The precracking structures developed in a creep-ruptured CrMo ferritic steel have been investigated metallographically using surface preparation methods that can be expected to produce truly representative surfaces. It is concluded that the structures are grain boundary carbides that have decohered from the matrix ferrite during creep straining. Traditional void cavities were not present. It is established, however, that several commonly used metallographic preparation procedures develop artifacts at the sites of decohered carbides, which could be mistaken for void cavities. The development of these artifacts greatly enhances the detectability of decohered carbides but gives a false impression of their size and destroys the evidence on which the nature of the precracking structures could be elucidated. Methods of reliably investigating the structure of decohered carbides, which is somewhat difficult, or of merely detecting their presence, which is comparatively simple, are discussed in the light of this new information, as is also the consequence of their presence instead of voids on the methods currently used in estimating the residual life of affected components. The implications to theoretical considerations of creep failure are also discussed, a number of questions being raised that require further investigation. Foremost among them is the frequency with which, and the reason why, carbide decohesion occurs without subsequent growth of significant voids. The morphology of the boundary carbides that grow during the earlier stages of creep appears to be central to these matters, and this, in turn, is in need of investigation. It apparently has to be accepted until these questions are settled that two mecanisms of creep failure are possible in these steels.