The chloride content at which reinforcement corrosion occurs (Ccrit) is an important parameter for the determination of the service life of a concrete structure. Many studies have been conducted on Portland cement concrete but few data are available concerning slag concrete. In this study, a method for determining the Ccrit value is proposed. It uses capillary suction and a diffusion phase with steel-potential monitoring during the progressive contamination of the concrete by chloride. The same experimental process, capillary suction then diffusion, was performed simultaneously with tap water and chloride solution to verify the specific behaviour of specimens due to the presence of chloride. During the phase of diffusion and potential monitoring, samples were kept immersed in saturated conditions. The criterion for detecting corrosion initiation was a variation of about -250 mV (Ghods et al., Corrosion Science 52:1649-1659, 2010), which corresponds to the difference usually noted between steel in the passive and active state. Two concrete compositions were investigated: one with only Portland cement as the binder and one incorporating slag with a substitution level of 60 %. As expected, in the Portland cement concrete, a drop in potential occurred when corrosion started. For slag concrete, no potential drop was recorded because the initial potential of the steel before the introduction of chlorides was lower. The method for determining Ccrit, based on the fall in potential recorded when corrosion started, would not be suitable to compare the different compositions of binder in saturated conditions. The presence of sulphides in the pore water solution of the slag concrete is a possible explanation of this result. Tests in a solution with added sulphides tended to confirm this hypothesis.