Models of memory propose that separate systems underpin the storage and recollection of specific events from our past (e.g., the first day at school), and of the generic structure of our experiences (e.g., how lonely I am), and that interplay between these systems serves to optimize everyday cognition. Specifically, it is proposed that memories of discrete events help define the circumstances (boundary conditions) in which our generalized knowledge applies, thereby enhancing accuracy of memory-dependent cognitive processes. However, in the domain of self-judgment, cognition is systematically biased, with a robust self-enhancement bias characterizing healthy individuals and a negativity bias characterizing the clinically depressed. We hypothesized that self-enhancement effects in the mentally healthy may partly rest on an impaired ability for specific memories to set appropriate boundary conditions on positive self-generalizations, while the opposite may be true for self-referred negative traits in the depressed. To assess this, we asked healthy and depressed individuals to think about the applicability of a trait to themselves, then to recall a specific memory that was inconsistent with that trait which would therefore index a boundary condition for its applicability. Healthy individuals showed faster recall only for specific positive memories following negative trait evaluations, while depressed individuals demonstrated faster recall only of specific negative memories following positive trait evaluations—the pattern expected given the respective self-enhancement and negativity biases. Results suggest that specific memories may serve to delimit self-generalizations in biased ways, and thus support systemic biases in trait judgments characteristic of healthy and depressed individuals.