Not every problem is complex. Metaheuristics stand out as the adequate rods for solving complex problems. This entails, from the outset, the capacity to identify the very nature of such problems. This paper presents and discusses metaheuristics as specific methodologies in the study of increasingly complex phenomena, and claims that the function of metaheu-ristics consists in solving complex problems. In order to do that, four steps are followed, thus: firstly the very nature of complex problems are identified; secondly, a general view of metaheuristics is set out; thirdly, a view of optimization is introduced; and fourthly, the theory of computational complexity is discussed aiming at clarifying furthermore what the sciences of complexity are all about. At the end some conclusions are drawn.