Abstract Individual differences in fear extinction learning are centrally involved in anxiety vulnerability. We here investigate individual extinction differences using a model-free, data-driven approach, by applying Latent Class Growth Modeling (LCGM) to four in-house data sets from altogether N = 234 healthy male participants. This revealed two distinct trajectory classes: fast extinguishers and slow extinguishers. This pattern was replicated in two independent public data sets (total N = 275, female and male). In a subset of the in-house samples with functional magnetic resonance imaging (fMRI) data (n = 122 males), we investigated the neural correlates of class membership, focusing on the ventral striatum (VS), a key area previously implicated in encoding extinction prediction errors (EPE). We found that fast extinguishers exhibited VS activity at the time of unconditioned stimulus omission early in extinction, consistent with an EPE signal, whereas this signal only appeared late in extinction in slow extinguishers. These findings suggest that extinction success is shaped by how the VS learns safety.
Andres et al. (2026) studied this question.