Psilocybin increases optimistic engagement over time: computational modelling of behaviour in rats

Abstract

Psilocybin has shown promise as a novel pharmacological intervention for treatment of depression, where post-acute effects of psilocybin treatment have been associated with increased positive mood and decreased pessimism. Although psilocybin is proving to be effective in clinical trials for treatment of psychiatric disorders, the information processing mechanisms affected by psilocybin are not well understood. Here, we fit active inference and reinforcement learning computational models to a novel two-armed bandit reversal learning task capable of capturing engagement behaviour in rats. The model revealed that after receiving psilocybin, rats achieve more rewards through increased task engagement, mediated by modification of forgetting rates and reduced loss aversion. These findings suggest that psilocybin may afford an optimism bias that arises through altered belief updating, with translational potential for clinical populations characterised by lack of optimism.

Fig. 1. Treatment and reversal learning task design.

Rats completed a reversal learning task over 14 experimental sessions on separate consecutive days in their cage, with the first session beginning 24 h after treatment with psilocybin (n = 12) or saline (n = 10). The reversal learning task had a deterministic reinforcement schedule and was designed so that 10 pokes on the rewarded side, with each poke resulting in a reward, triggered the reversal of the rewarded side. The rat could select from three actions: poke left, poke right, or stay in the cage. Created with biorender.com.

Fig. 2. Active inference model.

A graphical depiction of the active inference model. See Methods for more details on the model and equations.

Fig. 3. Effects of psilocybin on behavioural measures across days of reversal learning.

The first 5 panels in this figure illustrate pairwise comparisons of the behavioural measures across days including (A) rewards, (B) losses, (C) win-stay strategies (logit-scaled), (D) lose-shift strategies (logit-scaled), and (E) stay in cage behaviour (log-scaled). Estimated marginal means are represented with dotted lines (psilocybin group) or solid lines (saline group) with shading indicating SEM. Solid red lines indicate significant differences between groups (p < 0.05). Correlations between behavioural measures presented in F. *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 4. Effects of psilocybin on model parameters across days of reversal learning.

The first 5 panels in this figure illustrate pairwise comparisons of the model parameters across days including (A) forgetting rate for losses (logit-scaled), (B) forgetting rate for rewards (logit-scaled), (C) loss aversion (log-scaled), (D) action precision (log-scaled), and (E) prior reward probability (logit-scaled). Estimated marginal means are represented with dotted lines (psilocybin group) or solid lines (saline group) with shading indicating SEM. Solid red lines indicate significant differences between groups (p < 0.05 Correlations between parameters in the best-fit computational (AI) model and behavioural results measures (F). *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 5. Raw behavioural data.

A–D Nose-poke training results prior to psilocybin or saline administration. E–G Response rates for psilocybin and saline groups in the reversal learning task. H–K Reversal rates for psilocybin and saline groups during initial and stable performance periods. L–O Open field data for rats that received psilocybin and saline aged matched to cohort in current study. P–R Elevated plus maze data for rats that received psilocybin and saline aged matched to cohort in current study. PSI psilocybin, SAL saline. Data presented are mean ± SEM.