OBJECTIVE To assess the intraobserver reliability (in terms of hour-to-hour and day-to-day reliability) and the interobserver reliability with 3 palpation procedures for the detection of spinal biomechanic dysfunction in the upper 8 segments of the thoracic spine. DESIGN A repeated-measures design was used in all substudies. SETTING Department of Nuclear Medicine, Odense University Hospital, Denmark. PARTICIPANTS Two chiropractors examined 29 patients and 27 subjects in the interobserver part and 1 chiropractor examined 14 patients and 15 subjects in the intraobserver studies. INTERVENTION Three types of palpation were performed: Sitting motion palpation and prone motion palpation for biomechanic dysfunction and paraspinal palpation for tenderness. Each dimension was rated as "absent" or "present" for each segment. All examinations were performed according to a standard written procedure. RESULTS Using an "expanded" definition of agreement that accepts small inaccuracies (+/-1 segment) in the numbering of spinal segments, we found--based on the pooled data from the thoracic spine--kappa values of 0.59 to 0.77 for the hour-to-hour and the day-to-day intraobserver reliability with all 3 palpation procedures. Kappa coefficients were 0.24 and 0.22 for the interobserver reliability with prone and sitting motion palpation and 0.67 and 0.70, respectively, with paraspinal palpation for tenderness. CONCLUSION With expanded agreement we found good hour-to-hour and day-to-day intraobserver reliability with all 3 palpation procedures and good interobserver reliability for paraspinal tenderness. The interobserver reliability was unacceptably poor with prone and sitting motion palpation.