Easa suggests an excellent method for evaluating the bearing capacity of a shallow strip foundation in which both the effective friction angle + and the soil unit weight y are considered as random variables. This approach represents a step ahead of the method proposed by Cherubini (1990) by which only the variability of + could be considered. A few points, however, are worth noting concerning both the actual variability of y and + observable in laboratory testing and the variability of the same parameters actually involved in the computation model. ( I ) The variability of the soil unit weight is not generally high. The data of Smith (1986), cited by Easa, are integrated by Cherubini et al. (1993), and Fig. 1 shows the histogram of the coefficients of variation of y, i.e., CV,. Table 1 presents the main statistical indices of CV,. Note that most of the CV, values fall within the interval 2.5 < CV < 7.5%. I Both the mean value and the mode are within this interval. 1 The highest value of CV, is 16 and the lowest is 2. (2) The coefficients of variation of the effective friction angle (Cherubini et al. 1993) are shown in the histogram in Fig. 2 and are listed in Table 1. It is observed that the variability of the effective friction angle is greater than that of the soil unit weight. However, if the coefficients of variation of + are plotted versus the corresponding mean values (Fig. 3) then, apart from one outlier, the greatest variability is seen in soils having + < 30°, whereas for soils with 30 2 + 5 41°, the coefficients of variation are between 3 and 15%. (3) It should be noted that variabilities from either laboratory or in situ tests are in fact the highest observable ones, since they result from the sum of the soil's intrinsic variability and the variability induced by sampling and test procedures (Rethati 1988). (4) To be able to give a complete statistical or probabilistic description of a given soil, one must know the mean value of the property being studied, the standard deviation, and the fluctuation scales in the directions of the space under consideration (Vanmarcke 1977). Generally, it is the fluctuation scale in the vertical sense 6, or that in the horizontal sense 6, which plays a fundamental role in the reduction of variance, depending on the specific problem being investigated. Li (1991) believes that typical 6, values range between 1 and 2 m, whereas the observable values of 6, are normally much higher (5-50 m) (Keaveny et al. 1989; Quek et al. 1992). Favre and Genevois (1987) apply the finite-element method to the solution of a bearing-capacity problem for which they