Investigated the idea that since self-referent information is more accessible, has more connections, and is more deeply encoded that it should be more influential in the group estimates than other-referent information. Participants were given a random number to remember during the task. Used 24 trait adjectives and participants had to indicate if the term was indicative of themselves, a roommate, and a university student. At the end they had to recall the number and indicate how similar and how much they liked each person.