The objective of the study was to compare the precision of shoulder anatomical landmark palpation using a CAST-like method and a newly developed anatomical palpator device (called A-Palp) using the forefinger pulp directly. The repeated-measures experimental design included four examiners that twice repeated measurements on eleven scapula and humerus anatomical landmarks during two sessions. Inter-session and inter-examiner precision was determined on volunteers. A-Palp accuracy was obtained from in vitro measurements and using virtual palpation on 3D bone models. Error propagation on the motion representation was also analyzed for a continuous motion of abduction movement performed in the shoulder joint. Palpation results showed that CAST and A-Palp methods lead to similar precision with the Maximal A-Palp calibration error being 1.5mm. In vivo precision of the CAST and A-Palp methods varied between 4mm (inter-session) and 8mm (inter-examiner). Mean propagation of the palpation error on the motion graph representation was 2 degrees and 5 degrees for scapula and humerus, respectively. A-Palp accuracy was 3.6 and 8.1mm for scapula and humerus, respectively. The A-Palp seems promising and could probably become an additional method next to today's marker-based motion analysis systems (i.e., Helen-Hayes configuration, CAST method).