There can be many uncertainties in the application of soil mechanics in geotechnical engineering practice. Soil is a natural (not a manufactured) material, therefore some degree of heterogeneity can be expected within a deposit. A ground investigation may not detect all the variations and geological detail within soil strata so that the risk of encountering unexpected conditions during construction is always possible. Specimens of relatively small size, and subject to some degree of disturbance even with the most careful sampling technique, are tested to model the behaviour of large in-situ masses. Results obtained from in-situ tests can reflect uncertainties due to heterogeneity. Consequently, judgements must be made regarding the accuracy of soil parameters obtained for use in design. In clays the scatter normally apparent in plots of undrained shear strength against depth is an illustration of the problem of selecting parameters. A geotechnical design is based on an appropriate theory which inevitably involves simplification of real soil behaviour and a simplified soil profile. In general, however, such simplifications are of lesser significance than uncertainties in the values of the soil parameters necessary for the calculation of quantitative results. Details of construction procedure and the standard of workmanship can result in further uncertainties in the prediction of soil behaviour.