A sand–bentonite mix was used to form the liner of four lagoons for the La Tuque and Haute-Mauricie sewage water treatment project. This paper describes the site conditions, design, and construction of the lagoons. Since the results of the control tests on the sand–bentonite were satisfactory, it was expected that the liner would be as impervious as it was designed to be. Such control tests, however, provide local values that can give only an indirect prediction of the waterproofing of each lagoon. The global performance was evaluated in 1988, after filling the lagoons, by comparing their leakage rates with the design theoretical values. The performance of two of the lagoons was satisfactory, whereas the two other lagoons lost 10 times more water than the maximum leakage rate allowed in the specifications. An analysis of the various factors influencing the global performance is developed to explain these contradictory performances, which cannot be explained by the usual deterministic method used in design. The paper demonstrates how to analyze the graph of the total leakage rate of a lagoon to determine where the main hydraulic defects are located and whether or not they are deteriorating. Several techniques for repairing the lagoons were considered and are described. The technique selected was the underwater injection of a bentonite slurry of adequate viscosity. Repairs were successfully completed during the summer of 1989. Key words : impervious liner, soil, bentonite, permeability, case history.