Discography is the only available method of directly challenging the discs for pain sensitivity. However, it is invasive, expensive and there is a debate about its clinical value. There is a need to identify clinical signs that may indicate the need for a discography examination. Pain centralization (retreat of referred pain to towards the spinal midline) has been associated with positive discography and possesses specificity between 80 and 100% depending on levels of disablement and psychosocial distress. Less than half of positive discography patients reports centralization and there is a need to identify other potentially valuable predictors. This study estimated the diagnostic accuracy of clinical variables in relation to provocation discography. In a blinded, prospective reference standard design in a private interventional radiology clinic, chronic low back patients received a detailed clinical examination followed by discography. Patients were typically disabled with high levels of psychosocial distress. Logistic regression modelling identified potentially useful clinical variables, and sensitivity, specificity and likelihood ratios were calculated for promising specific variables and combinations of variables. Of 118 consecutive discography cases, 107 had unconfounded data on discography results. History of persistent pain between acute episodes, a significant loss of extension and a subjective report of 'vulnerability' in what is termed as the 'neutral zone' had specificities of 83-92% and likelihood ratios between 2.0 and 4.1. Two combinations of variables were highly specific to positive discography, but no suitable screening test useful for ruling out positive discography was established. Three clinical variables have modest predictive power in relation to lumbar discography results and two combinations of variables were highly specific to positive discography.