Life prediction of composite structures is still heavy in empiricism. In order to reduce the burden of extensive data generation and proof testing, and to thereby reduce cost, resort must be made to predictive, physically based models. A field with this as a central objective, and with the capability to merge elements of materials science and mechanics, is damage mechanics. This paper will address how specifically damage mechanics may be able to render life prediction methodologies a scientific basis. First, we shall discuss the essential, sequential and interactive elements of a life prediction methodology relevant to composite structures, identifying the role of damage mechanics in such a scheme. We shall then outline a philosophy (conceptual structure) of an approach to life prediction. A simple scenario will then be treated to illustrate the ideas involved. Finally, the multiple ways by which damage mechanics can be implemented will be discussed. The limitations inherent in isolationist (pure) forms of damage mechanics will be discussed, laying bare the need for synergistic approaches. One such approach, and the progress made in achieving it, will then be described. The references cited below report related work by this author and his associates.