This paper makes use of the randomized allocation of workers on sick leave in Denmark into self-management support, to examine the role of beliefs about control for prolonged absenteeism due to illness. Our results demonstrate that the ability of the intervention to lead sick-listed workers toward resuming employment crucially depends on workers' control beliefs. The intervention increases the perception of control among control pessimists and substantially accelerates the decision to return to work. Furthermore, we identify a group of control-optimist workers for whom "learning" about control beliefs is self-defeating, and leads them toward reduced capacity in terms of return-to-work performance.