|Abstract||Musical ensembles, such as a string quartet, are a clear case of music performance where a joint interpretation of the score as well as joint action during the performance is required by the musicians. Of the several explicit and implicit ways through which the musicians cooperate, we focus on the acoustic result of the performance – in this case in terms of dynamics and intonation - and attempt to detect evidence of interdependence among the musicians by performing a computational analysis. We have recorded a set of string quartet exercises whose challenge lies in achieving ensemble cohesion rather than correctly performing one’s individual task successfully, which serve as a ‘ground truth’ dataset; these exercises were recorded by a professional string quartet in two experimental conditions: solo, where each musician performs their part alone without having access to the full quartet score, and ensemble, where the musicians perform the exercise together following a short rehearsal period. Through an automatic analysis and post-processing of audio and motion capture data, we extract a set of low-level features, on which we apply several numerical methods of interdependence (such as Pearson correlation, Mutual Information, Granger causality, and Nonlinear coupling) in order to measure the interdependence -or lack thereof- among the musicians during the performance. Results show that, although dependent on the underlying musical score, this methodology can be used in order to automatically analyze the performance of a musical ensemble.