← Studies Suggest 🧠 Psychology

Listening Doesn't Change Minds. In 1,485 Video Conversations, the Story Did All the Persuading

A preregistered PNAS field experiment pitted high-quality listening against a persuasive personal narrative on immigration. Listening made people like the canvasser more. Only the story moved their views, and the effect was still there five weeks later.

By Mara Kessler, Persuasion & Political Psychology · September 28, 2026

The gospel of modern persuasion is simple: before you argue, listen. Conflict mediators train for it, management consultants sell it, and the deep-canvassing movement built its identity around open questions and the nonjudgmental nod. Ask open questions, then reflect their words without judgment. Make them feel heard, and the defenses fall. David Broockman, the Berkeley political scientist whose door-to-door experiments helped found deep canvassing, said it plainly in 2020: rather than presenting facts or value judgments, "you ask questions, dig in, make it a kind of collaborative dialogue where you're genuinely open-minded." In this telling, listening was not just courtesy but the active ingredient that lets persuasion work.

Erik Santoro's team decided to test whether the active ingredient works. Their PNAS experiment from February 2025 is the cleanest test anyone has run, and its design was built to give listening every advantage. The team recruited 1,485 American adults through Meta ads and paired each with a trained professional canvasser for a ten-minute video conversation about a genuinely contentious policy, in-state college tuition for unauthorized immigrants, while canvassers were randomly assigned to share a persuasive personal narrative or not, and to practice scripted high-quality, nonjudgmental listening or not. The design used random assignment, preregistered hypotheses, and attitude surveys both immediately after the call and five weeks later. No straw man: the strongest version of the listening claim, tested by researchers sympathetic to it.

The narrative worked, shifting attitudes by roughly three-tenths of a standard deviation right after the call, a large move by persuasion-research standards, with about two-tenths still present five weeks later. The listening manipulation also worked exactly as advertised: the people who received it felt more heard, liked the canvasser more, grew less defensive, and processed the conversation more deeply, so every mechanism listening was supposed to engage switched on. Then came the moment of truth. Nothing. The persuasive boost added by listening was statistically indistinguishable from zero, so a story told by a mediocre listener changed as many minds as the same story told by an expert listener, and although the people who felt most heard liked their conversation partner most, they did not change their minds one bit more than the people who were told the story.

Here is what the paper never quite states: a null result is most damning under the most favorable conditions. If you wanted listening to succeed, you would choose professional listeners, willing participants, a live disagreement, and ten full minutes of talk, which is what this test provided. A weak test can always be dismissed with "you didn't do it right," but a best-case test cannot. The authors themselves note that participants may have been motivated enough to process the message anyway, which would make the listening mechanism redundant rather than wrong. There is also a pattern across this lab's recent output that nobody has connected: a year earlier, Broockman and colleagues showed in PNAS that professional political practitioners predict persuasive messages no better than ordinary people do, and now the same research program finds the practitioners' favorite interpersonal technique adds nothing at all. Together, the two papers indict persuasion expertise: the people paid to change minds cannot predict what changes them and cannot improve on what does.

None of this makes listening worthless, and the paper is careful on the point. Listening warmed the conversation, lowered defensiveness, and improved the speaker's standing, real goods that may be what allows any conversation to happen at all. But the study draws a line the field had blurred, between being liked and being persuasive. For decades the literature treated them as points on one continuum, with liking as the lever that moves belief. This experiment suggests the lever is disconnected. You can be beloved and change zero minds while a clumsy storyteller with a true story moves the needle, because the mechanism that changes attitudes runs through the content of what is said rather than the quality of how it is received.

The Strongest Counterargument

Start with a paper the PNAS authors themselves cite. In 2024, Guy Itzchakov and colleagues published an experimental series in the Journal of Personality and Social Psychology finding that attentive, understanding listening did reduce defensiveness and support attitude depolarization during disagreements. Their designs were lab experiments rather than field conversations, and they measured depolarization rather than persuasion on a specific policy, so the honest reconciliation is that listening changes the temperature of a disagreement while narrative changes its content, and this experiment measured only content, which leaves the boundary conditions intact: one topic, one country, one medium. People who sign up for a video chat about immigration through a Meta ad are self-selected and open, already inclined to process what they hear, while listening might matter more for cold audiences who never pick up the phone. The canvassers were professionals, so even the "low listening" group received a polite ten-minute conversation, which compressed the contrast between conditions. And the manipulation tested one script of nonjudgmental listening, not reflective paraphrase with feedback, not repeated conversations over months, not listening inside a real relationship. A clean null, then, but a null about one defined kind of listening in one defined setting.

What We Didn't Prove

The study has not been independently replicated, and a single experiment from February 2025 cannot carry a field's reorientation. Everything ran through one issue, unauthorized immigration and tuition policy, in the United States, over video chat, with volunteers, which limits transfer to doorsteps, workplaces, dinner tables, and other causes. The reported effects are standardized and approximate; the headline finding is a null, and nulls stay hostage to statistical power even in a study this large. The "no listening" condition was still a polite ten-minute conversation rather than hostility, so the experiment compared excellent listening against decent conversation rather than against rudeness, and the acknowledgments disclose in-kind and financial support from Meta, which also supplied the recruitment channel, a funding note worth stating even with no evidence of interference. And five weeks, durable by persuasion-research standards, is not five years.

The Bottom Line

Two facts now coexist. Both are true. Listening is a genuine interpersonal good: it lowers defenses, warms the room, and makes people like you. But persuasion appears to run on different fuel. In the best-designed test the literature has produced, the story did the work and the listening did the decorating. Anyone whose job involves changing minds should know which is which. An entire industry has been charging for the decorations.

What You Can Do

Separate your goals before the conversation starts. If the goal is the relationship, listen hard, reflect, and be patient, because it works for warmth and de-escalation. If the goal is persuasion, lead with a concrete human story: content moved attitudes in this trial, and no empathic technique substitutes for it. If you train canvassers, audit where the training hours go: hours spent drilling listening scripts may be better spent collecting the true stories volunteers will tell. Measure the right outcome. Likability is not persuasion. Post-conversation likability ratings will flatter your listeners and tell you nothing about persuasion; only attitude measures, ideally taken weeks later, reveal whether minds actually moved. Stop using "they felt heard" as evidence of influence. This study is the receipt showing those two things can come completely apart.