Empirical evidence supports the provision of individual music therapy for the treatment of psychosis. Predictors of change in mental health care have been identified by outcomes research. Meanwhile the observation of musical micro-processes has been limited to qualitative research revealing fundamental difficulties in attunement capacities. The aim of this non-experimental follow-up study was to investigate rhythmic attunement processes within free improvisations in the early stage of psychodynamic music therapy in order to investigate its predictive value for therapeutic change. 21 in-patients diagnosed with psychosis (ICD-10 F20, 23, 25) were included, receiving five sessions of individual music therapy. Improvisations were audio recorded. The initial and final improvisations were analysed with the Rhythmic Attunement Scale for Psychosis (RAS-P). Clinical measures included external and self-report assessments (BPRS, Dührsen and Happach, TAS 20). Analysis of the audio recordings revealed early occurrence of rhythmic attunement on a stable and high level. BPRS scores showed a post-treatment decline of 20%, but no significant difference for the other measures. Statistical analysis revealed that when initial rhythmic attunement occurred further into the first session, it was a predictor for the decline of psychotic symptoms. Results are limited due to a small sample size and lack of controls.