National Survey of Family Growth · 2006–2010 | Brief
When Groups Help, and When They Harm
Executive summary
The most useful approach is showing the quiet majority that they are not alone.
Why this matters
As individuals, people feel they can stand their own when confronted with peer influence. They feel that their internal moral compass is bullet-proof against coercions of a mass of ill-intent. And the research actually supports this belief…with limits. The same forces that can push someone toward a dishonest or dangerous choice also make crowds smarter, bystanders braver, and teenagers more generous. Influence itself is neutral. The outcome depends on where the group's norm points, who holds sway, and whether anyone speaks up.
Almost everyone can bend
In 2023, researchers in Switzerland conducted an experiment, with 210 volunteers, testing how far people would go to agree with a group. Each sat with five actors posing as fellow participants, and everyone called out an answer in turn. The real volunteer answered fifth, and on six of ten rounds the actors all gave the same obviously wrong answer. Most of the time, participants held firm: 67% of their answers stayed independent. But 33% went along with the group.
When one group was paid a Swiss franc for each correct answer, conformity fell to 25% without disappearing. In a separate round, participants answered yes-or-no questions aloud, such as whether trade unions or the government should have more power, after hearing the actors answer in unison, and conformity reached 38%.
Personality offered little protection: intelligence, self-esteem, and need for approval did not predict who yielded. Conformity is a situational risk for almost everyone, not a flaw in a weak minority.
When researchers ran the same line task with a majority of small humanoid robots that deliberately gave wrong answers, adults held firm, but children aged seven to nine did not: 74% of children's wrong answers exactly matched the robots'.
Brain-imaging research shows that disagreeing with a group registers like making an error. In another line of experiments, people made quick visual judgments, such as which of two colors dominated a square of pixels, after seeing a group's answers. Their later judgments stayed tilted toward the group's answer even after the group was gone, a sign that the norm changed what they actually saw, and the effect lasted up to 7 days.
Sources: Franzen & Mader (2023); Vollmer et al. (2018); Wu et al. (2016); Germar & Mojzisch (2019); Germar et al. (2023, 2025).
When the group gets it right
Lab conformity studies are built so the majority is wrong, but in everyday life the majority is usually right, and following it is often the smart move. When 5,180 people answered questions alone and then debated in groups of five, the averaged answers of just four groups beat the entire crowd's individual estimates. Online groups estimating quantities grew more accurate after seeing one another's guesses, as long as no single person was linked to everyone. CCTV footage of 219 real public conflicts in three countries showed at least one bystander stepping in 90.9% of the time, and more bystanders made help more likely. Research on emergencies finds that strangers who share a sense of common fate help one another rather than panic, and teens in a giving game became more generous when peers approved of their choices.
When Groups Talk Themselves Into It
In an experiment with 273 participants, people watched a die roll on a screen and reported the number, earning more for higher numbers. In groups of three that could chat first and were paid only if everyone reported the same number, people lied more than they did alone, even when every member had been honest on their own. The driver was conversation: according to the journal publisher's summary, 51% of group chats explicitly made the case for lying. A meta-analysis of 87,771 decisions by 10,923 participants across 21 tasks found the same pattern at scale, with people lying more when their partners lied, partly out of loyalty to the group.
Sources: Kocher et al. (2018); Leib et al. (2021).
In driving simulators, young drivers crash more and take more risks when passengers favor risk and actively give advice. Passengers who quietly prefer risk have a weaker effect. Active encouragement, not mere presence, does most of the work, and some studies find no effect from presence alone.
In one experiment, young men played a computer driving game, deciding at each intersection whether to brake or run a yellow light, while three peers watched. When one of those peers was replaced by a man in his late twenties, the peer effect on risky decisions shrank. A single steadier voice changes the room.
Sources: Centifanti et al. (2016); Bingham et al. (2016); Silva et al. (2016).
When teens viewed a simulated Instagram feed inside a brain scanner, they were more likely to like photos that already had many likes, including photos of drinking and smoking, and activity dropped in the brain network responsible for self-control. Approval also teaches: across 7,331 users and 12.7 million tweets, likes and shares on outraged posts made users more likely to post outrage again, and in a game that imitates a social media feed, higher like and share counts made players more vulnerable to low-credibility news. On Twitter, false news was 70% more likely to be retweeted than the truth. The effects have limits, though: cutting exposure to like-minded sources by about a third for 23,377 Facebook users, or switching users to a newest-first feed, did not measurably change their political attitudes.
[Infographic. Sources: Vosoughi, Roy & Aral, Science, 2018; Brady et al., Science Advances, 2021; Nyhan et al., Nature, 2023]
Sources: Sherman et al. (2016); Brady et al. (2021); Avram et al. (2020); Vosoughi et al. (2018); Nyhan et al. (2023); Guess et al. (2023).
National data that followed students from high school into their thirties found that men who lived in fraternity houses had the highest binge drinking and marijuana use, and the pattern outlasted college. But men who join fraternities already drank more before college, and men who never attended college show rates close to non-member students. Much of the gap reflects who joins, with membership adding to it.
[Infographic. Sources: McCabe, Veliz & Schulenberg, Journal of Adolescent Health, 2018 · Monitoring the Future national panel, followed from age 18 to 35]
Source: McCabe SE, Veliz P, Schulenberg JE. How collegiate fraternity and sorority involvement relates to substance use during young adulthood and substance use disorders in early midlife: A national longitudinal study. Journal of Adolescent Health, 2018.
A study following 179 fraternity men across 38 campuses found little change in hazing attitudes over three years of membership, another sign that much of the pattern arrives with recruits. What keeps it alive is often a misreading of the room. Most students disapprove of hazing, yet many, especially men, underestimate how much their peers share that view. A recent national survey found fraternity men believed most of their brothers had been through hazing.
Service members binge drink more, but heavy drinking is close to civilian levels. Share reporting each pattern in the past 30 days. Not adjusted for age or sex; the survey's authors attribute part of the binge-drinking gap to the military's higher share of men and young adults.
Pattern, Active-duty service members, Civilian adults
Binge drinking, 34.0%, 26.5%
Heavy drinking, 9.8%, 8.9%
In the 2018 Health Related Behaviors Survey, a Defense Department survey of active-duty members run by RAND and the source of the table above, 28.2% of active-duty members agreed that military culture supports drinking. Service members who drink more tend to overestimate how much their peers drink, and in a longitudinal study of young Navy personnel, beliefs about heavy-drinking norms predicted later drinking.
Across the Army, RAND compared soldiers' survey ratings of their unit and leaders, known as climate, with the measured risk in each unit: better climate went with lower risk of sexual assault and harassment. And the group can pull people back: in a randomized trial with 739 National Guard members, a brief web-based program with peer follow-up reduced binge drinking.
Sources: McCabe et al. (2018); McCabe et al. (2005); McCready & Dahl (2023); NIC Project FAN (2026); Meadows et al. (2021); Lobach et al. (2026); Matthews et al. (2021); Blow et al. (2023).
Harm usually needs only a few to act and many to stay quiet. In the driving studies, active encouragement mattered more than passive presence. In the hazing research, most members privately disapproved but believed they were in the minority, so they stayed silent. A few set the norm. The many enable it through silence.
Some harms do not need to be common to be severe. One hazing death or one assault is catastrophic, even if most members never take part. A group can be mostly decent and still produce irreversible outcomes.
Spread multiplies small effects. False news outran the truth through ordinary human sharing. A modest tendency in each person, repeated across millions of people, can add up to large consequences.
To act against harm, a person must notice the harm, see it as harm, feel responsible, know what to do, and judge the cost worth paying. Silence can set in at any of those steps.
They believed they were alone. Most students disapprove of hazing, yet many underestimate how much their peers agree. Each person reads the others' silence as approval, so a norm few people hold survives because no one sees that others reject it too.
They did not see it as harm. Groups relabel harmful acts as tradition, bonding, or "just joking," and students often do not call their own experiences hazing. If it does not register as wrong, there is nothing to speak up about.
Loyalty outranked conscience. Research on fraternity men links membership to conformity to masculine norms and felt pressure to uphold them. When loyalty to the brotherhood is the highest value, incidents go unreported even when witnesses disapprove. Speaking up feels like betrayal.
They feared the cost. Speaking up risks exclusion, ridicule, lost status, or retaliation, especially where rank or seniority matters. When researchers interviewed forty employees about issues they had kept from their bosses, the most common reason, given by 30%, was fear of being labeled a troublemaker or complainer.
They were already complicit. People who went through hazing, or took part before, have a stake in believing it was acceptable. Speaking up now would mean admitting that what they endured or did was wrong.
They assumed someone else would act. Responsibility thins out in a group. Real-world footage shows bystanders usually do step in during sudden public conflicts, so this matters most when harm is ongoing, normalized, and inside a group the bystander belongs to.
They did not know what to do. Many assume the only option is dramatic confrontation. Without a script, freezing is the default.
They thought it was futile. When past complaints went nowhere, silence becomes the rational choice. The link between unit climate and harm fits here: in a poor climate, people learn that speaking up does not work.
Each reason for silence has a matching lever. The levers are described in the next section. The last one belongs to leaders, not individuals.
Reason for silence: Matching lever
Believing you are alone: Check the norm privately
Fear of what speaking up will cost: Recruit an ally ahead of time
Not knowing what to do: Learn a menu of low-cost moves
Loyalty and momentum in the moment: Set bright lines and if-then plans in advance
Belief that nothing will change: A leadership task: show that speaking up leads to action
A note on the evidence: these reasons come from surveys, interviews, and broader research on silence. They were not measured among the specific silent members in the studies above. They are the best-supported explanations, not direct observations.
Sources: NIC Project FAN (2026); Seabrook et al. (2018); Milliken et al. (2003); Philpot et al. (2020); Matthews et al. (2021).
The exercises with the best evidence work ahead of time, giving a person a plan to follow when the pressure hits instead of relying on courage in the moment.
Exercises with the strongest support
Write if-then plans in advance. Decide the response before the moment comes: "If the group starts planning to haze a new member, then I will say I am out and leave." Research on these plans, called implementation intentions, finds they reliably help people act on their goals, because the decision is already made before the pressure arrives.
Rehearse out loud. Practice the exact words with a friend until they come easily. Rehearsal is the core of bystander training. One such program, Green Dot, trains students, starting with socially influential peers, to recognize situations that could lead to violence and to step in safely. In a randomized trial across 26 Kentucky high schools, it was followed by lower rates of sexual violence perpetration once fully in place. Under stress, people fall back on what they have practiced.
Learn a menu of low-cost moves. Many people stay silent because they think confrontation is the only option. Bystander programs teach smaller ones: speak up directly, distract or interrupt, bring in someone with authority, check on the person harmed afterward, or keep a record. A small move a person will actually make beats a heroic one they will not.
Exercises built on how groups work
Recruit an ally ahead of time. Group pressure is strongest when a group appears to agree. Agree with one like-minded member in advance to back each other up. As the hazing research shows, many members privately disagree, and one visible dissenter can reveal that others share the doubt.
Check the norm privately. People overestimate how much peers approve of harmful behavior. Asking a few members one-on-one what they really think often reveals the silent majority, and a lone voice becomes several.
Draw bright lines in writing, while calm. List the few things you will not do no matter who asks. A rule set in advance is harder to argue away than a judgment made in the moment, which counters the way groups talk themselves into crossing lines.
Build in a pause. Keep a stalling phrase ready: "Let me think about it" or "I need to step out." Time lets the group's pull fade and individual judgment return.
Debrief the moments you stayed quiet. Write down what happened, what stopped you, and what you will do next time, then turn it into a new if-then plan. Each lapse becomes better preparation rather than a source of shame.
What doesn't work as well as people think
Moral reminders. In a well-known study, people listed either the Ten Commandments or books they had read, then solved puzzles and reported their own scores for cash, with a chance to overstate them. The Ten Commandments group appeared to cheat less, but a replication of the same design across 19 labs and 4,674 participants found no such effect. Willpower and character. Personality barely predicted who resisted in the 2023 conformity experiment. Planning and practice transfer better than resolve.
How strong is this evidence? The strongest support is for if-then planning and structured bystander programs. Exercises 4 through 8 apply well-established findings about groups but have not been widely tested as standalone exercises. Treat them as well-grounded advice, not proven interventions.
Sources: Gollwitzer & Sheeran (2006); Coker et al. (2017); Verschuere et al. (2018).
Groups amplify whatever direction their norm points. The same forces that sharpen a crowd's judgment and move bystanders to act can also supply the arguments for a lie or the cover for a hazing ritual. Most people resist most of the time, but almost no one is immune.
Taken together, the evidence suggests that harm rarely needs a majority. A few people acting, and many staying silent because they wrongly believe they are alone, can be enough. If that is right, the most useful lever is showing the quiet majority that they are not alone: make the real norm visible, add a steadier or dissenting voice, keep influence from concentrating in a few hands, and slow down the moment when a group talks itself into something.
This brief has been compiled by Rise X Shine under our own resources and initiatives. CDC nor any of its constituents contracted us to create the brief.