The steady state creep behaviour of both single phase and TiN dispersion strengthened 20% Cr-25% Ni stainless steel is examined with the object of determining the role of the second phase particles on dislocation network creep. A correlation between the stress/strain rate variation in the two materials can be made by writing. ε ss = A( σ − σ 0) p , where σ is ∼0 and ~80 MPa for single-phase and dispersion hardened steels respectively, and p is ∼5. An expression of this form leads to the prediction of threshold behaviour at σ 0, but the creep rate data indicate that steady state creep also occurs below this stress. Microstructural observations are described which suggest that σ 0 in the TiN bearing alloy can be identified with σ p , the Orowan stress. The Orowan mechanism cannot operate below σ p , but an alternative mode of dislocation bypass around particles is available, which leads to a mechanistic transition at a stress σ gt where the rates of the two processes are the same. The stress/dislocation density relationship is investigated, and it is found that the density p in the dispersion hardened alloy becomes constant below the Orowan stress at a density ( p 0) such that the average network link length is equal to the interparticle spacing. Further, the strain rate/dislocation density variation in single phase material at high stress appears to be such that ε α σ − σ s pα p p 2 while at low stresses there is an excess of dislocations. It is argued that writing ε α ρ − ρ 0 p 2 with ρp 0 the annealed-in density, allows the strain rate to be self-consistent function of the active dislocation content over the entire stress range. Moreover, treating the data from the two-phase steel in the same manner reduces the ε/( ρ − ρ 0) points from both alloys to a single curve. The density ρ 0 is suggested to arise as a result of stress relaxation where dislocations are held up at hard parts of the microstrucure. The stress relaxation produces an increase in the total dislocation line length, but a reduction in the stored energy, so that the driving force for recovery is reduced at a given total density. The relaxation structures themselves must recover, and this is likely to happen over the microstructural wavelength of the plastic inhomogeneity which is not only responsible for the presence of the geometrically necessary dislocations, but which also determines the content of annealed material. The spatial inhomogeneity of the dislocation density can be described in terms of a steady-state internal stress distribution which acts so as to compensate for variations in local flow stress. The threshold behaviour, however, is suggested to arise because the TiN particles are undeformable. The Orowan stress always therefore has to be exceeded before the network model of dislocation creep can be at all appropriate. However, the measured variations in dislocation density with stress imply that the creep rate can be correlated with the dislocation density, as in the network model, if the effects of plastic relaxation are taken into account.