Case wall · trace the string
Group Attribution Error
The tendency to assume that a group decision (or a single member's traits) reflects the dispositions of the individual members, ignoring the decision rule, sampling bias, or other structural constraints that produced it.
Group Attribution Error
The tendency to assume that a group decision (or a single member's traits) reflects the dispositions of the individual members, ignoring the decision…
Exhibit · study
Allison & Messick (1985), Exp. 1 — water-conservation law
Participants assumed community members held attitudes corresponding to the decision outcome regardless of how the decis…
Exhibit · study
Allison & Messick (1985), Exp. 2-3 — in-group/out-group asymmetry
Subjects inferred a greater correspondence between out-group decisions and out-group attitudes than between an in-group…
Exhibit · study
Allison & Messick (1985), Exp. 4-5 — recall election & jury vote
Subjects attended to the decision outcome roughly as much as to the actual proportion of voters, and jury-attitude infe…
Exhibit · field case · 1993
Jones & Harris (1967) Castro essay paradigm, revisited by Allison, MacKie, Muller & Worth (1993)
In the classic Castro paradigm, observers judged essay writers as personally pro- or anti-Castro even when told the wri…
Exhibit · field case · 1980
Hamill, Wilson & Nisbett (1980) welfare-recipient generalization
Experimental subjects who read one vivid welfare-recipient case formed strongly negative opinions of welfare recipients…
Person of interest
Group Attribution Error
The tendency to assume that a group decision (or a single member's traits) reflects the dispositions of the individual members, ignoring the decision rule, sampling bias, or other structural constraints that produced it.
Strings traced
- study · Allison & Messick (1985), Exp. 1 — water-conservation law
Participants assumed community members held attitudes corresponding to the decision outcome regardless of how the decision was actually made (the decision rule), demonstrating the core error.
- study · Allison & Messick (1985), Exp. 2-3 — in-group/out-group asymmetry
Subjects inferred a greater correspondence between out-group decisions and out-group attitudes than between an in-group decision and in-group attitudes — the error is asymmetric, larger for out-group…
- study · Allison & Messick (1985), Exp. 4-5 — recall election & jury vote
Subjects attended to the decision outcome roughly as much as to the actual proportion of voters, and jury-attitude inferences tracked the final decision (vote + rule) rather than isolating members' v…
- field case · Jones & Harris (1967) Castro essay paradigm, revisited by Allison, MacKie, Muller & Worth (1993)
In the classic Castro paradigm, observers judged essay writers as personally pro- or anti-Castro even when told the writers were assigned their position by coin flip.
- field case · Hamill, Wilson & Nisbett (1980) welfare-recipient generalization
Experimental subjects who read one vivid welfare-recipient case formed strongly negative opinions of welfare recipients in general, and these generalizations were essentially unmoved by statistics te…
The connective logic
Allison & Messick frame the GAE as a group-level analogue of the correspondence/fundamental attribution error: observers infer stable internal attitudes from an outcome (a group decision, or one member's behavior) while under-weighting the situational structure that generated it — the decision rule (single leader vs. supermajority vs. bare majority), the sampling process, or the non-representativeness of an observed member. The asymmetry (stronger for out-groups) connects it to out-group homogeneity and entitativity: out-groups are seen as more uniform and 'entity-like,' so one decision or one member is read as diagnostic of all members. Corneille et al. (2001) add perceived threat as an amplifier.
It is debated whether the GAE is a purely cold cognitive shortcut (under-correction for situational constraint, as with the correspondence bias) or partly motivated/identity-driven, as Pettigrew's (1979) 'ultimate attribution error' proposes for the intergroup case. Hewstone's (1990) review found the intergroup-attribution asymmetry is real but supports a strong motivational 'ultimate attribution error' only weakly, suggesting the group-level effect is more contingent than a unitary mechanism implies.
Case status — replication
Thin file
The foundational studies (Allison & Messick 1985; Hamill, Wilson & Nisbett 1980) are well-cited and the basic phenomenon was extended by Mackie & Allison (1987), the Castro-studies reanalysis (Allison, MacKie, Muller & Worth 1993), and Corneille et al. (2001). However, there is no large pre-registered replication or meta-analysis of the GAE itself. The closely related intergroup-attribution asymmetry was systematically reviewed by Hewstone (1990) across 19 studies, which found the in-group-favoring tendency in most studies but only LIMITED support for the strong 'ultimate attribution error' version. So the core effect is plausible and conceptually replicated within one research lineage, but it has not been stress-tested by the modern replication apparatus, and the motivational interpretation is contested.
How to break the chain
Separate the outcome from the rule that produced it. Before inferring 'the members believe X,' ask: what decision rule generated this (one leader? supermajority? bare majority?), and how representative is any member I actually observed? Force yourself to state the situational/structural explanation alongside the dispositional one, and check whether you'd apply the same generosity to your own in-group's decisions.
Allison & Messick (1985) showed the error arises precisely because people weight the decision outcome as heavily as the actual vote proportion and ignore the binding rule; Hamill, Wilson & Nisbett (1980) showed atypicality warnings barely help, so explicit structural reframing — not just 'remember they might be atypical' — is needed.
Catch it in the act
Watch for reasoning that slides from an outcome to people's hearts: 'The board voted to cut benefits, so the staff must be heartless' (ignoring that a single chair or a bare majority decided it), or 'I met one rude member of that team, so the whole team is rude' (ignoring that the one person may be atypical). The tell is treating a group-level result, or one observed member, as a clean readout of every member's disposition while the decision rule or sampling process goes unexamined — and it tends to be sharper when the group is an out-group.