The effect of clay content on the static filtration properties of purified Na-montmorillonite suspensions was studied at two different pressures. The results of filtering parameters are as expected. When the clay percentage increases, the thickness ( e) of the obtained cake increases, leading to the decrease of water loss (WL), and consequently, to the decrease of permeability ( k). These decreases of WL and k are more important for the low applied pressure than for the high one. The differences between values of the different experimental parameters (WL and e) obtained at applied pressures of 5.7×10 5 and 1.5×10 5 Pa, as a function of clay content in the initial samples, show some singular points. At the clay content corresponding to the sol–gel transition (SGT), already determined by rheological measurements, ΔWL presents a change in concavity and Δ e shows a maximum. Localisation of this maximum seems to be a simple way to detect the SGT by performing static filtration experiments of two series' of clay suspensions at two different pressures. According to the water contents obtained at the two pressures, all the cakes were localised in region (2), defined by Norrish (1954) as the “macroscopic” swelling paste of Na-montmorillonite. At low applied pressure, the texture of the cakes is identified by transmission electron microscopy (TEM). Generally, at low magnification, all samples are globally homogeneous. However, they present some regions more dense than others. At high magnification, the number of layers per particle seems to depend on SGT. Indeed, in the cake corresponding to 1% w/w of clay in the initial suspension, isolated layers are most frequent and particles with two to four layers per particle are scarce. In the cake obtained from the gel (3% w/w of clay) the most frequent stacking is about three layers per particle, and isolated layers almost disappear. In the cake obtained from gel, at the SGT (5.5% w/w of clay) which is the less oriented sample, the most frequent stacking is statistically also around three layers per particle but with a greater number of particles, some isolated layers appear again. After the sol–gel transition, in the cake obtained from the 7% sample (which is the most oriented sample), the number of particles increases but the number of layers in each particle decreases (two layers per particle most frequently occurs), and the number of isolated layers again rises. The frequent distances between the aggregates and particles were estimated on the TEM images of the cakes obtained at 1.5×10 5 Pa. The distances between the particles of the cakes obtained under 5.7×10 5 Pa was identified by X-ray diffraction (XRD). All these distances measured by the two techniques are compared with a linear extrapolation of Norrish's (1954) experimental data. At SGT, the cakes obtained at the two pressures correspond to aggregates connected with several distance configurations. However, the cakes obtained from the 3% initial clay content always show two distances (19 and 39 Å) at high pressure. At low pressure, one distance is about 300 Å, and the other is about 65 Å. This last distance disappears in the cake obtained from the more diluted suspension (1%) and only the higher distance persists. The 3% initial clay suspension seems to correspond to a “local osmotic transition”, which occurs between aggregated particles (or small aggregates) in the semi-diluted suspensions and isolated particles and/or layers in the more diluted suspensions.