Objective: Borderline personality disorder is the prototypical disorder of emotion reactivity and dysregulation, yet there remains limited understanding of its neurocognitive correlates. Two mechanisms that may underlie anomalous reactivity in response to negative stimuli among patients with borderline personality disorder are impairment in habituation and exaggerated sensitization of activity in the neural salience network, including the amygdala, anterior insula, and dorsal anterior cingulate cortex. The authors aimed to reveal the most plausible mechanism by examining the effect of repeated exposure to emotional images both within and across study sessions. Method: A total of 75 participants (patients with borderline personality disorder, N=26; patients with avoidant personality disorder included as a psychopathological control group, N=25; and healthy control subjects, N=24) were included in the study analyses. All participants viewed five presentations of the same set of negative and neutral images at each of two sessions, separated by approximately 3 days, while functional MRI data were acquired. Salience network activity, as measured by blood-oxygen-level-dependent signal in anatomically defined regions of interest across the salience network, was compared across the three groups for each presentation at each of the two study sessions. Self-reported negative affect was measured for each trial. Results: Salience network activity showed a main effect of within-session habituation across all groups and sessions. However, a group-by-session interaction was present, such that only patients with borderline personality disorder showed increased salience network activity in response to the images reencountered at the second session, and this increased salience network sensitization predicted greater sensitization in self-reported negative affect. Conclusions: These results elucidate the neural mechanisms by which patients with borderline personality disorder appraise negative social situations as exaggeratedly salient and suggest potential neurocognitive intervention targets.