Much research is ongoing into improving the accuracy of functional algorithms to determine joint centers (JC), but there has been limited testing using human movement data. This paper is in three parts: Part 1, errors in determining JCs from real human movement data using the SCoRE method; Part 2, variability of marker combinations during a punch; Part 3, variability in the JC due to reconstruction. Results indicate determining the JC of the shoulder or elbow with a triad of markers per segment with an accuracy greater than 20 mm is unlikely. Part 2 suggests conducting a pilot study with abundant markers to obtain triads, which are most stable due to differences of 300�400% in variability between triads. Variability due to the choice of reference frame for reconstruction during the punch ranged from 2.5 to 13.8 mm for the shoulder and 1.5 to 21.1 mm for the elbow. It would appear more pertinent to enhance the practical methods in situ than to further improve theoretical accuracy of functional methods.