Background. Chronic low back pain remains a major health problem in Australia and around the world. Unfortunately the majority of treatments for this condition produce small effects because not all patients respond to each treatment. It appears that only 25-50% of patients respond to exercise. The two most popular types of exercise for low back pain are graded activity and motor control exercises. At present however, there are no guidelines to help clinicians select the best treatment for a patient. As a result, time and money are wasted on treatments which ultimately fail to help the patient. Methods. This paper describes the protocol of a randomised clinical trial comparing the effects of motor control exercises with a graded activity program in the treatment of chronic non specific low back pain. Further analysis will identify clinical features that may predict a patient's response to each treatment. One hundred and seventy two participants will be randomly allocated to receive either a program of motor control exercises or graded activity. Measures of outcome will be obtained at 2, 6 and 12 months after randomisation. The primary outcomes are: pain (average pain intensity over the last week) and function (patient-specific functional scale) at 2 and 6 months. Potential treatment effect modifiers will be measured at baseline. Discussion. This trial will not only evaluate which exercise approach is more effective in general for patients will chronic low back pain, but will also determine which exercise approach is best for an individual patient. Trial registration number. ACTRN12607000432415.