Why Workplace Bullying Is So Hard to Prove

There is a particular moment that people describe almost identically, regardless of industry or country.

You have finally decided to raise it. You have thought about it for months. You sit down, and you discover that everything you have to offer sounds trivial when spoken aloud. He reassigned my project. She did not invite me to the meeting. He said my report was fine but in a way that was not fine. Each sentence, said out loud in a small room, sounds like someone complaining about ordinary office life.

You leave that meeting having confirmed the thing you most feared, which is that there is nothing to point at.

This article is about why that happens, and why it is a structural feature of the phenomenon rather than a failure of your preparation. It is a different question from what the conduct is called, which is covered on what counts as workplace bullying, and from whether your workplace is broadly healthy, which is is your workplace toxic or just difficult. Here the question is evidentiary: why is this so hard to show.

Why a Pattern Is Harder to Show Than an Incident

Most workplace processes are built to handle incidents. Something happened, on a date, involving named people, and the question is what it was and what follows.

Bullying as the research defines it is not an incident. The definition used by Einarsen, Hoel, Zapf and Cooper requires repeated negative acts, occurring weekly or more often, over six months or longer, in a relationship where the target cannot readily defend themselves. The pattern is the phenomenon. Individual acts are only its components.

This produces a mismatch that sits underneath almost every difficulty described in this article. You are asked to produce an incident. What you have is a distribution. And a distribution cannot be demonstrated by presenting one of its members, which is precisely what a meeting format forces you to do.

The decomposition problem

There is a further difficulty, which is that patterns decompose under examination in a way that destroys the information they carry.

Take twenty incidents over eight months. Examined individually, each one has an available innocent explanation, and in many cases that explanation is even plausible. The meeting invitation was an oversight. The reassignment was a resourcing decision. The comment was clumsy rather than pointed.

Now examine them together. Twenty oversights, all affecting the same person, all running in the same direction, none ever running the other way. The improbability lives entirely in the aggregate. It is a statistical property, not a property of any member.

Formal processes almost always examine allegations one at a time, because that is what fairness to the accused appears to require. Each one is assessed and found explainable. The conclusion is that no individual allegation is substantiated, which is true, and that therefore the complaint fails, which does not follow but is very hard to argue against in the room.

What Covert and Overt Actually Mean Here

Occupational psychology distinguishes between more and less visible forms of this conduct, and the distinction predicts how provable something will be.

Overt conduct is observable and usually witnessed. Shouting, public humiliation, explicit threats, openly hostile messages. It leaves traces. Others see it. When people imagine workplace bullying, this is generally the picture.

Covert conduct achieves comparable effects through means that do not register as events. Exclusion from information. Work quietly reallocated. Being left off distribution lists. Standards that apply differently to you. Opportunities that stop arriving without any decision ever being announced.

The critical asymmetry is that covert conduct can be every bit as damaging while being far less provable. Its damage comes from accumulation and from the fact that the target cannot form a stable account of what is happening. Its unprovability comes from the fact that its components are absences. Nothing occurred. A meeting did not include you. An opportunity did not arrive. And the costs of bullying in general are well documented. A 2012 meta analysis in Work & Stress, pooling 66 independent samples covering 77,721 people, linked exposure to mental and physical health problems, symptoms of post-traumatic stress, and burnout.

Absences leave no records. There is no document recording that you were not told something, and no timestamp for a conversation that did not take place. Almost all documentation systems record actions, which means the conduct that is hardest to bear is also the conduct those systems are structurally worst at capturing.

Ambiguity as a functional property

Andersson and Pearson's definition of workplace incivility, set out in the Academy of Management Review in 1999, includes a phrase worth isolating: low intensity deviant behavior with ambiguous intent to harm.

Ambiguity is built into the construct. This matters because the ambiguity is not a side effect that happens to make things difficult. It is doing work. Conduct with a ready alternative explanation is conduct that can continue indefinitely, because every instance of it can be answered without anyone having to lie.

This is also why targets so often report doubting themselves rather than doubting the other person. If every individual event has a reasonable explanation, and you continue to feel that something is wrong, the simplest available conclusion is that the problem is your interpretation. Our guide to gaslighting covers what happens when that doubt is actively encouraged, and DARVO describes a common structure that complaints run into when they are finally raised.

The Measurement Gap in the Research Itself

The clearest evidence that this is structural rather than personal comes from the research literature's own difficulty measuring it.

Nielsen, Matthiesen and Einarsen examined this in the Journal of Occupational and Organizational Psychology in 2010, pooling 102 prevalence estimates from 86 independent samples covering 130,973 people. Mean prevalence ranged from roughly 11 to 18 percent, and much of the variation was driven by measurement method and sampling rather than by differences between the people surveyed.

Two methods dominate. Self labeling presents a definition and asks whether it applies to you. The behavioral experience method lists specific negative acts and asks how often each occurred, applying the criteria afterward. In the 2010 meta analysis, self labeling with a definition supplied averaged 11.3 percent, against 14.8 percent for behavioral checklists.

The gap between the two is the finding that matters here. A substantial number of people report experiencing the acts, at the required frequency and duration, while declining to accept the label. If researchers with validated instruments and anonymous surveys struggle to get people to apply the word to their own experience, the difficulty an individual has articulating it in an HR meeting is not a shortcoming of that individual. It is the same effect, showing up in a different setting.

Why Witnesses Often Do Not Help

People assume colleagues will corroborate what they saw. This assumption fails more often than it holds, and understanding why is worth more than resenting it.

The Workplace Bullying Institute's 2021 US survey, conducted by Zogby Analytics with a national sample of 1,215 adults, found 30 percent reporting direct experience of abusive conduct and 49 percent either experiencing or witnessing it. The gap between those figures means witnesses are numerous. Their reluctance is not about absence.

Several things constrain them.

• They saw fragments. A colleague witnessed three of your twenty incidents. Those three, described in isolation, are as explainable as they were when you described them. The witness is being asked about individual data points and has no view of the distribution.
• The risk calculation is real. Corroborating a complaint against someone with organizational power carries genuine cost, and everyone can see what happened to the last person who did it.
• Interpretation differs from observation. A witness may confirm the event while declining to endorse your reading of it. That is usually honest, and it is frequently recorded as a failure to corroborate.
• Conduct is often calibrated to audience. Behavior that occurs in one to one settings and not in front of others produces witnesses who can honestly say they never saw anything.

There is a further problem specific to covert conduct. Its components are things that did not happen, and nobody witnesses an absence. A colleague cannot testify that you were not included in a conversation they were not part of either.

What Actually Constitutes Usable Documentation

Given all of the above, the conventional advice to document everything is correct but insufficiently specific. Most documentation people produce is not usable, because it records feelings rather than facts.

What tends to be usable has particular properties.

Contemporaneous. Written at the time, not reconstructed later. A note written that afternoon carries weight a summary written eight months later does not, and a detail recalled long after the fact is far easier for someone else to dispute.

Factual rather than interpretive. Record what was said and done, with date, time and who was present. The interpretation is the thing under dispute, so a record consisting mostly of interpretation puts the disputed conclusion into the evidence and weakens it.

Held somewhere the organization does not control. Notes on a work laptop or in a work account can become inaccessible at exactly the moment they matter, and access can be removed faster than most people expect.

Structured to reveal aggregate. This is the part most people miss and it is the most important. Because the pattern is the finding, documentation should make the pattern legible. A dated list, in one place, ordered chronologically, showing twenty entries running in one direction, does work that twenty separate accounts cannot. Do not present your best example. Present the distribution.

Including the absences. Record what did not happen, since for covert conduct that is most of the substance. Meetings you were not included in, requests that received no response, information that reached you after the decision. Absences are invisible unless you actively note them.

A realistic expectation

It is worth being straightforward about what documentation achieves. It rarely produces the outcome people hope for. Formal processes assess incidents, most incidents remain individually explainable, and the structural mismatch described above does not disappear because your records are good.

What documentation reliably does is different and still valuable. It protects your own account from erosion, which matters because that erosion is one of the most consistently reported effects of these situations. It gives you a stable factual record when your confidence in your memory is being undermined. It converts a diffuse sense of wrongness into something with dates attached, which changes how you are able to think about it. And if you do eventually need it, whether for a formal process, a solicitor or simply an honest conversation with someone outside the organization, it exists.

One final point that is easy to lose. The difficulty of proving this is not evidence that it is not happening. Those are separate questions, and conflating them is a very common error people make about their own situation. Plenty of things that are real are hard to demonstrate, and conduct that is specifically structured to be deniable is hard to demonstrate by design. If you have concluded that nothing is wrong because you could not prove anything, you have drawn a conclusion the evidence does not support.

And if the reason you are documenting at all is that leaving is not yet available to you, Corporate Hostage deals with that stretch directly, as a survival guide for people who have to keep showing up somewhere they have already seen clearly.

And if the thing you cannot demonstrate to anyone else is something you would still like laid out plainly for yourself, the Workplace Reality Check asks about frequency rather than severity across twelve questions, which is usually where this particular pattern lives.

What Else Do People Ask?

Why does every example sound trivial when I say it out loud?

Because you are describing one member of a pattern, and the pattern is where the significance lives. Twenty incidents all running in the same direction toward the same person is improbable. Any single one of them is ordinary. Spoken aloud a meeting format forces you to present individual instances, which strips out exactly the property that made the whole thing meaningful.

Is covert conduct less serious than open hostility?

Not necessarily. Covert conduct can be just as damaging, partly because it cannot be pinned down, which prevents the target forming a stable account of what is happening. What it is, is less provable, because its components are frequently absences such as being left out or not informed, and absences generate no records.

Why will my colleagues not back me up when they clearly saw it?

Usually a combination of things rather than cowardice. They witnessed a few of many incidents and cannot see the pattern. Corroborating against someone with organizational power carries real cost. And many will confirm an event occurred while honestly declining to endorse your interpretation of it, which is often recorded as a failure to corroborate.

What kind of documentation is actually worth keeping?

Contemporaneous notes, factual rather than interpretive, with dates, times and who was present, stored somewhere your employer does not control. Most importantly, keep it in one chronological list so the aggregate is visible, and record the absences too, since for covert conduct the meetings you were not included in are most of the substance.

If I cannot prove it, does that mean it is not happening?

No, and treating those as the same question is a very common mistake people make about their own situation. Provability and reality are separate. Conduct that is structured to be deniable is difficult to demonstrate by design, and the research literature has the same problem measuring it, which is why prevalence estimates vary so widely by method.

Sources: Nielsen, M. B., Matthiesen, S. B., and Einarsen, S. (2010), "The impact of methodological moderators on prevalence rates of workplace bullying: A meta-analysis," Journal of Occupational and Organizational Psychology, 83(4), 955-979, pooling 102 estimates from 86 independent samples (N = 130,973). Andersson, L. M., and Pearson, C. M. (1999), "Tit for Tat? The Spiraling Effect of Incivility in the Workplace," Academy of Management Review, 24(3), 452-471, defining incivility as low intensity deviant behavior with ambiguous intent to harm. Einarsen, S., Hoel, H., Zapf, D., and Cooper, C. L., Bullying and Harassment in the Workplace: Theory, Research and Practice (Routledge), setting out the repetition, frequency, duration and power imbalance criteria. Workplace Bullying Institute (2021), WBI U.S. Workplace Bullying Survey, conducted by Zogby Analytics, n = 1,215 (workplacebullying.org). Nielsen, M. B., and Einarsen, S. (2012), "Outcomes of exposure to workplace bullying: A meta-analytic review," Work & Stress, 26(4), 309-332.

For self-reflection and educational purposes only. Not a substitute for professional mental health support or qualified employment advice. If you feel unsafe, please contact a local support service.