Objective: Cognitive therapy (CT), mindfulness-based stress reduction (MBSR), and behavior therapy (BT) for chronic pain treatment produce outcome improvements. Evidence also suggests that changes in putative therapeutic mechanisms are associated with changes in outcomes. Nonetheless, methodological limitations preclude clear understanding of how psychosocial chronic pain treatments work. In this comparative mechanism study, we examined evidence for specific and shared mechanism effects across the three treatments. Method: CT, MBSR, BT, and treatment as usual (TAU) were compared in people with chronic low back pain (N = 521). Eight individual sessions were administered with weekly assessments of “specific” mechanisms (pain catastrophizing, mindfulness, behavior activation) and outcomes. Results: CT, MBSR, and BT produced similar pre- to posttreatment effects on all mechanism variables, and all three active treatments produced greater improvements than TAU. Participant ratings of expectations of benefit and working alliance were similar across treatments. Lagged and cross-lagged analyses revealed that prior week changes in both mechanism and outcome factors predicted next week changes in their counterparts. Analyses of variance contributions suggested that changes in pain catastrophizing and pain self-efficacy were consistent unique predictors of subsequent outcome changes. Conclusions: Findings support the operation of shared mechanisms over specific ones. Given significant lagged and cross-lagged effects, unidirectional conceptualizations—mechanism to outcome—need to be expanded to include reciprocal effects. Thus, prior week changes in pain-related cognitions could predict next week changes in pain interference which in turn could predict next week changes in pain-related cognitions, in what may be an upward spiral of improvement.