Following past work, we proposed that loneliness possesses a bidimensional structure, with dimensions corresponding to lack of pleasurable engagement and painful disconnection. Moreover, we hypothesized that the distinction between lack of pleasurable engagement and painful disconnection would emerge in confirmatory factor analyses, in the pattern of relations of each facet to social impairment, in differential risk for the development of depressive disorders, and in differential rates of recovery from depressive disorders. Data from the Oregon Adolescent Depression Project were used to address these issues. Adolescents from high schools in urban and rural Oregon completed interviews and questionnaires on psychiatric status, loneliness, social functioning, and so forth, at one point in time and then again at a second session approximately a year later. Consistent with predictions, we found that the proposed bidimensional structure of loneliness represented a good fit to the covariance structure of loneliness items; the relation of lack of pleasurable engagement to social impairment exceeded that between painful disconnection and social impairment; and lack of pleasurable engagement, but not painful disconnection, was significantly related (in the expected directions) to onset of and recovery from depressive disorders. We discuss theoretical and clinical implications.