Saturated groundwater flow and tracer experiments using fluorescent dye, chloride, and the herbicides mecoprop and simazine were carried out in the laboratory using three large-diameter (0.5 m) undisturbed columns of fractured clayey till. Hydraulic conductivity of the columns ranged from 10-5 m/s in the shallowest column (1 m depth) to 10-7 m/s in the deepest column (4 m depth) and were similar to field-measured values for these deposits. Results of the tracer experiments are consistent with a conceptual model of advective transport along the fractures combined with diffusion into the fine-grained matrix between the fractures. Arrival of the chloride tracer in the effluent was highly retarded relative to fracture flow velocities calculated on the basis of the cubic law and measured values of fracture spacing and hydraulic conductivity. The herbicides were more strongly retarded than the chloride at low flow rates, but at higher flow rates the herbicides arrived with the chloride, indicating the influence of nonequilibrium sorption of the herbicides to fracture walls and the matrix solids. The columns were dismantled following the tracer experiments and mapping under UV light showed that nearly all of the visible, weathered fractures (and the few root holes in the case of the shallowest sample) were active transport pathways, with the dye appearing mainly on the fracture surfaces and as a rim in the adjacent matrix. Concentration profiles measured perpendicular to the fracture surfaces showed that the herbicides had also moved into the matrix, apparently by diffusion. Simulations of solute transport with a discrete fracture flow/matrix diffusion model showed that the simulations could be fit to the data if all of the visible fractures were hydraulically active, but could not be fit if all or most of the flow was channelled through just the primary fractures (defined by prominent oxidation stains). Simulations with an equivalent porous media (EPM) model could not fit the data using the measured total porosity as the effective porosity. The simulations could likely be fit with a smaller value of effective porosity, but this would limit applicability to field situations because fitted effective porosity is expected to change with physical scale and residence time of the solute in the soil. ¿ 1998 American Geophysical Union