Laboratory column tests using $Cl^-$ tracer were conducted to study the correlation of soil particle distribution and hydrodynamic dispersion mechanism with three kinds of ununiformed soil samples, in which the ratio of gravel and sand versus silt and clay is 24.5 for S-1 soil, 4.48 for S-2 soil, and 0.4 for S-3 soil. Chloride breakthrough curves with time were fitted with gaussian functions. The relative concentrations of chloride were converged to 1.0 after 0.7 hours for S-1, 6.3 hours for S-2, and 389 hours for S-3. Average linear velocity, longitudinal dispersion coefficient, and longitudinal dispersivity were calculated by chloride breakthrough curves. Longitudinal dispersion coefficients were $1.20{ imes}10^{-4};m^2/sec$ for S-1, $8.87{ imes}10^{-7};m^2/sec$ for S-2, and $1.94{ imes}10^{-9};m^2/sec$ for S-3. Peclet numbers calculated by the molecular diffusion coefficient of chloride and the mean grain diameters of soils were $2.59{ imes}10^2$ for S-1, $6.27{ imes}10^0$ for S-2, and $1.35{ imes}10^{-4}$ for S-3. Mechanical dispersion was dominant for the hydrodynamic dispersion mechanism of S-1. Both mechanical dispersion and molecular diffusion were dominant for the hydrodynamic dispersion mechanism of S-2, but mechanical dispersion was ascendant over molecular diffusion. Hydrodynamic dispersion in S-3 was occurred mainly by molecular diffusion. When plotting three soils on the graph of $D_L/D_m$ versus Peclet number produced by Bijeljic and Blunt (2006), the values of $D_L/D_m$ for S-1 and S-2 were more than 2.0 order compared to their graph. S-3 was not plotted on their graph because the Peclet number was as small as $1.35{ imes}10^{-4}$.