This study addressed two questions: (a) What are the 30-month, 60-month, and 90-month test-retest reliabilities of the Parental Bonding Instrument (PBI); and (b) are recollections of parental bonding stable across changes in the level of depressed mood? Participants included 97 outpatients with primary early-onset dysthymic disorder and 45 outpatients with episodic major depressive disorder. Follow-ups were conducted at 30, 60, and 90 months after entry into the study. The PBI was scored using both the original two-factor model and a three-factor solution. As indexed by intraclass correlations, stability for the original two-factor model ranged from .64 to .88, with a median of .77. The intraclass correlations for the three-factor model ranged from .54 to .87, with a median of .73. The intraclass correlations were generally higher between adjacent assessments, but the decrease over increasingly longer intervals was quite small. Overall, the results indicated that the PBI is highly stable for intervals as long as 90 months. In addition, reports of parental bonding were relatively stable despite significant changes in the level of depressed mood. Thus, the present study provides evidence for the long-term stability of parental representations, as measured by the PBI, in depressed outpatients.