SEROTONIN MODULATES MOTIVATION STATE TRANSITIONS BY GATING INTERNAL CONTEXTUAL REPRESENTATIONS: A COMPUTATIONAL ACCOUNT

Recent work by Priestley et al. (2025) demonstrated that the dorsal raphe nucleus (DRN) controls motivation state transitions in primates, with DRN disruption impairing the ability to exit high-motivation states when environmental richness decreases. However, computational mechanisms underlying this serotonergic control remain unclear.We propose a dual-pathway model in which serotonin modulates the balance between an internal system tracking learned environmental context and an external system processing sensory evidence, with gating weights (β_int, β_ext) determining each pathway's influence. We tested this on a sequential decision-making task with varying reward environments (rich vs. poor blocks), applying a GLM-HMM to identify latent motivation states.
Under baseline serotonin, the model weighted the internal pathway more strongly in rich, predictable environments where stable reward statistics permit reliable context inference, suggesting the internal system encodes environmental richness. The GLM-HMM revealed clear environment-motivation coupling, with high-motivation states in rich blocks and low-motivation states in poor blocks, reproducing Priestley et al.'s findings. When serotonin was reduced, gating favoring the internal pathway was attenuated. Agents remained in high motivation even during poor blocks, suggesting without sufficient internal pathway weighting, the system cannot infer environmental poorness, preventing transitions to low motivation.
Analysis of internal pathway latent representations corroborated this: representational drift tracked motivation state transitions. Elevated serotonin strengthened environment-state coupling rather than biasing motivation directly, implicating serotonin's contextual inference gating. These findings suggest serotonin enables motivation state transitions by gating internal contextual and external sensory systems, providing a computational basis for Priestley et al.'s findings.