Fast evidence accumulation in social anxiety disorder enhances decision making in a probabilistic reward task.

Abstract

Choices and response times in two-alternative decision-making tasks can be modeled by assuming that individuals steadily accrue evidence in favor of each alternative until a response boundary for one of them is crossed, at which point that alternative is chosen. Prior studies have reported that evidence accumulation during decision-making tasks takes longer in adults with psychopathology than in healthy controls, indicating that slow evidence accumulation may be transdiagnostic. However, few studies have examined perceptual decision making in anxiety disorders, where hypervigilance might enhance performance. Therefore, this study used the Hierarchical Drift Diffusion model to investigate evidence accumulation in adults with social anxiety disorder (SAD) and healthy controls as they performed a probabilistic reward task (PRT), in which social rewards were delivered for correct perceptual judgments. Adults with SAD completed the PRT before and after gaze-contingent music reward therapy (GCMRT), which trains attention allocation and has shown efficacy for SAD. Healthy controls also completed the PRT twice. Results revealed excellent performance in adults with SAD, especially after GCMRT: relative to controls, they showed faster evidence accumulation, better discriminability, and earned more rewards. These data highlight a positive effect of attention training on performance in anxious adults and show how a behavioral trait that is typically problematic—hypervigilance in SAD—can nevertheless confer advantages in certain contexts. The data also indicate that, in contrast to other forms of psychopathology, SAD is not characterized by slow evidence accumulation, at least in the context of the social PRT.

Improved Perception of Aggression Under (un)Related Threat of Shock

“…That is, sensory evidence carries a bigger weight when we are trying to detect and judge the presence of threat in our visual environment, while under anxiety. This is in accordance with our third hypothesis, lending support to prior research showing a generalized increase in evidence accumulation rates during anxiety (Dillon et al, 2022;Gorka et al, 2023). Of relevance, this finding is also in line with a previous study by the authors , in which we showed the same increased evidence accumulation under threat when asking participants to identify aggressive interactions between two people.…”

Reward learning capacity in a community sample of individuals who use cannabis.

“…These criteria can include: the removal of trials where response time was faster than 150 ms and slower than 1,500 ms (Bogdan & Pizzagalli, 2006;Whitton et al, 2016) or 2,500 ms (AhnAllen et al, 2012;Dillon et al, 2022;, and additional outliers outside the range of ±3SD from the mean. Participants may be removed from analyses if they had more than 20 outlier trials per block based on reaction time (AhnAllen et al, 2012;Dillon et al, 2022;Liverant et al, 2014); others set this cutoff as 10 per block (Kaiser et al, 2018;Patel et al, 2020;Whitton et al, 2018Whitton et al, , 2021.…”

“…These criteria can include: the removal of trials where response time was faster than 150 ms and slower than 1,500 ms (Bogdan & Pizzagalli, 2006;Whitton et al, 2016) or 2,500 ms (AhnAllen et al, 2012;Dillon et al, 2022;, and additional outliers outside the range of ±3SD from the mean. Participants may be removed from analyses if they had more than 20 outlier trials per block based on reaction time (AhnAllen et al, 2012;Dillon et al, 2022;Liverant et al, 2014); others set this cutoff as 10 per block (Kaiser et al, 2018;Patel et al, 2020;Whitton et al, 2018Whitton et al, , 2021. Some studies specified the requirement of receiving at least 25 rich rewarded stimuli per block (AhnAllen et al, 2012;Liverant et al, 2014), no fewer than 6 lean rewarded stimuli (Dillon et al, 2022) and/or a rich/lean reward ratio of no less than 2 (Dillon et al, 2022;Kaiser et al, 2018).…”

“…Participants may be removed from analyses if they had more than 20 outlier trials per block based on reaction time (AhnAllen et al, 2012;Dillon et al, 2022;Liverant et al, 2014); others set this cutoff as 10 per block (Kaiser et al, 2018;Patel et al, 2020;Whitton et al, 2018Whitton et al, , 2021. Some studies specified the requirement of receiving at least 25 rich rewarded stimuli per block (AhnAllen et al, 2012;Liverant et al, 2014), no fewer than 6 lean rewarded stimuli (Dillon et al, 2022) and/or a rich/lean reward ratio of no less than 2 (Dillon et al, 2022;Kaiser et al, 2018). Others also removed participants with accuracy scores below 55% (Janes et al, 2015;Kaiser et al, 2018;Whitton et al, 2018Whitton et al, , 2021 or 60% (AhnAllen et al, 2012;Liverant et al, 2014), and some studies did not indicate the application of any quality control checks (Hou et al, 2020;.…”

Using Drift Diffusion and RL Models to Disentangle Effects of Depression On Decision-Making vs. Learning in the Probabilistic Reward Task

“…We investigated these topics using the probabilistic reward task (PRT; Pizzagalli et al, 2005 ). The PRT is commonly used to assess RL in depression, but recent data show that it can also provide insight into decision-making ( Dillon et al, 2022 ; Grange, 2022 ; Lawlor et al, 2020 ).…”