The ferritic steels used for reactor pressure vessels undergo a marked transition from ductile to brittle fracture behaviour over a relatively narrow temperature range. For most unirradiated mild steels the ductile to brittle transition temperature (d.b.t.t.) is between -50 degrees and 20 degrees C. The process of irradiation hardening, through the formation of clusters of intersitial or vacancy defects, increases the friction stress of these steels and thereby raises the transition temperature. Given the inherent tendency of these steels to fail in a brittle manner, the raising of the transition temperature under neutron irradiation poses a problem of considerable technological importance in the nuclear industry. At the time (1962) when the first of the Central Electricity Generating Board (C.E.G.B.) Magnox nuclear stations began operation the phenomenon of brittle fracture was already comparatively well understood. A theory of the process had already been developed and applied to the problem of radiation embrittlement. However, as the results from the Magnox pressure-vessel surveillance scheme accumulated, it gradually became evident that the measured changes in yield stress in the monitoring specimens could not be accounted for simply on the basis of irradiation hardening through the formation of damage clusters. By the late 1970s, sufficient data had been gathered from the surveillance programme to enable a detailed investigation of the processes occurring in the Magnox steels to be instituted. The form of the investigation was to subsequently evolve into two phases; an initial comprehensive microstructural study of the steels, followed by the formation of an interpretative model based on the observations. In this paper we present the Magnox yield-stress monitoring measurements and then briefly describe the principal findings from our microstructural studies. The Magnox pressure-vessel steels contain between 0.05 and 0.4% by mass of copper and we show that under certain conditions this element may precipitate as small spherical particles within the matrix of the steels. A review of previous work on copper precipitation in ferrite is then followed by a description of our model. This assumes that the changes in yield stress generally arise from the combined effects of irradiation damage loops and copper precipitates. The formation of the latter may be enhanced by irradiation and in some steels their contribution is dominant. It is shown that the model successfully accounts for the measurements made on both plate and weld steels in all the Magnox stations. Experimental support for the model comes from our own microstructural observations and from other studies, in the U.K. and elsewhere, using techniques which allow the detection of sub-microscopic particles in steels. The model may be applied to pressure-vessel steels in other reactor systems. Indeed, it predicts that the yield-stress changes in steels with a high copper content irradiated under p.w.r. (pressurized water reactor) conditions will be dominated by the contribution from copper precipitation.