Full Program »
Borderline personality is associated with disruptions to neurocomputational systems that support value-based decisions
Borderline personality disorder (BPD) is characterized by emotion dysregulation and interpersonal hypersensitivity. Although decision-making in emotional contexts is impaired in BPD, little is known about the neurocomputational systems that underlie such impairment. Here, we examined how in BPD, fronto-striatal-limbic circuits may fail to support convergence on high-value options due to aberrant learning from prediction errors and a limited representation of alternatives.
Ninety-two participants (47 with BPD symptoms, 45 matched controls) between the ages of 13 and 30 (M = 20.61) completed a reinforcement-based timing task during an fMRI scan. Runs consisted of fifty trials in which a dot revolved 360° in 4 seconds around a central stimulus (fearful, happy, or scrambled face). Participants pressed a button to obtain a probabilistic reward from a time-varying contingency. Behavioral data were fit using a novel computational model, SCEPTIC, that captures expected value, information content of the value distribution (entropy), and prediction error (PE). MRI analyses of decision signals employed FSL FLAME and voxelwise deconvolution of selected regions.
Behaviorally, whereas controls reverted to high-value actions after a negative PE in the fearful face condition, those with BPD symptoms did not (p < .001), suggesting a disruption of learning from PEs by negative emotion in BPD. In model-based voxelwise analyses along the longitudinal axis of the hippocampus, BPD participants showed greater modulation of the posterior hippocampus to increases in value entropy. This was mirrored by greater entropy-related sensitivity in the rostral and subgenual cingulate. This pattern of differential modulation by entropy in BPD was amplified by fear relative to happy cues and was associated with larger exploratory decision shifts following negative prediction errors.
Altogether, these findings suggest that negative emotional cues in BPD may disrupt neurocomputational systems that support value-based decisions in complex environments.