Abstract An improved understanding of the biopsychosocial influences that contribute to and maintain pain has promoted the development of new efficacious treatments for chronic low back pain (CLBP). This study aimed to investigate the mechanisms of a new treatment—education and graded sensorimotor retraining—on pain and disability. We conducted a preplanned causal mediation analysis of a randomized clinical trial which allocated 276 participants with CLBP to 12 weekly clinical sessions of education and graded sensorimotor retraining (n = 138) or a sham and attention control (n = 138). Outcomes were pain intensity and disability, both assessed at 18 weeks. Hypothesized mediators included tactile acuity, motor coordination, back self-perception, beliefs about the consequences of back pain, kinesiophobia, pain self-efficacy, and pain catastrophizing, all assessed at the end of treatment (12 weeks). Four of 7 mechanisms (57%) mediated the intervention effect on pain; the largest mediated effects were for beliefs about back pain consequences (−0.96 [−1.47 to −0.64]), pain catastrophizing (−0.49 [−0.61 to −0.24]), and pain self-efficacy (−0.37 [−0.66 to −0.22]). Five of 7 mechanisms (71%) mediated the intervention effect on disability; the largest mediated effects were for beliefs about back pain consequences (−1.66 [−2.62 to −0.87]), pain catastrophizing (−1.06 [−1.79 to −0.53]), and pain self-efficacy (−0.84 [−1.89 to −0.45]). When all 7 mechanisms were considered simultaneously, the joint mediation effect explained most of the intervention effect for both pain and disability. Optimizing interventions to target beliefs about the consequences of back pain, pain catastrophizing, and pain self-efficacy is likely to lead to improved outcomes for people with CLBP.