Based on the suggestion by R. A. Bagnold that the normal probability idea firmly associated with sand size distribution must be revised, the present author attempted to examine grain size distribution of sediment in transit: wind blown sand and water driven sand. The samples of blown sand were collected by trapping sand grains which had cleared the brink and fallen on the slip-face of a typical dune on the Nakatajima Coast, facing the Pacific Ocean. The samples of water driven sand were collected in the Shinano River just downstream of the junction with a main tributary, the Uono River; the collection work was conducted using a hand-made sampler lowered down from a boat during a small flood due to melting snow. Grain size analysis of the samples collected was carried out using a settling-tube measuring system installed at the Institute of Geoscience, University of Tsukuba. Results of the analysis are plotted in Fig. 2 (wind blown sand) and Fig. 3 (water driven sand). These two figures show that each of the analyzed samples contains sand-particle population and, except a few samples of blown sand, admixes finer-particle population. The sand-particle population shows conspicuous regularity in grain size distribution. The points of this population align along the two asymtotes (Figs. 2 and 3); this suggests that the grain size distribution may follow the log-hyperbolic distribution. However, there still remains the possibility that the grain size distribution belongs to the log-normal one. The author, then, examined this possibility applying the following method. At first, the size frequency data were plotted on normal probability paper (Fig. 4-A), and its phi-mean Mψand phi-standard deviation σψ. were obtained. For the case of samples admixed finer-particle population (Fig. 4-B), the sand-particle population was properly partitioned of from the finer-particle population by the method previously described (Inokuchi, 1977). Based on the values of Mψ and σψ, theoretical log-normal curves were drawn as dotted line in Fig. 5 for each sand-particle population except Nos. 1001 and 1002. In this figure in which size classes finer than 3ψand or 3.5ψand are not plotted, the original points seem to fit fairly well to the dotted curve, except size classes around 3ψand which may be affected by grains belonging to finer-particle population. However, a comparison of Figs. 2 and 3 with Fig. 5 indicates that the grain size distribution of the sand-particle population belongs to the log-hyperbolic distribution rather than the log-normal distribution. According to Bagnold (1953), the slope of the coarser wing is denoted by C, and that of the finer wing by S. Their values are shown in Table 3. For all the samples of wind blown sand, C is nearly equal to S, and for the samples of water driven sand, C > S. However, taking into account the effect of admixture of grains originally belonging to finerparticle population on size classes around 3tp, the proper slopes of sand-particle population may also be C _??_ S for water driven samples. It is found that values of C are not at random through all the samples but are grouped into discrete classes, and that there exists a definite ratio between average values of individual classes (Table 3).