For seven days, participants described the important interactions they had using a variant of the Rochester Interaction Record and reported their attachment style using Bartholomew's four-category system. A series of multilevel random coefficient analyses found that across all interactions securely attached participants, compared to those who were insecurely attached, found their interactions to be more intimate and more positive emotionally. Secure participants also felt that others were more responsive to them and their needs. Secure-insecure differences were most pronounced when secure and dismissive avoidant participants were compared. Differences between secure and fearful types were minimal. In contrast, differences in reactions to interactions with close and not close friends were more pronounced for fearful types than for secures, dismissing, or preoccupied types. These results highlight the importance of distinguishing fearful and dismissive avoidance. Copyright # 2002 John Wiley & Sons, Ltd.In the past decade and a half, a wealth of studies have thrown light on the cognitive, affective, and behavioural effects of attachment styles in people's relationships. Despite the considerable attention such individual differences in adult attachment have received, still relatively little is known about attachment style effects on people's everyday social lives. Much of the research concerning attachment and the nature of people's personal relationships has been either laboratory based or has used relatively broadly focused surveys or questionnaires. Although this existing research is informative, it needs to be complemented by studies in specific relational contexts that examine the links between attachment and interpersonal interaction, outside the confines of the laboratory and with more specificity than that provided by surveys and questionnaires.Studies of naturally occurring interaction are needed because of the difficulty in generalising the results of controlled laboratory research which typically suffer from limited external validity. By design, laboratory studies examine behaviour in restricted, perhaps artificial, settings to ensure internal validity and clarity of inferences. For example, some studies have examined attachment working models using reaction time methods (e.g.