Why you need to know this: if you have ever told someone they got it wrong, watched them nod, and watched nothing change, this explains why. The reason is not the one you are expecting.
Almost everyone can name someone who was magnetic for about three months. They arrived, they were impressive, and people who’d met them twice quoted them approvingly. Then, slowly, without anything you could put your finger on, the room changed, without anybody knowing when or why.
That’s not a personality clash. It’s a documented timetable. Two studies published since 2025 have explained why it’s so hard to interrupt.
The first is about mistakes. When any of us gets something wrong, the brain fires a small alarm about a tenth of a second later, before we have consciously noticed anything. Researchers at Southampton measured that alarm directly, using sensors on the scalp, while people did a task designed to make them slip up. In people higher in grandiose narcissism, the alarm was quieter. It stayed quieter even when the researchers told them, on every single attempt, that they had just made a mistake.
This is early work with two small groups of students, and another research team has found something that doesn't quite agree. Even so, it points to something nobody had explained properly before. The reason feedback bounces off may not be arrogance or denial. The alarm may simply be turned down.
The second is about being left out. Across nine studies and more than seventy-seven thousand people, including a fortnight in which participants logged on their phones every time they felt excluded, people higher in this trait reported being left out far more often than everybody else. Three reasons. They read unclear situations as rejection, though they read obvious ones normally. Other people genuinely do avoid them. And the effect runs both ways, so being left out appears to make the trait worse over time.
Now put the two together, because that’s where this becomes useful.
Here is somebody who doesn't register that they got it wrong but does register that you have gone quiet. The two commonest responses to a difficult colleague are to tell them more firmly and to withdraw politely. On this evidence, the first targets half the problem that was never the problem, and the second is the exact input that escalates it.
This produces the most useful and least obvious line in this article: the vague excuse costs you more than the clear no.
Three more findings sit alongside those. They can read you perfectly well; what is missing is being moved by what they read, which is why explaining how their behaviour makes people feel changes nothing. Across 259 chief executives, 359 managers and an experiment on 546 leaders, narcissism was the only trait that consistently predicted a leader objecting to remote work, ahead of any belief about productivity. And narcissism is falling, not rising. Pooling 1,105 studies and 546,225 people from 1982 to 2023, the epidemic everybody has heard about does not appear in the data.
The rest traces the two engines driving the pattern, what it costs the people working around it, who you will wrongly accuse if you stop reading too early, and what actually helps. Three corrections on the way will surprise anybody who has absorbed the popular version of this subject. The trait predicts who gets picked and not who turns out to be any good. Narcissism falls with age for almost everybody, while rankings between people barely shift. And the accelerating-decline story, the one the whole genre rests on, does not survive the longest test yet run on it.