Eighteen multiparous lactating grazing Holstein cows, 9 ruminally cannulated, average 136.1 +/- 14.6 days in milk, were randomly assigned to three treatments consisting of water containing different levels of total dissolved solids (TDS; mg/l): Treatment 1 = 1,000; Treatment 2 = 5,000 and Treatment 3 = 10,000, at the Experimental Dairy Unit at Rafaela Experimental Station (31 degrees 11'S latitude) during summer 2005. Animals were arranged in a randomized complete block design with three 28-day experimental periods, with 3 weeks for water adaptation and 1 week for measurements. Feed and water intake, milk production and composition, body weight and condition score and rumen parameters were evaluated. No treatment effects were observed in any of the variables evaluated, with the exception of water intake, which was higher for animals receiving 10,000 mg/l TDS in the drinking water (189 l/day vs. 106 and 122 l/day for cows receiving water with 1,000 and 5,000 mg/l TDS, respectively). Water intake was significantly higher for animals in treatment 10,000 (P < 0.05). It was concluded that the rumen presents a surprising buffer capacity and that consideration of TDS alone is insufficient to characterize drinking water quality.