In 3 experiments, supportive audiences were associated with unexpected perf
ormance decrements (i.e., "choking" under pressure). On a difficult, skill-
based task, participants were more likely to fail when observed by supporti
ve audiences than when observed by nonsupportive audiences. When the criter
ion for success was easy, supportive audiences had no effect. With a diffic
ult criterion, supportive audiences elicited cautious, protective strategie
s that were associated with poor performance: Speed decreased without impro
ving accuracy. Despite impairments caused by supportive audiences, performe
rs found supportive audiences more helpful and less stressful than neutral
or adversarial audiences, and participants believed (wrongly) that they per
formed better with a supportive audience. Results suggest that people are n
ot aware of debilitating effects of supportive audiences and may opt for em
otional comfort rather than objective success.