Ostracism

What it is

The experiment is almost insultingly small. Three figures on a screen toss a ball between them. After a few throws the other two stop throwing to you. They are not real people, participants are often told outright that they are not real people, and the whole thing lasts around two minutes. Then the questionnaires come out and the numbers are enormous.

Hartgerink, van Beest, Wicherts and Williams published the meta-analysis in PLOS ONE in 2015. One hundred and twenty studies, 11,869 participants, an average effect above d = 1.4. For scale, effect sizes in this part of psychology are usually reported in the range where a reader has to be told which direction is which. This one does not need the help. Two minutes of being left out by software moves people more than most things a laboratory can do to a person, and it does so across age, across gender, across country, and across every structural variation the authors could test.

The instrument is Williams's, and its design is the reason the number is trustworthy. Cyberball strips out everything except the exclusion itself. Nobody insults the participant. Nothing is taken from them. There is no audience, no consequence, no future interaction and nothing at stake. What remains when all of that has been removed is the bare fact of not being thrown to, and the bare fact is enough.

That same design decision is the honest limit, and it should travel with the number wherever it goes. The stakes are nil by construction. What the effect size measures is the sensitivity of the mechanism, how little it takes to set the thing off, rather than the magnitude of any real loss. Being frozen out by a team you have worked alongside for six years is a different event with a different shape, and it has not been measured this way for the obvious reason that it cannot be randomised. Anybody who quotes d = 1.4 as the size of what exclusion does to a person has swapped one claim for another.

What Cyberball does establish is more useful than the misreading anyway. It shows a detection system running underneath conscious appraisal, one that reacts before the question of whether the exclusion matters has been asked, and which cannot be talked out of reacting by information about the other players. That fits the account in Need to Belong, where belonging is treated as a fundamental motivation rather than a preference, and it supplies the other half of an exchange the belonging literature otherwise leaves unpriced. If exclusion has teeth, then membership has leverage, and the price of membership is what Conformity measured for entirely unrelated reasons in 1951.

In effect

Research: the meta-analysis that made the number hard to argue with (Hartgerink and colleagues)

Single Cyberball studies had reported strong effects for years, and a strong effect in a small literature is exactly the profile that has embarrassed social psychology elsewhere. The 2015 meta-analysis is what changed the standing of the finding. Pooling 120 studies and 11,869 participants, Hartgerink and colleagues found the effect held not just on average but across the moderators that usually break things: the age of the sample, the gender composition, the country it was run in, and the structural variations researchers had introduced to the paradigm. A result that survives that many different hands is a different kind of result from one that a laboratory can produce reliably.

Everyday life: the read receipt with nothing after it

The everyday version has no content, which is precisely what makes it recognisable. A message is delivered, marked read, and nothing follows. No refusal has been issued and nothing has been said, so there is nothing to respond to and no grievance that would survive being spoken aloud. Cyberball is a laboratory model of exactly this: exclusion with the accusation removed, which leaves the person on the receiving end holding a reaction they cannot justify in proportion to the event. The reaction is not out of proportion to the mechanism. It is out of proportion to the stakes, and those are two different measurements.

Commerce: the waiting list, the velvet rope and the closed beta

Products that want to be wanted manufacture visible exclusion on purpose. An invitation-only launch, a waiting list with a position number, a queue outside a venue that is not full, a members' tier displayed to people who do not have it: each of these makes non-membership perceptible to the people outside it, which is the whole mechanism. Nothing is taken from anyone. That is the point, and it is the same design logic Cyberball uses to produce its effect with no stakes at all. What a commercial version adds is a route out of the excluded state, priced. The research does not say these tactics work, and no Cyberball study has been run on a signup queue. It says the raw material they are working with is unusually responsive.

What it does not say

It does not say that two minutes of Cyberball is equivalent to real exclusion. The paradigm's strength is its emptiness, and its emptiness is exactly why the effect size cannot be read as the size of a real loss. It is a measure of how easily the mechanism fires.

It does not say the harm persists. Cyberball measures the immediate reaction, typically within minutes, and the literature distinguishes that reflexive stage from what a person does afterwards. Nothing on this page is evidence about what long-term exclusion does to anybody, and the studies that would establish it cannot be run by randomising people into years of isolation.

It does not establish that the reaction is uniform. An average effect above d = 1.4 across a pooled literature is a statement about distributions, not about individuals, and the meta-analysis reports the effect generalising across the moderators tested rather than every moderator that exists.

It is not evidence that exclusion is unjustified. The mechanism reacts to being left out and has nothing to say about whether leaving somebody out was the right call. Groups exclude people for reasons, and a page about the cost of exclusion is not an argument that the cost should never be imposed.

Who developed it

Kipling Williams developed the Cyberball paradigm and has run the research programme on ostracism that surrounds it; he is also fourth author on the 2015 meta-analysis. Chris Hartgerink is first author of that paper, working with Ilja van Beest and Jelte Wicherts. Crediting the meta-analysis to Williams alone, which happens often because his is the name attached to the paradigm, gets the authorship backwards.

---