The role which microstructural characteristics and clay mineralogy plays on the saturated hydraulic conductivity value (k) of marls from Northern Péloponnèse, measured in the laboratory, is studied. This value must be taken under consideration when severe geotechnical problems are investigated (e.g. embankment foundations, landslides phenomena involving clayey sediments, soil conditioning e.t.c.) or empirical functions are applied to predict hydraulic conductivity from basic geotechnical properties. The marly samples were analysed by x-ray diffraction (XRD) and by thermo-gravimetric methods (DTA, TG) for the determination of mineralogical composition of clay fraction and by scanning electron microscopy (SEM) for the study of their microstructure. Their basic physical characteristics (grain-size distribution, e, WL, WP, IP, GS, Yd, n) were also determined. The coefficient of permeability (or hydraulic conductivity, k) was measured by the falling head method and the values obtained range between 1.66 10~8 and 1.06 10"6 cm/s, with a few exceptions. Our results indicate that the occurrence of swelling clay minerals in these cohesive marly sediments influences the value of hydraulic conductivity. Because of the double-layer effect not all the pore space contributes to seepage. Furthermore, the aggregation or flocculation of clay minerals and other microstructural characteristics related to the packing of structural constituents (forming an open or tight microstructure), the shape and the distribution of micropores, and the cementation degree of the microstructure are influence factors that affect the value of k. Predicting k, using empirical functions reported by several researchers for cohesive materials, or from simple correlations, as this of k versus clay fraction which is reported in this paper, is not absolutely safe, especially for cemented sediments as marls. Physicochemical factors, as the above-mentioned, play a prevalent role on the hydraulic conductivity value and they cannot be quantified and accounted in existing models.