You have to raise something contentious with someone who will not like it. Do you call them, or do you write it out?
Almost everyone writes. Asked to imagine communicating with a counterpart they disagreed with, 83.9 percent of people chose writing, 7.3 percent chose speaking, and 8.9 percent chose video. They also predicted that writing would mean less conflict, and that speaking would be more awkward, more effortful and less enjoyable.
They had the direction wrong, and the experiments showing it are unusually large.
The Artful Age
A weekly letter on aging well, family across generations, and the creative life after the kids leave home.
The argument goes better out loud
Burint Bevis, Juliana Schroeder and Michael Yeomans, writing this year in Nature Communications, report a series of preregistered randomized experiments covering 1,576 conversation partners, 1,842 conversations and 1,432 people who later observed those conversations. Pairs who disagreed about something real, on topics from drug legalization to reparations, were randomly assigned to talk or to type.
Speaking produced more understanding and less conflict, the paper’s two named outcomes. In the first study, perceived understanding averaged 4.95 after spoken conversations and 4.30 after written ones, while perceived conflict ran 1.51 against 1.93. Impressions of the other person and enjoyment of the conversation improved as well.
They took it out of the laboratory, to three American campuses. Among pairs who genuinely disagreed, understanding came in at 4.27 for speech and 3.21 for writing, and conflict at 0.22 against 1.17. There was no difference between campuses. The effect also appeared among 127 pairs who agreed with each other, so this is not only a conflict-management finding.
Three details that sharpen it
It is not the camera. Speaking without video scored slightly higher on understanding than video-chatting did. The authors are careful there, noting the video quality may not have been ideal and that they never tested a genuine in-person conversation. But the voice, not the face, is where the effect sits.
It is not the word count. Spoken conversations ran to 616 words on average against 190 in writing. Control for length and the effects survive, and length does not mediate them.
And it works at both ends. A fresh set of observers read or listened to conversations that had originally been spoken or written. The originally spoken ones were judged to contain more understanding and less conflict, whichever way the observer took them in. So speaking changes the words people pick, and it changes how those same words land.
Language analysis showed spoken conversations carried more conversational receptiveness, driven by more tentative, subjective phrasing, more expressions of agreement, and fewer negative-emotion words. But receptiveness accounted for only around a tenth of the effect on understanding. Most of what is happening is still unexplained.
The authors also surface a puzzle they cannot resolve. Receptive language predicted understanding more strongly in writing than in speech. People use less of it in exactly the medium where it appears to do the most work.
The second move, from a different field entirely
Four years earlier, in a separate literature, Martin Reimann, Christoph Hüller, Oliver Schilke and Karen Cook tested something trust research usually treats as settled: that competence produces trust.
Across seven experiments in PNAS, they manipulated a person’s apparent ability and, separately, whether that person promoted themselves.
In an online shopping scenario, a capable-looking vendor who said nothing about himself drew a trust rating of 3.977 on a seven-point scale. The same vendor with a boastful message attached dropped to 2.415. The vendor who looked incapable went from 1.997 to 1.606.
Boasting cost everyone something. It cost the capable vendor about four times what it cost the incapable one.
The pattern repeated in a workplace scenario about a newly hired vice president, in a study where people decided how much of a $300 bonus to hand over as a down payment, and in a Facebook advertising experiment across 101,520 users, where the boastful seller drew clicks at 0.056 percent against 0.091 percent for the one who stayed quiet.
A mechanism study explains why. Apparent ability raised perceived benevolence and integrity, and it was those two, not ability itself, that produced trust. Self-promotion severed the link. When the vendor boasted, the trust gap between the highly capable and the barely capable disappeared entirely.
The authors read it as a penalty for knowing better. Capable people are held to a higher standard, so a breach of the modesty norm by someone who evidently understands the norm gets punished harder.
One blind spot, one known cost
The tidy version of this article would say people misjudge themselves twice. The papers will not allow it.
On the disagreement question, people genuinely cannot see it. They predict the wrong direction.
On self-promotion, they see it perfectly well. In a prediction market where participants had a real dollar to bet on which of four vendors other people would trust, they put 78 cents on the capable vendor who kept quiet and 17 cents on the capable vendor who boasted. They forecast the finding almost exactly.
And in the same paper, an exploratory study found that self-promotion is about as common among high-ability people as among low-ability ones.
Put those together and the second finding is the stranger of the two. People appear to know that boasting undercuts a capable person, and boast anyway, at ordinary rates. Whatever is going on there, it is not ignorance.
The limits, which are real
In the conversation work, conflict stayed low in every condition, below the midpoint of the scale, and nearly all the pairs were strangers. The authors say this may mean they underestimated the effect, or it may mean a genuinely bitter argument behaves differently out loud. They decline to guess. Studies one through three ran mostly among American college students. Nothing here shows attitudes changing, and the authors say so.
The misprediction result carries its own caveat, which the authors raise themselves: the people making predictions were a different population, surveyed at a different time, judging the options side by side rather than living through one.
On the self-promotion side, most of the evidence is vignettes. The one behavioral field test, the Facebook advertisement, is a borderline result. Effect sizes swing widely between studies. And the manipulation was strong, described by the authors themselves as “an inflated form of self-promotion”: a vendor claiming there had never been a single occasion when he failed to deliver. That is boasting, not mentioning a qualification. Their own closing position is that more work is needed “before any well-grounded practical recommendations for or against the use of impression management can be made.”
That caveat matters more than it might look, because the wrong lesson here is genuinely costly. This is not evidence that you should stay quiet about your work. Plenty of people, in job searches and in workplaces where their competence is not assumed by default, have to make their record visible or it does not get seen. Making a record visible and declaring yourself excellent are different acts, and only the second one was tested.
What is actually left
Two things, at the size the evidence supports.
When something has to be settled with a person who sees it differently, the pull toward putting it in writing is strong, close to universal, and appears to point away from the better outcome. The active ingredient is the voice, not the video.
And if you are genuinely good at something, the case for announcing it is weaker than it feels. Not because modesty is a virtue, but because on this evidence the announcement works against you specifically when there is something behind it.
Neither of these is charisma, and neither asks you to become a different sort of person. One is a choice about which button to press. The other is a sentence left unsaid.