Colloid transport through the vadose zone is of growing concern in recent years. Non-water-soluble contaminants can enter an aquifer very quickly (colloidal facilitated transport) or colloids can be pathogens itself (for example Cryptosporidium parvum), thus yielding the risk of polluting drinking water. Little is known about the complex mechanisms of transport and retention of colloids at the pore scale in soils. Measurements of colloid and microbial transport have been limited to the evaluation of breakthrough curves from column experiments in which only an integrated signal of all retention processes in the column is obtained or to the visualization in micromodels with limited applicability to realistic conditions. The objective in this presentation is to observe and model colloid transport and retention on the pore scale. Flow experiments were run in a horizontal flow chamber using clean quartz sand as porous medium and synthetic fluorescent microspheres or bacteria (Escherichia coli, containing a green florescent protein) as colloids. The water phase was stained with Rhodamine B. In order to establish unsaturated conditions, porous plates were mounted at the inlet and outlet of the chamber. The chamber was mounted under a Laser Scanning Confocal Microscope (Leica TCS SP2) which allows the acquisition of time series and 3D reconstruction of pore-scale images. Three spectral channels were used: a 488 nm (argon) line excites the colloid fluorescence, a 543 nm green (HeNe) line excites Rhodamine B fluorescence, and a transmitted light channel detects the reflectance of laser light at the grain surfaces. Thus, three 8 bit images were detected simultaneously for every time step. The system is also capable of obtaining image stacks in the z-direction, which allows the determination of the position of attached colloids relative to the interface between air, water menisci, and solid grains. The 3D z-stacks reveal that the colloids are attaching at the air/water meniscus/solid (AWmS) interface, where the water menisci diminish into a thin film covering the grains. Methods of digital image analysis are presented for quantification of the number and area of moving and retained colloids. After thresholding, binary images are obtained. Colloids that appear at the exact same position in two consecutive images are counted as attached. The results show that once the first colloid is attached at the AWmS interface, the attachment rate increases until the number of locations where the colloids can be attached near other colloids becomes limiting. The attachment continues until there is no space for the colloids to attach anymore. A theoretical model is presented that is capable of predicting the observed colloid attachment processes. Forces acting on the colloids are discussed.