Take 520 Americans, split them into six private communities, hand them twenty political fights to have over four weeks, and just watch who actually shows up to argue. That’s the design behind Disentangling participation in online political discussions with a collective field experiment, run by Lisa Oswald and Philipp Lorenz-Spreen of the Max Planck Institute for Human Development and TU Dresden, with William Schulz of Stanford’s Social Media Lab. It isn’t a survey asking people to guess how they’d behave. It’s a live experiment, tracking real posting behavior week over week, and it caught something a survey never could: perceiving a discussion as toxic doesn’t quiet everyone. It quiets some people and revs up others, in exactly opposite directions.

Elisabeth Noelle-Neumann didn’t have Reddit. She had 1974, a theory, and a hunch that would take fifty years and a randomized field experiment to actually test. Her spiral of silence theory argued that people who believe their opinion is losing go quiet — not because they changed their minds, but because they fear being isolated for holding them. Enough quiet people, and the loudest remaining voices start to look like consensus. This new study didn’t just confirm that people go quiet around perceived hostility. It found the other half of the mechanism Noelle-Neumann never had the data to see: the same perceived hostility that silences the middle also fuels the people who were never going to shut up.

The Quick Verdict

Question	What we found
What was studied?	Whether perceiving a political discussion as toxic predicts who lurks (posts nothing) versus who posts more
Who ran it?	Lisa Oswald and Philipp Lorenz-Spreen (Max Planck Institute for Human Development, TU Dresden) with William Schulz (Stanford)
Where and when?	Science Advances, published Dec 10, 2025; covered fresh this week by PsyPost and the Max Planck Society press office
Sample	520 US participants, six private Reddit communities, debating 20 political issues over 4 weeks (June–July 2024), tracked weekly
Core finding	Perceiving a discussion as toxic predicted more lurking among likely-quiet users and more posting among power users who didn’t lurk
What worked	A $2-per-day cash incentive for commenting narrowed the participation gap
What didn’t work	A “please stay respectful” civility-norm treatment produced no measurable change
What Noelle-Neumann already argued	The Spiral of Silence (1974): people who think their view is unpopular self-censor from fear of isolation, letting a vocal minority stand in for public opinion

The Quick Version

Researchers put 520 people into private Reddit communities for a month of real political arguing and tracked, week by week, who posted and who went quiet. Believing a discussion was toxic made moderate, quiet-leaning users lurk even harder. It made the loudest posters post even more. Telling people to be civil didn’t touch either pattern. Paying people to comment did. Elisabeth Noelle-Neumann called the first half of this in 1974, decades before the internet existed to prove it. This study is the mechanism, measured.

What the Study Actually Found

The setup wasn’t a poll asking “would you speak up if others disagreed?” It was six private subreddits, each running for four weeks, where recruited participants debated twenty real political issues — immigration, gun policy, universal basic income, among others — while researchers logged 5,819 comments and surveyed everyone at multiple points along the way. That distinction matters more than it sounds like it should. Surveys ask people to predict their own behavior, and people are bad at that. This experiment just watched what they actually did.

Two groups showed up in the data, and they didn’t respond to the same stimulus the same way. People who tended toward silence — the ones who, across the study, posted rarely or not at all — lurked more the more toxic or disrespectful they perceived the discussion to be. That’s the half that sounds obvious once you say it. The half that doesn’t sound obvious: among people who were already posting, perceiving the same discussion as toxic predicted more comments, not fewer. Same environment. Same perception of hostility. Opposite behavior, depending on who was already in the room talking.

The researchers have a name for that pairing: a spiral of toxicity. It’s not that toxic environments suppress participation across the board. It’s that they redistribute it — quietly editing out the people in the middle while handing the microphone, unintentionally, to whoever was already least bothered by the noise. The active posters here skewed male, skewed highly interested in politics, and skewed toward people who already described themselves as likely to comment online. The people doing the talking in a heated thread were not a random sample of what the group actually thought. They were a specific, self-selected slice of it.

Then came the interventions, and this is where the study stops being descriptive and starts being useful. Researchers tested whether they could close the gap between lurkers and power users. A civility-norm treatment — appeals to stay respectful, no enforcement behind them — produced no significant difference in who spoke up or how toxic the discussion felt. A $2-per-day cash incentive for writing at least one substantive comment did something a nicer tone couldn’t: it meaningfully narrowed the participation gap, pulling quieter people into the conversation without needing anyone to change how they felt about it.

What Did Noelle-Neumann Mean by the Spiral of Silence?

Elisabeth Noelle-Neumann’s spiral of silence theory (1974) holds that people scan their environment for which opinions are gaining ground and which are losing, then stay quiet about the losing ones — not from persuasion, but fear of isolation. The result: silence from the perceived minority makes the perceived majority look larger, producing still more silence.

She built the theory on television debates and door-to-door polling, in a media environment with no reply button and no comment count. What she couldn’t observe directly was the mechanism from the other side — what happens to the people who don’t go silent. This study fills in that missing half: the same perception of hostility that triggers Noelle-Neumann’s silence in some people appears to trigger the opposite in others. Not everyone is running the same calculation about isolation. Some are running a different one entirely, and toxicity feeds it.

Why Do People Lurk Instead of Comment?

1. They read the room as hostile, and they weren’t wrong to. Perceiving a discussion as toxic or disrespectful was the single clearest predictor of lurking in this study — not disagreement itself, but the tone of disagreement.
2. Fear of isolation still does the work Noelle-Neumann described in 1974. People don’t need to be personally attacked to go quiet. Sensing that the room has turned hostile is enough.
3. It’s not about having a minority opinion — it’s about perceived hostility, full stop. This experiment measured perceived toxicity, not perceived unpopularity, and found it moved lurking on its own.
4. Being asked nicely to participate doesn’t override that calculation. The civility-norm treatment gave people a reason to expect better treatment. It didn’t change whether they trusted that expectation enough to post.
5. Money changes the math in a way tone-policing doesn’t. A daily incentive for commenting measurably narrowed the gap — not by making the room feel safer, but by making participation worth the risk regardless.

The Fix That Worked Only Fixed Half the Split

Here’s the part worth sitting with: the one intervention that moved anything didn’t resolve the bifurcation. It closed one side of it and left the other untouched.

The civility-norm treatment — a plain ask to stay respectful, no enforcement behind it — produced no measurable shift in who posted, how much anyone posted, or how toxic the discussion felt afterward. That’s a null result on both halves of the split at once. It didn’t coax lurkers back into the conversation, and it didn’t quiet the power users either. A request for better tone doesn’t seem to touch a decision that’s really about risk-aversion on one end and appetite for a fight on the other.

The $2-per-day incentive is where it gets more interesting than a simple “cash beats civility” reading. The researchers describe it as narrowing the participation gap by pulling quieter people into commenting — it worked on the lurking half of the split, by making the math worth it regardless of how hostile the room still felt. Nothing in the findings suggests it also cooled off the power users, whose posting was climbing right alongside their perceived toxicity in the first place. Shrinking a gap by raising the floor on one side isn’t the same as addressing why the other side was flooding the room to begin with.

Which is really the finding underneath the finding: this was never one broken behavior with one lever to pull. It was two behaviors running in opposite directions off the identical trigger, and the fix that worked only ever had its hand on one of them.

Why the Study and the Philosophy Rhyme

The connection isn’t “a communications theorist predicted a Reddit study in 1974.” It’s narrower than that, and I think it holds up better for being narrower: Noelle-Neumann’s claim was that silence isn’t neutral — it’s a signal, generated by fear, that then gets mistaken by everyone else for the absence of disagreement. This study found the fear-driven silence exactly where she predicted it, and then found something she had no way to measure: the same fear, in people wired differently or invested differently, doesn’t produce silence. It produces volume.

There’s a sharper fit here in Sartre’s account of le regard — the look — from Being and Nothingness. Sartre argued that being seen by another consciousness does something specific to a person: it turns you from a subject moving freely through the world into an object, fixed and appraised from outside, whether or not the one looking ever says a word. The lurkers in this study read like people caught in that look with no way to return it — perceived hostility turns the room into a set of eyes doing the judging, and the only move that feels available is to stop being visible at all. The power users are doing something closer to the opposite: refusing to stay pinned as the object of the look, and answering it by posting into the room instead of shrinking out of it — turning exposure into action rather than retreat.

It also lines up with Hegel’s argument that being unseen is its own kind of wound. Hegel thought people need recognition from others to become fully themselves, and that being ignored isn’t a neutral non-event — it costs something. A hostile thread threatens exactly that recognition for people who are already unsure they’ll get it. For people who’ve already secured a foothold as commenters, the same threat reads differently: less like erasure, more like a fight worth having.

Habermas’s case for genuine dialogue is the uncomfortable one to sit with here. His whole account of legitimate public reasoning assumes something like equal access to the conversation — everyone’s argument getting weighed on its merits, not on who’s loudest or least afraid. A spiral of toxicity is close to a direct violation of that condition: it doesn’t just make discourse unpleasant, it systematically edits out who gets to participate in it, before anyone even gets to the merits.

And Nietzsche’s account of how being ignored breeds a meaner kind of retaliation offers a read on the power users worth taking seriously rather than dismissing. Nietzsche didn’t think people who lash out loudest are simply confident. He thought some of the loudest voices are compensating for a threat they feel more sharply than anyone watching would guess. Not every power user posting more into a hostile thread is thriving there. Some may be doing exactly what the lurkers are doing — responding to a threat — just with the opposite reflex.

What This Study Doesn’t Show

It’s worth being precise about the size of this, because it would be easy to overstate it.

The intervention findings — cash works, civility appeals don’t — come from a randomized experiment, so those comparisons carry real causal weight. The link between perceiving toxicity and lurking or posting more is different: it’s a relationship between self-reported perceptions and logged behavior within the same study, not something the researchers manipulated directly. That’s a meaningfully strong design compared to a one-time survey, but it isn’t the same as proving toxicity perception causes the behavior split in a lab-manipulation sense.

It also doesn’t generalize past the setting it was built in. Five hundred twenty US adults, recruited for a paid Reddit study, arguing about politics in private communities built for research — that’s a specific population with specific incentives already baked in, not a random cross-section of every comment section on the internet. Twitter’s quote-tweet dynamics, a Facebook group of neighbors, a Discord server for a niche hobby — each has its own structure, and this study doesn’t claim to speak for all of them.

And it doesn’t tell you that speaking up despite feeling the room is hostile is always the right call. Noelle-Neumann’s theory was descriptive, not a mandate. Sometimes staying quiet in a bad-faith thread is a reasonable read of the room, not a failure of courage.

What Actually Follows From This

1. Notice that your own silence is data, not nothing. If you’ve stopped commenting somewhere because it started feeling hostile, that’s not a personal failing to fix with more confidence. It’s the exact pattern this study found in hundreds of other people, and it’s telling you something true about the room.
2. Don’t mistake a loud thread for a representative one. The people still posting in a discussion you’ve perceived as toxic are, by this study’s own finding, a specific and skewed slice of who’s actually there. Read volume as volume, not as consensus.
3. If you’re trying to fix participation somewhere — a team, a group chat, a comment section — ask which half of the split you’re actually addressing. A fix aimed at coaxing quiet people back in, even a well-designed one, isn’t the same project as reining in the people whose posting spikes with the temperature of the room. Treating both as one problem is how a real fix ends up looking like it “didn’t work” when it just never touched the other half.

Noelle-Neumann never watched a subreddit fill up with comments, and she never had to. She was working from television audiences and door-to-door interviews, decades before anyone could log a single keystroke of who stayed quiet and who didn’t. But she was right that silence has a shape, that it comes from fear rather than agreement, and that it distorts what a room appears to believe. Oswald, Lorenz-Spreen, and Schulz just handed her the other half of the mechanism: the same fear that silences one person can be exactly what makes the person next to them type faster.

---

This post discusses a 2025 field experiment published in Science Advances and a 1974 theory of public opinion. The study’s causal claims apply to its randomized interventions (financial incentives, civility norms); the link between perceived toxicity and lurking or posting is an observed relationship within the experiment, not an independently manipulated one. Findings come from 520 US adults in a research-recruited Reddit setting and shouldn’t be read as a universal rule for every online platform.