Artificial capillary barriers are being used to divert water away from sensitive underground regions. Conversely, funneled flow over natural capillary barriers may increase the danger of groundwater contamination by decreasing the travel time and contact area. There have been relatively few experimental studies of capillary barrier flow patterns. In this study, water was applied uniformly across the top surface of a backlit tilting chamber, 1 cm thick, 110 cm high, and 180 cm long, in which a coarse sand layer was imbedded in a fine sand. Bedding slope and water application rates were varied between 0¿ and 12¿ and 1 and 3 cm h-1, respectively. After attaining steady state, matric potential was measured along the textural interface, and photos of dye traces were taken in order to visualize streamlines. The funneled flow was characterized by three discrete regions: an initial capillary diversion, a breakthrough region, and a toe diversion. The breakthrough region consisted of a significant zone of partial breakthrough where the vertical flux into the coarse layer was less than the water application rate. The lateral distance of the capillary diversion was explained well by previously published relationships when the water entry value at the textural interface was replaced by lower, observed matric potential at which breakthrough occurred at the most upslope point. The length of the capillary diversion was overpredicted using the air entry value. Finally, the toe of the coarse layer had significant, observed effects on funneled flow patterns, which have previously received little, if any, attention. The results of this study imply that the slope of the coarse layer and infiltration rate will largely govern the effectiveness of capillary barriers and that capillary barriers are less effective than previously assumed. ¿ 2000 American Geophysical Union