A consequence of the growing success in producing alloys that deform very slowly even at high stress and temperature is that the problem of creep fracture has loomed larger. There is a fracture mechanism quite distinctive to creep, in which tiny holes nucleate and grow by some means until they are so large, or sufficiently linked together, that the metal breaks. The speed of this fracture process increases with temperature and stress, and evidently also depends on composition in a complicated way, which has made possible a certain degree of manufacturing control if not of understanding. In essence, the formation and growth of the holes is a phase change in which stress provides the driving free energy since the holes enlarge the overall dimensions. Both the nucleation and growth rates help to determine the time to fracture, which is what really matters. There are several obscure points such as the nature of many of the nucleating sites and the strong influence of deformation rate during nucleation and growth. Both theoretically and practically, understanding of the youthful problem of creep fracture is less mature than that of the older problem of creep deformation.' Creep ' is the name given to the slow deformation and slow fracture that takes place in metals under sustained load, particularly when hot. I n the light of figure 1 it is no surprise that the real study of creep commenced shortly after the turn of the century with two papers by Andrade (1910, 1914), followed by one by Dickenson (1922). These three papers formed, as it were, the head-waters of a stream which swelled as it ran through time, until now it is a very large river of publications. The purpose of this article is to describe the state of physical understanding of creep processes which this river of information has produced.The river has two main channels: that concerned with creep deformation and that concerned with the special type of fracture peculiar to creep.