In 1966, forty-eight male students at the University of Minnesota filed into a room, sat in pairs, and were separated by a cardboard screen so they could not see each other. Each pair put on headphones and listened to what they were told was an audition tape for the College Bowl quiz team — one candidate, answering hard questions under pressure. Two of the tapes held a man who got 92% of the questions right and mentioned, modestly, that he had been an honor student, the yearbook editor, and a member of the track team. The other two held a man who got 30% right and had washed out of the track tryouts. Then, on one tape from each pile, something went wrong near the end: a clatter of crockery and chair legs, and an embarrassed voice. “Oh my goodness, I’ve spilled coffee all over my new suit.”
🧠 The Tape With Four Endings
Afterwards, the students rated how much they liked the man they had heard. The scores sorted themselves into a shape nobody expected. The most attractive candidate in the whole study was the brilliant one who knocked coffee into his lap. The least attractive was the mediocre one who did the same clumsy thing, at the same moment, in the same words — his score actually went negative. Elliot Aronson and his colleagues named it the pratfall effect, borrowing a vaudeville word for a comedian landing on his backside, and the split ran in opposite directions for the two men: up for the able one, down for the average one.
Aronson’s explanation was simple. Competence with no cracks in it starts to look inhuman, and remoteness is not the same thing as respect. Some evidence of fallibility closes the distance. A man who spills coffee on his new suit is a man, and a man is easier to like than a legend.
🤔 A Flaw Only Works After You’ve Earned It
The follow-up literature is where it gets interesting, because the effect turns out to be far more conditional than the headline version suggests. A 1970 study by Helmreich, Aronson and LeFan was titled “To err is humanizing — sometimes,” which tells you the finding: the observer’s own self-esteem helped decide whether the blunder charmed them, and people who rated themselves highly preferred the flawless ace. Mettee and Wilkins found that severity mattered too, and in a strange direction — when the interviewer reacted with open hostility and the capable candidate apologized profusely, liking for him rose even further while respect barely moved. Being scolded in public made the impressive man more lovable. For the unimpressive man, every version of the same mistake just made things worse.
Then there is the modern twist. When researchers started putting flawed robots in front of people, they got both answers. A 2017 study found that a faulty humanoid robot was liked significantly more than a flawless one. Other studies found the opposite: a robot that forgets things and apologizes was rated worse than one that never forgets, and an agent that made a performance error on a shared task was liked less than one that did not. The pattern underneath is that a blunder humanizes you only where your competence has already been established and is not the thing under review. Get the trust part wrong first, and the same stumble reads as evidence.
🔗 The Perfect Companion Problem
Anyone designing a digital companion eventually hits this wall. Build something that is always right, never hesitates, never admits it does not know, and the impression it leaves is impressive and far away. The instinct is to sand off every rough edge, and that sanding is exactly what makes it read as a machine. A small, honest, capability-level stumble — misremembering a detail and correcting itself, saying plainly that it does not know, having a preference it cannot fully justify — does more for likability than another round of polish, as long as the core promise is already settled: it knows you, it is reliable, it remembers.
The same logic wears a different costume in personalized reports and in advertising. A reading of someone’s year that fits perfectly, with no friction anywhere, announces itself as a template. One specific, mildly unflattering note — this stretch is hard, this part needs care — reads as honesty, and honesty raises the value of everything around it. Order matters here: the friction has to arrive after the good news, never before it.
🎲 The Broken Chocolate Bars
Marketing researchers gave this a name: the blemishing effect. Danit Ein-Gar, Baba Shiv and Zakary Tormala ran a field study around a real purchase, with students who were about to walk into an exam — distracted enough to be processing on autopilot. Those students bought more chocolate bars when they were told that some of the bars were broken. Relaxed students, with time to think, bought more when the description was purely positive. The blemish only worked under low attention, and only when it came after the pitch.
Which gives you something to try. Next time you present something — a project, a bio, a cover letter, a profile — say the good things first, then name one small, concrete, harmless flaw. Watch what happens to the room. Just do not lead with it, and do not try it before you have earned the floor.