Proxy reports of others' behaviors: when are they more accurate?

Surveys often ask people to report about someone else's behavior. Two approaches to constructing a behavioral frequency judgment are recalling specific episodes of the other person's behavior or using more general knowledge to infer a frequency judgment. A two-stage study was conducted with 40 roommate-pairs who maintained diaries for 7-days and then reported behavioral frequencies for themself and their roommate. Behavior regularity and relationship closeness were also measured. When general knowledge is cued, judgment error is negatively related to behavior regularity and relationship closeness, while cuing episodic-based recall resulted in less judgment error when regularity and relationship closeness are low.