Predictive Biases in Emotional Perception: Differential Influence of Mood and Affective Primes in Individuals with and without Mood Disorders

Introduction: Predictive processing models perception as an interplay between top-down predictions, shaped by prior experiences, and bottom-up sensory inputs. When signals are ambiguous, the brain further relies on prior expectations. In emotional contexts, mood and affective primes may act as endogenous and exogenous predictions, respectively, biasing emotional perception. This might be particularly relevant in mood disorders, where top-down predictions may become overly rigid or unstable. Methods: Fifty-three healthy participants (HC), 14 patients with bipolar disorder (BD), and 17 with depression (DEP) completed an online affective priming task on unambiguous (happy/sad) and ambiguous (morphs, i.e., both happy and sad) facial expressions. Targets were preceded by masked or unmasked happy, sad, morph primes, or none. Participants rated each target’s emotion and completed two mood questionnaires, from which latent dimensions were extracted with a Principal Component Analysis. Drift Diffusion Modeling (DDM) was used to estimate cognitive processes underlying ambiguous face perception. Results: In HC, mood congruently biased ambiguous face categorization, and masked primes accelerated responses. In BD, mood did not influence categorization, but both masked and unmasked primes induced faster responses. In DEP, mood congruently biased perception, with no effect of primes. DDM confirmed valence-congruent shifts in starting points – indicating pre-decisional affective predictions – for masked primes in HC, both masked and unmasked primes in BD, and no effect in DEP. Mood did not affect DDM parameters, suggesting that it biased the categorization of emotionally ambiguous stimuli after affective decision processes. Conclusions: Findings reveal group-specific predictive biases in emotional perception. HC used mood and affective cues, BD patients only relied on external cues (unstable mood priors), while DEP patients showed reduced sensitivity to primes (rigid mood priors). These prediction biases may underlie core clinical features of mood disorders and highlight new targets for interventions aiming to restore adaptive emotional inference.