Investigation of the behaviour of multiple stress corrosion cracks for two different metal-environment combinations has emphasised the uncertainties associated with the development of individual cracks, concerning whether cracks arrest or grow and what growth rate is relevant. The matter is further complicated by the coalescence of appropriately spaced cracks, the conditions for such having been determined and shown to be the same for both systems. A consequence of the variable behaviour of individual cracks is that it is difficult to derive a simple formula for describing the behaviour of multiple cracks, or to assess the overall response of all the cracks from a precise description of the behaviour of each individual crack. This difficulty, in turn, implies that it will inevitably be necessary to invoke some statistical approach in any model that is to be used reliably for dealing with multiple cracks when life prediction is the ultimate goal.