A commercially available salinity sensor (model EM-38 of Geonics Ltd., Canada) has been developed for diagnosing and quantifying soil salinity in the field. There is a need, however, to calibrate the instrument reading against a more commonly used measure of soil salinity. Linear regression equations are reported that allow the interpretation of field instrument readings in terms of electrical conductivity of the saturation extract (ECe) measured in the laboratory. Instrument measurements and soil samples were taken at some 110 sites on various irrigation schemes distributed across South Africa. Samples were analysed in the laboratory and soils categorized according to texture, water status and salinity distribution with depth. Calibration equations were developed firstly by relating field instrument readings, taken either in the vertical or horizontal position, to a depth-integrated ECevalue which was weighted for depth according to instrument response. Secondly, the mean of the readings taken in the vertical and horizontal positions at each site was related to the arithmetic mean ECefor the 0- to 1.2-m soil depth. A set of 12 linear regression equations was established in the first approach, and eight in the second. Two of the relationships showed statistical significance at the 5% level, the remainder at the 1% level. These equations enable the user of the EM-38 to derive a realistic index of soil salinity in terms of ECe, but calibration of the instrument for local soil conditions is preferable.