Threat expectancies in a VR fear conditioning paradigm follow non-linear extinction patterns but are not influenced by intolerance of uncertainty.

In exposure therapy, the degree to which individuals effectively correct their threat expectancies varies considerably. Using appropriate statistical models to capture this variability and investigating potential moderators (i.e. mechanisms of change) of the extinction learning process is therefore important. We hypothesized that a non-linear statistical model would better explain extinction learning trajectories indexed by unconditioned stimulus (US) expectancies than a linear model and that intolerance of uncertainty (IU) might negatively impact extinction. Seventy-one spider-fearful participants completed a two-day realistic virtual reality conditioning paradigm where a spider served as the US. Trial-by-trial US expectancy ratings were collected as the main outcome during acquisition and extinction. We used ordered beta regression in a Bayesian estimation framework to analyze data. As expected, trial-by-trial US expectancy ratings during extinction were substantially better explained by the non-linear ordered beta than a linear statistical model. Non-linear models can therefore more adequately capture interindividual differences in trial-by-trial extinction learning by providing a better fit to such data. However, IU did not moderate US expectancies in our study. Thus, our study supports the notion that US expectancies are largely insensitive to IU.