PTSD is often described as involving impaired fear extinction, yet reinforcement learning studies have produced mixed results, with evidence for both disrupted and intact extinction. To reconcile these findings, Cisler and colleagues proposed a latent state theory of PTSD. In this account, the core dysfunction of PTSD is not in fear learning per se, but in inferring and forming new, independent latent states (i.e., internal representations of environmental contexts such as “dangerous” vs “safe”). 

When the abstract structure of the environment changes, individuals with PTSD are hypothesized to be more reluctant to attribute these changes to a new latent state, instead preferring to assimilate new experiences into existing latent states. As a result, some experimental paradigms may fail to trigger state creation and yield impaired extinction, whereas others may have a strong enough change in environment to force PTSD individuals to create a new latent state and produce apparently normal extinction. To date, this remains a verbal theory. 

This thesis formalized the theory using an updated version of an existing latent state model and tested whether this model could reproduce both the significant and non-significant fear extinction findings reported in previous PTSD studies.