The purpose of this study was to determine the sensitivity and the specificity of 25 measures, known as the NIOSH Low Back Atlas (LEA), previously accepted as reliable and clinically useful for assessing persons with low back pain (LBP) due to sprain or strain. 114 patients who were referred to 5 different clinics agreed to participate in this study and were primarily tested by 7 different physical therapists. 112 control subjects were recruited through announcements, ads and word of mouth, and were primarily tested by another 3 physical therapists. We attempted to match the patient and control groups on age and gender.. The participating therapists were trained on every test item, checked out by videotaping procedures, and rechecked 2-3 times during the course of the study. Additionally 5% of the patients and 5% of the control subjects were tested as cross-over subjects by therapists not suspecting their true status. From the rechecks and the cross- over studies, reliability of the testing therapists was assessed. 80% of the items were acceptable and 20% were marginal - the same proportion as had been reported by the original study which developed the NIOSH Low Back Atlas. Al though our methods for achieving and assessing reliability differed, we bel ieve that the results reported here are based on test items which were performed with acceptable consistency by a group of practioners in different treatment settings. 24 of the 50 component items of the LEA test and measures were found to significantly discriminate between the patient and control groups (p<.03), with individual sensitivities ranging between 59.2 and 74.3%, and specificities ranging between 74.3 and 95.4%. Forward stepwise regression identified 7 LBA variables which, when used in conjunction, improved the overall sensitivity to 87%, the specificity to 93%, and correctly identified 90% of subjects with low back pain. The 7 items were related to pain changes with prone press up, symmetry of pelvic level, and limitations in hip joint internal rotation, lumbar mobility and lower abdominal strength. An additional purpose was to examine the heterogeneity of the LBP patient sample in order to describe homogeneous clusters of patients exhibiting similar signs. Cluster analysis was performed using all of the individual variables and then with certain combinations of the variables grouped because of significant correlations. Alternative clusterings were tried and evaluated by 2 therapists who agreed on a 4 cluster model which described patients who demonstrate: 1) weakness, tightness, restricted motion in hips and trunk in all planes 1 2) weakness, flexibility, increased lordosis, worsened pain after prone extension despite some centralizat_on: 3) no weakness, lowest number reporting pain with prone extension, restricted trunk motion and hip joint internal rotation, unequal pelvic heights, decreased lordosis: and 4) no weakness, highest number reporting central ized pain. and improvement with prone extension, reduced lumbar motion in sagittal plane and reduced straight leg raise. Analysis is continuing to further describe these groups, using additional data obtained through histories and questionnaires.