two is not the same across the range (Jauk et al., 2022).
Among people at ordinary levels of the loud form, the two sit apart. Among people at the extreme top, vulnerability rises alongside grandiosity rather than falling away from it. So the folk claim that underneath the confident one there is a fragile one turns out to be wrong about almost everybody and right about a small group at the far end, which is exactly the kind of result that gets flattened in the retelling.
The authors are careful about what this means, and I will be too. The reading they propose, that this reflects passing states leaking into a lasting score, is an interpretation rather than a demonstration, and we do not know yet whether it is the right one.
4.2 The Part They Share
What the two genuinely have in common is antagonism, meaning a settled disposition to treat other people as obstacles or instruments rather than as parties with claims of their own, and the field has increasingly organised itself around that shared core (Crowe, Lynam, Campbell & Miller, 2019; Miller, Back, Lynam & Wright, 2021). That is the piece worth carrying, because it explains why an article about the quiet form is not an article about a quiet person. Quietness may be a strategy, or in many cases simply a temperament. The antagonism is the part that produces the behaviour anybody else has to deal with.
It also explains why the earlier article in this series, on the single trait running underneath narcissism, psychopathy and manipulation alike, does not need rewriting to accommodate this one. That shared core is exactly what the earlier article covered.
Chapter 5: What It Costs the Organisation
5.1 Less Common in Managers, Not Absent
Start with the finding that most readers will get wrong, because it changes where they should be looking. When those 344 managers were compared with people drawn from the general population, they came out higher on the loud form and lower on the quiet one (Dåderman & Kajonius, 2024). Whatever selection and promotion are doing in organisations, they are not systematically pulling thin-skinned people upward. If anything, they appear to be doing the opposite.
This relocates the problem entirely. If you have spent this article picturing your boss, the probability is against you, and the person the description fits is more likely to be a peer, a colleague at the same level in another function, or somebody who reports to you.
There is a reasonable argument for why that might be, and it is worth stating as an argument rather than as a finding, since nobody has tested it. The behaviour that gets somebody promoted in most organisations involves visible confidence in front of an audience who do not know them well, a willingness to claim credit early and in public, and an appetite for the kind of exposure that a person who takes an offhand remark home with them at the end of the day is likely to find intolerable rather than energising.
Which would mean the promotion process is screening for comfort in a room rather than for character, and that anybody hoping the organisation will filter this pattern out on the way up has misunderstood what the filter is actually made of. Nobody has tested that either. That is a statement about averages across one group of Swedish managers, and it says nothing whatever about any particular organisation, let alone any particular person. I make that caution properly in Chapter 7.
5.2 The Trait That Predicted Nothing, and the One That Did
Inside that same group, the researchers asked what each form predicted about three different kinds of work performance: the job as written, the discretionary contribution around it, and the behaviour that actively damages the organisation. The loud form predicted none of the three. Not the good ones, and not the bad one either, which is a striking result in a literature that has spent two decades on it, and which sits alongside the finding from the previous article in this series that grandiose narcissism predicts who gets chosen far better than it predicts who turns out to be any good.
The quiet form predicted two of the three, in the directions you would expect: more of the damaging behaviour, and less of the job as written. Conscientiousness, the dullest quality in personality psychology, was the strongest predictor of doing the job you were hired to do. That result keeps appearing in this field and keeps failing to interest anybody.
One qualification, and it is a real one. This was a single group of managers answering questions about themselves, including questions about their own damaging behaviour, and self-report about counterproductive work behaviour, meaning conduct that harms the organisation or the people in it, is not the strongest evidence available on any question.
5.3 What the People Reporting to Them Notice
The programme in Chapter 2 asked leaders about themselves. A separate line of work asks the people underneath them, which is the harder and more useful question, and it is thinner than you would hope. One study followed leader behaviour of this kind through the early Covid period and found that where followers reported it, they subsequently reported more irritation themselves (Schyns, Gauglitz, Gilmore & Nieberle, 2023). Irritation is a modest outcome, and that could fairly be called a modest study, and I am reporting it as such deliberately.
A more recent study asked followers to rate specific sides of their leader’s narcissism and related those ratings to followers’ engagement and exhaustion (Gauglitz, Held, Schütz, Schütz et al., 2026). Its headline result concerns the loud form, where the hostile, rivalrous side was associated with lower follower engagement, and that association grew stronger the longer the two had worked together. I have the abstract and the author list for that study, not the full results, so I am not going to characterise what it found about the quiet side beyond one result that is stated plainly and is worth knowing. None of the sides of narcissism it examined was associated with follower exhaustion.
5.4 The Cost Is Withdrawal, Not Exhaustion
That last result is the useful one, and it points at the shape of the damage rather than its size.
Working near this pattern does not appear to burn people out in the way working for a genuinely abusive manager does, and the honest reading is that the cost lands somewhere quieter. People stop volunteering. They route around the person. They put things in writing that would have taken a minute to say, and they decide a particular piece of work isn't worth the conversation it would require.
None of that shows up in an engagement survey as distress. All of it shows up eventually as a team that is slower than it should be for reasons nobody can name, which is precisely why this pattern survives so long in organisations that are otherwise reasonably well run.
The signature to look for is not complaints, because complaints require somebody to have decided that a specific thing was bad enough to name, and nothing here is ever specific enough for that. It is the disappearance of small voluntary acts: the second draft nobody offers, the correction somebody notices and does not mention, the piece of work done to the specification rather than to the purpose. Those are hard to count and easy to lose, and by the time they are missed, the person who stopped offering them has usually stopped noticing that they stopped.
The reason this matters commercially, rather than merely being unpleasant, is that almost everything an organisation actually depends on in a difficult quarter sits in exactly that category of small voluntary act, since the work that is written down gets done under any conditions at all, and the work that is not written down is what separates a team which recovers from a bad month from a team which merely survives one. That is the part that goes first.
I am extending the evidence when I describe it that way, and the extension is mine rather than the studies’. The research supports that the follower outcome here is engagement rather than exhaustion; the account of how that shows up in a working week has never been tested and is a practitioner’s reading of it.
PART III. BEFORE YOU USE ANY OF THIS
Chapter 6: Half of It Is Not a Dark Trait at All
6.1 Ten Questions, Two Different People
Most of the research in this article rests on a ten-question survey, and a large study set out to check whether those ten questions really capture one thing, as everyone had assumed since it was published. They are not. Working with two large groups totalling more than 2,400 people, and testing the structure on one before confirming it on the other, the researchers found that the ten split into two halves that have surprisingly little to do with each other (Starlinger, Voracek & Tran, 2022).
The first half is sensitivity to being judged. It picks up the person who is easily wounded by criticism, who is preoccupied with what others think of them, and who takes an offhand remark home with them at the end of the day.
The second half is self-centredness. It describes the person whose attention constantly returns to their own concerns, and who finds other people’s difficulties less interesting than their own. Most of us know somebody who is very much the first without being remotely the second, and that is the whole practical point of the finding.
6.2 Which Half Does the Damage
The researchers then tested both halves against the dark factor of personality, meaning the common element running underneath the socially aversive traits, which is the subject of the fourth article in this series. Only the self-centred half loaded onto it. The sensitivity half did not, and on this evidence it should not be treated as a dark trait at all.
Sit with that, because it undoes a great deal of what circulates under the covert-narcissist heading. Thin skin is not the problem. Thin skin can be unpleasant to be around and is harder still on the person who has it, and it is not part of the antagonistic core that makes anybody a hazard to work with.
The behaviour this article describes may come from a combination. The sensitivity supplies the injury, and the self-centredness supplies the reason the injury becomes somebody else’s problem.
6.3 Why This Matters More Here Than Anywhere Else in the Series
Every article in this series warns about mislabelling. In the others it is a boundary question. Here it is closer to a structural problem with the popular vocabulary itself.
The term covert narcissist, as it is used outside the research, describes the sensitive half almost entirely, and attaches to it the moral weight that belongs to the other one. Those two halves have surprisingly little to do with each other. So the term as commonly used is not a strong description of a rare person; it is a weak description of a common one, carrying a heavy accusation.
I would go further, and say that the popular vocabulary here has been assembled backwards, because it starts from the experience of somebody who has been badly treated, works back to a set of observable behaviours which that person can recognise, and then attaches to those behaviours a name which asserts a motive that the behaviours themselves do not establish, with the result that the label travels perfectly well between people who have nothing in common except that somebody found them difficult. That describes a feeling, not a person. That is worth knowing before you use any of this on a colleague, and it is why the next chapter runs as long as it does.
Chapter 7: The Five People You Will Mislabel
7.1 The Anxious One Who Is Not Entitled
The most frequent error by a wide margin, and the one with the highest cost to the person on the receiving end, is applying this label to somebody who is simply anxious about their standing at work. The overlap in what you can observe is close to total. Both take criticism hard, both read unclear signals pessimistically, both go quiet after a difficult meeting, and both may apologise in a way that invites reassurance.
What separates them isn't visible from the outside and cannot be inferred from a single incident. The anxious colleague believes they may not be good enough. The pattern in this article involves believing you are owed something you have not been given, and these beliefs differ even when they produce the same silence.
If you cannot see evidence of the second belief, and you usually cannot, then you do not have grounds for the label. You have grounds for a conversation.
7.2 The Quiet One Who Is Simply Quiet
Introversion is a temperament and carries no antagonism at all, and yet an office that has absorbed the covert-narcissist vocabulary will reliably find the label attaching to whoever speaks least. Somebody who prefers to think before contributing, who dislikes large meetings, and who does their best work without an audience has told you about their preferences and nothing about their intentions towards anybody else. The mistake is usually made in good faith by somebody who has recently read about the quiet type and is now, quite naturally, looking for one.
7.3 The Person Being Badly Managed
Withdrawal, guardedness, unwillingness to volunteer, and a habit of putting things in writing are all reasonable responses to a manager who has previously used candour against somebody. They are also the observable signature of the pattern in this article. This mislabelling does institutional damage rather than personal damage, because it turns a sensible response to bad management into evidence of a personality problem in the person responding. It is also more likely to be made by the manager involved than by anybody else.
The test here is not a personality test at all. If several people behave that way around the same manager, the explanation is unlikely to be that several people share a rare trait.
It is worth being blunt about how this fails in practice, because the failure follows a recognisable sequence: a manager finds one member of the team difficult to read, encounters the covert-narcissist vocabulary somewhere and finds that it fits, begins to interpret every subsequent silence through it, stops asking the questions that might have produced a different explanation, and within about two quarters has assembled a documented case which is internally consistent, entirely sincere, and built on an interpretation that was never tested at any point. Nobody in that sequence acted in bad faith.
7.4 The Loud One From the Last Article
The reverse error is less common and worth naming anyway. In most groups studied, the loud and the quiet have very little to do with each other, and findings do not transfer between them. If the person you have in mind is confident, self-promoting, persuasive in a first meeting, and expensive by the twentieth, they belong to the previous article in this series, and almost nothing here applies to them.
7.5 The One Who Needs a Clinician
Narcissistic personality disorder, meaning the clinical condition rather than the everyday trait, is diagnosed by a qualified clinician with direct access to the person over time, and it is not what this article is about. Somebody in genuine distress, whose functioning is deteriorating, and whose difficulties extend well beyond work, is not somebody to be assessed by a colleague against a personality construct. They may need help. The useful thing a colleague can do is to know the difference between a person who is hard to work with and a person who is unwell.
7.6 Mostly, It Is Something Duller
The most deflating result in this literature is also one of the best evidenced, and it belongs here rather than in an appendix. When researchers examined what the quiet form is actually made of in terms of basic personality, roughly two-thirds of it was accounted for by neuroticism, meaning ordinary emotional instability, and roughly a fifth by disagreeableness (Miller et al., 2018). The way vulnerable narcissism relates to everything else in personality psychology turned out to be almost identical to the way neuroticism does.
What remains, once ordinary emotional instability is set aside, is specifically about distrust and a grandiosity the person does not display, and that residue is small. So, the honest summary is that most of what these ten questions capture is a quality nobody would think to make a video about, and the interesting part is the fraction left over. Treating the whole score as though it were the interesting part is the commonest error in the popular treatment of this subject, and it is not a subtle one.
7.7 And the Numbers Do Not Describe a Person Anyway
Everything in this article is a statement about groups. Every study cited here measured a population, worked out how the measurements moved together across that population, and reported the result, and not one of them licenses a conclusion about the person you have been thinking about while reading. The gap between those two things is not a technicality. A relationship that holds reliably across thousands of people is entirely compatible with the individual in front of you being an exception, and you have no way of knowing whether they are.
What research can do is change what you expect on average, which is genuinely useful, and change what you try first, which is more useful still. What it cannot do is tell you what is going on inside a particular colleague, and anybody selling you that certainty is selling you something the evidence does not contain. # PART IV. WHAT TO DO
Chapter 8: Four Moves and Three Refusals
A note on the status of everything in this chapter, which matters more here than in any other section of this article. The mechanism evidence in Chapters 2, 3 and 6 is peer-reviewed and reasonably strong, and the advice below is untested, in the sense that not one of these moves has ever been trialled against this pattern by anybody. I would rather label that honestly than dress reasoning up as evidence. Where a move follows from a finding, I say which finding, and where it comes from advisory practice, I say that instead.
8.1 Separate the Event From the Person, Out Loud
Status: reasoned from the mechanism in Chapter 2. Not tested.
The route the research describes runs from a failure, to a conclusion about the self, to shame, to hostility aimed downstream, and the earliest point at which anybody else can intervene is the second step. So when something goes wrong, name what happened and what caused it, in specific and boring terms, before anybody has had time to convert it into a verdict about a person. Not these things happen, which is a generality and settles nothing. Instead: the client changed the specification in week three, and we did not reprice, so the margin was gone before the pitch.
The reason to be specific is that a verdict about a person can only be displaced by an account of the situation that is more detailed than the verdict, and warmth is not detail. The reason to do it early is that the sequence in Chapter 2 is fast, and by the time somebody has decided what a failure proves about them, you are arguing with a conclusion rather than preventing one.
One way this goes wrong is worth naming, because it is the way most managers reach for first. Assembling a list of external causes so long that nobody could possibly be responsible for anything reads as room management rather than an honest explanation, and someone already inclined to turn the event into a verdict will hear the length of the list as evidence that somebody is being protected. What you say has to be true, and short enough to be believed.
8.2 Deliver Criticism Where the Audience Is Smallest
Status: reasoned from the mechanism. Partially supported by the experiment described below.
Shame is about being seen, which is what separates it from guilt, and the one experimental result bearing directly on this found that people high on this trait responded with more shame when they received disappointing feedback on a task they believed they had done well, while people low on it did not (Freis, Brown, Carroll & Arkin, 2015). The operational consequence is not to soften criticism, which helps nobody and leaves the recipient unclear about what was actually said. It is to hold the content steady and shrink the audience.
That means the correction happens in the one-to-one rather than in the team meeting, and it means resisting the modern managerial habit of using a group setting to correct one person efficiently. The habit may survive because it saves time, and the time is borrowed from the person who then spends the following fortnight deciding what it meant. Note the tension with 8.1, which asks for a public separation of the failure from the person. Both can hold: the account of what went wrong is public, and the criticism of an individual is private, and keeping those two on different channels is the whole skill.
8.3 Do Not Let Ambiguity Do Your Conflict-Avoidance For You
Status: reasoned from Chapter 3. Not tested.
Chapter 3 established that provocation roughly doubles the aggression link, and that the provocations available in an office are overwhelmingly ambiguous rather than explicit.
The natural response to a difficult colleague is to withdraw politely, because withdrawal is survivable and confrontation is not, and every available way of withdrawing is ambiguous by design. The vague excuse. The unanswered message. The meeting that quietly gets held without them.
Every one of those was chosen precisely because it avoids a scene, and on this evidence each of them is closer to the input that produces one. Which yields the least intuitive line in this article, and the same line the previous one arrived at from an entirely different direction: the vague excuse costs you more than the clear no.
8.4 Protect the People Downstream First
Status: practitioner judgement, drawn from advisory practice.
Everything above is addressed to the person managing the pattern. This one speaks to whoever holds authority over the situation, and it is the recommendation I would keep if I had to drop the other three. The finding in Chapter 2 concerns somebody having a hard time and then transferring it, under their own report, onto people with less power than they have. Those two things are not the same, and only the second one is anybody else’s business.
So the first question is rarely how to help the person in the middle. Ask instead who is currently absorbing the transfer, whether they have anywhere to take it, and whether the person causing the problem is writing the report on that team’s performance to the level above. Those are answerable questions; anybody with access to the right two or three people can work through them in an afternoon, and none requires a personality label or a conversation about whether somebody is a narcissist. In practice the answer to the third question is usually yes, which is the part that turns a personal difficulty into an organisational one, because the account that reaches the level above is written by the person whose behaviour is the subject of it and is generally a competent, plausible and well-evidenced account of somebody else’s shortcomings.
8.5 Do Not Reassure Harder
Status: reasoned from the mechanism, and stated as a refusal because the alternative is actively counterproductive.
Reassurance is the instinctive response to somebody visibly taking a failure personally; it costs nothing, and on the evidence in Chapter 2 it reaches neither of the two steps that matter. It does not supply the specific alternative explanation that 8.1 calls for, and it does not shrink the audience that 8.2 calls for. What it does supply is a reliable signal that this behaviour reliably produces attention, and I would not want to be the person supplying that signal three times a week for a year.
8.6 Do Not Screen Candidates For It
Status: follows from Chapter 6 and Chapter 7.
The measure that most of this research uses splits into two halves, only one of which behaves like a dark trait, and most of what it captures is ordinary emotional instability. A selection process built on that instrument would therefore screen out anxious people at a considerably higher rate than it screens out antagonistic ones, which is both unfair and useless, and would do so under a label that sounds far more rigorous than the instrument actually is. If an organisation wants to select against the thing that causes the damage, the evidence in this series points to conscientiousness and the antagonistic core rather than anything wearing the narcissism label, and the previous articles set out why.
8.7 Do Not Try to Treat It
Status: a limit on scope, not a finding.
Nothing in this article equips a manager, a colleague or a coach to intervene in somebody’s relationship with shame, and the attempt is likely to go badly for everybody involved. What is worth holding on to here is the difference between managing behaviour, which is squarely a manager’s job and is what the four moves above are for, and reaching whatever produces it, which is somebody else’s.
Chapter 9: What This Changes
9.1 The Short Version
The quiet form of this trait is not the mirror image of the loud one, and treating it as the same construct with the volume turned down produces the wrong expectations at every point. It is less common among managers than among people in general, and when present, it predicts worse performance rather than better, reversing the pattern the loud form shows. It is linked to aggression, and to hostile treatment of people further down, more directly than its loud counterpart is, and it gets there by a route almost nobody would guess: taking the blame, feeling the shame, and passing it on. Roughly two-thirds of what the standard ten questions capture is ordinary emotional instability, and those questions divide into a sensitive side that behaves nothing like a dark trait and a self-centred side that does.
9.2 What You Can Do on Monday
You can separate an account of what went wrong from a verdict about who somebody is, in specific terms, out loud, and early enough to matter. And the finding you will still be able to repeat next week, which is the harder test: across three studies of 938 supervisors, it was the quiet, self-blaming ones who reported treating their people worst, and not the confident ones.
9.3 The Honest Ending
There is a temptation in this territory to arrive at a villain, and the evidence does not support one, which I take to be the more interesting result rather than a disappointing one. What it supports is a loop involving two parties who are both behaving comprehensibly. One of them converts ordinary workplace failure into a verdict about themselves at a speed nobody around them can see, and relieves it in the only direction available. The other, faced with somebody hard to read, withdraws politely, and supplies the ambiguity that the first party is least equipped to interpret generously.
Somebody has to stop feeding that loop. As with the previous article, only one of the two is currently reading this, and it is not the one the loop is named after.
That is not a comfortable ending, and I am aware that it asks more of the reader than of the person the reader came here about, but it is where the evidence lands: the mechanism described in Chapter 2 runs inside somebody else’s head at a speed nobody around them can observe, the interventions that might reach it have never been tested, and the only variable in the whole system that is reliably under your control is how much interpretive work you leave a colleague to do about events that were never about them in the first place. That much is available on Monday.
9.4 What I Still Do Not Know
Four things, stated as questions rather than resolved by omission. Whether the self-blame route in Chapter 2 holds when the hostile treatment is reported by the people receiving it rather than by the leaders themselves, which is the study nobody has yet run and the one I would most like to see. Whether any of the moves in Chapter 8 improve anything, since not one has been tested against this problem, and whether the tension between 8.1 and 8.2 survives contact with an actual working week. Whether the division described in Chapter 6 appears again outside the German-speaking groups where it was found, and whether the sensitive side stays outside the dark core once better questions are written.
And whether the whole construct survives Chapter 7.6, or whether a decade from now this will be understood as emotional instability plus a small antagonistic residue, described at far greater length than it deserved because the narcissism label sells better than the neuroticism one does.
A closing note on how I have handled the material, because it applies to everything in this series. Vulnerable narcissism carries the widest gap in this entire literature between what has been demonstrated and what is confidently asserted about it, and almost all of the confident assertion happens outside the research. I have tried to keep the strong findings and the weak ones visibly apart, to label my own reasoning as reasoning rather than folding it quietly into the citations beside it, and to report the results that complicate the picture alongside the ones that make it. What I would most like to have made harder is a confident verdict about the quiet colleague you had in mind as you read.
