Given total endurances of low cycle fatigue specimens, several empirical formulae enable an estimate to be made of 'initiation' cycles to a specific crack depth in a zone ranging from 20 to 400 mm from the surface. For greater depths, the cyclic growth behaviour of cracks can be calculated directly from relations using established constants (including the effects of creep dwell) for a wide range of alloys. However, below a critical depth of 200 mm or thereabouts, these empirical growth laws, expressed in terms of total strain range, break down. Early crack growth can take place discontinuously, as the crack tip encounters grain boundaries or other obstacles. Such information arises from replica takings, or in situ microscopic observations of surfaces during continuous cycling and creep-fatigue tests. Striation spacings on the fracture surface and the first observed striation spacing provide valuable indications of early crack growth. In this region, the cyclic rate of growth depends on (i) crack shape which can change during advance; (ii) whether or not embryo cracks link up (coalescence); (iii) whether or not there is a small starter defect (deliberately introduced or otherwise); (iv) the degree of creep or oxidation damage; and (v) the repeat distance of obstacles or barriers. Creep effects can overtake these considerations by bulk damage accruing with consequent acceleration in growth. In this survey, crack growth characteristics in ferritic and austenitic steels up to 625uC and superalloys up to 1000uC are considered. Surface behaviour is connected to penetration into the depth. As part of a sensitivity analysis, various 'stop-start' models are examined and their related growth cycles calculated in the early region before the macroscopic growth law takes over. These are compared with an upper bound constant growth rate law. Alternatively, it is shown that crack growth can be described over the whole range by a single exponential function.