Although, several thorax models have been proposed for clinical gait analysis, none has received widespread acceptance nor been subject to any extensive validation work, especially for the marker set to use. The aim of this study was thus to determine the optimal and minimal makers' placement on the thorax for clinical gait analysis. Ten healthy subjects have performed a series of movements (arm, head, trunk) with large amplitude during walking. Reflective markers were taped on the thorax (C7, T2, T4, T6, T8, T10, T12, sternum, clavicles and ribs) and their 3D positions were captured with an opto-electronic system. Each combination of 3 markers has been tested. The global error of each model was computed with the estimated position of the markers considering the thorax segment as a solid segment. Two families of marker sets were identified with the lowest error. The first family was composed by two anterior and one posterior marker on the thorax (incisura jugularis (IJ), xiphoid process, and T8). The second family was composed by two posterior and one anterior maker (IJ, T2 and T8 or T10). Even, if these two families of marker sets presented a similar error for marker position, the angles obtained from these marker sets showed large differences especially for the axial rotation movement of the trunk (up to 40.1°). The optimal and minimal maker set identified with a variety of large movements of the trunk, head and arms was IJ, T2 and T8 or T10.