A series of field studies focused on the role of similarity as niche construction in friendships. Using a free-range dyad harvest method, we collected 11 independent samples with 1,523 interacting pairs, and compared dyad members’ personality traits, attitudes, values, recreational activities, and alcohol and drug use. Within-dyad similarity was statistically significant on 86% of variables measured. To determine whether similarity was primarily attributable to niche construction (i.e., selection) or social influence, we tested whether similarity increased as closeness, intimacy, discussion, length of relationship, and importance of the attitude increased. There were no effects on similarity of closeness, relationship length, or discussion of the attitude. There were quite modest effects of intimacy, and a reliable effect of the shared importance of the attitude. Because relationship length, intimacy, closeness, and discussion can all serve as markers of opportunity for, or potency of social influence, these data are consistent with the “niche construction” account of similarity. In 2 follow-up controlled longitudinal field studies, participants interacted with people they did not know from their large lecture classes, and at a later time completed a survey of attitudes, values, and personality traits. Interacting pairs were not more similar than chance, but for the 23% of dyads that interacted beyond the first meeting, there was significant similarity within dyad members. These 2 lines of inquiry converge to suggest that similarity is mainly due to niche construction, and is most important in the early stages of a relationship; its importance to further relationship development wanes.