This paper discusses the reasons for skill mix among health workers being important for health systems. It examines the evidence base (identifying its limitations), summarizes the main findings from a literature review, and highlights the evidence on skill mix that is available to inform health system managers, health professionals, health policy-makers and other stakeholders. Many published studies are merely descriptive accounts or have methodological weaknesses. With few exceptions, the published analytical studies were undertaken in the USA, and the findings may not be relevant to other health systems. The results from even the most rigorous of studies cannot necessarily be applied to a different setting. This reflects the basis on which skill mix should be examined--identifying the care needs of a specific patient population and using these to determine the required skills of staff. It is therefore not possible to prescribe in detail a "universal" ideal mix of health personnel. With these limitations in mind, the paper examines two main areas in which investigating current evidence can make a significant contribution to a better understanding of skill mix. For the mix of nursing staff, the evidence suggests that increased use of less qualified staff will not be effective in all situations, although in some cases increased use of care assistants has led to greater organizational effectiveness. Evidence on the doctor-nurse overlap indicates that there is unrealized scope in many systems for extending the use of nursing staff. The effectiveness of different skill mixes across other groups of health workers and professions, and the associated issue of developing new roles remain relatively unexplored.