Do we learn less about someone who always shows up with a friend?
People learned less about whether a partner was trustworthy when that partner always appeared alongside someone else, mirroring the 'overshadowing' effect seen in classic animal conditioning.
Source
Social overshadowing: Revisiting cue-competition in social interactions
Study at a glance
- Design
- Human experiment — Within-subject Trust Game learning phase in which face partners were shown alone or in a pair and were always cooperative or always non-cooperative; cooperation ratings before and after learning; Experiment 2 added mixed-gender pairs, Experiment 3 added a between-subjects instruction manipulation (pair decisions ambiguous vs explicitly joint).
- N
- Separate samples per experiment: 48 in Experiment 1, 48 recruited in Experiment 2 with 46 analysed after exclusions, and 96 in Experiment 3; no single pooled N is reported.
- Population
- UK-born adults recruited online via Prolific, half women
- Outcome
- Change in rated likelihood of cooperating with each partner (test minus baseline), comparing partners learned alone versus in pairs
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
Online participants played a repeated Trust Game: on each trial they saw a face (or a pair of faces) and chose whether to invest £1, which the partner could repay or keep. Some partners appeared alone and others always appeared in a fixed pair, and each partner was either always cooperative or always uncooperative; participants saw what the partner did even when they chose not to invest. Before and after this learning phase they rated how likely they were to cooperate with each individual face shown alone. Experiment 2 added mixed-gender pairs and Experiment 3 told half the participants that pair decisions were always made jointly.
What they found
In all three experiments, ratings shifted more for partners learned alone than for partners learned in a pair, both upward for cooperative partners and downward for uncooperative ones, which is the signature of overshadowing. The effect was large and was not changed by the partners' gender, by the participant's gender, or by whether participants were told pair members decided jointly, which argues against a reasoning explanation. At baseline, before any learning, participants were more willing to cooperate with women than with men, but this gender bias did not alter the overshadowing effect.
The limits
What it doesn't show
Participants received feedback about partners even on trials where they did not invest, which real life rarely provides, so learning may have been unusually accurate; the authors note that more realistic one-sided feedback might weaken the effect. Partners were photographs with fully scripted, perfectly consistent behaviour, so it is unclear how the effect holds with real, variable people. The absence of gender effects is a null result that does not rule out stereotypes mattering with more salient categories or harder tasks. The samples were UK-only online volunteers.
Key terms
- Overshadowing
- Weaker learning about a cue when it is always trained together with another cue than when it is trained alone.
- Cue competition
- A family of associative-learning effects, including blocking and overshadowing, in which cues presented together compete for a limited amount of learning about an outcome.
- Blocking
- Impaired learning about a new cue when it is paired with an outcome alongside a cue that already predicts that outcome.
- Trust Game
- An economic game where a player chooses whether to send money to a partner, who can return part of a multiplied amount (reciprocate) or keep it all (defect).
- Associative learning
- Learning the relationship between stimuli and outcomes through repeated pairings, as described by models such as Rescorla-Wagner.
Flashcards
0 of 11 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
Which comparison was critical for demonstrating overshadowing in this study?
Common questions
Why is it surprising that overshadowing shows up in social learning?
Overshadowing was first described in animal conditioning; finding it when people judge others' trustworthiness suggests social impressions may be formed by the same general-purpose learning machinery rather than special social processes.
Couldn't people just reason that a pair's choice tells you less about each member?
That was the alternative the authors tested in Experiment 3. Even when told pair members always decided together, so each member was fully responsible, overshadowing was just as strong, favouring an associative explanation.
Did gender stereotypes play any role?
Only at baseline: people started out more willing to cooperate with female faces. Gender did not change how much learning was overshadowed.
More on Learning