The concept of dispersive pressure was originally proposed by Bagnold (1954), who showed experimentally that sheared granular material develops a dispersive pressure that is proportional to shear stress and particle concentration. Legros (2002) presented a theoretical analysis suggesting that dispersive pressure cannot account for inverse grading commonly observed in grain flows, modified grain flows or traction carpets, contrary to what has been accepted for decades by many sedimentologists. He pointed out that, if dispersive pressure in a sheared granular mass increases and exceeds applied normal stress or static pressure, net upward force acting upon bed causes it to expand, as has in fact been observed in experiments. Bed expansion immediately causes a decrease in particle concentration and consequently in dispersive pressure, until latter balances static pressure and expansion stops. If dispersive pressure becomes less than static pressure, net downward force acting upon granular bed causes compaction, thus increasing particle concentration and dispersive pressure. This is an important concept enhancing our understanding of an as yet poorly understood process. However, Legros (2002) went further to state that the dispersive pressure must always equal applied normal shear stress at every level (emphasis mine) through shearing granular bed. This implies that whole sheared bed is in constant equilibrium, which is probably not case. In this discussion a modified process is proposed that accounts for bipartite traction carpet structure proposed by Sohn (1997). Traction carpets are moving layers of highly concentrated bedload material developed beneath and driven by an overlying turbulent flow (Dzulynski and Sanders 1962). Sohn (1997), on basis of numerical analyses, computer simulations, field measurements, and experimental studies (Bagnold 1955, 1963; Hanes and Bowen 1985; Wan …