We studied the synchronization dynamics of a therapist and patient during a psychotherapy session. This investigation was developed in order to explore a new possible perspective and methodology for studying the expression of emotions. More specifically, literature concerning synchronization of in-session non-verbal variables emphasises its positive correlation with empathy and therapeutic outcomes. We compared the dynamics of galvanic skin response (GSR) and linguistic prosody, chosen as indicators of emotional expression in different domains. We studied their synchronization through complementary methodologies: Recurrence Quantification Analysis (RQA) and Principal Component Analysis (PCA), Markov Transition Matrix (MTM) and Cross-Recurrence Quantification Analysis (CRQA). We investigated the nonlinearity of GSR in terms of self-similarity and power-law, as emerged in autocorrelation functions and signal variations. We considered time-lagged correlations as a measure of dynamical systems' memory. This article concludes by highlighting the importance of a deeper study of all variables related to the psychotherapeutic process and their synchronization in order to extend our knowledge of general human dynamics.