Transcranial direct current stimulation above the medial prefrontal cortex facilitates decision-making following periods of low outcome controllability

Abstract

Background: The arbitration between decision-making strategies is shaped by the degree of controllability over environmental events. Under low control, individuals might rely more heavily on Pavlovian bias (PB), which facilitates and inhibits actions when facing appetitive and aversive cues, respectively. More recently, extreme PB was implicated in learned helplessness (LH), which is typically induced by uncontrollable punishment. On the neural level, the medial prefrontal cortex (mPFC) was pinpointed as a region underlying both cognitive control over PB, and the pathogenesis of LH.Objective/Hypothesis: To test if high-definition transcranial direct current stimulation (HD-tDCS) targeting the mPFC counteracts with the deleterious behavioral effects of low controllability over rewards/losses (“yoking”) during reinforcement learning.Methods: In a pre-registered, between-group, double-blind study (N = 103, healthy adults), we tested the interaction of low controllability and HD-tDCS on performance in a Go/NoGo task. Yoking was implemented by presenting random outcomes following responses, while matching reward/loss frequencies between control and yoked groups. HD-tDCS was delivered for 15 minutes at 2 mA using a 4x1 montage centered at position Fz.Results: HD-tDCS improved response accuracy by the end of the task only when applied simultaneously with yoking. The beneficial consequences of active stimulation in yoked participants were more pronounced in reward-predictive trials. Finally, computational modeling revealed that parameter estimates of learning rate and choice randomness were modulated by yoking and HD-tDCS in an interactive manner.Conclusions: These results highlight the potential of our HD-tDCS protocol for interfering with choice arbitration in volatile environments, resulting in more adaptive behavior.

Transcranial direct-current stimulation enhances Pavlovian tendencies during intermittent loss of control

“…Altogether, the mPFC/dACC seems to be crucial for mediating both Pavlovian-instrumental interaction and the effect of outcome controllability on task performance. In line with this assumption, healthy humans receiving anodal high-density transcranial direct current stimulation (HD-tDCS) over the mPFC/dACC while facing uncontrollable outcomes during RL improved task performance after control was regained (Csifcsák, Bjørkøy, et al, 2020). In addition, PB was reduced in participants receiving active HD-tDCS during diminished outcome controllability but not in participants undergoing either active HD-tDCS or the controllability manipulation alone, suggesting that the level of control over rewards/losses interferes with the degree to which mPFC/dACC suppresses PB.…”

“…We used a computerized version of an orthogonalized Go-NoGo RL task (Cavanagh et al, 2013;Csifcsák, Bjørkøy, et al, 2020;Csifcsák, Melsaeter, et al, 2020;Guitart-Masip et al, 2012) designed to investigate PB during instrumental learning. The participants should learn to maximize the rewards and minimize losses by trial-and-error, through correctly responding ('Go') or not responding ('NoGo') to a given stimulus (card).…”

“…In contrast with earlier studies (Cavanagh et al, 2013;Csifcsák et al, 2021;Csifcsák, Melsaeter, et al, 2020), we decided to increase the contingency level in the controllable blocks to 80/20% (compared to the original 70/30%) in order to contrast the "control" vs. "no control" blocks even further. During the second block, rewards were delivered at random (response-feedback contingency was set to 50/50%).…”

“…In order to estimate the latent parameters of RL, we used the approach used in previous studies (Cavanagh et al, 2013;Csifcsák, Bjørkøy, et al, 2020;Csifcsák, Melsaeter, et al, 2020;Swart et al, 2018) and fitted an RL model that includes an estimate of the Pavlovian bias (parameter π) to the data.…”

“…The previously reported protocols for manipulating outcome controllability during RL consisted of pairing participants in terms of reward/loss frequency, similar to "yoking" paradigms in animal studies of LH (Csifcsák, Bjørkøy, et al, 2020;Csifcsák, Melsaeter, et al, 2020;Maier & Seligman, 2016).…”