In this paper, we have developed a novel and simple method to quantify the ability to selectively activate our muscles in an effective pattern to achieve a particular task. In the context of this study, we define an effective pattern as that in which muscles whose mechanical contribution to the task is greatest, are mostly active, while the antagonist muscles are mostly silent. This new method uses biomechanical parameters to project the multi-channel EMGs into a three-dimensional artificial torque space, where the EMGs are represented as muscle activation vectors. Using the muscle activation vectors we defined a simple scalar, the muscle selection index, to quantify muscle selectivity. We demonstrate that by using this index we are able to quantify the muscle selectivity during the generation of isometric shoulder or elbow torques in brain-injured and able-bodied subjects. This method can be used during both static and dynamic motor tasks in a multi-articular musculoskeletal system.