Abstract.The laboratory, spatial and temporal components of variation associated with sampling soil for the measurement of pH, organic matter and extractable P, K, Mg, S, Cu, and Co were studied over two years using soil samples from 15 farms in S.E. Scotland. On each farm a selected field was divided into 4–8 sectors, and sampled three times each year, in June, August and October, by bulking 25 cores taken in a ‘W’ pattern. Analysis of variance showed that inter‐field variation was greater than that between sampling dates for most of the properties measured. Restricted Maximum Likelihood Estimation showed that for all elements except K and S the variation between fields was greater than that within a field. Temporal variation was usually smaller than spatial, but K and Co showed similarly small temporal and spatial variations. Variation associated with laboratory procedures was much less than either spatial or temporal variation except for S, most of the total observed variation of which resulted from laboratory error. It is suggested that the most cost effective field sampling technique is to split a field into sectors, sample each individually and analyse a bulked sample made up from the