Vadose zone soils in arid regions often contain a high gravel fraction (>2 mm size). This paper examines, for coarse-textured soils, variability of the slope &agr; of Gardner's <1958> unsaturated hydraulic conductivity K versus matric potential &psgr; relation, K=Kse&agr;&psgr;. The steady state head control method was used to obtain unsaturated K for tensions as high as 386 cm. A total of 79 samples were analyzed in the laboratory; the gravel fraction for the 41 gravelly samples ranged from 20 to 71% (by weight); the remaining 38 samples were sandy with no gravel fraction. The measured K values for the gravelly samples fall within a narrow range and are well within the range of measured K values for the sandy samples. The bulk of both gravelly and sandy samples can be described using a single-slope exponential model; Gardner's &agr; and the intercept K0 are based on a least squares fit to the late time measurements at lower &psgr;. The mean and variance of &agr; for the gravelly samples are lower compared with the sandy type. The &agr; values for the gravelly soils follow a normal distribution, whereas the K0 estimates and the laboratory-measured Ks follow lognormal distributions. For the sandy type, &agr;, K0, and Ks follow lognormal distributions. The &agr; values appear to be related to the median particle diameter d50 and the slope of the particle size distribution. A clear reduction in variability of &agr; is apparent with an increase in d50, suggesting that the unsaturated K and &agr; variability for the gravelly soils can be estimated within narrow ranges. The flow-weighted characteristic pore size &lgr;m is clearly dependent on the soil type; &lgr;m values for the gravelly soils are characterized by a smaller mean and variance than for the sandy type. This helps explain the finding that the range in unsaturated K for the gravelly soils is less than for the sandy type. At higher tensions the mean and variance of &agr; and ln K0 as well as the cross correlation between &agr; and ln K0 are important inputs to stochastic models of unsaturated flow. The implications of their variability on flow in heterogeneous media are discussed. ¿ 2001 American Geophysical Union