Testing the theory of stress as a cumulative prediction error

Abstract

Stress exerts a multitude of effects, including effects on cognition. Recently, a theory was put forward to explain these effects, proposing that stress should be reconceptualized as a cumulative prediction error. This theory posits that large prediction errors, such as how bad a public speaking task in the lab could possibly be, trigger biological responses to manage unpredictable conditions. To date, however, no study has tested whether stressor-related prediction errors are associated with biological responses to that stressor. We addressed that gap. After acclimation, participants (N=61) were told about an upcoming Trier Social Stress Test, and completed questionnaires during anticipation that assessed how difficult they expected the stressor to be, as well as their expected affect after the stressor. Participants then completed the Trier Social Stress Test. After stress induction, participants reported how difficult they found stressor, and their current affect. Saliva samples were taken before and after the stressor to quantify changes in cortisol and dehydroepiandrosterone (DHEA). Cortisol and DHEA increased from baseline to post-stressor, but prediction errors in actual-expected difficulty of the stressor or actual-expected negative affect were not associated with hormone changes. However, absolute stressor difficulty prediction errors were associated with greater negative affect post-stressor, even controlling for baseline negative affect. Although these results provide support for the idea that prediction errors may play a role in subjective stress responses, they do not provide strong support for the idea that biological stress responses occur as a result of prediction errors.