Experimental water content profiles established during infiltration into a horizontal sand column were compared with theoretical ones calculated from the soil-water diffusivity, which was determined independently of the infiltration process. Discrepancies between the theoretical and experimental profiles were found. In particular, the experimental profiles for small times showed a retardation of the penetration which became more pronounced as the water content increased, and no unique relationship between the water content and the variable λ = x/(t)½ (x being the distance and t the time) was found. Thus the experimental profiles gave incorrect and varying soil-water diffusivities. The comparison of the actual flow rate q with the rate q′ given by the product of the hydraulic conductivity and the observed potential gradient showed that q = q′, where 0 < α ≤ 1, a varying with the water content and the potential gradient. The effect of the ‘less than proportional’ flow on the profile development is discussed. It is argued that the infiltration process at relatively late times was conditioned by the profile development during the very early stages. Previous experimental infiltration work is discussed, and it is argued that two mechanisms able to produce non-Darcian behavior may be responsible for the various reported infiltration patterns. Finally, the results of the present work are discussed in terms of a critical Reynolds number where the flow becomes non-Darcian.