Creep constitutive equations are discussed in their role as material descriptions forming one of the governing conditions of an inelastic stress analysis. The view taken is that they must not only give an adequate picture of the material behaviour itself, but must also be suitable from the numerical point of view. Furthermore, in practical engineering contexts, constitutive equations must be principally aimed at giving information for structural assessment under realistic loading conditions and subject to various uncertainties. The choice of model is, therefore, not aimed at an exact material description, and the practical choice may be problem-dependent for a given material. Both the classical formulation and the more recent unified formulation are outlined, with discussion of advantages and disadvantages. Some unified formulations are regarded as practically useful but great use is still to be expected of classical forms in which creep and plasticity are regarded as separate. In the latter class descriptions are given of the widely-employed time- and strain-hardening descriptions, of back-stress models, and of certain tertiary creep models including damage variables. The difficulties of anisotropy are also noted. Finally the paper discusses the process of fitting material data to the classical time- and strain-hardening models, and some hints are given for the analyst who may be faced with performing this task for materials of interest.