BackgroundUnavoidable stress can lead to perceived lack of control and learned helplessness, a risk factor for depression. Avoiding punishment and gaining rewards involve updating the values of actions based on experience. We examined whether self-reported stress uncontrollability during the first-wave of the COVID-19 pandemic predicted impaired reward-learning.MethodsIn a preregistered study, 427 online participants completed a three-armed-bandit probabilistic reward-learning task to maximise monetary reward. Participants chose between three composite-stimuli comprising three images each. Choosing the target image, which changed every 20-30 trials without warning, resulted in reward on 80% of trials (20% otherwise). Participants completed self-reported measures of depression, anxiety, stress uncontrollability, and risk of COVID-19, which we used to predict accuracy across trials. We tested various reinforcement learning and hidden Markov models (HMMs) to examine cognitive mechanisms underlying the effect of perceived uncontrollability on choices. We hypothesized that stress uncontrollability would predict worse reward-learning.ResultsHigher state anxiety and stress uncontrollability separately predicted impaired reward-learning (z=-2.66, p=.008; z=-3.05, p=.002). The effect of stress uncontrollability was mediated by anxiety (β=0.61, p=.052). State anxiety was positively associated with a parameter from the winning HMM that estimates how much beliefs about the identity of the target diffuse among images (r=.12, p=.0152).ConclusionsGreater perceived stress uncontrollability during the COVID-19 pandemic was associated with impaired reward-learning, and this effect was explained by current anxiety. These results suggest that anxiety promotes imprecision in the representation of the reward structure of the environment, mediating the observed effect of lack of control on reward-learning.Supported BySwedish Research CouncilKeywordsStress Controllability, Reward Learning, Computational Modelling, State Anxiety BackgroundUnavoidable stress can lead to perceived lack of control and learned helplessness, a risk factor for depression. Avoiding punishment and gaining rewards involve updating the values of actions based on experience. We examined whether self-reported stress uncontrollability during the first-wave of the COVID-19 pandemic predicted impaired reward-learning. Unavoidable stress can lead to perceived lack of control and learned helplessness, a risk factor for depression. Avoiding punishment and gaining rewards involve updating the values of actions based on experience. We examined whether self-reported stress uncontrollability during the first-wave of the COVID-19 pandemic predicted impaired reward-learning. MethodsIn a preregistered study, 427 online participants completed a three-armed-bandit probabilistic reward-learning task to maximise monetary reward. Participants chose between three composite-stimuli comprising three images each. Choosing the target image, which changed every 20-30 trials without warning, resulted in reward on 80% of trials (20% otherwise). Participants completed self-reported measures of depression, anxiety, stress uncontrollability, and risk of COVID-19, which we used to predict accuracy across trials. We tested various reinforcement learning and hidden Markov models (HMMs) to examine cognitive mechanisms underlying the effect of perceived uncontrollability on choices. We hypothesized that stress uncontrollability would predict worse reward-learning. In a preregistered study, 427 online participants completed a three-armed-bandit probabilistic reward-learning task to maximise monetary reward. Participants chose between three composite-stimuli comprising three images each. Choosing the target image, which changed every 20-30 trials without warning, resulted in reward on 80% of trials (20% otherwise). Participants completed self-reported measures of depression, anxiety, stress uncontrollability, and risk of COVID-19, which we used to predict accuracy across trials. We tested various reinforcement learning and hidden Markov models (HMMs) to examine cognitive mechanisms underlying the effect of perceived uncontrollability on choices. We hypothesized that stress uncontrollability would predict worse reward-learning. ResultsHigher state anxiety and stress uncontrollability separately predicted impaired reward-learning (z=-2.66, p=.008; z=-3.05, p=.002). The effect of stress uncontrollability was mediated by anxiety (β=0.61, p=.052). State anxiety was positively associated with a parameter from the winning HMM that estimates how much beliefs about the identity of the target diffuse among images (r=.12, p=.0152). Higher state anxiety and stress uncontrollability separately predicted impaired reward-learning (z=-2.66, p=.008; z=-3.05, p=.002). The effect of stress uncontrollability was mediated by anxiety (β=0.61, p=.052). State anxiety was positively associated with a parameter from the winning HMM that estimates how much beliefs about the identity of the target diffuse among images (r=.12, p=.0152). ConclusionsGreater perceived stress uncontrollability during the COVID-19 pandemic was associated with impaired reward-learning, and this effect was explained by current anxiety. These results suggest that anxiety promotes imprecision in the representation of the reward structure of the environment, mediating the observed effect of lack of control on reward-learning. Greater perceived stress uncontrollability during the COVID-19 pandemic was associated with impaired reward-learning, and this effect was explained by current anxiety. These results suggest that anxiety promotes imprecision in the representation of the reward structure of the environment, mediating the observed effect of lack of control on reward-learning.
Read full abstract