RATIONALE: Depressed patients robustly exhibit affective biases in emotional processing which are altered by SSRIs and predict clinical outcome. OBJECTIVES: The objective of this study is to investigate whether psilocybin, recently shown to rapidly improve mood in treatment-resistant depression (TRD), alters patients' emotional processing biases. METHODS: Seventeen patients with treatment-resistant depression completed a dynamic emotional face recognition task at baseline and 1 month later after two doses of psilocybin with psychological support. Sixteen controls completed the emotional recognition task over the same time frame but did not receive psilocybin. RESULTS: We found evidence for a group × time interaction on speed of emotion recognition (p = .035). At baseline, patients were slower at recognising facial emotions compared with controls (p < .001). After psilocybin, this difference was remediated (p = .208). Emotion recognition was faster at follow-up compared with baseline in patients (p = .004, d = .876) but not controls (p = .263, d = .302). In patients, this change was significantly correlated with a reduction in anhedonia over the same time period (r = .640, p = .010). CONCLUSIONS: Psilocybin with psychological support appears to improve processing of emotional faces in treatment-resistant depression, and this correlates with reduced anhedonia. Placebo-controlled studies are warranted to follow up these preliminary findings.