Behavioural Science & Sociocultural Psychiatry · Paper A
Group polarisation
Put a handful of people who already lean the same way on something in a room together, and the conversation does not average them out to where they started -- it manufactures certainty. The group leaves more convinced, in the SAME direction, than any one of them was walking in. Nothing about this requires anonymity, a crowd, or suppressed dissent: it happens in named, face-to-face committees where everyone knows exactly who said what.
Imagine four friends who each mildly think a risky investment is a good idea. Put them in a room to discuss it for twenty minutes, and something odd happens: they don't average out to the same mild opinion they started with. They leave more excited about it than any one of them was walking in. Nobody argued them out of anything -- arguing FOR the idea is exactly what pushed them further into it. This is group polarisation (a shift, through group discussion, toward a MORE EXTREME version of whatever the group already leaned toward before it talked) -- and once you can spot it, you will see it is not the same thing as several other 'the group changed' findings the exam loves to mix it up with.
Group polarisation is what happens when people who already share a leaning on a topic discuss it together: afterwards, the group's shared position is a stronger, more extreme version of that SAME original leaning -- never a watered-down compromise, and never a flip to the opposite side. A group that starts out slightly cautious becomes noticeably more cautious; a group that starts out slightly bold becomes noticeably bolder. The direction never changes. Only the degree does.
Here is the trap almost everyone falls into: assuming group polarisation means groups make riskier, or worse, decisions than individuals. It does not say that at all. It says groups make MORE EXTREME decisions in whichever direction they already leaned -- and if that direction was the sensible, cautious one, the group ends up MORE sensible, not less. The word 'polarisation' describes a direction sharpening, not a quality judgement. A second, equally common trap: assuming this needs anonymity, a crowd, or hidden identities. It does not -- polarisation happens in named, face-to-face committees where everyone knows exactly who said what.
Suppose you privately think a policy is a fairly good idea, rating it 6 out of 10. Everyone else in your discussion group independently landed on roughly a 6 too, though you don't know that yet. Two things happen when you talk.
First, arguments come out that you had not thought of. Each person has their own private reasons for their 6/10, and discussion pools those reasons together. You hear two or three genuinely new arguments FOR the policy that never crossed your mind alone, and hearing more supporting arguments than you had before nudges your own rating up. This is persuasive arguments theory (informational influence): the group's total argument-weight, not just your own, is now shaping your opinion.
Second, you learn something about your social position. You discover your 6/10 was actually the most cautious rating in the room -- everyone else was at 7 or 8. Nobody wants to be the most timid person in a room that broadly agrees with them, so several members nudge their stated position up toward the group's more confident end. This is social comparison theory (normative influence): people shift toward the position that looks most favourable relative to the group's perceived norm. Both effects push the average rating the same way, and most research finds real evidence for both mechanisms working together, not one alone.
The finding actually started narrower than it ended up. In 1961, a graduate student named James Stoner ran a simple experiment: get people's private ratings on decisions with a risky option and a safe option, then get their group's rating after discussion. Groups consistently chose the riskier option more strongly than the average of their own members beforehand. This got the catchy name 'risky shift', and for a while people assumed group discussion simply makes everyone bolder.
Then, in 1969, Serge Moscovici and Marisa Zavalloni ran the same kind of study on topics where people started out cautious rather than risk-tolerant -- and found groups became even MORE cautious after discussion, not riskier. That result could not be explained by 'groups love risk' at all. It could only be explained by 'groups intensify whatever they already lean toward', and that bidirectional rule is what they renamed group polarisation. Risky shift, it turned out, was just the one-directional special case researchers happened to notice first, because most of the early study topics quietly leaned risky to begin with.
Group polarisation sits in a crowded family of 'something about the group changed a person's behaviour or opinion' findings, and the exam's whole game is making you tell them apart.
| Looks similar | How it actually differs |
|---|---|
| Groupthink | Dissent is actively silenced under threat, time pressure, or a dominant leader; the poor outcome comes from debate that never happened, the opposite of polarisation's real argument exchange |
| Conformity | An individual's private view stays the same; only their public behaviour matches an unchanged group norm -- the norm itself never shifts |
| Social loafing | A drop in individual effort on a shared task with unidentifiable contribution -- no opinion content at all |
| Deindividuation | Disinhibited action caused by anonymity or crowd immersion, needing no discussion or shared prior opinion whatsoever |
| Minority influence | A small, consistent minority converts the majority -- the direction of persuasion runs the opposite way from polarisation |
There is no test or criterion checklist here -- group polarisation is established through a specific research design, not a diagnostic instrument. Researchers measure each member's private opinion before any discussion, let the group discuss freely, then measure the group's or each member's opinion again afterwards. If the post-discussion position sits further along the SAME direction as the pre-discussion average, that before-and-after comparison, not a single snapshot of the group's final view, is what counts as evidence of polarisation. A single after-the-fact observation that 'the group agreed on something extreme' proves nothing on its own without the comparison.
Practical debiasing techniques used in real committees, juries, and boardrooms directly target the two mechanisms above. Introducing a genuine devil's advocate role, or deliberately assigning someone to argue the opposite case, works against persuasive arguments theory by forcing the group to hear arguments it would not otherwise generate for itself. Anonymous voting or private written opinions collected before open discussion, and structured methods such as the nominal group technique, work against social comparison theory by removing the social pressure to visibly out-do the room. NICE and RCPsych do not publish clinical guidance on this process, because it is a normal group-decision phenomenon studied across everyday committees and juries, not a clinical entity -- the countermeasures above come from organisational psychology and decision science, not a medical guideline.
Whenever a vignette shows a group ending up MORE unanimous and MORE confident than any individual started, the decisive question is never 'did the group agree' -- it is 'was real disagreement genuinely voiced and debated (polarisation), or was it actively silenced under pressure (groupthink)?' Before you move on: could you explain to a friend why a jury that starts out leaning mildly toward acquittal can polarise toward a MORE confident acquittal, using the exact same process that in a different jury, leaning mildly toward conviction, would produce a more confident conviction?
Group polarisation is discussion sharpening a group's shared lean into a more extreme version of the SAME direction it already had -- never a reversal, never proof the decision got worse, and never dependent on anonymity or silenced dissent.