Water diffusion is an important transport property necessary for the design an optimisation of food processes. Its determination is very difficult especially from drying experiments which involved other transport mechanisms. This paper presents the effect of temperature on the water diffusion coefficient in a concentrated casein solution measured by Nuclear Magnetic Resonance (NMR). A dairy concentrated protein solution was obtained by ultrafiltration. The proteins content is 19,9 g proteins/g of water. In order to avoid any bacterial development, sodium azide was added (0,02% v/v). The measurements of water self-diffusion coefficients was performed at various temperatures (5, 20, 30 and 40°C) with a NMR spectrometer operating at 20MHz (The Minispec, Bruker). This spectrometer is equipped with gradients of 4T/m. The method described by Stejskal and Tanner (1965) was used. This method makes it possible to highlight two different behaviours describing the water diffusion : either the diffusion is slowed down by the presence of obstacles or the tortuosity of a gel for example, or the diffusion is restricted by porous structures with walls characterised by a more or less important permeability to diffusion. Whatever the studied temperature, our results show that in the case of solution, there is no restriction on the water diffusion at a distance between 6 and 10 µm approximately. Thus, the reduction of the water diffusion compared to the diffusion of pure water is explained mainly by the presence of casein micelles. The variation of the measured coefficient of water diffusion at various temperatures follows a variation described by a relation of Arrhenius. Between 5 and 40°C, the mechanisms which control the water diffusion are thus not modified. The energy of activation deduced from the law of Arrhenius and the coefficient of water diffusion are discussed compared to the values published for other biological systems.