The aim of this study was to evaluate the spatial and morphological configuration of the pore space in 3D images of Latosols with different mineralogy under a conservation tillage system in a coffee crop area. The visualisation and quantification of pore size distribution by data mining and spatial variability by semi-variogams were investigated in 3D images with 60-[micro]m spatial resolution generated by X-ray CT scan (EVS/GE MS8x-130) in soil core samples collected at different depths of a kaolinitic Red-Yellow Latosol (RYL) and a gibbsitic Red Latosol (RL) from Brazil. Greater spatial variability occurred in the horizontal direction of the 3D image, a novel finding in this area of research. The pores detected were different between the Latosols studied, mainly at 0.20-0.34 m depth. The largest number ( 4000) and volume ([+ or -] 30 [mm.sup.3]) of pores was found in the RL. The soil classes differed in 3D pore characteristics, and this aspect may be important in the characterisation of causes of pore variability. Sphericity was similar for both soils, with greater emphasis on pore classes with a diameter 15%). Additional keywords: Brazil, CT scan, coffee crop, geostatistics, regression models, soil pore configuration.