There is growing evidence for the efficacy and effectiveness of psychotherapy in patients with personality disorder (PD), but very little is known about the factors underlying these effects. Two-polarities models of personality development provide an empirically supported approach to studying therapeutic change. Briefly, these models argue that personality pathology is characterized by an imbalance between development of the capacity for self-definition and for relatedness, with an exaggerated emphasis on issues regarding self-definition and relatedness being expressed in high levels of self-critical perfectionism (SCP) and dependency, respectively. This study used data from a study of 111 patients with PD who received long-term hospitalization-based psychodynamic treatment to investigate whether (a) treatment was related to changes in SCP, dependency, and symptomatic distress; (b) these changes could be explained by pretreatment levels of SCP, dependency, and/or symptomatic distress; and (c) changes in these personality dimensions over time were associated with symptomatic improvement. SCP, dependency, and symptomatic distress were assessed at admission (baseline), at 12 and 24 weeks into treatment, and at discharge. Parallel process multilevel growth modeling showed that (a) treatment was associated with a significant decrease in levels of SCP, dependency, and symptomatic distress, whereas (b) pretreatment levels of each of these three factors did not predict the decreases observed, and (c) changes in SCP, but not dependency, were associated with the rate of decrease in symptomatic distress over time. Implications of these findings for our understanding of therapeutic change in the treatment of PD are discussed.