Many soil mapping units (MU) have not been adequately sampled to provide a true measure of their variability; therefore, their descriptions must be regarded as incomplete, and valid statistical comparisons cannot be made with other closely related MU. The number of samples required to detect the differences in means of 18 soil properties between Brantford and Beverly Silt Loam MU were calculated and they ranged from 4 at the 80% probability level (10 at the 95% probability level) for organic matter content of the Ap horizon to several thousand for pH of the Ap horizon. Calculation of required sample numbers indicated that sufficient samples had been collected to make valid statistical comparisons between seven of the soil properties. All seven properties were found to be significantly different between the two MU at the 95% probability level. However, only two of the properties, hue and organic matter content of the Ap horizon, had distinctly different modal values between the two MU and neither of these properties is easily measured in the field. Therefore, it was concluded that the 18 soil properties examined were impractical and unreliable criteria for separating the MU in the field. But, the MU separations can be readily and validly made on the basis of landscape position.