This paper discusses ongoing work towards a theoretical basis intended to facilitate the development of self-regulating adaptive systems. Self-regulation refers to the capacity of the system to assess the effects of, and modify, its own adaptive behaviour in prescribed ways at run-time. Although not new, the concept of self-regulation is largely missing from existing adaptive systems, arguably due to the perceived complexity involved in its theoretical grounding and practical implementation. The paper addresses in particular the following two questions: What are the operational requirements of self-regulating adaptive systems? What implications does self-regulation impose on the modelling- and decision making- approaches used? The theoretical benefits of "clusters" of self-regulating systems, and the role of human experts in the self-regulation process are also briefly discussed.