Category: Adult Anxiety - GAD
Research suggests that those with GAD may have differential learning tendencies compared to controls, such as deficits in learning outcome probabilities and greater learning by avoidance of punishment over approach of rewards. Those with GAD may have deficits in learning the likelihood of outcomes over time. Several lines of research have demonstrated that those with GAD have problems with estimating the probability of future events based on past consequences. Excessive worrying is linked to higher rated likelihoods for the occurrence of negative events (MacLeod, Williams, & Bekerian, 1991) and such probabilities were rated higher when both GAD analogues and those who met full criteria were compared to controls (Berenbaum, Thompson, & Pomerantz, 2007). These estimated probabilities of event occurrence are often highly inaccurate (85-91% of their worrisome predictions do not come true; Borkovec et al., 1999). Furthermore, the use of worry to increase distress and avoid negative emotional shifts (as in the Contrast Avoidance Model; Newman & Llera, 2011) suggests better learning by avoidance of punishment in GAD than approach of rewards. The current study tests the hypothesis that those with GAD show worse probabilistic learning and learn better by avoidance over approach compared to controls.
This study used a computerized implicit cognitive task with probabilistic reinforcement to determine if such tendencies exist. Thus far, 59 GAD and 98 non-GAD participants have taken the Probabilistic Selection Task (PST; Frank et al., 2007). In the PST, participants chose between stimuli with specific probabilities of reinforcement, learning over time which have the highest probability. Then in a testing phase they are assessed on their accuracy choosing the most rewarding stimuli over others (approach) vs. not choosing the most punishing stimuli over others (avoidance). Correct choices led to positive reinforcement while incorrect choices led to punishment. Analyses used multilevel modeling and factorial ANOVAs.
Aligning with our hypotheses, those with GAD had a significantly lower percentage of accurate trials compared to non-GADs overall (t(138) = 1.988, p = 0.049). Over time, there was a significant difference in quadratic trajectory of learning between GADs and non-GADs (t(4271.44) = 2.23, p = 0.022) where those with GAD learned more slowly and to a lesser degree than non-GADs. Yet unlike expectations, analyses examining differences between groups on the primary outcome of the PST (difference between accuracy on approach and avoidance trials) showed no significant findings with this limited sample (t(154) = -0.897, p = 0.371). Findings presented at ABCT 2017 will use an increased sample size for adequate power and present results of additional PST indices.
The Pennsylvania State University
State College, Pennsylvania