The dependencies on the mobile phase flow velocity of the efficiency of a column packed with shell particles of neat porous silica (Halo) was measured under two different sets of experimental conditions. These conditions corresponded to the retention mechanisms of per aqueous liquid chromatography (PALC) at low acetonitrile concentrations and of hydrophilic interaction chromatography (HILIC) at high acetonitrile concentrations. The results are compared. Small amounts of a diluted solution of caffeine were injected in order to record the chromatograms under strictly linear conditions. These efficiencies were measured in both water-rich (PALC retention mechanism) and acetonitrile-rich (HILIC mechanism) mobile phases for the same retention factors, between 0.25 and 2.5. The mobile phases were mixtures of acetonitrile and water containing neither supporting salt nor buffer component. At low retention factors, the efficiency of caffeine is better in the PALC than in the HILIC mode. For k(')=0.5, the minimum reduced height equivalent to a theoretical plate (HETP) is close to 2.5 in PALC while it exceeds 5 in HILIC. The converse is true for high retention factors. For k(')>2.5, the HETP is lower in HILIC than in PALC, because the major contribution to band broadening and peak tailing in this latter mode originates from the heterogeneous thermodynamics of retention and eventually restricts column performance in PALC. Most interestingly, the reduced HETP measured in HILIC for caffeine never falls below 4. This suggests that the mass transfer of caffeine between the multilayer adsorbed phase (due to the interactions of the strong solvent and the silanol groups) and the acetonitrile-rich bulk eluent is slow.