A series of studies tested whether people underestimate the likelihood that others will comply with their direct requests for help. In the first 3 studies, people underestimated by as much as 50% the likelihood that others would agree to a direct request for help, across a range of requests occurring in both experimental and natural field settings. Studies 4 and 5 demonstrated that experimentally manipulating a person's perspective (as help seeker or potential helper) could elicit this underestimation effect. Finally, in Study 6, the authors explored the source of the bias, finding that help seekers were less willing than potential helpers were to appreciate the social costs of refusing a direct request for help (the costs of saying "no"), attending instead to the instrumental costs of helping (the costs of saying "yes").