The importance of classifying chronic low back pain (LBP) patients into homogeneous sub-groups has recently been emphasized. This paper reports on two studies examining clinicians ability to agree independently on patients' chronic LBP classification, using a novel classification system (CS) proposed by O'Sullivan. In the first study, a sub-group of 35 patients with non-specific chronic LBP were independently classified by two 'expert' clinicians. Almost perfect agreement (kappa-coefficient 0.96; %-of-agreement 97%) was demonstrated. In the second study, 13 clinicians from Australia and Norway were given 25 cases (patients' subjective information and videotaped functional tests) to classify. Kappa-coefficients (mean 0.61, range 0.47-0.80) and %-of-agreement (mean 70%, range 60-84%) indicated substantial reliability. Increased familiarity with the CS improved reliability. These studies demonstrate the reliability of this multi-dimensional mechanism-based CS and provide essential evidence in a multi-step validation process. A fully validated CS will have significant research and clinical application.