Dissociable effects of psilocybin and escitalopram for depression on processing of musical surprises

Major Depressive Disorder (MDD) is marked by disrupted emotional and hedonic processing, with anhedonia — a reduced capacity to experience pleasure — as a core symptom tied to dysfunction in mesocorticolimbic reward circuitry. Harding and colleagues frame this dysfunction around nodes such as the nucleus accumbens (NAc) and the ventromedial prefrontal cortex (vmPFC), and situate their hypotheses within a Bayesian predictive coding account (the REBUS model) in which psychedelics are proposed to reduce the precision or weighting of top-down priors encoded by deep cortical pyramidal cells. In contrast, selective serotonin reuptake inhibitors (SSRIs) such as escitalopram are widely used but are often associated with limited effects on anhedonia and with reports of emotional blunting. Music is introduced as a naturalistic, hedonic stimulus that evokes reward-related responses and expectancy violations (‘musical surprises’), providing a probe to examine changes in both subjective pleasure and neural responses linked to prediction error and reward. This study aimed to compare the effects of psilocybin therapy (PT) versus escitalopram on behavioural and fMRI responses to musical surprises in patients with MDD. The investigators predicted that PT would produce greater improvement in anhedonia than escitalopram, and that the two treatments would differentially modulate surprise-related pleasantness and neural activation, notably anticipating increased NAc activation to surprising events after PT and distinct surprise-related effects in the vmPFC and superior temporal gyrus (STG). The design used musical surprises together with music-evoked emotion scales to provide a combined behavioural and neurobiological assessment of treatment effects on hedonic and predictive-processing mechanisms.

This report presents a specific fMRI analysis derived from a Phase II, double-blind, randomised, controlled trial in which participants with MDD were allocated to either PT or escitalopram. Of 59 patients enrolled in the parent trial, 50 were included in this analysis; after exclusions for excessive movement or protocol deviation, 22 participants in the PT arm and 19 in the escitalopram arm remained for the present analyses. MRI scanning was performed before any therapeutic intervention (one day prior to first dosing) and again six weeks plus one day after the first dosing day (three weeks after the second psilocybin dose). Scanning-day procedures were identical at both time points and were followed by a continuous music rating task and self-report measures. During fMRI, participants listened with eyes closed to a single piano arrangement (“The Hours” by Philip Glass; 423 s) interleaved with rest blocks (total task 510 s). Data were acquired on a Siemens TIM Trio 3T scanner with high-resolution anatomical MPRAGE and multiband EPI functional sequences (TR = 1250 ms, TE = 30 ms, 3 mm isotropic voxels, 408 volumes). After scanning, participants completed continuous valence and arousal ratings (down-sampled to 1 Hz) and the Geneva Emotional Music Scale (GEMS-25). Anhedonia was assessed using the Snaith–Hamilton Pleasure Scale (SHAPS). Surprising and unsurprising events in the musical stimulus were annotated and ranked by professional musicians in a separate study. For first-level fMRI modelling, the investigators used a general linear model (GLM) with three regressors: surprising events (n = 17), unsurprising events (n = 17), and the remaining music, convolved with a canonical haemodynamic response. Pre-whitening and autocorrelation correction were applied, and contrasts comparing surprising versus unsurprising events were obtained. Mid-level fixed-effects analyses generated within-subject pre-to-post contrasts, and group-level mixed-effects analyses (FLAME 1 + 2) tested a 2 (pre vs post) by 2 (treatment) interaction via unpaired t-tests comparing PT versus escitalopram. Group maps were cluster-corrected (Z = 2.3, p < 0.05). Region-of-interest (ROI) analyses targeted three a priori sites: right NAc (6 mm sphere, MNI 11,9,-1), vmPFC (10 mm sphere, MNI -2,46,-8), and right STG (task-constrained mask). Continuous ratings were processed to derive event-related affective change indices: the first derivative of valence and arousal time series produced valence-increase, valence-decrease, arousal-increase and arousal-decrease traces; mean changes during 1–4 s post-event were extracted and standardised within subjects. Behavioural validation compared surprising with unsurprising events using paired t-tests with alpha adjusted to 0.01 for multiple comparisons. Correlations between surprise-related BOLD in ROIs and subjective measures were also tested.

The final analytic sample comprised 41 participants: 22 in the PT group and 19 in the escitalopram group. On the SHAPS measure of anhedonia both groups improved from pre- to post-treatment, but PT produced a larger mean reduction (PT mean change = -5.273, SEM = 0.7845; escitalopram mean change = -3.211, SEM = 0.5952). A mixed-effects model showed a significant treatment-by-time interaction on SHAPS scores (F(1,39) = 4.170, p = 0.0480). Post hoc paired tests indicated significant pre-to-post decreases in anhedonia within both escitalopram (t(18) = 5.394, p < 0.0001) and PT (t(21) = 6.721, p < 0.0001), consistent with overall efficacy but with a greater effect size for PT. Analysis of music-evoked emotion (GEMS) revealed a treatment-by-time interaction for the vitality factor (F(1,39) = 7.967, p = 0.0075). Escitalopram showed a significant decrease in vitality from pre- to post-treatment (mean = -0.5702, SEM = 0.2292; t(18) = 2.488, p = 0.0229), while PT demonstrated a non-significant numerical increase (mean = 0.333, SEM = 0.222; t(21) = 1.50, p = 0.1482), implying the between-group difference was driven by a decline under escitalopram. No significant treatment or time effects emerged for sublimity or unease. Behavioural validation of the surprise paradigm showed that, at pre-treatment, surprising events produced a transient increase in valence relative to unsurprising events in both groups (escitalopram: t(18) = 3.011, p = 0.0075; PT: t(21) = 4.134, p = 0.0005). Post-treatment, the PT group retained a significant surprise-related valence increase (t(21) = 3.818, p = 0.0010). By contrast, the escitalopram group’s pattern changed after treatment: valence rose for unsurprising events and decreased for surprising ones such that the pre-existing difference between event types was no longer significant (p = 0.2202). No significant surprise-related valence decreases were observed at either time point in either group. Results for arousal were reported in supplementary materials. ROI fMRI analyses identified a significant treatment-by-time interaction in the vmPFC for the surprise>unsurprise contrast (F(1,39) = 7.074, p = 0.0113). Within-group tests showed a significant decrease in surprise-related vmPFC activation post-PT (t(21) = 2.195, p = 0.0395), whereas escitalopram produced a non-significant numerical increase (t(18) = -1.767, p = 0.0941). Baseline (pre-treatment) vmPFC activation did not significantly differ between groups (t(39) = 1.791, p = 0.0810). No significant interaction, main effect of treatment, or time effect was observed in the right NAc (p = 0.7655 for interaction), and STG surprise-related activation was likewise non-significant (p = 0.3575). Surprise-related BOLD in these ROIs did not correlate significantly with SHAPS or music-evoked ratings. Whole-brain group comparisons of pre-to-post changes (escitalopram > PT) showed greater increases under escitalopram (relative to PT) in regions including the right angular gyrus, right posterior supramarginal gyrus, frontal pole, middle frontal gyrus, frontal medial cortex and paracingulate gyrus. There were no significant regions where PT exceeded escitalopram in the between-group contrast. Within-group paired tests revealed no significant within-group effect in escitalopram, but in the PT group widespread post-treatment increases in BOLD to surprising versus unsurprising events were observed in bilateral lateral occipital cortex, occipital fusiform gyrus, occipital pole, right postcentral and precentral gyri, central opercular cortex, and bilateral clusters in the superior temporal gyrus. PT also produced post-treatment reductions in activity in the left lateral occipital cortex extending into the angular gyrus. All group-level maps were cluster-corrected at Z = 2.3, p < 0.05.

Harding and colleagues interpret their findings as evidence of dissociable mechanisms between escitalopram and PT in how musical surprises are processed behaviourally and neurally in MDD. Behaviourally, both treatments reduced anhedonia but PT had a larger effect. Escitalopram reduced music-evoked vitality and attenuated the distinction in valence responses between surprising and unsurprising events post-treatment, consistent with a blunting or redistribution of affective responses. By contrast, PT preserved surprise-related positive valence and produced a modest, non-significant increase in vitality. Neurobiologically, the study found reduced surprise-related vmPFC activation after PT relative to escitalopram, alongside post-PT increases in sensory and occipital regions and decreases in angular gyrus activity. The investigators link decreased vmPFC responses to a possible enduring reduction in the precision weighting of top-down predictions, a mechanism hypothesised in the REBUS account whereby 5-HT2A agonism acutely relaxes priors and could lead to lasting changes in predictive processing. Alternatively, the authors note that PT might increase hedonic priors or positive expectations, which could also explain sustained positive valence to surprises. The increase in sensory-region activation is discussed as a possible shift from internally focused processing to a more externally driven, sensory-dominant mode that might accompany recovery from depression. The absence of any NAc surprise-related effect was unexpected given the literature linking the NAc to musical pleasure; the authors suggest several possible reasons, including diminished NAc responsiveness in depression and individual differences in musical taste and reward. They further highlight that surprise annotation was performed by professional musicians rather than subjectively by participants and that the single-song design, small sample size, repeated exposure to the same piece, lack of physiological arousal measures, and a predominance of positive-valence surprises constrain generalisability. The study team recommends future work using a wider variety of stimuli (including those inducing negative valence), objective arousal measures, stratification by baseline music-evoked pleasantness, and directed connectivity analyses to test hypotheses about increased bottom-up information flow after PT. Overall, the authors conclude that their data provide convergent behavioural and neural evidence for distinct treatment effects of PT versus escitalopram on hedonic and predictive-processing aspects of music-evoked experience in MDD, while acknowledging the exploratory nature and limitations of the findings.